Autoconf

Table of Contents


Next: , Up: (dir)

Autoconf

This manual is for GNU Autoconf (version 2.60, 23 June 2006), a package for creating scripts to configure source code packages using templates and an M4 macro package.

Copyright © 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, with the Front-Cover texts being “A GNU Manual,” and with the Back-Cover Texts as in (a) below. A copy of the license is included in the section entitled “GNU Free Documentation License.”

(a) The FSF's Back-Cover Text is: “You have freedom to copy and modify this GNU Manual, like GNU software. Copies published by the Free Software Foundation raise funds for GNU development.”

--- The Detailed Node Listing ---

The GNU Build System

Making configure Scripts

Writing configure.ac

Initialization and Output Files

Substitutions in Makefiles

Configuration Header Files

Existing Tests

Common Behavior

Alternative Programs

Library Functions

Header Files

Declarations

Structures

Types

Compilers and Preprocessors

Writing Tests

Writing Test Programs

Results of Tests

Caching Results

Programming in M4

M4 Quotation

Using autom4te

Programming in M4sugar

Writing Autoconf Macros

Dependencies Between Macros

Portable Shell Programming

Portable Make Programming

VPATH and Make

Portable C and C++ Programming

Manual Configuration

Site Configuration

Transforming Program Names When Installing

Running configure Scripts

Obsolete Constructs

Upgrading From Version 1

Upgrading From Version 2.13

Generating Test Suites with Autotest

Using an Autotest Test Suite

Frequent Autoconf Questions, with answers

History of Autoconf

Copying This Manual

Indices


Next: , Previous: Top, Up: Top

1 Introduction

A physicist, an engineer, and a computer scientist were discussing the

nature of God. “Surely a Physicist,” said the physicist, “because

early in the Creation, God made Light; and you know, Maxwell's

equations, the dual nature of electromagnetic waves, the relativistic

consequences...” “An Engineer!,” said the engineer, “because

before making Light, God split the Chaos into Land and Water; it takes a

hell of an engineer to handle that big amount of mud, and orderly

separation of solids from liquids...” The computer scientist

shouted: “And the Chaos, where do you think it was coming from, hmm?”

—Anonymous

Autoconf is a tool for producing shell scripts that automatically configure software source code packages to adapt to many kinds of Posix-like systems. The configuration scripts produced by Autoconf are independent of Autoconf when they are run, so their users do not need to have Autoconf.

The configuration scripts produced by Autoconf require no manual user intervention when run; they do not normally even need an argument specifying the system type. Instead, they individually test for the presence of each feature that the software package they are for might need. (Before each check, they print a one-line message stating what they are checking for, so the user doesn't get too bored while waiting for the script to finish.) As a result, they deal well with systems that are hybrids or customized from the more common Posix variants. There is no need to maintain files that list the features supported by each release of each variant of Posix.

For each software package that Autoconf is used with, it creates a configuration script from a template file that lists the system features that the package needs or can use. After the shell code to recognize and respond to a system feature has been written, Autoconf allows it to be shared by many software packages that can use (or need) that feature. If it later turns out that the shell code needs adjustment for some reason, it needs to be changed in only one place; all of the configuration scripts can be regenerated automatically to take advantage of the updated code.

The Metaconfig package is similar in purpose to Autoconf, but the scripts it produces require manual user intervention, which is quite inconvenient when configuring large source trees. Unlike Metaconfig scripts, Autoconf scripts can support cross-compiling, if some care is taken in writing them.

Autoconf does not solve all problems related to making portable software packages—for a more complete solution, it should be used in concert with other GNU build tools like Automake and Libtool. These other tools take on jobs like the creation of a portable, recursive makefile with all of the standard targets, linking of shared libraries, and so on. See The GNU Build System, for more information.

Autoconf imposes some restrictions on the names of macros used with #if in C programs (see Preprocessor Symbol Index).

Autoconf requires GNU M4 in order to generate the scripts. It uses features that some versions of M4, including GNU M4 1.3, do not have. You should use version 1.4.4 or later of GNU M4.

See Autoconf 1, for information about upgrading from version 1. See History, for the story of Autoconf's development. See FAQ, for answers to some common questions about Autoconf.

See the Autoconf web page for up-to-date information, details on the mailing lists, pointers to a list of known bugs, etc.

Mail suggestions to the Autoconf mailing list. Past suggestions are archived.

Mail bug reports to the Autoconf Bugs mailing list. Past bug reports are archived.

If possible, first check that your bug is not already solved in current development versions, and that it has not been reported yet. Be sure to include all the needed information and a short configure.ac that demonstrates the problem.

Autoconf's development tree is accessible via anonymous CVS; see the Autoconf Summary for details. Patches relative to the current CVS version can be sent for review to the Autoconf Patches mailing list. Past patches are archived.

Because of its mission, the Autoconf package itself includes only a set of often-used macros that have already demonstrated their usefulness. Nevertheless, if you wish to share your macros, or find existing ones, see the Autoconf Macro Archive, which is kindly run by Peter Simons.


Next: , Previous: Introduction, Up: Top

2 The GNU Build System

Autoconf solves an important problem—reliable discovery of system-specific build and runtime information—but this is only one piece of the puzzle for the development of portable software. To this end, the GNU project has developed a suite of integrated utilities to finish the job Autoconf started: the GNU build system, whose most important components are Autoconf, Automake, and Libtool. In this chapter, we introduce you to those tools, point you to sources of more information, and try to convince you to use the entire GNU build system for your software.


Next: , Up: The GNU Build System

2.1 Automake

The ubiquity of make means that a makefile is almost the only viable way to distribute automatic build rules for software, but one quickly runs into its numerous limitations. Its lack of support for automatic dependency tracking, recursive builds in subdirectories, reliable timestamps (e.g., for network file systems), and so on, mean that developers must painfully (and often incorrectly) reinvent the wheel for each project. Portability is non-trivial, thanks to the quirks of make on many systems. On top of all this is the manual labor required to implement the many standard targets that users have come to expect (make install, make distclean, make uninstall, etc.). Since you are, of course, using Autoconf, you also have to insert repetitive code in your Makefile.in to recognize @CC@, @CFLAGS@, and other substitutions provided by configure. Into this mess steps Automake. Automake allows you to specify your build needs in a Makefile.am file with a vastly simpler and more powerful syntax than that of a plain makefile, and then generates a portable Makefile.in for use with Autoconf. For example, the Makefile.am to build and install a simple “Hello world” program might look like:

     bin_PROGRAMS = hello
     hello_SOURCES = hello.c

The resulting Makefile.in (~400 lines) automatically supports all the standard targets, the substitutions provided by Autoconf, automatic dependency tracking, VPATH building, and so on. make builds the hello program, and make install installs it in /usr/local/bin (or whatever prefix was given to configure, if not /usr/local).

The benefits of Automake increase for larger packages (especially ones with subdirectories), but even for small programs the added convenience and portability can be substantial. And that's not all...


Next: , Previous: Automake, Up: The GNU Build System

2.2 Gnulib

GNU software has a well-deserved reputation for running on many different types of systems. While our primary goal is to write software for the GNU system, many users and developers have been introduced to us through the systems that they were already using.

Gnulib is a central location for common GNU code, intended to be shared among free software packages. Its components are typically shared at the source level, rather than being a library that gets built, installed, and linked against. The idea is to copy files from Gnulib into your own source tree. There is no distribution tarball; developers should just grab source modules from the repository. The source files are available online, under various licenses, mostly GNU GPL or GNU LGPL.

Gnulib modules typically contain C source code along with Autoconf macros used to configure the source code. For example, the Gnulib stdbool module implements a stdbool.h header that nearly conforms to C99, even on old-fashioned hosts that lack stdbool.h. This module contains a source file for the replacement header, along with an Autoconf macro that arranges to use the replacement header on old-fashioned systems.


Next: , Previous: Gnulib, Up: The GNU Build System

2.3 Libtool

Often, one wants to build not only programs, but libraries, so that other programs can benefit from the fruits of your labor. Ideally, one would like to produce shared (dynamically linked) libraries, which can be used by multiple programs without duplication on disk or in memory and can be updated independently of the linked programs. Producing shared libraries portably, however, is the stuff of nightmares—each system has its own incompatible tools, compiler flags, and magic incantations. Fortunately, GNU provides a solution: Libtool. Libtool handles all the requirements of building shared libraries for you, and at this time seems to be the only way to do so with any portability. It also handles many other headaches, such as: the interaction of Make rules with the variable suffixes of shared libraries, linking reliably with shared libraries before they are installed by the superuser, and supplying a consistent versioning system (so that different versions of a library can be installed or upgraded without breaking binary compatibility). Although Libtool, like Autoconf, can be used without Automake, it is most simply utilized in conjunction with Automake—there, Libtool is used automatically whenever shared libraries are needed, and you need not know its syntax.


Previous: Libtool, Up: The GNU Build System

2.4 Pointers

Developers who are used to the simplicity of make for small projects on a single system might be daunted at the prospect of learning to use Automake and Autoconf. As your software is distributed to more and more users, however, you otherwise quickly find yourself putting lots of effort into reinventing the services that the GNU build tools provide, and making the same mistakes that they once made and overcame. (Besides, since you're already learning Autoconf, Automake is a piece of cake.)

There are a number of places that you can go to for more information on the GNU build tools.


Next: , Previous: The GNU Build System, Up: Top

3 Making configure Scripts

The configuration scripts that Autoconf produces are by convention called configure. When run, configure creates several files, replacing configuration parameters in them with appropriate values. The files that configure creates are:

To create a configure script with Autoconf, you need to write an Autoconf input file configure.ac (or configure.in) and run autoconf on it. If you write your own feature tests to supplement those that come with Autoconf, you might also write files called aclocal.m4 and acsite.m4. If you use a C header file to contain #define directives, you might also run autoheader, and you can distribute the generated file config.h.in with the package.

Here is a diagram showing how the files that can be used in configuration are produced. Programs that are executed are suffixed by ‘*’. Optional files are enclosed in square brackets (‘[]’). autoconf and autoheader also read the installed Autoconf macro files (by reading autoconf.m4).

Files used in preparing a software package for distribution:

     your source files --> [autoscan*] --> [configure.scan] --> configure.ac
     
     configure.ac --.
                    |   .------> autoconf* -----> configure
     [aclocal.m4] --+---+
                    |   `-----> [autoheader*] --> [config.h.in]
     [acsite.m4] ---'
     
     Makefile.in -------------------------------> Makefile.in

Files used in configuring a software package:

                            .-------------> [config.cache]
     configure* ------------+-------------> config.log
                            |
     [config.h.in] -.       v            .-> [config.h] -.
                    +--> config.status* -+               +--> make*
     Makefile.in ---'                    `-> Makefile ---'


Next: , Up: Making configure Scripts

3.1 Writing configure.ac

To produce a configure script for a software package, create a file called configure.ac that contains invocations of the Autoconf macros that test the system features your package needs or can use. Autoconf macros already exist to check for many features; see Existing Tests, for their descriptions. For most other features, you can use Autoconf template macros to produce custom checks; see Writing Tests, for information about them. For especially tricky or specialized features, configure.ac might need to contain some hand-crafted shell commands; see Portable Shell. The autoscan program can give you a good start in writing configure.ac (see autoscan Invocation, for more information).

Previous versions of Autoconf promoted the name configure.in, which is somewhat ambiguous (the tool needed to process this file is not described by its extension), and introduces a slight confusion with config.h.in and so on (for which ‘.in’ means “to be processed by configure”). Using configure.ac is now preferred.


Next: , Up: Writing configure.ac

3.1.1 A Shell Script Compiler

Just as for any other computer language, in order to properly program configure.ac in Autoconf you must understand what problem the language tries to address and how it does so.

The problem Autoconf addresses is that the world is a mess. After all, you are using Autoconf in order to have your package compile easily on all sorts of different systems, some of them being extremely hostile. Autoconf itself bears the price for these differences: configure must run on all those systems, and thus configure must limit itself to their lowest common denominator of features.

Naturally, you might then think of shell scripts; who needs autoconf? A set of properly written shell functions is enough to make it easy to write configure scripts by hand. Sigh! Unfortunately, shell functions do not belong to the least common denominator; therefore, where you would like to define a function and use it ten times, you would instead need to copy its body ten times.

So, what is really needed is some kind of compiler, autoconf, that takes an Autoconf program, configure.ac, and transforms it into a portable shell script, configure.

How does autoconf perform this task?

There are two obvious possibilities: creating a brand new language or extending an existing one. The former option is attractive: all sorts of optimizations could easily be implemented in the compiler and many rigorous checks could be performed on the Autoconf program (e.g., rejecting any non-portable construct). Alternatively, you can extend an existing language, such as the sh (Bourne shell) language.

Autoconf does the latter: it is a layer on top of sh. It was therefore most convenient to implement autoconf as a macro expander: a program that repeatedly performs macro expansions on text input, replacing macro calls with macro bodies and producing a pure sh script in the end. Instead of implementing a dedicated Autoconf macro expander, it is natural to use an existing general-purpose macro language, such as M4, and implement the extensions as a set of M4 macros.


Next: , Previous: Shell Script Compiler, Up: Writing configure.ac

3.1.2 The Autoconf Language

The Autoconf language differs from many other computer languages because it treats actual code the same as plain text. Whereas in C, for instance, data and instructions have different syntactic status, in Autoconf their status is rigorously the same. Therefore, we need a means to distinguish literal strings from text to be expanded: quotation.

When calling macros that take arguments, there must not be any white space between the macro name and the open parenthesis. Arguments should be enclosed within the M4 quote characters ‘[’ and ‘]’, and be separated by commas. Any leading blanks or newlines in arguments are ignored, unless they are quoted. You should always quote an argument that might contain a macro name, comma, parenthesis, or a leading blank or newline. This rule applies recursively for every macro call, including macros called from other macros.

For instance:

     AC_CHECK_HEADER([stdio.h],
                     [AC_DEFINE([HAVE_STDIO_H], [1],
                        [Define to 1 if you have <stdio.h>.])],
                     [AC_MSG_ERROR([Sorry, can't do anything for you])])

is quoted properly. You may safely simplify its quotation to:

     AC_CHECK_HEADER([stdio.h],
                     [AC_DEFINE([HAVE_STDIO_H], 1,
                        [Define to 1 if you have <stdio.h>.])],
                     [AC_MSG_ERROR([Sorry, can't do anything for you])])

because ‘1’ cannot contain a macro call. Here, the argument of AC_MSG_ERROR must be quoted; otherwise, its comma would be interpreted as an argument separator. Also, the second and third arguments of ‘AC_CHECK_HEADER’ must be quoted, since they contain macro calls. The three arguments ‘HAVE_STDIO_H’, ‘stdio.h’, and ‘Define to 1 if you have <stdio.h>.’ do not need quoting, but if you unwisely defined a macro with a name like ‘Define’ or ‘stdio’ then they would need quoting. Cautious Autoconf users would keep the quotes, but many Autoconf users find such precautions annoying, and would rewrite the example as follows:

     AC_CHECK_HEADER(stdio.h,
                     [AC_DEFINE(HAVE_STDIO_H, 1,
                        [Define to 1 if you have <stdio.h>.])],
                     [AC_MSG_ERROR([Sorry, can't do anything for you])])

This is safe, so long as you adopt good naming conventions and do not define macros with names like ‘HAVE_STDIO_H’, ‘stdio’, or ‘h’. Though it is also safe here to omit the quotes around ‘Define to 1 if you have <stdio.h>.’ this is not recommended, as message strings are more likely to inadvertently contain commas.

The following example is wrong and dangerous, as it is underquoted:

     AC_CHECK_HEADER(stdio.h,
                     AC_DEFINE(HAVE_STDIO_H, 1,
                        Define to 1 if you have <stdio.h>.),
                     AC_MSG_ERROR([Sorry, can't do anything for you]))

In other cases, you may have to use text that also resembles a macro call. You must quote that text even when it is not passed as a macro argument:

     echo "Hard rock was here!  --[AC_DC]"

which results in:

     echo "Hard rock was here!  --AC_DC"

When you use the same text in a macro argument, you must therefore have an extra quotation level (since one is stripped away by the macro substitution). In general, then, it is a good idea to use double quoting for all literal string arguments:

     AC_MSG_WARN([[AC_DC stinks  --Iron Maiden]])

You are now able to understand one of the constructs of Autoconf that has been continually misunderstood... The rule of thumb is that whenever you expect macro expansion, expect quote expansion; i.e., expect one level of quotes to be lost. For instance:

     AC_COMPILE_IFELSE([char b[10];], [], [AC_MSG_ERROR([you lose])])

is incorrect: here, the first argument of AC_COMPILE_IFELSE is ‘char b[10];’ and is expanded once, which results in ‘char b10;’. (There was an idiom common in Autoconf's past to address this issue via the M4 changequote primitive, but do not use it!) Let's take a closer look: the author meant the first argument to be understood as a literal, and therefore it must be quoted twice:

     AC_COMPILE_IFELSE([[char b[10];]], [], [AC_MSG_ERROR([you lose])])

Voilà, you actually produce ‘char b[10];’ this time!

On the other hand, descriptions (e.g., the last parameter of AC_DEFINE or AS_HELP_STRING) are not literals—they are subject to line breaking, for example—and should not be double quoted. Even if these descriptions are short and are not actually broken, double quoting them yields weird results.

Some macros take optional arguments, which this documentation represents as [arg] (not to be confused with the quote characters). You may just leave them empty, or use ‘[]’ to make the emptiness of the argument explicit, or you may simply omit the trailing commas. The three lines below are equivalent:

     AC_CHECK_HEADERS([stdio.h], [], [], [])
     AC_CHECK_HEADERS([stdio.h],,,)
     AC_CHECK_HEADERS([stdio.h])

It is best to put each macro call on its own line in configure.ac. Most of the macros don't add extra newlines; they rely on the newline after the macro call to terminate the commands. This approach makes the generated configure script a little easier to read by not inserting lots of blank lines. It is generally safe to set shell variables on the same line as a macro call, because the shell allows assignments without intervening newlines.

You can include comments in configure.ac files by starting them with the ‘#’. For example, it is helpful to begin configure.ac files with a line like this:

     # Process this file with autoconf to produce a configure script.


Previous: Autoconf Language, Up: Writing configure.ac

3.1.3 Standard configure.ac Layout

The order in which configure.ac calls the Autoconf macros is not important, with a few exceptions. Every configure.ac must contain a call to AC_INIT before the checks, and a call to AC_OUTPUT at the end (see Output). Additionally, some macros rely on other macros having been called first, because they check previously set values of some variables to decide what to do. These macros are noted in the individual descriptions (see Existing Tests), and they also warn you when configure is created if they are called out of order.

To encourage consistency, here is a suggested order for calling the Autoconf macros. Generally speaking, the things near the end of this list are those that could depend on things earlier in it. For example, library functions could be affected by types and libraries.

     Autoconf requirements
     AC_INIT(package, version, bug-report-address)
     information on the package
     checks for programs
     checks for libraries
     checks for header files
     checks for types
     checks for structures
     checks for compiler characteristics
     checks for library functions
     checks for system services
     AC_CONFIG_FILES([file...])
     AC_OUTPUT


Next: , Previous: Writing configure.ac, Up: Making configure Scripts

3.2 Using autoscan to Create configure.ac

The autoscan program can help you create and/or maintain a configure.ac file for a software package. autoscan examines source files in the directory tree rooted at a directory given as a command line argument, or the current directory if none is given. It searches the source files for common portability problems and creates a file configure.scan which is a preliminary configure.ac for that package, and checks a possibly existing configure.ac for completeness.

When using autoscan to create a configure.ac, you should manually examine configure.scan before renaming it to configure.ac; it probably needs some adjustments. Occasionally, autoscan outputs a macro in the wrong order relative to another macro, so that autoconf produces a warning; you need to move such macros manually. Also, if you want the package to use a configuration header file, you must add a call to AC_CONFIG_HEADERS (see Configuration Headers). You might also have to change or add some #if directives to your program in order to make it work with Autoconf (see ifnames Invocation, for information about a program that can help with that job).

When using autoscan to maintain a configure.ac, simply consider adding its suggestions. The file autoscan.log contains detailed information on why a macro is requested.

autoscan uses several data files (installed along with Autoconf) to determine which macros to output when it finds particular symbols in a package's source files. These data files all have the same format: each line consists of a symbol, one or more blanks, and the Autoconf macro to output if that symbol is encountered. Lines starting with ‘#’ are comments.

autoscan accepts the following options:

--help
-h
Print a summary of the command line options and exit.
--version
-V
Print the version number of Autoconf and exit.
--verbose
-v
Print the names of the files it examines and the potentially interesting symbols it finds in them. This output can be voluminous.
--include=dir
-I dir
Append dir to the include path. Multiple invocations accumulate.
--prepend-include=dir
-B dir
Prepend dir to the include path. Multiple invocations accumulate.


Next: , Previous: autoscan Invocation, Up: Making configure Scripts

3.3 Using ifnames to List Conditionals

ifnames can help you write configure.ac for a software package. It prints the identifiers that the package already uses in C preprocessor conditionals. If a package has already been set up to have some portability, ifnames can thus help you figure out what its configure needs to check for. It may help fill in some gaps in a configure.ac generated by autoscan (see autoscan Invocation).

ifnames scans all of the C source files named on the command line (or the standard input, if none are given) and writes to the standard output a sorted list of all the identifiers that appear in those files in #if, #elif, #ifdef, or #ifndef directives. It prints each identifier on a line, followed by a space-separated list of the files in which that identifier occurs.

ifnames accepts the following options:

--help
-h
Print a summary of the command line options and exit.
--version
-V
Print the version number of Autoconf and exit.


Next: , Previous: ifnames Invocation, Up: Making configure Scripts

3.4 Using autoconf to Create configure

To create configure from configure.ac, run the autoconf program with no arguments. autoconf processes configure.ac with the M4 macro processor, using the Autoconf macros. If you give autoconf an argument, it reads that file instead of configure.ac and writes the configuration script to the standard output instead of to configure. If you give autoconf the argument -, it reads from the standard input instead of configure.ac and writes the configuration script to the standard output.

The Autoconf macros are defined in several files. Some of the files are distributed with Autoconf; autoconf reads them first. Then it looks for the optional file acsite.m4 in the directory that contains the distributed Autoconf macro files, and for the optional file aclocal.m4 in the current directory. Those files can contain your site's or the package's own Autoconf macro definitions (see Writing Autoconf Macros, for more information). If a macro is defined in more than one of the files that autoconf reads, the last definition it reads overrides the earlier ones.

autoconf accepts the following options:

--help
-h
Print a summary of the command line options and exit.
--version
-V
Print the version number of Autoconf and exit.
--verbose
-v
Report processing steps.
--debug
-d
Don't remove the temporary files.
--force
-f
Remake configure even if newer than its input files.
--include=dir
-I dir
Append dir to the include path. Multiple invocations accumulate.
--prepend-include=dir
-B dir
Prepend dir to the include path. Multiple invocations accumulate.
--output=file
-o file
Save output (script or trace) to file. The file - stands for the standard output.
--warnings=category
-W category
Report the warnings related to category (which can actually be a comma separated list). See Reporting Messages, macro AC_DIAGNOSE, for a comprehensive list of categories. Special values include:
all
report all the warnings
none
report none
error
treats warnings as errors
no-category
disable warnings falling into category

Warnings about ‘syntax’ are enabled by default, and the environment variable WARNINGS, a comma separated list of categories, is honored as well. Passing -W category actually behaves as if you had passed --warnings=syntax,$WARNINGS,category. If you want to disable the defaults and WARNINGS, but (for example) enable the warnings about obsolete constructs, you would use -W none,obsolete.

Because autoconf uses autom4te behind the scenes, it displays a back trace for errors, but not for warnings; if you want them, just pass -W error. See autom4te Invocation, for some examples.

--trace=macro[:format]
-t macro[:format]
Do not create the configure script, but list the calls to macro according to the format. Multiple --trace arguments can be used to list several macros. Multiple --trace arguments for a single macro are not cumulative; instead, you should just make format as long as needed.

The format is a regular string, with newlines if desired, and several special escape codes. It defaults to ‘$f:$l:$n:$%’; see autom4te Invocation, for details on the format.

--initialization
-i
By default, --trace does not trace the initialization of the Autoconf macros (typically the AC_DEFUN definitions). This results in a noticeable speedup, but can be disabled by this option.

It is often necessary to check the content of a configure.ac file, but parsing it yourself is extremely fragile and error-prone. It is suggested that you rely upon --trace to scan configure.ac. For instance, to find the list of variables that are substituted, use:

     $ autoconf -t AC_SUBST
     configure.ac:2:AC_SUBST:ECHO_C
     configure.ac:2:AC_SUBST:ECHO_N
     configure.ac:2:AC_SUBST:ECHO_T
     More traces deleted

The example below highlights the difference between ‘$@’, ‘$*’, and ‘$%’.

     $ cat configure.ac
     AC_DEFINE(This, is, [an
     [example]])
     $ autoconf -t 'AC_DEFINE:@: $@
     *: $*
     %: $%'
     @: [This],[is],[an
     [example]]
     *: This,is,an
     [example]
     %: This:is:an [example]

The format gives you a lot of freedom:

     $ autoconf -t 'AC_SUBST:$$ac_subst{"$1"} = "$f:$l";'
     $ac_subst{"ECHO_C"} = "configure.ac:2";
     $ac_subst{"ECHO_N"} = "configure.ac:2";
     $ac_subst{"ECHO_T"} = "configure.ac:2";
     More traces deleted

A long separator can be used to improve the readability of complex structures, and to ease their parsing (for instance when no single character is suitable as a separator):

     $ autoconf -t 'AM_MISSING_PROG:${|:::::|}*'
     ACLOCAL|:::::|aclocal|:::::|$missing_dir
     AUTOCONF|:::::|autoconf|:::::|$missing_dir
     AUTOMAKE|:::::|automake|:::::|$missing_dir
     More traces deleted


Previous: autoconf Invocation, Up: Making configure Scripts

3.5 Using autoreconf to Update configure Scripts

Installing the various components of the GNU Build System can be tedious: running autopoint for Gettext, automake for Makefile.in etc. in each directory. It may be needed either because some tools such as automake have been updated on your system, or because some of the sources such as configure.ac have been updated, or finally, simply in order to install the GNU Build System in a fresh tree.

autoreconf runs autoconf, autoheader, aclocal, automake, libtoolize, and autopoint (when appropriate) repeatedly to update the GNU Build System in the specified directories and their subdirectories (see Subdirectories). By default, it only remakes those files that are older than their sources.

If you install a new version of some tool, you can make autoreconf remake all of the files by giving it the --force option.

See Automatic Remaking, for Make rules to automatically remake configure scripts when their source files change. That method handles the timestamps of configuration header templates properly, but does not pass --autoconf-dir=dir or --localdir=dir.

Gettext supplies the autopoint command to add translation infrastructure to a source package. If you use autopoint, your configure.ac should invoke both AM_GNU_GETTEXT and AM_GNU_GETTEXT_VERSION(gettext-version). See Invoking the autopoint Program, for further details.

autoreconf accepts the following options:

--help
-h
Print a summary of the command line options and exit.
--version
-V
Print the version number of Autoconf and exit.
--verbose
Print the name of each directory autoreconf examines and the commands it runs. If given two or more times, pass --verbose to subordinate tools that support it.
--debug
-d
Don't remove the temporary files.
--force
-f
Remake even configure scripts and configuration headers that are newer than their input files (configure.ac and, if present, aclocal.m4).
--install
-i
Install the missing auxiliary files in the package. By default, files are copied; this can be changed with --symlink.

If deemed appropriate, this option triggers calls to ‘automake --add-missing’, ‘libtoolize’, ‘autopoint’, etc.

--no-recursive
Do not rebuild files in subdirectories to configure (see Subdirectories, macro AC_CONFIG_SUBDIRS).
--symlink
-s
When used with --install, install symbolic links to the missing auxiliary files instead of copying them.
--make
-m
When the directories were configured, update the configuration by running ‘./config.status --recheck && ./config.status’, and then run ‘make’.
--include=dir
-I dir
Append dir to the include path. Multiple invocations accumulate. Passed on to autoconf and autoheader internally.
--prepend-include=dir
-B dir
Prepend dir to the include path. Multiple invocations accumulate. Passed on to autoconf and autoheader internally.
--warnings=category
-W category
Report the warnings related to category (which can actually be a comma separated list).
cross
related to cross compilation issues.
obsolete
report the uses of obsolete constructs.
portability
portability issues
syntax
dubious syntactic constructs.
all
report all the warnings
none
report none
error
treats warnings as errors
no-category
disable warnings falling into category

Warnings about ‘syntax’ are enabled by default, and the environment variable WARNINGS, a comma separated list of categories, is honored as well. Passing -W category actually behaves as if you had passed --warnings=syntax,$WARNINGS,category. If you want to disable the defaults and WARNINGS, but (for example) enable the warnings about obsolete constructs, you would use -W none,obsolete.

If you want autoreconf to pass flags that are not listed here on to aclocal, set ACLOCAL_AMFLAGS in your Makefile.am.


Next: , Previous: Making configure Scripts, Up: Top

4 Initialization and Output Files

Autoconf-generated configure scripts need some information about how to initialize, such as how to find the package's source files and about the output files to produce. The following sections describe the initialization and the creation of output files.


Next: , Up: Setup

4.1 Initializing configure

Every configure script must call AC_INIT before doing anything else. The only other required macro is AC_OUTPUT (see Output).

— Macro: AC_INIT (package, version, [bug-report], [tarname])

Process any command-line arguments and perform various initializations and verifications.

Set the name of the package and its version. These are typically used in --version support, including that of configure. The optional argument bug-report should be the email to which users should send bug reports. The package tarname differs from package: the latter designates the full package name (e.g., ‘GNU Autoconf’), while the former is meant for distribution tar ball names (e.g., ‘autoconf’). It defaults to package with ‘GNU ’ stripped, lower-cased, and all characters other than alphanumerics and underscores are changed to ‘-’.

It is preferable that the arguments of AC_INIT be static, i.e., there should not be any shell computation, but they can be computed by M4.

The following M4 macros (e.g., AC_PACKAGE_NAME), output variables (e.g., PACKAGE_NAME), and preprocessor symbols (e.g., PACKAGE_NAME) are defined by AC_INIT:

AC_PACKAGE_NAME, PACKAGE_NAME
Exactly package.
AC_PACKAGE_TARNAME, PACKAGE_TARNAME
Exactly tarname.
AC_PACKAGE_VERSION, PACKAGE_VERSION
Exactly version.
AC_PACKAGE_STRING, PACKAGE_STRING
Exactly ‘package version’.
AC_PACKAGE_BUGREPORT, PACKAGE_BUGREPORT
Exactly bug-report.

If your configure script does its own option processing, it should inspect ‘$@’ or ‘$*’ immediately after calling AC_INIT, because other Autoconf macros liberally use the set command to process strings, and this has the side effect of updating ‘$@’ and ‘$*’. However, we suggest that you use standard macros like AC_ARG_ENABLE instead of attempting to implement your own option processing. See Site Configuration.


Next: , Previous: Initializing configure, Up: Setup

4.2 Notices in configure

The following macros manage version numbers for configure scripts. Using them is optional.

— Macro: AC_PREREQ (version)

Ensure that a recent enough version of Autoconf is being used. If the version of Autoconf being used to create configure is earlier than version, print an error message to the standard error output and exit with failure (exit status is 63). For example:

          AC_PREREQ([2.60])
     

This macro is the only macro that may be used before AC_INIT, but for consistency, you are invited not to do so.

— Macro: AC_COPYRIGHT (copyright-notice)

State that, in addition to the Free Software Foundation's copyright on the Autoconf macros, parts of your configure are covered by the copyright-notice.

The copyright-notice shows up in both the head of configure and in ‘configure --version’.

— Macro: AC_REVISION (revision-info)

Copy revision stamp revision-info into the configure script, with any dollar signs or double-quotes removed. This macro lets you put a revision stamp from configure.ac into configure without RCS or CVS changing it when you check in configure. That way, you can determine easily which revision of configure.ac a particular configure corresponds to.

For example, this line in configure.ac:

          AC_REVISION([$Revision: 1.1 $])
     

produces this in configure:

          #!/bin/sh
          # From configure.ac Revision: 1.30
     


Next: , Previous: Notices, Up: Setup

4.3 Finding configure Input

— Macro: AC_CONFIG_SRCDIR (unique-file-in-source-dir)

unique-file-in-source-dir is some file that is in the package's source directory; configure checks for this file's existence to make sure that the directory that it is told contains the source code in fact does. Occasionally people accidentally specify the wrong directory with --srcdir; this is a safety check. See configure Invocation, for more information.

Packages that do manual configuration or use the install program might need to tell configure where to find some other shell scripts by calling AC_CONFIG_AUX_DIR, though the default places it looks are correct for most cases.

— Macro: AC_CONFIG_AUX_DIR (dir)

Use the auxiliary build tools (e.g., install-sh, config.sub, config.guess, Cygnus configure, Automake and Libtool scripts, etc.) that are in directory dir. These are auxiliary files used in configuration. dir can be either absolute or relative to srcdir. The default is srcdir or srcdir/.. or srcdir/../.., whichever is the first that contains install-sh. The other files are not checked for, so that using AC_PROG_INSTALL does not automatically require distributing the other auxiliary files. It checks for install.sh also, but that name is obsolete because some make have a rule that creates install from it if there is no makefile.

The auxiliary directory is commonly named build-aux. If you need portability to DOS variants, do not name the auxiliary directory aux. See File System Conventions.

— Macro: AC_REQUIRE_AUX_FILE (file)

Declares that file is expected in the directory defined above. In Autoconf proper, this macro does nothing: its sole purpose is to be traced by third-party tools to produce a list of expected auxiliary files. For instance it is called by macros like AC_PROG_INSTALL (see Particular Programs) or AC_CANONICAL_BUILD (see Canonicalizing) to register the auxiliary files they need.

Similarly, packages that use aclocal should declare where local macros can be found using AC_CONFIG_MACRO_DIR.

— Macro: AC_CONFIG_MACRO_DIR (dir)

Future versions of autopoint, libtoolize, aclocal and autoreconf will use directory dir as the location of additional local Autoconf macros. Be sure to call this macro directly from configure.ac so that tools that install macros for aclocal can find the declaration before --trace can be called safely.


Next: , Previous: Input, Up: Setup

4.4 Outputting Files

Every Autoconf script, e.g., configure.ac, should finish by calling AC_OUTPUT. That is the macro that generates and runs config.status, which in turn creates the makefiles and any other files resulting from configuration. This is the only required macro besides AC_INIT (see Input).

— Macro: AC_OUTPUT

Generate config.status and launch it. Call this macro once, at the end of configure.ac.

config.status performs all the configuration actions: all the output files (see Configuration Files, macro AC_CONFIG_FILES), header files (see Configuration Headers, macro AC_CONFIG_HEADERS), commands (see Configuration Commands, macro AC_CONFIG_COMMANDS), links (see Configuration Links, macro AC_CONFIG_LINKS), subdirectories to configure (see Subdirectories, macro AC_CONFIG_SUBDIRS) are honored.

The location of your AC_OUTPUT invocation is the exact point where configuration actions are taken: any code afterwards is executed by configure once config.status was run. If you want to bind actions to config.status itself (independently of whether configure is being run), see Running Arbitrary Configuration Commands.

Historically, the usage of AC_OUTPUT was somewhat different. See Obsolete Macros, for a description of the arguments that AC_OUTPUT used to support.

If you run make in subdirectories, you should run it using the make variable MAKE. Most versions of make set MAKE to the name of the make program plus any options it was given. (But many do not include in it the values of any variables set on the command line, so those are not passed on automatically.) Some old versions of make do not set this variable. The following macro allows you to use it even with those versions.

— Macro: AC_PROG_MAKE_SET

If the Make command, $MAKE if set or else ‘make’, predefines $(MAKE), define output variable SET_MAKE to be empty. Otherwise, define SET_MAKE to a macro definition that sets $(MAKE), such as ‘MAKE=make’. Calls AC_SUBST for SET_MAKE.

If you use this macro, place a line like this in each Makefile.in that runs MAKE on other directories:

     @SET_MAKE@


Next: , Previous: Output, Up: Setup

4.5 Performing Configuration Actions

configure is designed so that it appears to do everything itself, but there is actually a hidden slave: config.status. configure is in charge of examining your system, but it is config.status that actually takes the proper actions based on the results of configure. The most typical task of config.status is to instantiate files.

This section describes the common behavior of the four standard instantiating macros: AC_CONFIG_FILES, AC_CONFIG_HEADERS, AC_CONFIG_COMMANDS and AC_CONFIG_LINKS. They all have this prototype:

     AC_CONFIG_FOOS(tag..., [commands], [init-cmds])

where the arguments are:

tag...
A blank-or-newline-separated list of tags, which are typically the names of the files to instantiate.

You are encouraged to use literals as tags. In particular, you should avoid

          ... && my_foos="$my_foos fooo"
          ... && my_foos="$my_foos foooo"
          AC_CONFIG_FOOS([$my_foos])
     

and use this instead:

          ... && AC_CONFIG_FOOS([fooo])
          ... && AC_CONFIG_FOOS([foooo])
     

The macros AC_CONFIG_FILES and AC_CONFIG_HEADERS use special tag values: they may have the form ‘output’ or ‘output:inputs’. The file output is instantiated from its templates, inputs (defaulting to ‘output.in’).

AC_CONFIG_FILES([Makefile:boiler/top.mk:boiler/bot.mk)]’, for example, asks for the creation of the file Makefile that contains the expansion of the output variables in the concatenation of boiler/top.mk and boiler/bot.mk.

The special value ‘-’ might be used to denote the standard output when used in output, or the standard input when used in the inputs. You most probably don't need to use this in configure.ac, but it is convenient when using the command line interface of ./config.status, see config.status Invocation, for more details.

The inputs may be absolute or relative file names. In the latter case they are first looked for in the build tree, and then in the source tree.

commands
Shell commands output literally into config.status, and associated with a tag that the user can use to tell config.status which the commands to run. The commands are run each time a tag request is given to config.status, typically each time the file tag is created.

The variables set during the execution of configure are not available here: you first need to set them via the init-cmds. Nonetheless the following variables are precomputed:

srcdir
The name of the top source directory, assuming that the working directory is the top build directory. This is what the configure option --srcdir sets.
ac_top_srcdir
The name of the top source directory, assuming that the working directory is the current build directory.
ac_top_build_prefix
The name of the top build directory, assuming that the working directory is the current build directory. It can be empty, or else ends with a slash, so that you may concatenate it.
ac_srcdir
The name of the corresponding source directory, assuming that the working directory is the current build directory.

The current directory refers to the directory (or pseudo-directory) containing the input part of tags. For instance, running

          AC_CONFIG_COMMANDS([deep/dir/out:in/in.in], [...], [...])
     

with --srcdir=../package produces the following values:

          # Argument of --srcdir
          srcdir='../package'
          # Reversing deep/dir
          ac_top_build_prefix='../../'
          # Concatenation of $ac_top_build_prefix and srcdir
          ac_top_srcdir='../../../package'
          # Concatenation of $ac_top_srcdir and deep/dir
          ac_srcdir='../../../package/deep/dir'
     

independently of ‘in/in.in’.

init-cmds
Shell commands output unquoted near the beginning of config.status, and executed each time config.status runs (regardless of the tag). Because they are unquoted, for example, ‘$var’ is output as the value of var. init-cmds is typically used by configure to give config.status some variables it needs to run the commands.

You should be extremely cautious in your variable names: all the init-cmds share the same name space and may overwrite each other in unpredictable ways. Sorry...

All these macros can be called multiple times, with different tag values, of course!


Next: , Previous: Configuration Actions, Up: Setup

4.6 Creating Configuration Files

Be sure to read the previous section, Configuration Actions.

— Macro: AC_CONFIG_FILES (file..., [cmds], [init-cmds])

Make AC_OUTPUT create each file by copying an input file (by default file.in), substituting the output variable values. This macro is one of the instantiating macros; see Configuration Actions. See Makefile Substitutions, for more information on using output variables. See Setting Output Variables, for more information on creating them. This macro creates the directory that the file is in if it doesn't exist. Usually, makefiles are created this way, but other files, such as .gdbinit, can be specified as well.

Typical calls to AC_CONFIG_FILES look like this:

          AC_CONFIG_FILES([Makefile src/Makefile man/Makefile X/Imakefile])
          AC_CONFIG_FILES([autoconf], [chmod +x autoconf])
     

You can override an input file name by appending to file a colon-separated list of input files. Examples:

          AC_CONFIG_FILES([Makefile:boiler/top.mk:boiler/bot.mk]
                          [lib/Makefile:boiler/lib.mk])
     

Doing this allows you to keep your file names acceptable to DOS variants, or to prepend and/or append boilerplate to the file.


Next: , Previous: Configuration Files, Up: Setup

4.7 Substitutions in Makefiles

Each subdirectory in a distribution that contains something to be compiled or installed should come with a file Makefile.in, from which configure creates a file Makefile in that directory. To create Makefile, configure performs a simple variable substitution, replacing occurrences of ‘@variable@’ in Makefile.in with the value that configure has determined for that variable. Variables that are substituted into output files in this way are called output variables. They are ordinary shell variables that are set in configure. To make configure substitute a particular variable into the output files, the macro AC_SUBST must be called with that variable name as an argument. Any occurrences of ‘@variable@’ for other variables are left unchanged. See Setting Output Variables, for more information on creating output variables with AC_SUBST.

A software package that uses a configure script should be distributed with a file Makefile.in, but no makefile; that way, the user has to properly configure the package for the local system before compiling it.

See Makefile Conventions, for more information on what to put in makefiles.


Next: , Up: Makefile Substitutions

4.7.1 Preset Output Variables

Some output variables are preset by the Autoconf macros. Some of the Autoconf macros set additional output variables, which are mentioned in the descriptions for those macros. See Output Variable Index, for a complete list of output variables. See Installation Directory Variables, for the list of the preset ones related to installation directories. Below are listed the other preset ones. They all are precious variables (see Setting Output Variables, AC_ARG_VAR).

— Variable: CFLAGS

Debugging and optimization options for the C compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_CC (or empty if you don't). configure uses this variable when compiling programs to test for C features.

— Variable: configure_input

A comment saying that the file was generated automatically by configure and giving the name of the input file. AC_OUTPUT adds a comment line containing this variable to the top of every makefile it creates. For other files, you should reference this variable in a comment at the top of each input file. For example, an input shell script should begin like this:

          #!/bin/sh
          # @configure_input@
     

The presence of that line also reminds people editing the file that it needs to be processed by configure in order to be used.

— Variable: CPPFLAGS

Header file search directory (-Idir) and any other miscellaneous options for the C and C++ preprocessors and compilers. If it is not set in the environment when configure runs, the default value is empty. configure uses this variable when compiling or preprocessing programs to test for C and C++ features. See Special Chars in Variables, for limitations that CPPFLAGS might run into.

— Variable: CXXFLAGS

Debugging and optimization options for the C++ compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_CXX (or empty if you don't). configure uses this variable when compiling programs to test for C++ features.

— Variable: DEFS

-D options to pass to the C compiler. If AC_CONFIG_HEADERS is called, configure replaces ‘@DEFS@’ with -DHAVE_CONFIG_H instead (see Configuration Headers). This variable is not defined while configure is performing its tests, only when creating the output files. See Setting Output Variables, for how to check the results of previous tests.

— Variable: ECHO_C
— Variable: ECHO_N
— Variable: ECHO_T

How does one suppress the trailing newline from echo for question-answer message pairs? These variables provide a way:

          echo $ECHO_N "And the winner is... $ECHO_C"
          sleep 100000000000
          echo "${ECHO_T}dead."
     

Some old and uncommon echo implementations offer no means to achieve this, in which case ECHO_T is set to tab. You might not want to use it.

— Variable: ERLCFLAGS

Debugging and optimization options for the Erlang compiler. If it is not set in the environment when configure runs, the default value is empty. configure uses this variable when compiling programs to test for Erlang features.

— Variable: FCFLAGS

Debugging and optimization options for the Fortran compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_FC (or empty if you don't). configure uses this variable when compiling programs to test for Fortran features.

— Variable: FFLAGS

Debugging and optimization options for the Fortran 77 compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_F77 (or empty if you don't). configure uses this variable when compiling programs to test for Fortran 77 features.

— Variable: LDFLAGS

Stripping (-s), path (-L), and any other miscellaneous options for the linker. Don't use this variable to pass library names (-l) to the linker, use LIBS instead. If it is not set in the environment when configure runs, the default value is empty. configure uses this variable when linking programs to test for C, C++, and Fortran features.

— Variable: LIBS

-l options to pass to the linker. The default value is empty, but some Autoconf macros may prepend extra libraries to this variable if those libraries are found and provide necessary functions, see Libraries. configure uses this variable when linking programs to test for C, C++, and Fortran features.

— Variable: OBJCFLAGS

Debugging and optimization options for the Objective C compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_OBJC (or empty if you don't). configure uses this variable when compiling programs to test for Objective C features.

— Variable: builddir

Rigorously equal to ‘.’. Added for symmetry only.

— Variable: abs_builddir

Absolute name of builddir.

— Variable: top_builddir

The relative name of the top level of the current build tree. In the top-level directory, this is the same as builddir.

— Variable: abs_top_builddir

Absolute name of top_builddir.

— Variable: srcdir

The relative name of the directory that contains the source code for that makefile.

— Variable: abs_srcdir

Absolute name of srcdir.

— Variable: top_srcdir

The relative name of the top-level source code directory for the package. In the top-level directory, this is the same as srcdir.

— Variable: abs_top_srcdir

Absolute name of top_srcdir.


Next: , Previous: Preset Output Variables, Up: Makefile Substitutions

4.7.2 Installation Directory Variables

The following variables specify the directories for package installation, see Variables for Installation Directories, for more information. See the end of this section for details on when and how to use these variables.

— Variable: bindir

The directory for installing executables that users run.

— Variable: datadir

The directory for installing idiosyncratic read-only architecture-independent data.

— Variable: datarootdir

The root of the directory tree for read-only architecture-independent data files.

— Variable: docdir

The directory for installing documentation files (other than Info and man).

— Variable: dvidir

The directory for installing documentation files in DVI format.

— Variable: exec_prefix

The installation prefix for architecture-dependent files. By default it's the same as prefix. You should avoid installing anything directly to exec_prefix. However, the default value for directories containing architecture-dependent files should be relative to exec_prefix.

— Variable: htmldir

The directory for installing HTML documentation.

— Variable: includedir

The directory for installing C header files.

— Variable: infodir

The directory for installing documentation in Info format.

— Variable: libdir

The directory for installing object code libraries.

— Variable: libexecdir

The directory for installing executables that other programs run.

— Variable: localedir

The directory for installing locale-dependent but architecture-independent data, such as message catalogs. This directory usually has a subdirectory per locale.

— Variable: localstatedir

The directory for installing modifiable single-machine data.

— Variable: mandir

The top-level directory for installing documentation in man format.

— Variable: oldincludedir

The directory for installing C header files for non-GCC compilers.

— Variable: pdfdir

The directory for installing PDF documentation.

— Variable: prefix

The common installation prefix for all files. If exec_prefix is defined to a different value, prefix is used only for architecture-independent files.

— Variable: psdir

The directory for installing PostScript documentation.

— Variable: sbindir

The directory for installing executables that system administrators run.

— Variable: sharedstatedir

The directory for installing modifiable architecture-independent data.

— Variable: sysconfdir

The directory for installing read-only single-machine data.

Most of these variables have values that rely on prefix or exec_prefix. It is deliberate that the directory output variables keep them unexpanded: typically ‘@datarootdir@’ is replaced by ‘${prefix}/share’, not ‘/usr/local/share’, and ‘@datadir@’ is replaced by ‘${datarootdir}’.

This behavior is mandated by the GNU coding standards, so that when the user runs:

make
she can still specify a different prefix from the one specified to configure, in which case, if needed, the package should hard code dependencies corresponding to the make-specified prefix.
make install
she can specify a different installation location, in which case the package must still depend on the location which was compiled in (i.e., never recompile when ‘make install’ is run). This is an extremely important feature, as many people may decide to install all the files of a package grouped together, and then install links from the final locations to there.

In order to support these features, it is essential that datarootdir remains being defined as ‘${prefix}/share’ to depend upon the current value of prefix.

A corollary is that you should not use these variables except in makefiles. For instance, instead of trying to evaluate datadir in configure and hard-coding it in makefiles using e.g., ‘AC_DEFINE_UNQUOTED([DATADIR], ["$datadir"], [Data directory.])’, you should add -DDATADIR='$(datadir)' to your makefile's definition of CPPFLAGS (AM_CPPFLAGS if you are also using Automake).

Similarly, you should not rely on AC_CONFIG_FILES to replace datadir and friends in your shell scripts and other files; instead, let make manage their replacement. For instance Autoconf ships templates of its shell scripts ending with ‘.in’, and uses a makefile snippet similar to the following to build scripts like autoheader and autom4te:

     edit = sed \
             -e 's|@datadir[@]|$(pkgdatadir)|g' \
             -e 's|@prefix[@]|$(prefix)|g'
     
     autoheader autom4te: Makefile
             rm -f $@ $@.tmp
             $(edit) '$(srcdir)/$@.in' >$@.tmp
             chmod +x $@.tmp
             chmod a-w $@.tmp
             mv $@.tmp $@
     
     autoheader: $(srcdir)/autoheader.in
     autom4te: $(srcdir)/autom4te.in

Some details are noteworthy:

@datadir[@]
The brackets prevent configure from replacing ‘@datadir@’ in the Sed expression itself. Brackets are preferable to a backslash here, since Posix says ‘\@’ is not portable.
$(pkgdatadir)
Don't use ‘@pkgdatadir@’! Use the matching makefile variable instead.
/
Don't use ‘/’ in the Sed expressions that replace file names since most likely the variables you use, such as ‘$(pkgdatadir)’, contain ‘/’. Use a shell metacharacter instead, such as ‘|’.
special characters
File names, file name components, and the value of VPATH should not contain shell metacharacters or white space. See Special Chars in Variables.
dependency on Makefile
Since edit uses values that depend on the configuration specific values (prefix, etc.) and not only on VERSION and so forth, the output depends on Makefile, not configure.ac.
$@
The main rule is generic, and uses ‘$@’ extensively to avoid the need for multiple copies of the rule.
Separated dependencies and single suffix rules
You can't use them! The above snippet cannot be (portably) rewritten as:
          autoconf autoheader: Makefile
          .in:
                  rm -f $@ $@.tmp
                  $(edit) $< >$@.tmp
                  chmod +x $@.tmp
                  mv $@.tmp $@
     

See Single Suffix Rules, for details.

$(srcdir)
Be sure to specify the name of the source directory, otherwise the package won't support separated builds.

For the more specific installation of Erlang libraries, the following variables are defined:

— Variable: ERLANG_INSTALL_LIB_DIR

The common parent directory of Erlang library installation directories. This variable is set by calling the AC_ERLANG_SUBST_INSTALL_LIB_DIR macro in configure.ac.

— Variable: ERLANG_INSTALL_LIB_DIR_library

The installation directory for Erlang library library. This variable is set by calling the ‘AC_ERLANG_SUBST_INSTALL_LIB_SUBDIR(library, version’ macro in configure.ac.

See Erlang Libraries, for details.


Next: , Previous: Installation Directory Variables, Up: Makefile Substitutions

4.7.3 Changed Directory Variables

In Autoconf 2.60, the set of directory variables has changed, and the defaults of some variables have been adjusted (see Installation Directory Variables) to changes in the GNU Coding Standards. Notably, datadir, infodir, and mandir are now expressed in terms of datarootdir. If you are upgrading from an earlier Autoconf version, you may need to adjust your files to ensure that the directory variables are substituted correctly (see Defining Directories), and that a definition of datarootdir is in place. For example, in a Makefile.in, adding

     datarootdir = @datarootdir@

is usually sufficient. If you use Automake to create Makefile.in, it will add this for you.

To help with the transition, Autoconf warns about files that seem to use datarootdir without defining it. In some cases, it then expands the value of $datarootdir in substitutions of the directory variables. The following example shows such a warning:

     $ cat configure.ac
     AC_INIT
     AC_CONFIG_FILES([Makefile])
     AC_OUTPUT
     $ cat Makefile.in
     prefix = @prefix@
     datadir = @datadir@
     $ autoconf
     $ configure
     configure: creating ./config.status
     config.status: creating Makefile
     config.status: WARNING:
                    Makefile.in seems to ignore the --datarootdir setting
     $ cat Makefile
     prefix = /usr/local
     datadir = ${prefix}/share

Usually one can easily change the file to accommodate both older and newer Autoconf releases:

     $ cat Makefile.in
     prefix = @prefix@
     datarootdir = @datarootdir@
     datadir = @datadir@
     $ configure
     configure: creating ./config.status
     config.status: creating Makefile
     $ cat Makefile
     prefix = /usr/local
     datarootdir = ${prefix}/share
     datadir = ${datarootdir}

In some cases, however, the checks may not be able to detect that a suitable initialization of datarootdir is in place, or they may fail to detect that such an initialization is necessary in the output file. If, after auditing your package, there are still spurious configure warnings about datarootdir, you may add the line

     AC_DEFUN([AC_DATAROOTDIR_CHECKED])

to your configure.ac to disable the warnings. This is an exception to the usual rule that you should not define a macro whose name begins with AC_ (see Macro Names).


Next: , Previous: Changed Directory Variables, Up: Makefile Substitutions

4.7.4 Build Directories

You can support compiling a software package for several architectures simultaneously from the same copy of the source code. The object files for each architecture are kept in their own directory.

To support doing this, make uses the VPATH variable to find the files that are in the source directory. GNU Make and most other recent make programs can do this. Older make programs do not support VPATH; when using them, the source code must be in the same directory as the object files.

To support VPATH, each Makefile.in should contain two lines that look like:

     srcdir = @srcdir@
     VPATH = @srcdir@

Do not set VPATH to the value of another variable, for example ‘VPATH = $(srcdir)’, because some versions of make do not do variable substitutions on the value of VPATH.

configure substitutes the correct value for srcdir when it produces Makefile.

Do not use the make variable $<, which expands to the file name of the file in the source directory (found with VPATH), except in implicit rules. (An implicit rule is one such as ‘.c.o’, which tells how to create a .o file from a .c file.) Some versions of make do not set $< in explicit rules; they expand it to an empty value.

Instead, Make command lines should always refer to source files by prefixing them with ‘$(srcdir)/’. For example:

     time.info: time.texinfo
             $(MAKEINFO) '$(srcdir)/time.texinfo'


Previous: Build Directories, Up: Makefile Substitutions

4.7.5 Automatic Remaking

You can put rules like the following in the top-level Makefile.in for a package to automatically update the configuration information when you change the configuration files. This example includes all of the optional files, such as aclocal.m4 and those related to configuration header files. Omit from the Makefile.in rules for any of these files that your package does not use.

The ‘$(srcdir)/’ prefix is included because of limitations in the VPATH mechanism.

The stamp- files are necessary because the timestamps of config.h.in and config.h are not changed if remaking them does not change their contents. This feature avoids unnecessary recompilation. You should include the file stamp-h.in your package's distribution, so that make considers config.h.in up to date. Don't use touch (see Limitations of Usual Tools); instead, use echo (using date would cause needless differences, hence CVS conflicts, etc.).

     $(srcdir)/configure: configure.ac aclocal.m4
             cd '$(srcdir)' && autoconf
     
     # autoheader might not change config.h.in, so touch a stamp file.
     $(srcdir)/config.h.in: stamp-h.in
     $(srcdir)/stamp-h.in: configure.ac aclocal.m4
             cd '$(srcdir)' && autoheader
             echo timestamp > '$(srcdir)/stamp-h.in'
     
     config.h: stamp-h
     stamp-h: config.h.in config.status
             ./config.status
     
     Makefile: Makefile.in config.status
             ./config.status
     
     config.status: configure
             ./config.status --recheck

(Be careful if you copy these lines directly into your makefile, as you need to convert the indented lines to start with the tab character.)

In addition, you should use

     AC_CONFIG_FILES([stamp-h], [echo timestamp > stamp-h])

so config.status ensures that config.h is considered up to date. See Output, for more information about AC_OUTPUT.

See config.status Invocation, for more examples of handling configuration-related dependencies.


Next: , Previous: Makefile Substitutions, Up: Setup

4.8 Configuration Header Files

When a package contains more than a few tests that define C preprocessor symbols, the command lines to pass -D options to the compiler can get quite long. This causes two problems. One is that the make output is hard to visually scan for errors. More seriously, the command lines can exceed the length limits of some operating systems. As an alternative to passing -D options to the compiler, configure scripts can create a C header file containing ‘#define’ directives. The AC_CONFIG_HEADERS macro selects this kind of output. Though it can be called anywhere between AC_INIT and AC_OUTPUT, it is customary to call it right after AC_INIT.

The package should ‘#include’ the configuration header file before any other header files, to prevent inconsistencies in declarations (for example, if it redefines const).

To provide for VPATH builds, remember to pass the C compiler a -I. option (or -I..; whichever directory contains config.h). Even if you use ‘#include "config.h"’, the preprocessor searches only the directory of the currently read file, i.e., the source directory, not the build directory.

With the appropriate -I option, you can use ‘#include <config.h>’. Actually, it's a good habit to use it, because in the rare case when the source directory contains another config.h, the build directory should be searched first.

— Macro: AC_CONFIG_HEADERS (header ..., [cmds], [init-cmds])

This macro is one of the instantiating macros; see Configuration Actions. Make AC_OUTPUT create the file(s) in the blank-or-newline-separated list header containing C preprocessor #define statements, and replace ‘@DEFS@’ in generated files with -DHAVE_CONFIG_H instead of the value of DEFS. The usual name for header is config.h.

If header already exists and its contents are identical to what AC_OUTPUT would put in it, it is left alone. Doing this allows making some changes in the configuration without needlessly causing object files that depend on the header file to be recompiled.

Usually the input file is named header.in; however, you can override the input file name by appending to header a colon-separated list of input files. Examples:

          AC_CONFIG_HEADERS([config.h:config.hin])
          AC_CONFIG_HEADERS([defines.h:defs.pre:defines.h.in:defs.post])
     

Doing this allows you to keep your file names acceptable to DOS variants, or to prepend and/or append boilerplate to the file.

— Macro: AH_HEADER

This macro is defined as the name of the first declared config header and undefined if no config headers have been declared up to this point. A third-party macro may, for example, require use of a config header without invoking AC_CONFIG_HEADERS twice, like this:

          AC_CONFIG_COMMANDS_PRE(
                  [m4_ifndef([AH_HEADER], [AC_CONFIG_HEADERS([config.h])])])
     

See Configuration Actions, for more details on header.


Next: , Up: Configuration Headers

4.8.1 Configuration Header Templates

Your distribution should contain a template file that looks as you want the final header file to look, including comments, with #undef statements which are used as hooks. For example, suppose your configure.ac makes these calls:

     AC_CONFIG_HEADERS([conf.h])
     AC_CHECK_HEADERS([unistd.h])

Then you could have code like the following in conf.h.in. On systems that have unistd.h, configure defines ‘HAVE_UNISTD_H’ to 1. On other systems, the whole line is commented out (in case the system predefines that symbol).

     /* Define as 1 if you have unistd.h.  */
     #undef HAVE_UNISTD_H

Pay attention that ‘#undef’ is in the first column, and there is nothing after ‘HAVE_UNISTD_H’, not even white space. You can then decode the configuration header using the preprocessor directives:

     #include <conf.h>
     
     #if HAVE_UNISTD_H
     # include <unistd.h>
     #else
     /* We are in trouble.  */
     #endif

The use of old form templates, with ‘#define’ instead of ‘#undef’ is strongly discouraged. Similarly with old templates with comments on the same line as the ‘#undef’. Anyway, putting comments in preprocessor macros has never been a good idea.

Since it is a tedious task to keep a template header up to date, you may use autoheader to generate it, see autoheader Invocation.


Next: , Previous: Header Templates, Up: Configuration Headers

4.8.2 Using autoheader to Create config.h.in

The autoheader program can create a template file of C ‘#define’ statements for configure to use. If configure.ac invokes AC_CONFIG_HEADERS(file), autoheader creates file.in; if multiple file arguments are given, the first one is used. Otherwise, autoheader creates config.h.in.

In order to do its job, autoheader needs you to document all of the symbols that you might use. Typically this is done via an AC_DEFINE or AC_DEFINE_UNQUOTED call whose first argument is a literal symbol and whose third argument describes the symbol (see Defining Symbols). Alternatively, you can use AH_TEMPLATE (see Autoheader Macros), or you can supply a suitable input file for a subsequent configuration header file. Symbols defined by Autoconf's builtin tests are already documented properly; you need to document only those that you define yourself.

You might wonder why autoheader is needed: after all, why would configure need to “patch” a config.h.in to produce a config.h instead of just creating config.h from scratch? Well, when everything rocks, the answer is just that we are wasting our time maintaining autoheader: generating config.h directly is all that is needed. When things go wrong, however, you'll be thankful for the existence of autoheader.

The fact that the symbols are documented is important in order to check that config.h makes sense. The fact that there is a well-defined list of symbols that should be defined (or not) is also important for people who are porting packages to environments where configure cannot be run: they just have to fill in the blanks.

But let's come back to the point: the invocation of autoheader...

If you give autoheader an argument, it uses that file instead of configure.ac and writes the header file to the standard output instead of to config.h.in. If you give autoheader an argument of -, it reads the standard input instead of configure.ac and writes the header file to the standard output.

autoheader accepts the following options:

--help
-h
Print a summary of the command line options and exit.
--version
-V
Print the version number of Autoconf and exit.
--verbose
-v
Report processing steps.
--debug
-d
Don't remove the temporary files.
--force
-f
Remake the template file even if newer than its input files.
--include=dir
-I dir
Append dir to the include path. Multiple invocations accumulate.
--prepend-include=dir
-B dir
Prepend dir to the include path. Multiple invocations accumulate.
--warnings=category
-W category
Report the warnings related to category (which can actually be a comma separated list). Current categories include:
obsolete
report the uses of obsolete constructs
all
report all the warnings
none
report none
error
treats warnings as errors
no-category
disable warnings falling into category


Previous: autoheader Invocation, Up: Configuration Headers

4.8.3 Autoheader Macros

autoheader scans configure.ac and figures out which C preprocessor symbols it might define. It knows how to generate templates for symbols defined by AC_CHECK_HEADERS, AC_CHECK_FUNCS etc., but if you AC_DEFINE any additional symbol, you must define a template for it. If there are missing templates, autoheader fails with an error message.

The simplest way to create a template for a symbol is to supply the description argument to an ‘AC_DEFINE(symbol)’; see Defining Symbols. You may also use one of the following macros.

— Macro: AH_VERBATIM (key, template)

Tell autoheader to include the template as-is in the header template file. This template is associated with the key, which is used to sort all the different templates and guarantee their uniqueness. It should be a symbol that can be defined via AC_DEFINE.

For example:

          AH_VERBATIM([_GNU_SOURCE],
          [/* Enable GNU extensions on systems that have them.  */
          #ifndef _GNU_SOURCE
          # define _GNU_SOURCE
          #endif])
     
— Macro: AH_TEMPLATE (key, description)

Tell autoheader to generate a template for key. This macro generates standard templates just like AC_DEFINE when a description is given.

For example:

          AH_TEMPLATE([CRAY_STACKSEG_END],
                      [Define to one of _getb67, GETB67, getb67
                       for Cray-2 and Cray-YMP systems.  This
                       function is required for alloca.c support
                       on those systems.])
     

generates the following template, with the description properly justified.

          /* Define to one of _getb67, GETB67, getb67 for Cray-2 and
             Cray-YMP systems.  This function is required for alloca.c
             support on those systems.  */
          #undef CRAY_STACKSEG_END
     
— Macro: AH_TOP (text)

Include text at the top of the header template file.

— Macro: AH_BOTTOM (text)

Include text at the bottom of the header template file.


Next: , Previous: Configuration Headers, Up: Setup

4.9 Running Arbitrary Configuration Commands

You can execute arbitrary commands before, during, and after config.status is run. The three following macros accumulate the commands to run when they are called multiple times. AC_CONFIG_COMMANDS replaces the obsolete macro AC_OUTPUT_COMMANDS; see Obsolete Macros, for details.

— Macro: AC_CONFIG_COMMANDS (tag..., [cmds], [init-cmds])

Specify additional shell commands to run at the end of config.status, and shell commands to initialize any variables from configure. Associate the commands with tag. Since typically the cmds create a file, tag should naturally be the name of that file. If needed, the directory hosting tag is created. This macro is one of the instantiating macros; see Configuration Actions.

Here is an unrealistic example:

          fubar=42
          AC_CONFIG_COMMANDS([fubar],
                             [echo this is extra $fubar, and so on.],
                             [fubar=$fubar])
     

Here is a better one:

          AC_CONFIG_COMMANDS([timestamp], [date >timestamp])
     

The following two macros look similar, but in fact they are not of the same breed: they are executed directly by configure, so you cannot use config.status to rerun them.

— Macro: AC_CONFIG_COMMANDS_PRE (cmds)

Execute the cmds right before creating config.status.

This macro presents the last opportunity to call AC_SUBST, AC_DEFINE, or AC_CONFIG_FOOS macros.

— Macro: AC_CONFIG_COMMANDS_POST (cmds)

Execute the cmds right after creating config.status.


Next: , Previous: Configuration Commands, Up: Setup

4.10 Creating Configuration Links

You may find it convenient to create links whose destinations depend upon results of tests. One can use AC_CONFIG_COMMANDS but the creation of relative symbolic links can be delicate when the package is built in a directory different from the source directory.

— Macro: AC_CONFIG_LINKS (dest:source..., [cmds], [init-cmds])

Make AC_OUTPUT link each of the existing files source to the corresponding link name dest. Makes a symbolic link if possible, otherwise a hard link if possible, otherwise a copy. The dest and source names should be relative to the top level source or build directory. This macro is one of the instantiating macros; see Configuration Actions.

For example, this call:

          AC_CONFIG_LINKS([host.h:config/$machine.h
                          object.h:config/$obj_format.h])
     

creates in the current directory host.h as a link to srcdir/config/$machine.h, and object.h as a link to srcdir/config/$obj_format.h.

The tempting value ‘.’ for dest is invalid: it makes it impossible for ‘config.status’ to guess the links to establish.

One can then run:

          ./config.status host.h object.h
     

to create the links.


Next: , Previous: Configuration Links, Up: Setup

4.11 Configuring Other Packages in Subdirectories

In most situations, calling AC_OUTPUT is sufficient to produce makefiles in subdirectories. However, configure scripts that control more than one independent package can use AC_CONFIG_SUBDIRS to run configure scripts for other packages in subdirectories.

— Macro: AC_CONFIG_SUBDIRS (dir ...)

Make AC_OUTPUT run configure in each subdirectory dir in the given blank-or-newline-separated list. Each dir should be a literal, i.e., please do not use:

          if test "$package_foo_enabled" = yes; then
            $my_subdirs="$my_subdirs foo"
          fi
          AC_CONFIG_SUBDIRS([$my_subdirs])
     

because this prevents ‘./configure --help=recursive’ from displaying the options of the package foo. Instead, you should write:

          if test "$package_foo_enabled" = yes; then
            AC_CONFIG_SUBDIRS([foo])
          fi
     

If a given dir is not found, an error is reported: if the subdirectory is optional, write:

          if test -d "$srcdir/foo"; then
            AC_CONFIG_SUBDIRS([foo])
          fi
     

If a given dir contains configure.gnu, it is run instead of configure. This is for packages that might use a non-Autoconf script Configure, which can't be called through a wrapper configure since it would be the same file on case-insensitive file systems. Likewise, if a dir contains configure.in but no configure, the Cygnus configure script found by AC_CONFIG_AUX_DIR is used.

The subdirectory configure scripts are given the same command line options that were given to this configure script, with minor changes if needed, which include:

This macro also sets the output variable subdirs to the list of directories ‘dir ...’. Make rules can use this variable to determine which subdirectories to recurse into.

This macro may be called multiple times.


Previous: Subdirectories, Up: Setup

4.12 Default Prefix

By default, configure sets the prefix for files it installs to /usr/local. The user of configure can select a different prefix using the --prefix and --exec-prefix options. There are two ways to change the default: when creating configure, and when running it.

Some software packages might want to install in a directory other than /usr/local by default. To accomplish that, use the AC_PREFIX_DEFAULT macro.

— Macro: AC_PREFIX_DEFAULT (prefix)

Set the default installation prefix to prefix instead of /usr/local.

It may be convenient for users to have configure guess the installation prefix from the location of a related program that they have already installed. If you wish to do that, you can call AC_PREFIX_PROGRAM.

— Macro: AC_PREFIX_PROGRAM (program)

If the user did not specify an installation prefix (using the --prefix option), guess a value for it by looking for program in PATH, the way the shell does. If program is found, set the prefix to the parent of the directory containing program, else default the prefix as described above (/usr/local or AC_PREFIX_DEFAULT). For example, if program is gcc and the PATH contains /usr/local/gnu/bin/gcc, set the prefix to /usr/local/gnu.


Next: , Previous: Setup, Up: Top

5 Existing Tests

These macros test for particular system features that packages might need or want to use. If you need to test for a kind of feature that none of these macros check for, you can probably do it by calling primitive test macros with appropriate arguments (see Writing Tests).

These tests print messages telling the user which feature they're checking for, and what they find. They cache their results for future configure runs (see Caching Results).

Some of these macros set output variables. See Makefile Substitutions, for how to get their values. The phrase “define name” is used below as a shorthand to mean “define the C preprocessor symbol name to the value 1”. See Defining Symbols, for how to get those symbol definitions into your program.


Next: , Up: Existing Tests

5.1 Common Behavior

Much effort has been expended to make Autoconf easy to learn. The most obvious way to reach this goal is simply to enforce standard interfaces and behaviors, avoiding exceptions as much as possible. Because of history and inertia, unfortunately, there are still too many exceptions in Autoconf; nevertheless, this section describes some of the common rules.


Next: , Up: Common Behavior

5.1.1 Standard Symbols

All the generic macros that AC_DEFINE a symbol as a result of their test transform their argument values to a standard alphabet. First, argument is converted to upper case and any asterisks (‘*’) are each converted to ‘P’. Any remaining characters that are not alphanumeric are converted to underscores.

For instance,

     AC_CHECK_TYPES([struct $Expensive*])

defines the symbol ‘HAVE_STRUCT__EXPENSIVEP’ if the check succeeds.


Previous: Standard Symbols, Up: Common Behavior

5.1.2 Default Includes

Several tests depend upon a set of header files. Since these headers are not universally available, tests actually have to provide a set of protected includes, such as:

     #if TIME_WITH_SYS_TIME
     # include <sys/time.h>
     # include <time.h>
     #else
     # if HAVE_SYS_TIME_H
     #  include <sys/time.h>
     # else
     #  include <time.h>
     # endif
     #endif

Unless you know exactly what you are doing, you should avoid using unconditional includes, and check the existence of the headers you include beforehand (see Header Files).

Most generic macros use the following macro to provide the default set of includes:

— Macro: AC_INCLUDES_DEFAULT ([include-directives])

Expand to include-directives if defined, otherwise to:

          #include <stdio.h>
          #if HAVE_SYS_TYPES_H
          # include <sys/types.h>
          #endif
          #if HAVE_SYS_STAT_H
          # include <sys/stat.h>
          #endif
          #if STDC_HEADERS
          # include <stdlib.h>
          # include <stddef.h>
          #else
          # if HAVE_STDLIB_H
          #  include <stdlib.h>
          # endif
          #endif
          #if HAVE_STRING_H
          # if !STDC_HEADERS && HAVE_MEMORY_H
          #  include <memory.h>
          # endif
          # include <string.h>
          #endif
          #if HAVE_STRINGS_H
          # include <strings.h>
          #endif
          #if HAVE_INTTYPES_H
          # include <inttypes.h>
          #endif
          #if HAVE_STDINT_H
          # include <stdint.h>
          #endif
          #if HAVE_UNISTD_H
          # include <unistd.h>
          #endif
     

If the default includes are used, then check for the presence of these headers and their compatibility, i.e., you don't need to run AC_HEADER_STDC, nor check for stdlib.h etc.

These headers are checked for in the same order as they are included. For instance, on some systems string.h and strings.h both exist, but conflict. Then HAVE_STRING_H is defined, not HAVE_STRINGS_H.


Next: , Previous: Common Behavior, Up: Existing Tests

5.2 Alternative Programs

These macros check for the presence or behavior of particular programs. They are used to choose between several alternative programs and to decide what to do once one has been chosen. If there is no macro specifically defined to check for a program you need, and you don't need to check for any special properties of it, then you can use one of the general program-check macros.


Next: , Up: Alternative Programs

5.2.1 Particular Program Checks

These macros check for particular programs—whether they exist, and in some cases whether they support certain features.

— Macro: AC_PROG_AWK

Check for gawk, mawk, nawk, and awk, in that order, and set output variable AWK to the first one that is found. It tries gawk first because that is reported to be the best implementation.

— Macro: AC_PROG_GREP

Look for the best available grep or ggrep that accepts the longest input lines possible, and that supports multiple -e options. Set the output variable GREP to whatever is chosen. See Limitations of Usual Tools, for more information about portability problems with the grep command family.

— Macro: AC_PROG_EGREP

Check whether $GREP -E works, or else look for the best available egrep or gegrep that accepts the longest input lines possible. Set the output variable EGREP to whatever is chosen.

— Macro: AC_PROG_FGREP

Check whether $GREP -F works, or else look for the best available fgrep or gfgrep that accepts the longest input lines possible. Set the output variable FGREP to whatever is chosen.

— Macro: AC_PROG_INSTALL

Set output variable INSTALL to the name of a BSD-compatible install program, if one is found in the current PATH. Otherwise, set INSTALL to ‘dir/install-sh -c’, checking the directories specified to AC_CONFIG_AUX_DIR (or its default directories) to determine dir (see Output). Also set the variables INSTALL_PROGRAM and INSTALL_SCRIPT to ‘${INSTALL}’ and INSTALL_DATA to ‘${INSTALL} -m 644’.

This macro screens out various instances of install known not to work. It prefers to find a C program rather than a shell script, for speed. Instead of install-sh, it can also use install.sh, but that name is obsolete because some make programs have a rule that creates install from it if there is no makefile.

Autoconf comes with a copy of install-sh that you can use. If you use AC_PROG_INSTALL, you must include either install-sh or install.sh in your distribution; otherwise configure produces an error message saying it can't find them—even if the system you're on has a good install program. This check is a safety measure to prevent you from accidentally leaving that file out, which would prevent your package from installing on systems that don't have a BSD-compatible install program.

If you need to use your own installation program because it has features not found in standard install programs, there is no reason to use AC_PROG_INSTALL; just put the file name of your program into your Makefile.in files.

— Macro: AC_PROG_MKDIR_P

Set output variable MKDIR_P to a program that ensures that for each argument, a directory named by this argument exists, creating it and its parent directories if needed, and without race conditions when two instances of the program attempt to make the same directory at nearly the same time.

This macro uses the ‘mkdir -p’ command if possible. Otherwise, it falls back on invoking install-sh with the -d option, so your package should contain install-sh as described under AC_PROG_INSTALL. An install-sh file that predates Autoconf 2.60 or Automake 1.10 is vulnerable to race conditions, so if you want to support parallel installs from different packages into the same directory you need to make sure you have an up-to-date install-sh. In particular, be careful about using ‘autoreconf -if’ if your Automake predates Automake 1.10.

This macro is related to the AS_MKDIR_P macro (see Programming in M4sh), but it sets an output variable intended for use in other files, whereas AS_MKDIR_P is intended for use in scripts like configure. Also, AS_MKDIR_P does not accept options, but MKDIR_P supports the -m option, e.g., a makefile might invoke $(MKDIR_P) -m 0 dir to create an inaccessible directory, and conversely a makefile should use $(MKDIR_P) -- $(FOO) if FOO might yield a value that begins with ‘-’. Finally, AS_MKDIR_P does not check for race condition vulnerability, whereas AC_PROG_MKDIR_P does.

— Macro: AC_PROG_LEX

If flex is found, set output variable LEX to ‘flex’ and LEXLIB to -lfl, if that library is in a standard place. Otherwise set LEX to ‘lex’ and LEXLIB to -ll.

Define YYTEXT_POINTER if yytext is a ‘char *’ instead of a ‘char []’. Also set output variable LEX_OUTPUT_ROOT to the base of the file name that the lexer generates; usually lex.yy, but sometimes something else. These results vary according to whether lex or flex is being used.

You are encouraged to use Flex in your sources, since it is both more pleasant to use than plain Lex and the C source it produces is portable. In order to ensure portability, however, you must either provide a function yywrap or, if you don't use it (e.g., your scanner has no ‘#include’-like feature), simply include a ‘%noyywrap’ statement in the scanner's source. Once this done, the scanner is portable (unless you felt free to use nonportable constructs) and does not depend on any library. In this case, and in this case only, it is suggested that you use this Autoconf snippet:

          AC_PROG_LEX
          if test "$LEX" != flex; then
            LEX="$SHELL $missing_dir/missing flex"
            AC_SUBST([LEX_OUTPUT_ROOT], [lex.yy])
            AC_SUBST([LEXLIB], [''])
          fi
     

The shell script missing can be found in the Automake distribution.

To ensure backward compatibility, Automake's AM_PROG_LEX invokes (indirectly) this macro twice, which causes an annoying but benign “AC_PROG_LEX invoked multiple times” warning. Future versions of Automake will fix this issue; meanwhile, just ignore this message.

As part of running the test, this macro may delete any file in the configuration directory named lex.yy.c or lexyy.c.

— Macro: AC_PROG_LN_S

If ‘ln -s’ works on the current file system (the operating system and file system support symbolic links), set the output variable LN_S to ‘ln -s’; otherwise, if ‘ln’ works, set LN_S to ‘ln’, and otherwise set it to ‘cp -p’.

If you make a link in a directory other than the current directory, its meaning depends on whether ‘ln’ or ‘ln -s’ is used. To safely create links using ‘$(LN_S)’, either find out which form is used and adjust the arguments, or always invoke ln in the directory where the link is to be created.

In other words, it does not work to do:

          $(LN_S) foo /x/bar
     

Instead, do:

          (cd /x && $(LN_S) foo bar)
     
— Macro: AC_PROG_RANLIB

Set output variable RANLIB to ‘ranlib’ if ranlib is found, and otherwise to ‘:’ (do nothing).

— Macro: AC_PROG_SED

Set output variable SED to a Sed implementation that conforms to Posix and does not have arbitrary length limits. Report an error if no acceptable Sed is found. See Limitations of Usual Tools, for more information about portability problems with Sed.

— Macro: AC_PROG_YACC

If bison is found, set output variable YACC to ‘bison -y’. Otherwise, if byacc is found, set YACC to ‘byacc’. Otherwise set YACC to ‘yacc’.


Previous: Particular Programs, Up: Alternative Programs

5.2.2 Generic Program and File Checks

These macros are used to find programs not covered by the “particular” test macros. If you need to check the behavior of a program as well as find out whether it is present, you have to write your own test for it (see Writing Tests). By default, these macros use the environment variable PATH. If you need to check for a program that might not be in the user's PATH, you can pass a modified path to use instead, like this:

     AC_PATH_PROG([INETD], [inetd], [/usr/libexec/inetd],
                  [$PATH:/usr/libexec:/usr/sbin:/usr/etc:/etc])

You are strongly encouraged to declare the variable passed to AC_CHECK_PROG etc. as precious, See Setting Output Variables, AC_ARG_VAR, for more details.

— Macro: AC_CHECK_PROG (variable, prog-to-check-for, value-if-found, [value-if-not-found], [path], [reject])

Check whether program prog-to-check-for exists in PATH. If it is found, set variable to value-if-found, otherwise to value-if-not-found, if given. Always pass over reject (an absolute file name) even if it is the first found in the search path; in that case, set variable using the absolute file name of the prog-to-check-for found that is not reject. If variable was already set, do nothing. Calls AC_SUBST for variable.

— Macro: AC_CHECK_PROGS (variable, progs-to-check-for, [value-if-not-found], [path])

Check for each program in the blank-separated list progs-to-check-for existing in the PATH. If one is found, set variable to the name of that program. Otherwise, continue checking the next program in the list. If none of the programs in the list are found, set variable to value-if-not-found; if value-if-not-found is not specified, the value of variable is not changed. Calls AC_SUBST for variable.

— Macro: AC_CHECK_TARGET_TOOL (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_PROG, but first looks for prog-to-check-for with a prefix of the target type as determined by AC_CANONICAL_TARGET, followed by a dash (see Canonicalizing). If the tool cannot be found with a prefix, and if the build and target types are equal, then it is also searched for without a prefix.

As noted in Specifying the system type, the target is rarely specified, because most of the time it is the same as the host: it is the type of system for which any compiler tool in the package produces code. What this macro looks for is, for example, a tool (assembler, linker, etc.) that the compiler driver (gcc for the GNU C Compiler) uses to produce objects, archives or executables.

— Macro: AC_CHECK_TOOL (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_PROG, but first looks for prog-to-check-for with a prefix of the host type as determined by AC_CANONICAL_HOST, followed by a dash (see Canonicalizing). For example, if the user runs ‘configure --host=i386-gnu’, then this call:

          AC_CHECK_TOOL([RANLIB], [ranlib], [:])
     

sets RANLIB to i386-gnu-ranlib if that program exists in PATH, or otherwise to ‘ranlib’ if that program exists in PATH, or to ‘:’ if neither program exists.

In the future, when cross-compiling this macro will only accept program names that are prefixed with the host type. For more information, see Specifying the system type.

— Macro: AC_CHECK_TARGET_TOOLS (variable, progs-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_TARGET_TOOL, each of the tools in the list progs-to-check-for are checked with a prefix of the target type as determined by AC_CANONICAL_TARGET, followed by a dash (see Canonicalizing). If none of the tools can be found with a prefix, and if the build and target types are equal, then the first one without a prefix is used. If a tool is found, set variable to the name of that program. If none of the tools in the list are found, set variable to value-if-not-found; if value-if-not-found is not specified, the value of variable is not changed. Calls AC_SUBST for variable.

— Macro: AC_CHECK_TOOLS (variable, progs-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_TOOL, each of the tools in the list progs-to-check-for are checked with a prefix of the host type as determined by AC_CANONICAL_HOST, followed by a dash (see Canonicalizing). If none of the tools can be found with a prefix, then the first one without a prefix is used. If a tool is found, set variable to the name of that program. If none of the tools in the list are found, set variable to value-if-not-found; if value-if-not-found is not specified, the value of variable is not changed. Calls AC_SUBST for variable.

In the future, when cross-compiling this macro will not accept program names that are not prefixed with the host type.

— Macro: AC_PATH_PROG (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_PROG, but set variable to the absolute name of prog-to-check-for if found.

— Macro: AC_PATH_PROGS (variable, progs-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_PROGS, but if any of progs-to-check-for are found, set variable to the absolute name of the program found.

— Macro: AC_PATH_TARGET_TOOL (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_TARGET_TOOL, but set variable to the absolute name of the program if it is found.

— Macro: AC_PATH_TOOL (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_TOOL, but set variable to the absolute name of the program if it is found.

In the future, when cross-compiling this macro will not accept program names that are not prefixed with the host type.


Next: , Previous: Alternative Programs, Up: Existing Tests

5.3 Files

You might also need to check for the existence of files. Before using these macros, ask yourself whether a runtime test might not be a better solution. Be aware that, like most Autoconf macros, they test a feature of the host machine, and therefore, they die when cross-compiling.

— Macro: AC_CHECK_FILE (file, [action-if-found], [action-if-not-found])

Check whether file file exists on the native system. If it is found, execute action-if-found, otherwise do action-if-not-found, if given.

— Macro: AC_CHECK_FILES (files, [action-if-found], [action-if-not-found])

Executes AC_CHECK_FILE once for each file listed in files. Additionally, defines ‘HAVE_file’ (see Standard Symbols) for each file found.


Next: , Previous: Files, Up: Existing Tests

5.4 Library Files

The following macros check for the presence of certain C, C++, or Fortran library archive files.

— Macro: AC_CHECK_LIB (library, function, [action-if-found], [action-if-not-found], [other-libraries])

Test whether the library library is available by trying to link a test program that calls function function with the library. function should be a function provided by the library. Use the base name of the library; e.g., to check for -lmp, use ‘mp’ as the library argument.

action-if-found is a list of shell commands to run if the link with the library succeeds; action-if-not-found is a list of shell commands to run if the link fails. If action-if-found is not specified, the default action prepends -llibrary to LIBS and defines ‘HAVE_LIBlibrary’ (in all capitals). This macro is intended to support building LIBS in a right-to-left (least-dependent to most-dependent) fashion such that library dependencies are satisfied as a natural side effect of consecutive tests. Linkers are sensitive to library ordering so the order in which LIBS is generated is important to reliable detection of libraries.

If linking with library results in unresolved symbols that would be resolved by linking with additional libraries, give those libraries as the other-libraries argument, separated by spaces: e.g., -lXt -lX11. Otherwise, this macro fails to detect that library is present, because linking the test program always fails with unresolved symbols. The other-libraries argument should be limited to cases where it is desirable to test for one library in the presence of another that is not already in LIBS.

AC_CHECK_LIB requires some care in usage, and should be avoided in some common cases. Many standard functions like gethostbyname appear the standard C library on some hosts, and in special libraries like nsl on other hosts. On some hosts the special libraries contain variant implementations that you may not want to use. These days it is normally better to use AC_SEARCH_LIBS([gethostbyname], [nsl]) instead of AC_CHECK_LIB([nsl], [gethostbyname]).

— Macro: AC_SEARCH_LIBS (function, search-libs, [action-if-found], [action-if-not-found], [other-libraries])

Search for a library defining function if it's not already available. This equates to calling ‘AC_LINK_IFELSE([AC_LANG_CALL([], [function])])’ first with no libraries, then for each library listed in search-libs.

Add -llibrary to LIBS for the first library found to contain function, and run action-if-found. If the function is not found, run action-if-not-found.

If linking with library results in unresolved symbols that would be resolved by linking with additional libraries, give those libraries as the other-libraries argument, separated by spaces: e.g., -lXt -lX11. Otherwise, this macro fails to detect that function is present, because linking the test program always fails with unresolved symbols.


Next: , Previous: Libraries, Up: Existing Tests

5.5 Library Functions

The following macros check for particular C library functions. If there is no macro specifically defined to check for a function you need, and you don't need to check for any special properties of it, then you can use one of the general function-check macros.


Next: , Up: Library Functions

5.5.1 Portability of C Functions

Most usual functions can either be missing, or be buggy, or be limited on some architectures. This section tries to make an inventory of these portability issues. By definition, this list always requires additions. Please help us keeping it as complete as possible.

exit
On ancient hosts, exit returned int. This is because exit predates void, and there was a long tradition of it returning int.

On current hosts, the problem more likely is that exit is not declared, due to C++ problems of some sort or another. For this reason we suggest that test programs not invoke exit, but return from main instead.

free
The C standard says a call free (NULL) does nothing, but some old systems don't support this (e.g., NextStep).
isinf
isnan
The C99 standard says that isinf and isnan are macros. On some systems just macros are available (e.g., HP-UX and Solaris 10), on some systems both macros and functions (e.g., glibc 2.3.2), and on some systems only functions (e.g., IRIX 6 and Solaris 9). In some cases these functions are declared in nonstandard headers like <sunmath.h> and defined in non-default libraries like -lm or -lsunmath.

The C99 isinf and isnan macros work correctly with long double arguments, but pre-C99 systems that use functions typically assume double arguments. On such a system, isinf incorrectly returns true for a finite long double argument that is outside the range of double.

To work around this porting mess, you can use code like the following.

          #include <math.h>
          
          #ifndef isnan
          # define isnan(x) \
              (sizeof (x) == sizeof (long double) ? isnan_ld (x) \
               : sizeof (x) == sizeof (double) ? isnan_d (x) \
               : isnan_f (x))
          static inline int isnan_f  (float       x) { return x != x; }
          static inline int isnan_d  (double      x) { return x != x; }
          static inline int isnan_ld (long double x) { return x != x; }
          #endif
          
          #ifndef isinf
          # define isinf(x) \
              (sizeof (x) == sizeof (long double) ? isinf_ld (x) \
               : sizeof (x) == sizeof (double) ? isinf_d (x) \
               : isinf_f (x))
          static inline int isinf_f  (float       x) { return isnan (x - x); }
          static inline int isinf_d  (double      x) { return isnan (x - x); }
          static inline int isinf_ld (long double x) { return isnan (x - x); }
          #endif
     

Use AC_C_INLINE (see C Compiler) so that this code works on compilers that lack the inline keyword. Some optimizing compilers mishandle these definitions, but systems with that bug typically have missing or broken isnan functions anyway, so it's probably not worth worrying about.

malloc
The C standard says a call malloc (0) is implementation dependent. It may either return NULL (e.g., OSF 4) or non-NULL (e.g., GNU C Library). AC_FUNC_MALLOC can be used to insist on non-NULL (see Particular Functions).
putenv
Posix prefers setenv to putenv; among other things, putenv is not required of all Posix implementations, but setenv is.

Posix specifies that putenv puts the given string directly in environ, but some systems make a copy of it instead (e.g., glibc 2.0, or BSD). And when a copy is made, unsetenv might not free it, causing a memory leak (e.g., FreeBSD 4).

On some systems putenv ("FOO") removes ‘FOO’ from the environment, but this is not standard usage and it dumps core on some systems (e.g., AIX).

On MinGW, a call putenv ("FOO=") removes ‘FOO’ from the environment, rather than inserting it with an empty value.

realloc
The C standard says a call realloc (NULL, size) is equivalent to malloc (size), but some old systems don't support this (e.g., NextStep).
signal handler
Normally signal takes a handler function with a return type of void, but some old systems required int instead. Any actual int value returned is not used; this is only a difference in the function prototype demanded.

All systems we know of in current use return void. The int was to support K&R C, where of course void is not available. AC_TYPE_SIGNAL (see Particular Types) can be used to establish the correct type in all cases.

snprintf
The C99 standard says that if the output array isn't big enough and if no other errors occur, snprintf and vsnprintf truncate the output and return the number of bytes that ought to have been produced. Some older systems return the truncated length (e.g., GNU C Library 2.0.x or irix 6.5), some a negative value (e.g., earlier GNU C Library versions), and some the buffer length without truncation (e.g., 32-bit Solaris 7). Also, some buggy older systems ignore the length and overrun the buffer (e.g., 64-bit Solaris 7).
sprintf
The C standard says sprintf and vsprintf return the number of bytes written. On some ancient systems (SunOS 4 for instance) they return the buffer pointer instead, but these no longer need to be worried about.
sscanf
On various old systems, e.g., HP-UX 9, sscanf requires that its input string be writable (though it doesn't actually change it). This can be a problem when using gcc since it normally puts constant strings in read-only memory (see Incompatibilities of GCC). Apparently in some cases even having format strings read-only can be a problem.
strerror_r
Posix specifies that strerror_r returns an int, but many systems (e.g., GNU C Library version 2.2.4) provide a different version returning a char *. AC_FUNC_STRERROR_R can detect which is in use (see Particular Functions).
strnlen
AIX 4.3 provides a broken version which produces the following results:
          strnlen ("foobar", 0) = 0
          strnlen ("foobar", 1) = 3
          strnlen ("foobar", 2) = 2
          strnlen ("foobar", 3) = 1
          strnlen ("foobar", 4) = 0
          strnlen ("foobar", 5) = 6
          strnlen ("foobar", 6) = 6
          strnlen ("foobar", 7) = 6
          strnlen ("foobar", 8) = 6
          strnlen ("foobar", 9) = 6
     

sysconf
_SC_PAGESIZE is standard, but some older systems (e.g., HP-UX 9) have _SC_PAGE_SIZE instead. This can be tested with #ifdef.
unlink
The Posix spec says that unlink causes the given file to be removed only after there are no more open file handles for it. Some non-Posix hosts have trouble with this requirement, though, and some DOS variants even corrupt the file system.
unsetenv
On MinGW, unsetenv is not available, but a variable ‘FOO’ can be removed with a call putenv ("FOO="), as described under putenv above.
va_copy
The C99 standard provides va_copy for copying va_list variables. It may be available in older environments too, though possibly as __va_copy (e.g., gcc in strict pre-C99 mode). These can be tested with #ifdef. A fallback to memcpy (&dst, &src, sizeof (va_list)) gives maximum portability.
va_list
va_list is not necessarily just a pointer. It can be a struct (e.g., gcc on Alpha), which means NULL is not portable. Or it can be an array (e.g., gcc in some PowerPC configurations), which means as a function parameter it can be effectively call-by-reference and library routines might modify the value back in the caller (e.g., vsnprintf in the GNU C Library 2.1).
Signed >>
Normally the C >> right shift of a signed type replicates the high bit, giving a so-called “arithmetic” shift. But care should be taken since Standard C doesn't require that behavior. On those few processors without a native arithmetic shift (for instance Cray vector systems) zero bits may be shifted in, the same as a shift of an unsigned type.
Integer /
C divides signed integers by truncating their quotient toward zero, yielding the same result as Fortran. However, before C99 the standard allowed C implementations to take the floor or ceiling of the quotient in some cases. Hardly any implementations took advantage of this freedom, though, and it's probably not worth worrying about this issue nowadays.


Next: , Previous: Function Portability, Up: Library Functions

5.5.2 Particular Function Checks

These macros check for particular C functions—whether they exist, and in some cases how they respond when given certain arguments.

— Macro: AC_FUNC_ALLOCA

Check how to get alloca. Tries to get a builtin version by checking for alloca.h or the predefined C preprocessor macros __GNUC__ and _AIX. If this macro finds alloca.h, it defines HAVE_ALLOCA_H.

If those attempts fail, it looks for the function in the standard C library. If any of those methods succeed, it defines HAVE_ALLOCA. Otherwise, it sets the output variable ALLOCA to ‘${LIBOBJDIR}alloca.o’ and defines C_ALLOCA (so programs can periodically call ‘alloca (0)’ to garbage collect). This variable is separate from LIBOBJS so multiple programs can share the value of ALLOCA without needing to create an actual library, in case only some of them use the code in LIBOBJS. The ‘${LIBOBJDIR}’ prefix serves the same purpose as in LIBOBJS (see AC_LIBOBJ vs LIBOBJS).

This macro does not try to get alloca from the System V R3 libPW or the System V R4 libucb because those libraries contain some incompatible functions that cause trouble. Some versions do not even contain alloca or contain a buggy version. If you still want to use their alloca, use ar to extract alloca.o from them instead of compiling alloca.c.

Source files that use alloca should start with a piece of code like the following, to declare it properly.

          #if HAVE_ALLOCA_H
          # include <alloca.h>
          #elif defined __GNUC__
          # define alloca __builtin_alloca
          #elif defined _AIX
          # define alloca __alloca
          #elif defined _MSC_VER
          # include <malloc.h>
          # define alloca _alloca
          #else
          # include <stddef.h>
          # ifdef  __cplusplus
          extern "C"
          # endif
          void *alloca (size_t);
          #endif
     
— Macro: AC_FUNC_CHOWN

If the chown function is available and works (in particular, it should accept -1 for uid and gid), define HAVE_CHOWN.

— Macro: AC_FUNC_CLOSEDIR_VOID

If the closedir function does not return a meaningful value, define CLOSEDIR_VOID. Otherwise, callers ought to check its return value for an error indicator.

Currently this test is implemented by running a test program. When cross compiling the pessimistic assumption that closedir does not return a meaningful value is made.

This macro is obsolescent, as closedir returns a meaningful value on current systems. New programs need not use this macro.

— Macro: AC_FUNC_ERROR_AT_LINE

If the error_at_line function is not found, require an AC_LIBOBJ replacement of ‘error’.

— Macro: AC_FUNC_FNMATCH

If the fnmatch function conforms to Posix, define HAVE_FNMATCH. Detect common implementation bugs, for example, the bugs in Solaris 2.4.

Unlike the other specific AC_FUNC macros, AC_FUNC_FNMATCH does not replace a broken/missing fnmatch. This is for historical reasons. See AC_REPLACE_FNMATCH below.

— Macro: AC_FUNC_FNMATCH_GNU

Behave like AC_REPLACE_FNMATCH (replace) but also test whether fnmatch supports GNU extensions. Detect common implementation bugs, for example, the bugs in the GNU C Library 2.1.

— Macro: AC_FUNC_FORK

This macro checks for the fork and vfork functions. If a working fork is found, define HAVE_WORKING_FORK. This macro checks whether fork is just a stub by trying to run it.

If vfork.h is found, define HAVE_VFORK_H. If a working vfork is found, define HAVE_WORKING_VFORK. Otherwise, define vfork to be fork for backward compatibility with previous versions of autoconf. This macro checks for several known errors in implementations of vfork and considers the system to not have a working vfork if it detects any of them. It is not considered to be an implementation error if a child's invocation of signal modifies the parent's signal handler, since child processes rarely change their signal handlers.

Since this macro defines vfork only for backward compatibility with previous versions of autoconf you're encouraged to define it yourself in new code:

          #if !HAVE_WORKING_VFORK
          # define vfork fork
          #endif
     
— Macro: AC_FUNC_FSEEKO

If the fseeko function is available, define HAVE_FSEEKO. Define _LARGEFILE_SOURCE if necessary to make the prototype visible on some systems (e.g., glibc 2.2). Otherwise linkage problems may occur when compiling with AC_SYS_LARGEFILE on largefile-sensitive systems where off_t does not default to a 64bit entity.

— Macro: AC_FUNC_GETGROUPS

If the getgroups function is available and works (unlike on Ultrix 4.3, where ‘getgroups (0, 0)’ always fails), define HAVE_GETGROUPS. Set GETGROUPS_LIBS to any libraries needed to get that function. This macro runs AC_TYPE_GETGROUPS.

— Macro: AC_FUNC_GETLOADAVG

Check how to get the system load averages. To perform its tests properly, this macro needs the file getloadavg.c; therefore, be sure to set the AC_LIBOBJ replacement directory properly (see Generic Functions, AC_CONFIG_LIBOBJ_DIR).

If the system has the getloadavg function, define HAVE_GETLOADAVG, and set GETLOADAVG_LIBS to any libraries necessary to get that function. Also add GETLOADAVG_LIBS to LIBS. Otherwise, require an AC_LIBOBJ replacement for ‘getloadavg’ with source code in dir/getloadavg.c, and possibly define several other C preprocessor macros and output variables:

  1. Define C_GETLOADAVG.
  2. Define SVR4, DGUX, UMAX, or UMAX4_3 if on those systems.
  3. If nlist.h is found, define HAVE_NLIST_H.
  4. If ‘struct nlist’ has an ‘n_un.n_name’ member, define HAVE_STRUCT_NLIST_N_UN_N_NAME. The obsolete symbol NLIST_NAME_UNION is still defined, but do not depend upon it.
  5. Programs may need to be installed set-group-ID (or set-user-ID) for getloadavg to work. In this case, define GETLOADAVG_PRIVILEGED, set the output variable NEED_SETGID to ‘true’ (and otherwise to ‘false’), and set KMEM_GROUP to the name of the group that should own the installed program.

— Macro: AC_FUNC_GETMNTENT

Check for getmntent in the standard C library, and then in the sun, seq, and gen libraries, for unicos, irix 4, ptx, and UnixWare, respectively. Then, if getmntent is available, define HAVE_GETMNTENT.

— Macro: AC_FUNC_GETPGRP

Define GETPGRP_VOID if it is an error to pass 0 to getpgrp; this is the Posix behavior. On older BSD systems, you must pass 0 to getpgrp, as it takes an argument and behaves like Posix's getpgid.

          #if GETPGRP_VOID
            pid = getpgrp ();
          #else
            pid = getpgrp (0);
          #endif
     

This macro does not check whether getpgrp exists at all; if you need to work in that situation, first call AC_CHECK_FUNC for getpgrp.

This macro is obsolescent, as current systems have a getpgrp whose signature conforms to Posix. New programs need not use this macro.

— Macro: AC_FUNC_LSTAT_FOLLOWS_SLASHED_SYMLINK

If link is a symbolic link, then lstat should treat link/ the same as link/.. However, many older lstat implementations incorrectly ignore trailing slashes.

It is safe to assume that if lstat incorrectly ignores trailing slashes, then other symbolic-link-aware functions like unlink also incorrectly ignore trailing slashes.

If lstat behaves properly, define LSTAT_FOLLOWS_SLASHED_SYMLINK, otherwise require an AC_LIBOBJ replacement of lstat.

— Macro: AC_FUNC_MALLOC

If the malloc function is compatible with the GNU C library malloc (i.e., ‘malloc (0)’ returns a valid pointer), define HAVE_MALLOC to 1. Otherwise define HAVE_MALLOC to 0, ask for an AC_LIBOBJ replacement for ‘malloc’, and define malloc to rpl_malloc so that the native malloc is not used in the main project.

Typically, the replacement file malloc.c should look like (note the ‘#undef malloc’):

     
     #if HAVE_CONFIG_H
     # include <config.h>
     #endif
     #undef malloc
     
     #include <sys/types.h>
     
     void *malloc ();
     
     /* Allocate an N-byte block of memory from the heap.
        If N is zero, allocate a 1-byte block.  */
     
     void *
     rpl_malloc (size_t n)
     {
       if (n == 0)
         n = 1;
       return malloc (n);
     }

— Macro: AC_FUNC_MEMCMP

If the memcmp function is not available, or does not work on 8-bit data (like the one on SunOS 4.1.3), or fails when comparing 16 bytes or more and with at least one buffer not starting on a 4-byte boundary (such as the one on NeXT x86 OpenStep), require an AC_LIBOBJ replacement for ‘memcmp’.

This macro is obsolescent, as current systems have a working memcmp. New programs need not use this macro.

— Macro: AC_FUNC_MBRTOWC

Define HAVE_MBRTOWC to 1 if the function mbrtowc and the type mbstate_t are properly declared.

— Macro: AC_FUNC_MKTIME

If the mktime function is not available, or does not work correctly, require an AC_LIBOBJ replacement for ‘mktime’. For the purposes of this test, mktime should conform to the Posix standard and should be the inverse of localtime.

— Macro: AC_FUNC_MMAP

If the mmap function exists and works correctly, define HAVE_MMAP. This checks only private fixed mapping of already-mapped memory.

— Macro: AC_FUNC_OBSTACK

If the obstacks are found, define HAVE_OBSTACK, else require an AC_LIBOBJ replacement for ‘obstack’.

— Macro: AC_FUNC_REALLOC

If the realloc function is compatible with the GNU C library realloc (i.e., ‘realloc (NULL, 0)’ returns a valid pointer), define HAVE_REALLOC to 1. Otherwise define HAVE_REALLOC to 0, ask for an AC_LIBOBJ replacement for ‘realloc’, and define realloc to rpl_realloc so that the native realloc is not used in the main project. See AC_FUNC_MALLOC for details.

— Macro: AC_FUNC_SELECT_ARGTYPES

Determines the correct type to be passed for each of the select function's arguments, and defines those types in SELECT_TYPE_ARG1, SELECT_TYPE_ARG234, and SELECT_TYPE_ARG5 respectively. SELECT_TYPE_ARG1 defaults to ‘int’, SELECT_TYPE_ARG234 defaults to ‘int *’, and SELECT_TYPE_ARG5 defaults to ‘struct timeval *’.

This macro is obsolescent, as current systems have a select whose signature conforms to Posix. New programs need not use this macro.

— Macro: AC_FUNC_SETPGRP

If setpgrp takes no argument (the Posix version), define SETPGRP_VOID. Otherwise, it is the BSD version, which takes two process IDs as arguments. This macro does not check whether setpgrp exists at all; if you need to work in that situation, first call AC_CHECK_FUNC for setpgrp.

This macro is obsolescent, as current systems have a setpgrp whose signature conforms to Posix. New programs need not use this macro.

— Macro: AC_FUNC_STAT
— Macro: AC_FUNC_LSTAT

Determine whether stat or lstat have the bug that it succeeds when given the zero-length file name as argument. The stat and lstat from SunOS 4.1.4 and the Hurd (as of 1998-11-01) do this.

If it does, then define HAVE_STAT_EMPTY_STRING_BUG (or HAVE_LSTAT_EMPTY_STRING_BUG) and ask for an AC_LIBOBJ replacement of it.

These macros are obsolescent, as no current systems have the bug. New programs need not use these macros.

— Macro: AC_FUNC_SETVBUF_REVERSED

If setvbuf takes the buffering type as its second argument and the buffer pointer as the third, instead of the other way around, define SETVBUF_REVERSED.

This macro is obsolescent, as no current systems have the bug. New programs need not use this macro.

— Macro: AC_FUNC_STRCOLL

If the strcoll function exists and works correctly, define HAVE_STRCOLL. This does a bit more than ‘AC_CHECK_FUNCS(strcoll)’, because some systems have incorrect definitions of strcoll that should not be used.

— Macro: AC_FUNC_STRERROR_R

If strerror_r is available, define HAVE_STRERROR_R, and if it is declared, define HAVE_DECL_STRERROR_R. If it returns a char * message, define STRERROR_R_CHAR_P; otherwise it returns an int error number. The Thread-Safe Functions option of Posix requires strerror_r to return int, but many systems (including, for example, version 2.2.4 of the GNU C Library) return a char * value that is not necessarily equal to the buffer argument.

— Macro: AC_FUNC_STRFTIME

Check for strftime in the intl library, for SCO Unix. Then, if strftime is available, define HAVE_STRFTIME.

This macro is obsolescent, as no current systems require the intl library for strftime. New programs need not use this macro.

— Macro: AC_FUNC_STRTOD

If the strtod function does not exist or doesn't work correctly, ask for an AC_LIBOBJ replacement of ‘strtod’. In this case, because strtod.c is likely to need ‘pow’, set the output variable POW_LIB to the extra library needed.

— Macro: AC_FUNC_STRTOLD

If the strtold function exists and conforms to C99, define HAVE_STRTOLD.

— Macro: AC_FUNC_STRNLEN

If the strnlen function is not available, or is buggy (like the one from AIX 4.3), require an AC_LIBOBJ replacement for it.

— Macro: AC_FUNC_UTIME_NULL

If ‘utime (file, NULL)’ sets file's timestamp to the present, define HAVE_UTIME_NULL.

This macro is obsolescent, as all current systems have a utime that behaves this way. New programs need not use this macro.

— Macro: AC_FUNC_VPRINTF

If vprintf is found, define HAVE_VPRINTF. Otherwise, if _doprnt is found, define HAVE_DOPRNT. (If vprintf is available, you may assume that vfprintf and vsprintf are also available.)

This macro is obsolescent, as all current systems have vprintf. New programs need not use this macro.

— Macro: AC_REPLACE_FNMATCH

If the fnmatch function does not conform to Posix (see AC_FUNC_FNMATCH), ask for its AC_LIBOBJ replacement.

The files fnmatch.c, fnmatch_loop.c, and fnmatch_.h in the AC_LIBOBJ replacement directory are assumed to contain a copy of the source code of GNU fnmatch. If necessary, this source code is compiled as an AC_LIBOBJ replacement, and the fnmatch_.h file is linked to fnmatch.h so that it can be included in place of the system <fnmatch.h>.


Previous: Particular Functions, Up: Library Functions

5.5.3 Generic Function Checks

These macros are used to find functions not covered by the “particular” test macros. If the functions might be in libraries other than the default C library, first call AC_CHECK_LIB for those libraries. If you need to check the behavior of a function as well as find out whether it is present, you have to write your own test for it (see Writing Tests).

— Macro: AC_CHECK_FUNC (function, [action-if-found], [action-if-not-found])

If C function function is available, run shell commands action-if-found, otherwise action-if-not-found. If you just want to define a symbol if the function is available, consider using AC_CHECK_FUNCS instead. This macro checks for functions with C linkage even when AC_LANG(C++) has been called, since C is more standardized than C++. (see Language Choice, for more information about selecting the language for checks.)

— Macro: AC_CHECK_FUNCS (function..., [action-if-found], [action-if-not-found])

For each function enumerated in the blank-or-newline-separated argument list, define HAVE_function (in all capitals) if it is available. If action-if-found is given, it is additional shell code to execute when one of the functions is found. You can give it a value of ‘break’ to break out of the loop on the first match. If action-if-not-found is given, it is executed when one of the functions is not found.

— Macro: AC_CHECK_FUNCS_ONCE (function...)

For each function enumerated in the blank-or-newline-separated argument list, define HAVE_function (in all capitals) if it is available. This is a once-only variant of AC_CHECK_FUNCS. It generates the checking code at most once, so that configure is smaller and faster; but the checks cannot be conditionalized and are always done once, early during the configure run.


Autoconf follows a philosophy that was formed over the years by those who have struggled for portability: isolate the portability issues in specific files, and then program as if you were in a Posix environment. Some functions may be missing or unfixable, and your package must be ready to replace them.

Suitable replacements for many such problem functions are available from Gnulib (see Gnulib).

— Macro: AC_LIBOBJ (function)

Specify that ‘function.c’ must be included in the executables to replace a missing or broken implementation of function.

Technically, it adds ‘function.$ac_objext’ to the output variable LIBOBJS if it is not already in, and calls AC_LIBSOURCE for ‘function.c’. You should not directly change LIBOBJS, since this is not traceable.

— Macro: AC_LIBSOURCE (file)

Specify that file might be needed to compile the project. If you need to know what files might be needed by a configure.ac, you should trace AC_LIBSOURCE. file must be a literal.

This macro is called automatically from AC_LIBOBJ, but you must call it explicitly if you pass a shell variable to AC_LIBOBJ. In that case, since shell variables cannot be traced statically, you must pass to AC_LIBSOURCE any possible files that the shell variable might cause AC_LIBOBJ to need. For example, if you want to pass a variable $foo_or_bar to AC_LIBOBJ that holds either "foo" or "bar", you should do:

          AC_LIBSOURCE([foo.c])
          AC_LIBSOURCE([bar.c])
          AC_LIBOBJ([$foo_or_bar])
     

There is usually a way to avoid this, however, and you are encouraged to simply call AC_LIBOBJ with literal arguments.

Note that this macro replaces the obsolete AC_LIBOBJ_DECL, with slightly different semantics: the old macro took the function name, e.g., foo, as its argument rather than the file name.

— Macro: AC_LIBSOURCES (files)

Like AC_LIBSOURCE, but accepts one or more files in a comma-separated M4 list. Thus, the above example might be rewritten:

          AC_LIBSOURCES([foo.c, bar.c])
          AC_LIBOBJ([$foo_or_bar])
     
— Macro: AC_CONFIG_LIBOBJ_DIR (directory)

Specify that AC_LIBOBJ replacement files are to be found in directory, a name relative to the top level of the source tree. The replacement directory defaults to ., the top level directory, and the most typical value is lib, corresponding to ‘AC_CONFIG_LIBOBJ_DIR([lib])’.

configure might need to know the replacement directory for the following reasons: (i) some checks use the replacement files, (ii) some macros bypass broken system headers by installing links to the replacement headers (iii) when used in conjunction with Automake, within each makefile, directory is used as a relative path from $(top_srcdir) to each object named in LIBOBJS and LTLIBOBJS, etc.


It is common to merely check for the existence of a function, and ask for its AC_LIBOBJ replacement if missing. The following macro is a convenient shorthand.
— Macro: AC_REPLACE_FUNCS (function...)

Like AC_CHECK_FUNCS, but uses ‘AC_LIBOBJ(function)’ as action-if-not-found. You can declare your replacement function by enclosing the prototype in ‘#if !HAVE_function’. If the system has the function, it probably declares it in a header file you should be including, so you shouldn't redeclare it lest your declaration conflict.


Next: , Previous: Library Functions, Up: Existing Tests

5.6 Header Files

The following macros check for the presence of certain C header files. If there is no macro specifically defined to check for a header file you need, and you don't need to check for any special properties of it, then you can use one of the general header-file check macros.


Next: , Up: Header Files

5.6.1 Portability of Headers

This section tries to collect knowledge about common headers, and the problems they cause. By definition, this list always requires additions. Please help us keeping it as complete as possible.

limits.h
C99 says that limits.h defines LLONG_MIN, LLONG_MAX, and ULLONG_MAX, but many almost-C99 environments (e.g., default GCC 4.0.2 + glibc 2.4) do not define them.
inttypes.h vs. stdint.h
The C99 standard says that inttypes.h includes stdint.h, so there's no need to include stdint.h separately in a standard environment. Some implementations have inttypes.h but not stdint.h (e.g., Solaris 7), but we don't know of any implementation that has stdint.h but not inttypes.h.
linux/irda.h
It requires linux/types.h and sys/socket.h.
linux/random.h
It requires linux/types.h.
net/if.h
On Darwin, this file requires that sys/socket.h be included beforehand. One should run:
          AC_CHECK_HEADERS([sys/socket.h])
          AC_CHECK_HEADERS([net/if.h], [], [],
          [#include <stdio.h>
          #if STDC_HEADERS
          # include <stdlib.h>
          # include <stddef.h>
          #else
          # if HAVE_STDLIB_H
          #  include <stdlib.h>
          # endif
          #endif
          #if HAVE_SYS_SOCKET_H
          # include <sys/socket.h>
          #endif
          ])
     

netinet/if_ether.h
On Darwin, this file requires that stdio.h and sys/socket.h be included beforehand. One should run:
          AC_CHECK_HEADERS([sys/socket.h])
          AC_CHECK_HEADERS([netinet/if_ether.h], [], [],
          [#include <stdio.h>
          #if STDC_HEADERS
          # include <stdlib.h>
          # include <stddef.h>
          #else
          # if HAVE_STDLIB_H
          #  include <stdlib.h>
          # endif
          #endif
          #if HAVE_SYS_SOCKET_H
          # include <sys/socket.h>
          #endif
          ])
     

stdint.h
See above, item inttypes.h vs. stdint.h.
stdlib.h
On many systems (e.g., Darwin), stdio.h is a prerequisite.
sys/mount.h
On FreeBSD 4.8 on ia32 and using gcc version 2.95.4, sys/params.h is a prerequisite.
sys/ptem.h
On Solaris 8, sys/stream.h is a prerequisite.
sys/socket.h
On Darwin, stdlib.h is a prerequisite.
sys/ucred.h
On HP Tru64 5.1, sys/types.h is a prerequisite.
X11/extensions/scrnsaver.h
Using XFree86, this header requires X11/Xlib.h, which is probably so required that you might not even consider looking for it.
          AC_CHECK_HEADERS([X11/extensions/scrnsaver.h], [], [],
          [[#include <X11/Xlib.h>
          ]])
     


Next: , Previous: Header Portability, Up: Header Files

5.6.2 Particular Header Checks

These macros check for particular system header files—whether they exist, and in some cases whether they declare certain symbols.

— Macro: AC_HEADER_ASSERT

Check whether to enable assertions in the style of assert.h. Assertions are enabled by default, but the user can override this by invoking configure with the --disable-assert option.

— Macro: AC_HEADER_DIRENT

Check for the following header files. For the first one that is found and defines ‘DIR’, define the listed C preprocessor macro:

dirent.h HAVE_DIRENT_H
sys/ndir.h HAVE_SYS_NDIR_H
sys/dir.h HAVE_SYS_DIR_H
ndir.h HAVE_NDIR_H

The directory-library declarations in your source code should look something like the following:

          #include <sys/types.h>
          #ifdef HAVE_DIRENT_H
          # include <dirent.h>
          # define NAMLEN(dirent) strlen ((dirent)->d_name)
          #else
          # define dirent direct
          # define NAMLEN(dirent) ((dirent)->d_namlen)
          # if HAVE_SYS_NDIR_H
          #  include <sys/ndir.h>
          # endif
          # if HAVE_SYS_DIR_H
          #  include <sys/dir.h>
          # endif
          # if HAVE_NDIR_H
          #  include <ndir.h>
          # endif
          #endif
     

Using the above declarations, the program would declare variables to be of type struct dirent, not struct direct, and would access the length of a directory entry name by passing a pointer to a struct dirent to the NAMLEN macro.

This macro also checks for the SCO Xenix dir and x libraries.

This macro is obsolescent, as all current systems with directory libraries have <dirent.h>. New programs need not use this macro.

Also see AC_STRUCT_DIRENT_D_INO and AC_STRUCT_DIRENT_D_TYPE (see Particular Structures).

— Macro: AC_HEADER_MAJOR

If sys/types.h does not define major, minor, and makedev, but sys/mkdev.h does, define MAJOR_IN_MKDEV; otherwise, if sys/sysmacros.h does, define MAJOR_IN_SYSMACROS.

— Macro: AC_HEADER_RESOLV

Checks for header resolv.h, checking for prerequisites first. To properly use resolv.h, your code should contain something like the following:

     
     #if HAVE_SYS_TYPES_H
     #  include <sys/types.h>
     #endif
     #ifdef HAVE_NETINET_IN_H
     #  include <netinet/in.h>   /* inet_ functions / structs */
     #endif
     #ifdef HAVE_ARPA_NAMESER_H
     #  include <arpa/nameser.h> /* DNS HEADER struct */
     #endif
     #ifdef HAVE_NETDB_H
     #  include <netdb.h>
     #endif
     #include <resolv.h>

— Macro: AC_HEADER_STAT

If the macros S_ISDIR, S_ISREG, etc. defined in sys/stat.h do not work properly (returning false positives), define STAT_MACROS_BROKEN. This is the case on Tektronix UTekV, Amdahl UTS and Motorola System V/88.

This macro is obsolescent, as no current systems have the bug. New programs need not use this macro.

— Macro: AC_HEADER_STDBOOL

If stdbool.h exists and conforms to C99, define HAVE_STDBOOL_H to 1; if the type _Bool is defined, define HAVE__BOOL to 1. To fulfill the C99 requirements, your system.h could contain the following code:

     
     #if HAVE_STDBOOL_H
     # include <stdbool.h>
     #else
     # if ! HAVE__BOOL
     #  ifdef __cplusplus
     typedef bool _Bool;
     #  else
     #   define _Bool signed char
     #  endif
     # endif
     # define bool _Bool
     # define false 0
     # define true 1
     # define __bool_true_false_are_defined 1
     #endif

Alternatively you can use the ‘stdbool’ package of Gnulib (see Gnulib); it packages the above code into a replacement header and contains a few other bells and whistles.

— Macro: AC_HEADER_STDC

Define STDC_HEADERS if the system has C header files conforming to ANSI C89 (ISO C90). Specifically, this macro checks for stdlib.h, stdarg.h, string.h, and float.h; if the system has those, it probably has the rest of the C89 header files. This macro also checks whether string.h declares memchr (and thus presumably the other mem functions), whether stdlib.h declare free (and thus presumably malloc and other related functions), and whether the ctype.h macros work on characters with the high bit set, as the C standard requires.

If you use this macro, your code can refer to STDC_HEADERS to determine whether the system has conforming header files (and probably C library functions).

This macro is obsolescent, as current systems have conforming header files. New programs need not use this macro.

Nowadays string.h is part of the C standard and declares functions like strcpy, and strings.h is standardized by Posix and declares BSD functions like bcopy; but historically, string functions were a major sticking point in this area. If you still want to worry about portability to ancient systems without standard headers, there is so much variation that it is probably easier to declare the functions you use than to figure out exactly what the system header files declare. Some ancient systems contained a mix of functions from the C standard and from BSD; some were mostly standard but lacked ‘memmove’; some defined the BSD functions as macros in string.h or strings.h; some had only the BSD functions but string.h; some declared the memory functions in memory.h, some in string.h; etc. It is probably sufficient to check for one string function and one memory function; if the library had the standard versions of those then it probably had most of the others. If you put the following in configure.ac:

          # This example is obsolescent.
          # Nowadays you can omit these macro calls.
          AC_HEADER_STDC
          AC_CHECK_FUNCS([strchr memcpy])
     

then, in your code, you can use declarations like this:

          /* This example is obsolescent.
             Nowadays you can just #include <string.h>.  */
          #if STDC_HEADERS
          # include <string.h>
          #else
          # if !HAVE_STRCHR
          #  define strchr index
          #  define strrchr rindex
          # endif
          char *strchr (), *strrchr ();
          # if !HAVE_MEMCPY
          #  define memcpy(d, s, n) bcopy ((s), (d), (n))
          #  define memmove(d, s, n) bcopy ((s), (d), (n))
          # endif
          #endif
     

If you use a function like memchr, memset, strtok, or strspn, which have no BSD equivalent, then macros don't suffice to port to ancient hosts; you must provide an implementation of each function. An easy way to incorporate your implementations only when needed (since the ones in system C libraries may be hand optimized) is to, taking memchr for example, put it in memchr.c and use ‘AC_REPLACE_FUNCS([memchr])’.

— Macro: AC_HEADER_SYS_WAIT

If sys/wait.h exists and is compatible with Posix, define HAVE_SYS_WAIT_H. Incompatibility can occur if sys/wait.h does not exist, or if it uses the old BSD union wait instead of int to store a status value. If sys/wait.h is not Posix compatible, then instead of including it, define the Posix macros with their usual interpretations. Here is an example:

          #include <sys/types.h>
          #if HAVE_SYS_WAIT_H
          # include <sys/wait.h>
          #endif
          #ifndef WEXITSTATUS
          # define WEXITSTATUS(stat_val) ((unsigned int) (stat_val) >> 8)
          #endif
          #ifndef WIFEXITED
          # define WIFEXITED(stat_val) (((stat_val) & 255) == 0)
          #endif
     

This macro is obsolescent, as current systems are compatible with Posix. New programs need not use this macro.

_POSIX_VERSION is defined when unistd.h is included on Posix systems. If there is no unistd.h, it is definitely not a Posix system. However, some non-Posix systems do have unistd.h.

The way to check whether the system supports Posix is:

     #if HAVE_UNISTD_H
     # include <sys/types.h>
     # include <unistd.h>
     #endif
     
     #ifdef _POSIX_VERSION
     /* Code for Posix systems.  */
     #endif
— Macro: AC_HEADER_TIME

If a program may include both time.h and sys/time.h, define TIME_WITH_SYS_TIME. On some ancient systems, sys/time.h included time.h, but time.h was not protected against multiple inclusion, so programs could not explicitly include both files. This macro is useful in programs that use, for example, struct timeval as well as struct tm. It is best used in conjunction with HAVE_SYS_TIME_H, which can be checked for using AC_CHECK_HEADERS([sys/time.h]).

          #if TIME_WITH_SYS_TIME
          # include <sys/time.h>
          # include <time.h>
          #else
          # if HAVE_SYS_TIME_H
          #  include <sys/time.h>
          # else
          #  include <time.h>
          # endif
          #endif
     

This macro is obsolescent, as current systems can include both files when they exist. New programs need not use this macro.

— Macro: AC_HEADER_TIOCGWINSZ

If the use of TIOCGWINSZ requires <sys/ioctl.h>, then define GWINSZ_IN_SYS_IOCTL. Otherwise TIOCGWINSZ can be found in <termios.h>.

Use:

          #if HAVE_TERMIOS_H
          # include <termios.h>
          #endif
          
          #if GWINSZ_IN_SYS_IOCTL
          # include <sys/ioctl.h>
          #endif
     


Previous: Particular Headers, Up: Header Files

5.6.3 Generic Header Checks

These macros are used to find system header files not covered by the “particular” test macros. If you need to check the contents of a header as well as find out whether it is present, you have to write your own test for it (see Writing Tests).

— Macro: AC_CHECK_HEADER (header-file, [action-if-found], [action-if-not-found], [includes = ‘default-includes])

If the system header file header-file is compilable, execute shell commands action-if-found, otherwise execute action-if-not-found. If you just want to define a symbol if the header file is available, consider using AC_CHECK_HEADERS instead.

For compatibility issues with older versions of Autoconf, please read below.

— Macro: AC_CHECK_HEADERS (header-file..., [action-if-found], [action-if-not-found], [includes = ‘default-includes])

For each given system header file header-file in the blank-separated argument list that exists, define HAVE_header-file (in all capitals). If action-if-found is given, it is additional shell code to execute when one of the header files is found. You can give it a value of ‘break’ to break out of the loop on the first match. If action-if-not-found is given, it is executed when one of the header files is not found.

For compatibility issues with older versions of Autoconf, please read below.

Previous versions of Autoconf merely checked whether the header was accepted by the preprocessor. This was changed because the old test was inappropriate for typical uses. Headers are typically used to compile, not merely to preprocess, and the old behavior sometimes accepted headers that clashed at compile-time. If you need to check whether a header is preprocessable, you can use AC_PREPROC_IFELSE (see Running the Preprocessor).

This scheme, which improves the robustness of the test, also requires that you make sure that headers that must be included before the header-file be part of the includes, (see Default Includes). If looking for bar.h, which requires that foo.h be included before if it exists, we suggest the following scheme:

AC_CHECK_HEADERS([foo.h])
AC_CHECK_HEADERS([bar.h], [], [],
[#if HAVE_FOO_H
# include <foo.h>
# endif
])

The following variant generates smaller, faster configure files if you do not need the full power of AC_CHECK_HEADERS.

— Macro: AC_CHECK_HEADERS_ONCE (header-file...)

For each given system header file header-file in the blank-separated argument list that exists, define HAVE_header-file (in all capitals). This is a once-only variant of AC_CHECK_HEADERS. It generates the checking code at most once, so that configure is smaller and faster; but the checks cannot be conditionalized and are always done once, early during the configure run.


Next: , Previous: Header Files, Up: Existing Tests

5.7 Declarations

The following macros check for the declaration of variables and functions. If there is no macro specifically defined to check for a symbol you need, then you can use the general macros (see Generic Declarations) or, for more complex tests, you may use AC_COMPILE_IFELSE (see Running the Compiler).


Next: , Up: Declarations

5.7.1 Particular Declaration Checks

There are no specific macros for declarations.


Previous: Particular Declarations, Up: Declarations

5.7.2 Generic Declaration Checks

These macros are used to find declarations not covered by the “particular” test macros.

— Macro: AC_CHECK_DECL (symbol, [action-if-found], [action-if-not-found], [includes = ‘default-includes])

If symbol (a function or a variable) is not declared in includes and a declaration is needed, run the shell commands action-if-not-found, otherwise action-if-found. If no includes are specified, the default includes are used (see Default Includes).

This macro actually tests whether it is valid to use symbol as an r-value, not if it is really declared, because it is much safer to avoid introducing extra declarations when they are not needed.

— Macro: AC_CHECK_DECLS (symbols, [action-if-found], [action-if-not-found], [includes = ‘default-includes])

For each of the symbols (comma-separated list), define HAVE_DECL_symbol (in all capitals) to ‘1’ if symbol is declared, otherwise to ‘0’. If action-if-not-found is given, it is additional shell code to execute when one of the function declarations is needed, otherwise action-if-found is executed.

This macro uses an M4 list as first argument:

          AC_CHECK_DECLS([strdup])
          AC_CHECK_DECLS([strlen])
          AC_CHECK_DECLS([malloc, realloc, calloc, free])
     

Unlike the other ‘AC_CHECK_*S’ macros, when a symbol is not declared, HAVE_DECL_symbol is defined to ‘0’ instead of leaving HAVE_DECL_symbol undeclared. When you are sure that the check was performed, use HAVE_DECL_symbol just like any other result of Autoconf:

          #if !HAVE_DECL_SYMBOL
          extern char *symbol;
          #endif
     

If the test may have not been performed, however, because it is safer not to declare a symbol than to use a declaration that conflicts with the system's one, you should use:

          #if defined HAVE_DECL_MALLOC && !HAVE_DECL_MALLOC
          void *malloc (size_t *s);
          #endif
     

You fall into the second category only in extreme situations: either your files may be used without being configured, or they are used during the configuration. In most cases the traditional approach is enough.

— Macro: AC_CHECK_DECLS_ONCE (symbols)

For each of the symbols (comma-separated list), define HAVE_DECL_symbol (in all capitals) to ‘1’ if symbol is declared in the default include files, otherwise to ‘0’. This is a once-only variant of AC_CHECK_DECLS. It generates the checking code at most once, so that configure is smaller and faster; but the checks cannot be conditionalized and are always done once, early during the configure run.


Next: , Previous: Declarations, Up: Existing Tests

5.8 Structures

The following macros check for the presence of certain members in C structures. If there is no macro specifically defined to check for a member you need, then you can use the general structure-member macros (see Generic Structures) or, for more complex tests, you may use AC_COMPILE_IFELSE (see Running the Compiler).


Next: , Up: Structures

5.8.1 Particular Structure Checks

The following macros check for certain structures or structure members.

— Macro: AC_STRUCT_DIRENT_D_INO

Perform all the actions of AC_HEADER_DIRENT (see Particular Headers). Then, if struct dirent contains a d_ino member, define HAVE_STRUCT_DIRENT_D_INO.

HAVE_STRUCT_DIRENT_D_INO indicates only the presence of d_ino, not whether its contents are always reliable. Traditionally, a zero d_ino indicated a deleted directory entry, though current systems hide this detail from the user and never return zero d_ino values. Many current systems report an incorrect d_ino for a directory entry that is a mount point.

— Macro: AC_STRUCT_DIRENT_D_TYPE

Perform all the actions of AC_HEADER_DIRENT (see Particular Headers). Then, if struct dirent contains a d_type member, define HAVE_STRUCT_DIRENT_D_TYPE.

— Macro: AC_STRUCT_ST_BLKSIZE

If struct stat contains an st_blksize member, define HAVE_STRUCT_STAT_ST_BLKSIZE. The former name, HAVE_ST_BLKSIZE is to be avoided, as its support will cease in the future. This macro is obsoleted, and should be replaced by

          AC_CHECK_MEMBERS([struct stat.st_blksize])
     
— Macro: AC_STRUCT_ST_BLOCKS

If struct stat contains an st_blocks member, define HAVE_STRUCT_STAT_ST_BLOCKS. Otherwise, require an AC_LIBOBJ replacement of ‘fileblocks’. The former name, HAVE_ST_BLOCKS is to be avoided, as its support will cease in the future.

— Macro: AC_STRUCT_ST_RDEV

If struct stat contains an st_rdev member, define HAVE_STRUCT_STAT_ST_RDEV. The former name for this macro, HAVE_ST_RDEV, is to be avoided as it will cease to be supported in the future. Actually, even the new macro is obsolete and should be replaced by:

          AC_CHECK_MEMBERS([struct stat.st_rdev])
     
— Macro: AC_STRUCT_TM

If time.h does not define struct tm, define TM_IN_SYS_TIME, which means that including sys/time.h had better define struct tm.

This macro is obsolescent, as time.h defines struct tm in current systems. New programs need not use this macro.

— Macro: AC_STRUCT_TIMEZONE

Figure out how to get the current timezone. If struct tm has a tm_zone member, define HAVE_STRUCT_TM_TM_ZONE (and the obsoleted HAVE_TM_ZONE). Otherwise, if the external array tzname is found, define HAVE_TZNAME; if it is declared, define HAVE_DECL_TZNAME.


Previous: Particular Structures, Up: Structures

5.8.2 Generic Structure Checks

These macros are used to find structure members not covered by the “particular” test macros.

— Macro: AC_CHECK_MEMBER (aggregate.member, [action-if-found], [action-if-not-found], [includes = ‘default-includes])

Check whether member is a member of the aggregate aggregate. If no includes are specified, the default includes are used (see Default Includes).

          AC_CHECK_MEMBER([struct passwd.pw_gecos], [],
                          [AC_MSG_ERROR([We need `passwd.pw_gecos'!])],
                          [#include <pwd.h>])
     

You can use this macro for submembers:

          AC_CHECK_MEMBER(struct top.middle.bot)
     
— Macro: AC_CHECK_MEMBERS (members, [action-if-found], [action-if-not-found], [includes = ‘default-includes])

Check for the existence of each ‘aggregate.member’ of members using the previous macro. When member belongs to aggregate, define HAVE_aggregate_member (in all capitals, with spaces and dots replaced by underscores). If action-if-found is given, it is executed for each of the found members. If action-if-not-found is given, it is executed for each of the members that could not be found.

This macro uses M4 lists:

          AC_CHECK_MEMBERS([struct stat.st_rdev, struct stat.st_blksize])
     


Next: , Previous: Structures, Up: Existing Tests

5.9 Types

The following macros check for C types, either builtin or typedefs. If there is no macro specifically defined to check for a type you need, and you don't need to check for any special properties of it, then you can use a general type-check macro.


Next: , Up: Types

5.9.1 Particular Type Checks

These macros check for particular C types in sys/types.h, stdlib.h, stdint.h, inttypes.h and others, if they exist.

The Gnulib stdint module is an alternate way to define many of these symbols; it is useful if you prefer your code to assume a C99-or-better environment. See Gnulib.

— Macro: AC_TYPE_GETGROUPS

Define GETGROUPS_T to be whichever of gid_t or int is the base type of the array argument to getgroups.

— Macro: AC_TYPE_INT8_T

If stdint.h or inttypes.h defines the type int8_t, define HAVE_INT8_T. Otherwise, define int8_t to a signed integer type that is exactly 8 bits wide and that uses two's complement representation, if such a type exists.

— Macro: AC_TYPE_INT16_T

This is like AC_TYPE_INT8_T, except for 16-bit integers.

— Macro: AC_TYPE_INT32_T

This is like AC_TYPE_INT8_T, except for 32-bit integers.

— Macro: AC_TYPE_INT64_T

This is like AC_TYPE_INT8_T, except for 64-bit integers.

— Macro: AC_TYPE_INTMAX_T

If stdint.h or inttypes.h defines the type intmax_t, define HAVE_INTMAX_T. Otherwise, define intmax_t to the widest signed integer type.

— Macro: AC_TYPE_INTPTR_T

If stdint.h or inttypes.h defines the type intptr_t, define HAVE_INTPTR_T. Otherwise, define intptr_t to a signed integer type wide enough to hold a pointer, if such a type exists.

— Macro: AC_TYPE_LONG_DOUBLE

If the C compiler supports a working long double type, define HAVE_LONG_DOUBLE. The long double type might have the same range and precision as double.

— Macro: AC_TYPE_LONG_DOUBLE_WIDER

If the C compiler supports a working long double type with more range or precision than the double type, define HAVE_LONG_DOUBLE_WIDER.

— Macro: AC_TYPE_LONG_LONG_INT

If the C compiler supports a working long long int type, define HAVE_LONG_LONG_INT.

— Macro: AC_TYPE_MBSTATE_T

Define HAVE_MBSTATE_T if <wchar.h> declares the mbstate_t type. Also, define mbstate_t to be a type if <wchar.h> does not declare it.

— Macro: AC_TYPE_MODE_T

Define mode_t to a suitable type, if standard headers do not define it.

— Macro: AC_TYPE_OFF_T

Define off_t to a suitable type, if standard headers do not define it.

— Macro: AC_TYPE_PID_T

Define pid_t to a suitable type, if standard headers do not define it.

— Macro: AC_TYPE_SIGNAL

If signal.h declares signal as returning a pointer to a function returning void, define RETSIGTYPE to be void; otherwise, define it to be int.

Define signal handlers as returning type RETSIGTYPE:

          RETSIGTYPE
          hup_handler ()
          {
          ...
          }
     
— Macro: AC_TYPE_SIZE_T

Define size_t to a suitable type, if standard headers do not define it.

— Macro: AC_TYPE_SSIZE_T

Define ssize_t to a suitable type, if standard headers do not define it.

— Macro: AC_TYPE_UID_T

Define uid_t and gid_t to suitable types, if standard headers do not define them.

— Macro: AC_TYPE_UINT8_T

If stdint.h or inttypes.h defines the type uint8_t, define HAVE_UINT8_T. Otherwise, define uint8_t to an unsigned integer type that is exactly 8 bits wide, if such a type exists.

— Macro: AC_TYPE_UINT16_T

This is like AC_TYPE_UINT8_T, except for 16-bit unsigned integers.

— Macro: AC_TYPE_UINT32_T

This is like AC_TYPE_UINT8_T, except for 32-bit unsigned integers.

— Macro: AC_TYPE_UINT64_T

This is like AC_TYPE_UINT8_T, except for 64-bit unsigned integers.

— Macro: AC_TYPE_UINTMAX_T

If stdint.h or inttypes.h defines the type uintmax_t, define HAVE_UINTMAX_T. Otherwise, define uintmax_t to the widest unsigned integer type.

— Macro: AC_TYPE_UINTPTR_T

If stdint.h or inttypes.h defines the type uintptr_t, define HAVE_UINTPTR_T. Otherwise, define uintptr_t to an unsigned integer type wide enough to hold a pointer, if such a type exists.

— Macro: AC_TYPE_UNSIGNED_LONG_LONG_INT

If the C compiler supports a working unsigned long long int type, define HAVE_UNSIGNED_LONG_LONG_INT.


Previous: Particular Types, Up: Types

5.9.2 Generic Type Checks

These macros are used to check for types not covered by the “particular” test macros.

— Macro: AC_CHECK_TYPE (type, [action-if-found], [action-if-not-found], [includes = ‘default-includes])

Check whether type is defined. It may be a compiler builtin type or defined by the includes (see Default Includes).

— Macro: AC_CHECK_TYPES (types, [action-if-found], [action-if-not-found], [includes = ‘default-includes])

For each type of the types that is defined, define HAVE_type (in all capitals). If no includes are specified, the default includes are used (see Default Includes). If action-if-found is given, it is additional shell code to execute when one of the types is found. If action-if-not-found is given, it is executed when one of the types is not found.

This macro uses M4 lists:

          AC_CHECK_TYPES([ptrdiff_t])
          AC_CHECK_TYPES([unsigned long long int, uintmax_t])
     

Autoconf, up to 2.13, used to provide to another version of AC_CHECK_TYPE, broken by design. In order to keep backward compatibility, a simple heuristics, quite safe but not totally, is implemented. In case of doubt, read the documentation of the former AC_CHECK_TYPE, see Obsolete Macros.


Next: , Previous: Types, Up: Existing Tests

5.10 Compilers and Preprocessors

All the tests for compilers (AC_PROG_CC, AC_PROG_CXX, AC_PROG_F77) define the output variable EXEEXT based on the output of the compiler, typically to the empty string if Posix and ‘.exe’ if a DOS variant.

They also define the output variable OBJEXT based on the output of the compiler, after .c files have been excluded, typically to ‘o’ if Posix, ‘obj’ if a DOS variant.

If the compiler being used does not produce executables, the tests fail. If the executables can't be run, and cross-compilation is not enabled, they fail too. See Manual Configuration, for more on support for cross compiling.


Next: , Up: Compilers and Preprocessors

5.10.1 Specific Compiler Characteristics

Some compilers exhibit different behaviors.

Static/Dynamic Expressions
Autoconf relies on a trick to extract one bit of information from the C compiler: using negative array sizes. For instance the following excerpt of a C source demonstrates how to test whether ‘int’ objects are 4 bytes wide:
          int
          main (void)
          {
            static int test_array [sizeof (int) == 4 ? 1 : -1];
            test_array [0] = 0;
            return 0;
          }
     

To our knowledge, there is a single compiler that does not support this trick: the HP C compilers (the real one, not only the “bundled”) on HP-UX 11.00. They incorrectly reject the above program with the diagnostic “Variable-length arrays cannot have static storage.” This bug comes from HP compilers' mishandling of sizeof (int), not from the ? 1 : -1, and Autoconf works around this problem by casting sizeof (int) to long int before comparing it.


Next: , Previous: Specific Compiler Characteristics, Up: Compilers and Preprocessors

5.10.2 Generic Compiler Characteristics

— Macro: AC_CHECK_SIZEOF (type, [unused], [includes = ‘default-includes])

Define SIZEOF_type (see Standard Symbols) to be the size in bytes of type. If ‘type’ is unknown, it gets a size of 0. If no includes are specified, the default includes are used (see Default Includes). If you provide include, be sure to include stdio.h which is required for this macro to run.

This macro now works even when cross-compiling. The unused argument was used when cross-compiling.

For example, the call

          AC_CHECK_SIZEOF([int *])
     

defines SIZEOF_INT_P to be 8 on DEC Alpha AXP systems.

— Macro: AC_CHECK_ALIGNOF (type, [includes = ‘default-includes])

Define ALIGNOF_type (see Standard Symbols) to be the alignment in bytes of type. If ‘type’ is unknown, it gets a size of 0. If no includes are specified, the default includes are used (see Default Includes). If you provide include, be sure to include stddef.h and stdio.h which are required for this macro to work correctly.

— Macro: AC_LANG_WERROR

Normally Autoconf ignores warnings generated by the compiler, linker, and preprocessor. If this macro is used, warnings count as fatal errors for the current language. This macro is useful when the results of configuration are used where warnings are unacceptable; for instance, if parts of a program are built with the GCC -Werror option. If the whole program is built using -Werror it is often simpler to put -Werror in the compiler flags (CFLAGS, etc.).


Next: , Previous: Generic Compiler Characteristics, Up: Compilers and Preprocessors

5.10.3 C Compiler Characteristics

The following macros provide ways to find and exercise a C Compiler. There are a few constructs that ought to be avoided, but do not deserve being checked for, since they can easily be worked around.

Don't use lines containing solitary backslashes
They tickle a bug in the HP-UX C compiler (checked on HP-UX 10.20, 11.00, and 11i). When given the following source:
          #ifdef __STDC__
          /\
          * A comment with backslash-newlines in it.  %{ %} *\
          \
          /
          char str[] = "\\
          " A string with backslash-newlines in it %{ %} \\
          "";
          char apostrophe = '\\
          \
          '\
          ';
          #endif
     

the compiler incorrectly fails with the diagnostics “Non-terminating comment at end of file” and “Missing ‘#endif’ at end of file.” Removing the lines with solitary backslashes solves the problem.

Don't compile several files at once if output matters to you
Some compilers, such as the HP's, reports the name of the file it is compiling when they are several. For instance:
          $ cc a.c b.c
          a.c:
          b.c:
     

This can cause problems if you observe the output of the compiler to detect failures. Invoking ‘cc -c a.c && cc -c b.c && cc -o c a.o b.o’ solves the issue.

Don't rely on #error failing
The irix C compiler does not fail when #error is preprocessed; it simply emits a diagnostic and continues, exiting successfully. So, instead of an error directive like #error "Unsupported word size" it is more portable to use an invalid directive like #Unsupported word size in Autoconf tests. In ordinary source code, #error is OK, since installers with inadequate compilers like irix can simply examine these compilers' diagnostic output.
Don't rely on correct #line support
On Solaris, c89 (at least Sun C 5.3 through 5.8) diagnoses #line directives whose line numbers are greater than 32767. Nothing in Posix makes this invalid. That is why Autoconf stopped issuing #line directives.
— Macro: AC_PROG_CC ([compiler-search-list])

Determine a C compiler to use. If CC is not already set in the environment, check for gcc and cc, then for other C compilers. Set output variable CC to the name of the compiler found.

This macro may, however, be invoked with an optional first argument which, if specified, must be a blank-separated list of C compilers to search for. This just gives the user an opportunity to specify an alternative search list for the C compiler. For example, if you didn't like the default order, then you could invoke AC_PROG_CC like this:

          AC_PROG_CC([gcc cl cc])
     

If the C compiler does not handle function prototypes correctly by default, try to add an option to output variable CC to make it so. This macro tries various options that select standard-conformance modes on various systems.

After calling this macro you can check whether the C compiler has been set to accept ANSI C89 (ISO C90); if not, the shell variable ac_cv_prog_cc_c89 is set to ‘no’. See also AC_C_PROTOTYPES below.

If using the GNU C compiler, set shell variable GCC to ‘yes’. If output variable CFLAGS was not already set, set it to -g -O2 for the GNU C compiler (-O2 on systems where GCC does not accept -g), or -g for other compilers.

— Macro: AC_PROG_CC_C_O

If the C compiler does not accept the -c and -o options simultaneously, define NO_MINUS_C_MINUS_O. This macro actually tests both the compiler found by AC_PROG_CC, and, if different, the first cc in the path. The test fails if one fails. This macro was created for GNU Make to choose the default C compilation rule.

— Macro: AC_PROG_CPP

Set output variable CPP to a command that runs the C preprocessor. If ‘$CC -E’ doesn't work, /lib/cpp is used. It is only portable to run CPP on files with a .c extension.

Some preprocessors don't indicate missing include files by the error status. For such preprocessors an internal variable is set that causes other macros to check the standard error from the preprocessor and consider the test failed if any warnings have been reported. For most preprocessors, though, warnings do not cause include-file tests to fail unless AC_PROG_CPP_WERROR is also specified.

— Macro: AC_PROG_CPP_WERROR

This acts like AC_PROG_CPP, except it treats warnings from the preprocessor as errors even if the preprocessor exit status indicates success. This is useful for avoiding headers that generate mandatory warnings, such as deprecation notices.

The following macros check for C compiler or machine architecture features. To check for characteristics not listed here, use AC_COMPILE_IFELSE (see Running the Compiler) or AC_RUN_IFELSE (see Runtime).

— Macro: AC_PROG_CC_STDC

If the C compiler cannot compile ISO Standard C (currently C99), try to add an option to output variable CC to make it work. If the compiler does not support C99, fall back to supporting ANSI C89 (ISO C90).

After calling this macro you can check whether the C compiler has been set to accept Standard C; if not, the shell variable ac_cv_prog_cc_stdc is set to ‘no’.

— Macro: AC_PROG_CC_C89

If the C compiler is not in ANSI C89 (ISO C90) mode by default, try to add an option to output variable CC to make it so. This macro tries various options that select ANSI C89 on some system or another. It considers the compiler to be in ANSI C89 mode if it handles function prototypes correctly.

After calling this macro you can check whether the C compiler has been set to accept ANSI C89; if not, the shell variable ac_cv_prog_cc_c89 is set to ‘no’.

This macro is called automatically by AC_PROG_CC.

— Macro: AC_PROG_CC_C99

If the C compiler is not in C99 mode by default, try to add an option to output variable CC to make it so. This macro tries various options that select C99 on some system or another. It considers the compiler to be in C99 mode if it handles _Bool, flexible arrays, inline, long long int, mixed code and declarations, named initialization of structs, restrict, varargs macros, variable declarations in for loops and variable length arrays.

After calling this macro you can check whether the C compiler has been set to accept C99; if not, the shell variable ac_cv_prog_cc_c99 is set to ‘no’.

— Macro: AC_C_BACKSLASH_A

Define ‘HAVE_C_BACKSLASH_A’ to 1 if the C compiler understands ‘\a’.

This macro is obsolescent, as current C compilers understand ‘\a’. New programs need not use this macro.

— Macro: AC_C_BIGENDIAN ([action-if-true], [action-if-false], [action-if-unknown])

If words are stored with the most significant byte first (like Motorola and SPARC CPUs), execute action-if-true. If words are stored with the least significant byte first (like Intel and VAX CPUs), execute action-if-false.

This macro runs a test-case if endianness cannot be determined from the system header files. When cross-compiling, the test-case is not run but grep'ed for some magic values. action-if-unknown is executed if the latter case fails to determine the byte sex of the host system.

The default for action-if-true is to define ‘WORDS_BIGENDIAN’. The default for action-if-false is to do nothing. And finally, the default for action-if-unknown is to abort configure and tell the installer which variable he should preset to bypass this test.

— Macro: AC_C_CONST

If the C compiler does not fully support the const keyword, define const to be empty. Some C compilers that do not define __STDC__ do support const; some compilers that define __STDC__ do not completely support const. Programs can simply use const as if every C compiler supported it; for those that don't, the makefile or configuration header file defines it as empty.

Occasionally installers use a C++ compiler to compile C code, typically because they lack a C compiler. This causes problems with const, because C and C++ treat const differently. For example:

          const int foo;
     

is valid in C but not in C++. These differences unfortunately cannot be papered over by defining const to be empty.

If autoconf detects this situation, it leaves const alone, as this generally yields better results in practice. However, using a C++ compiler to compile C code is not recommended or supported, and installers who run into trouble in this area should get a C compiler like GCC to compile their C code.

This macro is obsolescent, as current C compilers support const. New programs need not use this macro.

— Macro: AC_C_RESTRICT

If the C compiler recognizes the restrict keyword, don't do anything. If it recognizes only a variant spelling (__restrict, __restrict__, or _Restrict), then define restrict to that. Otherwise, define restrict to be empty. Thus, programs may simply use restrict as if every C compiler supported it; for those that do not, the makefile or configuration header defines it away.

Although support in C++ for the restrict keyword is not required, several C++ compilers do accept the keyword. This macro works for them, too.

— Macro: AC_C_VOLATILE

If the C compiler does not understand the keyword volatile, define volatile to be empty. Programs can simply use volatile as if every C compiler supported it; for those that do not, the makefile or configuration header defines it as empty.

If the correctness of your program depends on the semantics of volatile, simply defining it to be empty does, in a sense, break your code. However, given that the compiler does not support volatile, you are at its mercy anyway. At least your program compiles, when it wouldn't before.

In general, the volatile keyword is a standard C feature, so you might expect that volatile is available only when __STDC__ is defined. However, Ultrix 4.3's native compiler does support volatile, but does not define __STDC__.

This macro is obsolescent, as current C compilers support volatile. New programs need not use this macro.

— Macro: AC_C_INLINE

If the C compiler supports the keyword inline, do nothing. Otherwise define inline to __inline__ or __inline if it accepts one of those, otherwise define inline to be empty.

— Macro: AC_C_CHAR_UNSIGNED

If the C type char is unsigned, define __CHAR_UNSIGNED__, unless the C compiler predefines it.

— Macro: AC_C_STRINGIZE

If the C preprocessor supports the stringizing operator, define HAVE_STRINGIZE. The stringizing operator is ‘#’ and is found in macros such as this:

          #define x(y) #y
     

This macro is obsolescent, as current C compilers support the stringizing operator. New programs need not use this macro.

— Macro: AC_C_TYPEOF

If the C compiler supports GCC's typeof syntax either directly or through a different spelling of the keyword (e.g., __typeof__), define HAVE_TYPEOF. If the support is available only through a different spelling, define typeof to that spelling.

— Macro: AC_C_PROTOTYPES

If function prototypes are understood by the compiler (as determined by AC_PROG_CC), define PROTOTYPES and __PROTOTYPES. Defining __PROTOTYPES is for the benefit of header files that cannot use macros that infringe on user name space.

This macro is obsolescent, as current C compilers support prototypes. New programs need not use this macro.

— Macro: AC_PROG_GCC_TRADITIONAL

Add -traditional to output variable CC if using the GNU C compiler and ioctl does not work properly without -traditional. That usually happens when the fixed header files have not been installed on an old system.

This macro is obsolescent, since current versions of the GNU C compiler fix the header files automatically when installed.


Next: , Previous: C Compiler, Up: Compilers and Preprocessors

5.10.4 C++ Compiler Characteristics

— Macro: AC_PROG_CXX ([compiler-search-list])

Determine a C++ compiler to use. Check whether the environment variable CXX or CCC (in that order) is set; if so, then set output variable CXX to its value.

Otherwise, if the macro is invoked without an argument, then search for a C++ compiler under the likely names (first g++ and c++ then other names). If none of those checks succeed, then as a last resort set CXX to g++.

This macro may, however, be invoked with an optional first argument which, if specified, must be a blank-separated list of C++ compilers to search for. This just gives the user an opportunity to specify an alternative search list for the C++ compiler. For example, if you didn't like the default order, then you could invoke AC_PROG_CXX like this:

          AC_PROG_CXX([gcc cl KCC CC cxx cc++ xlC aCC c++ g++])
     

If using the GNU C++ compiler, set shell variable GXX to ‘yes’. If output variable CXXFLAGS was not already set, set it to -g -O2 for the GNU C++ compiler (-O2 on systems where G++ does not accept -g), or -g for other compilers.

— Macro: AC_PROG_CXXCPP

Set output variable CXXCPP to a command that runs the C++ preprocessor. If ‘$CXX -E’ doesn't work, /lib/cpp is used. It is portable to run CXXCPP only on files with a .c, .C, .cc, or .cpp extension.

Some preprocessors don't indicate missing include files by the error status. For such preprocessors an internal variable is set that causes other macros to check the standard error from the preprocessor and consider the test failed if any warnings have been reported. However, it is not known whether such broken preprocessors exist for C++.

— Macro: AC_PROG_CXX_C_O

Test whether the C++ compiler accepts the options -c and -o simultaneously, and define CXX_NO_MINUS_C_MINUS_O, if it does not.


Next: , Previous: C++ Compiler, Up: Compilers and Preprocessors

5.10.5 Objective C Compiler Characteristics

— Macro: AC_PROG_OBJC ([compiler-search-list])

Determine an Objective C compiler to use. If OBJC is not already set in the environment, check for Objective C compilers. Set output variable OBJC to the name of the compiler found.

This macro may, however, be invoked with an optional first argument which, if specified, must be a blank-separated list of Objective C compilers to search for. This just gives the user an opportunity to specify an alternative search list for the Objective C compiler. For example, if you didn't like the default order, then you could invoke AC_PROG_OBJC like this:

          AC_PROG_OBJC([gcc objcc objc])
     

If using the GNU Objective C compiler, set shell variable GOBJC to ‘yes’. If output variable OBJCFLAGS was not already set, set it to -g -O2 for the GNU Objective C compiler (-O2 on systems where gcc does not accept -g), or -g for other compilers.

— Macro: AC_PROG_OBJCCPP

Set output variable OBJCCPP to a command that runs the Objective C preprocessor. If ‘$OBJC -E’ doesn't work, /lib/cpp is used.


Next: , Previous: Objective C Compiler, Up: Compilers and Preprocessors

5.10.6 Erlang Compiler and Interpreter Characteristics

Autoconf defines the following macros for determining paths to the essential Erlang/OTP programs:

— Macro: AC_ERLANG_PATH_ERLC ([value-if-not-found], [path])

Determine an Erlang compiler to use. If ERLC is not already set in the environment, check for erlc. Set output variable ERLC to the complete path of the compiler command found. In addition, if ERLCFLAGS is not set in the environment, set it to an empty value.

The two optional arguments have the same meaning as the two last arguments of macro AC_PROG_PATH for looking for the erlc program. For example, to look for erlc only in the /usr/lib/erlang/bin directory:

          AC_ERLANG_PATH_ERLC([not found], [/usr/lib/erlang/bin])
     
— Macro: AC_ERLANG_NEED_ERLC ([path])

A simplified variant of the AC_ERLANG_PATH_ERLC macro, that prints an error message and exits the configure script if the erlc program is not found.

— Macro: AC_ERLANG_PATH_ERL ([value-if-not-found], [path])

Determine an Erlang interpreter to use. If ERL is not already set in the environment, check for erl. Set output variable ERL to the complete path of the interpreter command found.

The two optional arguments have the same meaning as the two last arguments of macro AC_PROG_PATH for looking for the erl program. For example, to look for erl only in the /usr/lib/erlang/bin directory:

          AC_ERLANG_PATH_ERL([not found], [/usr/lib/erlang/bin])
     
— Macro: AC_ERLANG_NEED_ERL ([path])

A simplified variant of the AC_ERLANG_PATH_ERL macro, that prints an error message and exits the configure script if the erl program is not found.


Previous: Erlang Compiler and Interpreter, Up: Compilers and Preprocessors

5.10.7 Fortran Compiler Characteristics

The Autoconf Fortran support is divided into two categories: legacy Fortran 77 macros (F77), and modern Fortran macros (FC). The former are intended for traditional Fortran 77 code, and have output variables like F77, FFLAGS, and FLIBS. The latter are for newer programs that can (or must) compile under the newer Fortran standards, and have output variables like FC, FCFLAGS, and FCLIBS.

Except for two new macros AC_FC_SRCEXT and AC_FC_FREEFORM (see below), the FC and F77 macros behave almost identically, and so they are documented together in this section.

— Macro: AC_PROG_F77 ([compiler-search-list])

Determine a Fortran 77 compiler to use. If F77 is not already set in the environment, then check for g77 and f77, and then some other names. Set the output variable F77 to the name of the compiler found.

This macro may, however, be invoked with an optional first argument which, if specified, must be a blank-separated list of Fortran 77 compilers to search for. This just gives the user an opportunity to specify an alternative search list for the Fortran 77 compiler. For example, if you didn't like the default order, then you could invoke AC_PROG_F77 like this:

          AC_PROG_F77([fl32 f77 fort77 xlf g77 f90 xlf90])
     

If using g77 (the GNU Fortran 77 compiler), then set the shell variable G77 to ‘yes’. If the output variable FFLAGS was not already set in the environment, then set it to -g -02 for g77 (or -O2 where g77 does not accept -g). Otherwise, set FFLAGS to -g for all other Fortran 77 compilers.

— Macro: AC_PROG_FC ([compiler-search-list], [dialect])

Determine a Fortran compiler to use. If FC is not already set in the environment, then dialect is a hint to indicate what Fortran dialect to search for; the default is to search for the newest available dialect. Set the output variable FC to the name of the compiler found.

By default, newer dialects are preferred over older dialects, but if dialect is specified then older dialects are preferred starting with the specified dialect. dialect can currently be one of Fortran 77, Fortran 90, or Fortran 95. However, this is only a hint of which compiler name to prefer (e.g., f90 or f95), and no attempt is made to guarantee that a particular language standard is actually supported. Thus, it is preferable that you avoid the dialect option, and use AC_PROG_FC only for code compatible with the latest Fortran standard.

This macro may, alternatively, be invoked with an optional first argument which, if specified, must be a blank-separated list of Fortran compilers to search for, just as in AC_PROG_F77.

If the output variable FCFLAGS was not already set in the environment, then set it to -g -02 for GNU g77 (or -O2 where g77 does not accept -g). Otherwise, set FCFLAGS to -g for all other Fortran compilers.

— Macro: AC_PROG_F77_C_O
— Macro: AC_PROG_FC_C_O

Test whether the Fortran compiler accepts the options -c and -o simultaneously, and define F77_NO_MINUS_C_MINUS_O or FC_NO_MINUS_C_MINUS_O, respectively, if it does not.

The following macros check for Fortran compiler characteristics. To check for characteristics not listed here, use AC_COMPILE_IFELSE (see Running the Compiler) or AC_RUN_IFELSE (see Runtime), making sure to first set the current language to Fortran 77 or Fortran via AC_LANG([Fortran 77]) or AC_LANG(Fortran) (see Language Choice).

— Macro: AC_F77_LIBRARY_LDFLAGS
— Macro: AC_FC_LIBRARY_LDFLAGS

Determine the linker flags (e.g., -L and -l) for the Fortran intrinsic and runtime libraries that are required to successfully link a Fortran program or shared library. The output variable FLIBS or FCLIBS is set to these flags (which should be included after LIBS when linking).

This macro is intended to be used in those situations when it is necessary to mix, e.g., C++ and Fortran source code in a single program or shared library (see Mixing Fortran 77 With C and C++).

For example, if object files from a C++ and Fortran compiler must be linked together, then the C++ compiler/linker must be used for linking (since special C++-ish things need to happen at link time like calling global constructors, instantiating templates, enabling exception support, etc.).

However, the Fortran intrinsic and runtime libraries must be linked in as well, but the C++ compiler/linker doesn't know by default how to add these Fortran 77 libraries. Hence, this macro was created to determine these Fortran libraries.

The macros AC_F77_DUMMY_MAIN and AC_FC_DUMMY_MAIN or AC_F77_MAIN and AC_FC_MAIN are probably also necessary to link C/C++ with Fortran; see below.

— Macro: AC_F77_DUMMY_MAIN ([action-if-found], [action-if-not-found])
— Macro: AC_FC_DUMMY_MAIN ([action-if-found], [action-if-not-found])

With many compilers, the Fortran libraries detected by AC_F77_LIBRARY_LDFLAGS or AC_FC_LIBRARY_LDFLAGS provide their own main entry function that initializes things like Fortran I/O, and which then calls a user-provided entry function named (say) MAIN__ to run the user's program. The AC_F77_DUMMY_MAIN and AC_FC_DUMMY_MAIN or AC_F77_MAIN and AC_FC_MAIN macros figure out how to deal with this interaction.

When using Fortran for purely numerical functions (no I/O, etc.) often one prefers to provide one's own main and skip the Fortran library initializations. In this case, however, one may still need to provide a dummy MAIN__ routine in order to prevent linking errors on some systems. AC_F77_DUMMY_MAIN or AC_FC_DUMMY_MAIN detects whether any such routine is required for linking, and what its name is; the shell variable F77_DUMMY_MAIN or FC_DUMMY_MAIN holds this name, unknown when no solution was found, and none when no such dummy main is needed.

By default, action-if-found defines F77_DUMMY_MAIN or FC_DUMMY_MAIN to the name of this routine (e.g., MAIN__) if it is required. action-if-not-found defaults to exiting with an error.

In order to link with Fortran routines, the user's C/C++ program should then include the following code to define the dummy main if it is needed:

          #ifdef F77_DUMMY_MAIN
          #  ifdef __cplusplus
               extern "C"
          #  endif
             int F77_DUMMY_MAIN() { return 1; }
          #endif
     

(Replace F77 with FC for Fortran instead of Fortran 77.)

Note that this macro is called automatically from AC_F77_WRAPPERS or AC_FC_WRAPPERS; there is generally no need to call it explicitly unless one wants to change the default actions.

— Macro: AC_F77_MAIN
— Macro: AC_FC_MAIN

As discussed above, many Fortran libraries allow you to provide an entry point called (say) MAIN__ instead of the usual main, which is then called by a main function in the Fortran libraries that initializes things like Fortran I/O. The AC_F77_MAIN and AC_FC_MAIN macros detect whether it is possible to utilize such an alternate main function, and defines F77_MAIN and FC_MAIN to the name of the function. (If no alternate main function name is found, F77_MAIN and FC_MAIN are simply defined to main.)

Thus, when calling Fortran routines from C that perform things like I/O, one should use this macro and name the "main" function F77_MAIN or FC_MAIN instead of main.

— Macro: AC_F77_WRAPPERS
— Macro: AC_FC_WRAPPERS

Defines C macros F77_FUNC (name, NAME), FC_FUNC (name, NAME), F77_FUNC_(name, NAME), and FC_FUNC_(name, NAME) to properly mangle the names of C/C++ identifiers, and identifiers with underscores, respectively, so that they match the name-mangling scheme used by the Fortran compiler.

Fortran is case-insensitive, and in order to achieve this the Fortran compiler converts all identifiers into a canonical case and format. To call a Fortran subroutine from C or to write a C function that is callable from Fortran, the C program must explicitly use identifiers in the format expected by the Fortran compiler. In order to do this, one simply wraps all C identifiers in one of the macros provided by AC_F77_WRAPPERS or AC_FC_WRAPPERS. For example, suppose you have the following Fortran 77 subroutine:

                subroutine foobar (x, y)
                double precision x, y
                y = 3.14159 * x
                return
                end
     

You would then declare its prototype in C or C++ as:

          #define FOOBAR_F77 F77_FUNC (foobar, FOOBAR)
          #ifdef __cplusplus
          extern "C"  /* prevent C++ name mangling */
          #endif
          void FOOBAR_F77(double *x, double *y);
     

Note that we pass both the lowercase and uppercase versions of the function name to F77_FUNC so that it can select the right one. Note also that all parameters to Fortran 77 routines are passed as pointers (see Mixing Fortran 77 With C and C++).

(Replace F77 with FC for Fortran instead of Fortran 77.)

Although Autoconf tries to be intelligent about detecting the name-mangling scheme of the Fortran compiler, there may be Fortran compilers that it doesn't support yet. In this case, the above code generates a compile-time error, but some other behavior (e.g., disabling Fortran-related features) can be induced by checking whether F77_FUNC or FC_FUNC is defined.

Now, to call that routine from a C program, we would do something like:

          {
              double x = 2.7183, y;
              FOOBAR_F77 (&x, &y);
          }
     

If the Fortran identifier contains an underscore (e.g., foo_bar), you should use F77_FUNC_ or FC_FUNC_ instead of F77_FUNC or FC_FUNC (with the same arguments). This is because some Fortran compilers mangle names differently if they contain an underscore.

— Macro: AC_F77_FUNC (name, [shellvar])
— Macro: AC_FC_FUNC (name, [shellvar])

Given an identifier name, set the shell variable shellvar to hold the mangled version name according to the rules of the Fortran linker (see also AC_F77_WRAPPERS or AC_FC_WRAPPERS). shellvar is optional; if it is not supplied, the shell variable is simply name. The purpose of this macro is to give the caller a way to access the name-mangling information other than through the C preprocessor as above, for example, to call Fortran routines from some language other than C/C++.

— Macro: AC_FC_SRCEXT (ext, [action-if-success], [action-if-failure])

By default, the FC macros perform their tests using a .f extension for source-code files. Some compilers, however, only enable newer language features for appropriately named files, e.g., Fortran 90 features only for .f90 files. On the other hand, some other compilers expect all source files to end in .f and require special flags to support other file name extensions. The AC_FC_SRCEXT macro deals with both of these issues.

The AC_FC_SRCEXT tries to get the FC compiler to accept files ending with the extension .ext (i.e., ext does not contain the dot). If any special compiler flags are needed for this, it stores them in the output variable FCFLAGS_ext. This extension and these flags are then used for all subsequent FC tests (until AC_FC_SRCEXT is called again).

For example, you would use AC_FC_SRCEXT(f90) to employ the .f90 extension in future tests, and it would set a FCFLAGS_f90 output variable with any extra flags that are needed to compile such files.

The FCFLAGS_ext can not be simply absorbed into FCFLAGS, for two reasons based on the limitations of some compilers. First, only one FCFLAGS_ext can be used at a time, so files with different extensions must be compiled separately. Second, FCFLAGS_ext must appear immediately before the source-code file name when compiling. So, continuing the example above, you might compile a foo.f90 file in your makefile with the command:

          foo.o: foo.f90
               $(FC) -c $(FCFLAGS) $(FCFLAGS_f90) '$(srcdir)/foo.f90'
     

If AC_FC_SRCEXT succeeds in compiling files with the ext extension, it calls action-if-success (defaults to nothing). If it fails, and cannot find a way to make the FC compiler accept such files, it calls action-if-failure (defaults to exiting with an error message).

— Macro: AC_FC_FREEFORM ([action-if-success], [action-if-failure])

The AC_FC_FREEFORM tries to ensure that the Fortran compiler ($FC) allows free-format source code (as opposed to the older fixed-format style from Fortran 77). If necessary, it may add some additional flags to FCFLAGS.

This macro is most important if you are using the default .f extension, since many compilers interpret this extension as indicating fixed-format source unless an additional flag is supplied. If you specify a different extension with AC_FC_SRCEXT, such as .f90 or .f95, then AC_FC_FREEFORM ordinarily succeeds without modifying FCFLAGS.

If AC_FC_FREEFORM succeeds in compiling free-form source, it calls action-if-success (defaults to nothing). If it fails, it calls action-if-failure (defaults to exiting with an error message).


Next: , Previous: Compilers and Preprocessors, Up: Existing Tests

5.11 System Services

The following macros check for operating system services or capabilities.

— Macro: AC_PATH_X

Try to locate the X Window System include files and libraries. If the user gave the command line options --x-includes=dir and --x-libraries=dir, use those directories.

If either or both were not given, get the missing values by running xmkmf (or an executable pointed to by the XMKMF environment variable) on a trivial Imakefile and examining the makefile that it produces. Setting XMKMF to ‘false’ disables this method.

If this method fails to find the X Window System, configure looks for the files in several directories where they often reside. If either method is successful, set the shell variables x_includes and x_libraries to their locations, unless they are in directories the compiler searches by default.

If both methods fail, or the user gave the command line option --without-x, set the shell variable no_x to ‘yes’; otherwise set it to the empty string.

— Macro: AC_PATH_XTRA

An enhanced version of AC_PATH_X. It adds the C compiler flags that X needs to output variable X_CFLAGS, and the X linker flags to X_LIBS. Define X_DISPLAY_MISSING if X is not available.

This macro also checks for special libraries that some systems need in order to compile X programs. It adds any that the system needs to output variable X_EXTRA_LIBS. And it checks for special X11R6 libraries that need to be linked with before -lX11, and adds any found to the output variable X_PRE_LIBS.

— Macro: AC_SYS_INTERPRETER

Check whether the system supports starting scripts with a line of the form ‘#!/bin/sh’ to select the interpreter to use for the script. After running this macro, shell code in configure.ac can check the shell variable interpval; it is set to ‘yes’ if the system supports ‘#!’, ‘no’ if not.

— Macro: AC_SYS_LARGEFILE

Arrange for large-file support. On some hosts, one must use special compiler options to build programs that can access large files. Append any such options to the output variable CC. Define _FILE_OFFSET_BITS and _LARGE_FILES if necessary.

Large-file support can be disabled by configuring with the --disable-largefile option.

If you use this macro, check that your program works even when off_t is wider than long int, since this is common when large-file support is enabled. For example, it is not correct to print an arbitrary off_t value X with printf ("%ld", (long int) X).

The LFS introduced the fseeko and ftello functions to replace their C counterparts fseek and ftell that do not use off_t. Take care to use AC_FUNC_FSEEKO to make their prototypes available when using them and large-file support is enabled.

— Macro: AC_SYS_LONG_FILE_NAMES

If the system supports file names longer than 14 characters, define HAVE_LONG_FILE_NAMES.

— Macro: AC_SYS_POSIX_TERMIOS

Check to see if the Posix termios headers and functions are available on the system. If so, set the shell variable ac_cv_sys_posix_termios to ‘yes’. If not, set the variable to ‘no’.


Next: , Previous: System Services, Up: Existing Tests

5.12 Posix Variants

The following macros check for certain operating systems that need special treatment for some programs, due to exceptional oddities in their header files or libraries. These macros are warts; they will be replaced by a more systematic approach, based on the functions they make available or the environments they provide.

— Macro: AC_AIX

If on AIX, define _ALL_SOURCE. Allows the use of some BSD functions. Should be called before any macros that run the C compiler.

— Macro: AC_GNU_SOURCE

If using the GNU C library, define _GNU_SOURCE. Allows the use of some GNU functions. Should be called before any macros that run the C compiler.

— Macro: AC_ISC_POSIX

For interactive Systems Corporation Unix, add -lcposix to output variable LIBS if necessary for Posix facilities. Call this after AC_PROG_CC and before any other macros that use Posix interfaces. interactive Unix is no longer sold, and Sun says that they will drop support for it on 2006-07-23, so this macro is becoming obsolescent.

— Macro: AC_MINIX

If on Minix, define _MINIX and _POSIX_SOURCE and define _POSIX_1_SOURCE to be 2. This allows the use of Posix facilities. Should be called before any macros that run the C compiler.

— Macro: AC_USE_SYSTEM_EXTENSIONS

If possible, enable extensions to Posix on hosts that normally disable the extensions, typically due to standards-conformance namespace issues. This may involve defining __EXTENSIONS__ and _POSIX_PTHREAD_SEMANTICS, which are macros used by Solaris. This macro also has the combined effects of AC_GNU_SOURCE, AC_AIX, and AC_MINIX.


Previous: Posix Variants, Up: Existing Tests

5.13 Erlang Libraries

The following macros check for an installation of Erlang/OTP, and for the presence of certain Erlang libraries. All those macros require the configuration of an Erlang interpreter and an Erlang compiler (see Erlang Compiler and Interpreter).

— Macro: AC_ERLANG_SUBST_ROOT_DIR

Set the output variable ERLANG_ROOT_DIR to the path to the base directory in which Erlang/OTP is installed (as returned by Erlang's code:root_dir/0 function). The result of this test is cached if caching is enabled when running configure.

— Macro: AC_ERLANG_SUBST_LIB_DIR

Set the output variable ERLANG_LIB_DIR to the path of the library directory of Erlang/OTP (as returned by Erlang's code:lib_dir/0 function), which subdirectories each contain an installed Erlang/OTP library. The result of this test is cached if caching is enabled when running configure.

— Macro: AC_ERLANG_CHECK_LIB (library, [action-if-found], [action-if-not-found])

Test whether the Erlang/OTP library library is installed by calling Erlang's code:lib_dir/1 function. The result of this test is cached if caching is enabled when running configure. action-if-found is a list of shell commands to run if the library is installed; action-if-not-found is a list of shell commands to run if it is not. Additionally, if the library is installed, the output variable ‘ERLANG_LIB_DIR_library’ is set to the path to the library installation directory. For example, to check if library stdlib is installed:

          AC_ERLANG_CHECK_LIB([stdlib],
            [echo "stdlib is installed in $ERLANG_LIB_DIR_stdlib"],
            [AC_MSG_ERROR([stdlib was not found!])])
     

In addition to the above macros, which test installed Erlang libraries, the following macros determine the paths to the directories into which newly built Erlang libraries are to be installed:

— Macro: AC_ERLANG_SUBST_INSTALL_LIB_DIR

Set the ERLANG_INSTALL_LIB_DIR output variable to the directory into which every built Erlang library should be installed in a separate subdirectory. If this variable is not set in the environment when configure runs, its default value is $ERLANG_LIB_DIR, which value is set by the AC_ERLANG_SUBST_LIB_DIR macro.

— Macro: AC_ERLANG_SUBST_INSTALL_LIB_SUBDIR (library, version)

Set the ‘ERLANG_INSTALL_LIB_DIR_library’ output variable to the directory into which the built Erlang library library version version should be installed. If this variable is not set in the environment when configure runs, its default value is ‘$ERLANG_INSTALL_LIB_DIR/library-version’, the value of the ERLANG_INSTALL_LIB_DIR variable being set by the AC_ERLANG_SUBST_INSTALL_LIB_DIR macro.


Next: , Previous: Existing Tests, Up: Top

6 Writing Tests

If the existing feature tests don't do something you need, you have to write new ones. These macros are the building blocks. They provide ways for other macros to check whether various kinds of features are available and report the results.

This chapter contains some suggestions and some of the reasons why the existing tests are written the way they are. You can also learn a lot about how to write Autoconf tests by looking at the existing ones. If something goes wrong in one or more of the Autoconf tests, this information can help you understand the assumptions behind them, which might help you figure out how to best solve the problem.

These macros check the output of the compiler system of the current language (see Language Choice). They do not cache the results of their tests for future use (see Caching Results), because they don't know enough about the information they are checking for to generate a cache variable name. They also do not print any messages, for the same reason. The checks for particular kinds of features call these macros and do cache their results and print messages about what they're checking for.

When you write a feature test that could be applicable to more than one software package, the best thing to do is encapsulate it in a new macro. See Writing Autoconf Macros, for how to do that.


Next: , Up: Writing Tests

6.1 Language Choice

Autoconf-generated configure scripts check for the C compiler and its features by default. Packages that use other programming languages (maybe more than one, e.g., C and C++) need to test features of the compilers for the respective languages. The following macros determine which programming language is used in the subsequent tests in configure.ac.

— Macro: AC_LANG (language)

Do compilation tests using the compiler, preprocessor, and file extensions for the specified language.

Supported languages are:

C
Do compilation tests using CC and CPP and use extension .c for test programs. Use compilation flags: CPPFLAGS with CPP, and both CPPFLAGS and CFLAGS with CC.
C++
Do compilation tests using CXX and CXXCPP and use extension .C for test programs. Use compilation flags: CPPFLAGS with CXXPP, and both CPPFLAGS and CXXFLAGS with CXX.
Fortran 77
Do compilation tests using F77 and use extension .f for test programs. Use compilation flags: FFLAGS.
Fortran
Do compilation tests using FC and use extension .f (or whatever has been set by AC_FC_SRCEXT) for test programs. Use compilation flags: FCFLAGS.
Erlang
Compile and execute tests using ERLC and ERL and use extension .erl for test Erlang modules. Use compilation flags: ERLCFLAGS.
Objective C
Do compilation tests using OBJC and OBJCCPP and use extension .m for test programs. Use compilation flags: CPPFLAGS with OBJCPP, and both CPPFLAGS and OBJCFLAGS with OBJC.

— Macro: AC_LANG_PUSH (language)

Remember the current language (as set by AC_LANG) on a stack, and then select the language. Use this macro and AC_LANG_POP in macros that need to temporarily switch to a particular language.

— Macro: AC_LANG_POP ([language])

Select the language that is saved on the top of the stack, as set by AC_LANG_PUSH, and remove it from the stack.

If given, language specifies the language we just quit. It is a good idea to specify it when it's known (which should be the case...), since Autoconf detects inconsistencies.

          AC_LANG_PUSH([Fortran 77])
          # Perform some tests on Fortran 77.
          # ...
          AC_LANG_POP([Fortran 77])
     
— Macro: AC_LANG_ASSERT (language)

Check statically that the current language is language. You should use this in your language specific macros to avoid that they be called with an inappropriate language.

This macro runs only at autoconf time, and incurs no cost at configure time. Sadly enough and because Autoconf is a two layer language 2, the macros AC_LANG_PUSH and AC_LANG_POP cannot be “optimizing”, therefore as much as possible you ought to avoid using them to wrap your code, rather, require from the user to run the macro with a correct current language, and check it with AC_LANG_ASSERT. And anyway, that may help the user understand she is running a Fortran macro while expecting a result about her Fortran 77 compiler...

— Macro: AC_REQUIRE_CPP

Ensure that whichever preprocessor would currently be used for tests has been found. Calls AC_REQUIRE (see Prerequisite Macros) with an argument of either AC_PROG_CPP or AC_PROG_CXXCPP, depending on which language is current.


Next: , Previous: Language Choice, Up: Writing Tests

6.2 Writing Test Programs

Autoconf tests follow a common scheme: feed some program with some input, and most of the time, feed a compiler with some source file. This section is dedicated to these source samples.


Next: , Up: Writing Test Programs

6.2.1 Guidelines for Test Programs

The most important rule to follow when writing testing samples is:

Look for realism.

This motto means that testing samples must be written with the same strictness as real programs are written. In particular, you should avoid “shortcuts” and simplifications.

Don't just play with the preprocessor if you want to prepare a compilation. For instance, using cpp to check whether a header is functional might let your configure accept a header which causes some compiler error. Do not hesitate to check a header with other headers included before, especially required headers.

Make sure the symbols you use are properly defined, i.e., refrain for simply declaring a function yourself instead of including the proper header.

Test programs should not write to standard output. They should exit with status 0 if the test succeeds, and with status 1 otherwise, so that success can be distinguished easily from a core dump or other failure; segmentation violations and other failures produce a nonzero exit status. Unless you arrange for exit to be declared, test programs should return, not exit, from main, because on many systems exit is not declared by default.

Test programs can use #if or #ifdef to check the values of preprocessor macros defined by tests that have already run. For example, if you call AC_HEADER_STDBOOL, then later on in configure.ac you can have a test program that includes stdbool.h conditionally:

     #if HAVE_STDBOOL_H
     # include <stdbool.h>
     #endif

If a test program needs to use or create a data file, give it a name that starts with conftest, such as conftest.data. The configure script cleans up by running ‘rm -f -r conftest*’ after running test programs and if the script is interrupted.


Next: , Previous: Guidelines, Up: Writing Test Programs

6.2.2 Test Functions

These days it's safe to assume support for function prototypes (introduced in C89).

Functions that test programs declare should also be conditionalized for C++, which requires ‘extern "C"’ prototypes. Make sure to not include any header files containing clashing prototypes.

     #ifdef __cplusplus
     extern "C"
     #endif
     void *valloc (size_t);

If a test program calls a function with invalid parameters (just to see whether it exists), organize the program to ensure that it never invokes that function. You can do this by calling it in another function that is never invoked. You can't do it by putting it after a call to exit, because GCC version 2 knows that exit never returns and optimizes out any code that follows it in the same block.

If you include any header files, be sure to call the functions relevant to them with the correct number of arguments, even if they are just 0, to avoid compilation errors due to prototypes. GCC version 2 has internal prototypes for several functions that it automatically inlines; for example, memcpy. To avoid errors when checking for them, either pass them the correct number of arguments or redeclare them with a different return type (such as char).


Previous: Test Functions, Up: Writing Test Programs

6.2.3 Generating Sources

Autoconf provides a set of macros that can be used to generate test source files. They are written to be language generic, i.e., they actually depend on the current language (see Language Choice) to “format” the output properly.

— Macro: AC_LANG_CONFTEST (source)

Save the source text in the current test source file: conftest.extension where the extension depends on the current language.

Note that the source is evaluated exactly once, like regular Autoconf macro arguments, and therefore (i) you may pass a macro invocation, (ii) if not, be sure to double quote if needed.

— Macro: AC_LANG_SOURCE (source)

Expands into the source, with the definition of all the AC_DEFINE performed so far.

For instance executing (observe the double quotation!):

     AC_INIT([Hello], [1.0], [bug-hello@example.org])
     AC_DEFINE([HELLO_WORLD], ["Hello, World\n"],
       [Greetings string.])
     AC_LANG(C)
     AC_LANG_CONFTEST(
        [AC_LANG_SOURCE([[const char hw[] = "Hello, World\n";]])])
     gcc -E -dD -o - conftest.c

results in:

     ...
     # 1 "conftest.c"
     
     #define PACKAGE_NAME "Hello"
     #define PACKAGE_TARNAME "hello"
     #define PACKAGE_VERSION "1.0"
     #define PACKAGE_STRING "Hello 1.0"
     #define PACKAGE_BUGREPORT "bug-hello@example.org"
     #define HELLO_WORLD "Hello, World\n"
     
     const char hw[] = "Hello, World\n";

When the test language is Fortran or Erlang, the AC_DEFINE definitions are not automatically translated into constants in the source code by this macro.

— Macro: AC_LANG_PROGRAM (prologue, body)

Expands into a source file which consists of the prologue, and then body as body of the main function (e.g., main in C). Since it uses AC_LANG_SOURCE, the features of the latter are available.

For instance:

     AC_INIT([Hello], [1.0], [bug-hello@example.org])
     AC_DEFINE([HELLO_WORLD], ["Hello, World\n"],
       [Greetings string.])
     AC_LANG_CONFTEST(
     [AC_LANG_PROGRAM([[const char hw[] = "Hello, World\n";]],
                      [[fputs (hw, stdout);]])])
     gcc -E -dD -o - conftest.c

results in:

     ...
     # 1 "conftest.c"
     
     #define PACKAGE_NAME "Hello"
     #define PACKAGE_TARNAME "hello"
     #define PACKAGE_VERSION "1.0"
     #define PACKAGE_STRING "Hello 1.0"
     #define PACKAGE_BUGREPORT "bug-hello@example.org"
     #define HELLO_WORLD "Hello, World\n"
     
     const char hw[] = "Hello, World\n";
     int
     main ()
     {
     fputs (hw, stdout);
       ;
       return 0;
     }

In Erlang tests, the created source file is that of an Erlang module called conftest (conftest.erl). This module defines and exports at least one start/0 function, which is called to perform the test. The prologue is optional code that is inserted between the module header and the start/0 function definition. body is the body of the start/0 function without the final period (see Runtime, about constraints on this function's behaviour).

For instance:

     AC_INIT([Hello], [1.0], [bug-hello@example.org])
     AC_LANG(Erlang)
     AC_LANG_CONFTEST(
     [AC_LANG_PROGRAM([[-define(HELLO_WORLD, "Hello, world!").]],
                      [[io:format("~s~n", [?HELLO_WORLD])]])])
     cat conftest.erl

results in:

     -module(conftest).
     -export([start/0]).
     -define(HELLO_WORLD, "Hello, world!").
     start() ->
     io:format("~s~n", [?HELLO_WORLD])
     .
— Macro: AC_LANG_CALL (prologue, function)

Expands into a source file which consists of the prologue, and then a call to the function as body of the main function (e.g., main in C). Since it uses AC_LANG_PROGRAM, the feature of the latter are available.

This function will probably be replaced in the future by a version which would enable specifying the arguments. The use of this macro is not encouraged, as it violates strongly the typing system.

This macro cannot be used for Erlang tests.

— Macro: AC_LANG_FUNC_LINK_TRY (function)

Expands into a source file which uses the function in the body of the main function (e.g., main in C). Since it uses AC_LANG_PROGRAM, the features of the latter are available.

As AC_LANG_CALL, this macro is documented only for completeness. It is considered to be severely broken, and in the future will be removed in favor of actual function calls (with properly typed arguments).

This macro cannot be used for Erlang tests.


Next: , Previous: Writing Test Programs, Up: Writing Tests

6.3 Running the Preprocessor

Sometimes one might need to run the preprocessor on some source file. Usually it is a bad idea, as you typically need to compile your project, not merely run the preprocessor on it; therefore you certainly want to run the compiler, not the preprocessor. Resist the temptation of following the easiest path.

Nevertheless, if you need to run the preprocessor, then use AC_PREPROC_IFELSE.

The macros described in this section cannot be used for tests in Erlang or Fortran, since those languages require no preprocessor.

— Macro: AC_PREPROC_IFELSE (input, [action-if-true], [action-if-false])

Run the preprocessor of the current language (see Language Choice) on the input, run the shell commands action-if-true on success, action-if-false otherwise. The input can be made by AC_LANG_PROGRAM and friends.

This macro uses CPPFLAGS, but not CFLAGS, because -g, -O, etc. are not valid options to many C preprocessors.

It is customary to report unexpected failures with AC_MSG_FAILURE.

For instance:

     AC_INIT([Hello], [1.0], [bug-hello@example.org])
     AC_DEFINE([HELLO_WORLD], ["Hello, World\n"],
       [Greetings string.])
     AC_PREPROC_IFELSE(
        [AC_LANG_PROGRAM([[const char hw[] = "Hello, World\n";]],
                         [[fputs (hw, stdout);]])],
        [AC_MSG_RESULT([OK])],
        [AC_MSG_FAILURE([unexpected preprocessor failure])])

results in:

     checking for gcc... gcc
     checking for C compiler default output file name... a.out
     checking whether the C compiler works... yes
     checking whether we are cross compiling... no
     checking for suffix of executables...
     checking for suffix of object files... o
     checking whether we are using the GNU C compiler... yes
     checking whether gcc accepts -g... yes
     checking for gcc option to accept ISO C89... none needed
     checking how to run the C preprocessor... gcc -E
     OK

The macro AC_TRY_CPP (see Obsolete Macros) used to play the role of AC_PREPROC_IFELSE, but double quotes its argument, making it impossible to use it to elaborate sources. You are encouraged to get rid of your old use of the macro AC_TRY_CPP in favor of AC_PREPROC_IFELSE, but, in the first place, are you sure you need to run the preprocessor and not the compiler?
— Macro: AC_EGREP_HEADER (pattern, header-file, action-if-found, [action-if-not-found])

If the output of running the preprocessor on the system header file header-file matches the extended regular expression pattern, execute shell commands action-if-found, otherwise execute action-if-not-found.

— Macro: AC_EGREP_CPP (pattern, program, [action-if-found], [action-if-not-found])

program is the text of a C or C++ program, on which shell variable, back quote, and backslash substitutions are performed. If the output of running the preprocessor on program matches the extended regular expression pattern, execute shell commands action-if-found, otherwise execute action-if-not-found.


Next: , Previous: Running the Preprocessor, Up: Writing Tests

6.4 Running the Compiler

To check for a syntax feature of the current language's (see Language Choice) compiler, such as whether it recognizes a certain keyword, or simply to try some library feature, use AC_COMPILE_IFELSE to try to compile a small program that uses that feature.

— Macro: AC_COMPILE_IFELSE (input, [action-if-true], [action-if-false])

Run the compiler and compilation flags of the current language (see Language Choice) on the input, run the shell commands action-if-true on success, action-if-false otherwise. The input can be made by AC_LANG_PROGRAM and friends.

It is customary to report unexpected failures with AC_MSG_FAILURE. This macro does not try to link; use AC_LINK_IFELSE if you need to do that (see Running the Linker).

For tests in Erlang, the input must be the source code of a module named conftest. AC_COMPILE_IFELSE generates a conftest.beam file that can be interpreted by the Erlang virtual machine (ERL). It is recommended to use AC_LANG_PROGRAM to specify the test program, to ensure that the Erlang module has the right name.


Next: , Previous: Running the Compiler, Up: Writing Tests

6.5 Running the Linker

To check for a library, a function, or a global variable, Autoconf configure scripts try to compile and link a small program that uses it. This is unlike Metaconfig, which by default uses nm or ar on the C library to try to figure out which functions are available. Trying to link with the function is usually a more reliable approach because it avoids dealing with the variations in the options and output formats of nm and ar and in the location of the standard libraries. It also allows configuring for cross-compilation or checking a function's runtime behavior if needed. On the other hand, it can be slower than scanning the libraries once, but accuracy is more important than speed.

AC_LINK_IFELSE is used to compile test programs to test for functions and global variables. It is also used by AC_CHECK_LIB to check for libraries (see Libraries), by adding the library being checked for to LIBS temporarily and trying to link a small program.

— Macro: AC_LINK_IFELSE (input, [action-if-true], [action-if-false])

Run the compiler (and compilation flags) and the linker of the current language (see Language Choice) on the input, run the shell commands action-if-true on success, action-if-false otherwise. The input can be made by AC_LANG_PROGRAM and friends.

LDFLAGS and LIBS are used for linking, in addition to the current compilation flags.

It is customary to report unexpected failures with AC_MSG_FAILURE. This macro does not try to execute the program; use AC_RUN_IFELSE if you need to do that (see Runtime).

The AC_LINK_IFELSE macro cannot be used for Erlang tests, since Erlang programs are interpreted and do not require linking.


Next: , Previous: Running the Linker, Up: Writing Tests

6.6 Checking Runtime Behavior

Sometimes you need to find out how a system performs at runtime, such as whether a given function has a certain capability or bug. If you can, make such checks when your program runs instead of when it is configured. You can check for things like the machine's endianness when your program initializes itself.

If you really need to test for a runtime behavior while configuring, you can write a test program to determine the result, and compile and run it using AC_RUN_IFELSE. Avoid running test programs if possible, because this prevents people from configuring your package for cross-compiling.

— Macro: AC_RUN_IFELSE (input, [action-if-true], [action-if-false], [action-if-cross-compiling])

If program compiles and links successfully and returns an exit status of 0 when executed, run shell commands action-if-true. Otherwise, run shell commands action-if-false.

The input can be made by AC_LANG_PROGRAM and friends. LDFLAGS and LIBS are used for linking, in addition to the compilation flags of the current language (see Language Choice).

If the compiler being used does not produce executables that run on the system where configure is being run, then the test program is not run. If the optional shell commands action-if-cross-compiling are given, they are run instead. Otherwise, configure prints an error message and exits.

In the action-if-false section, the failing exit status is available in the shell variable ‘$?’. This exit status might be that of a failed compilation, or it might be that of a failed program execution.

It is customary to report unexpected failures with AC_MSG_FAILURE.

Try to provide a pessimistic default value to use when cross-compiling makes runtime tests impossible. You do this by passing the optional last argument to AC_RUN_IFELSE. autoconf prints a warning message when creating configure each time it encounters a call to AC_RUN_IFELSE with no action-if-cross-compiling argument given. You may ignore the warning, though users cannot configure your package for cross-compiling. A few of the macros distributed with Autoconf produce this warning message.

To configure for cross-compiling you can also choose a value for those parameters based on the canonical system name (see Manual Configuration). Alternatively, set up a test results cache file with the correct values for the host system (see Caching Results).

To provide a default for calls of AC_RUN_IFELSE that are embedded in other macros, including a few of the ones that come with Autoconf, you can test whether the shell variable cross_compiling is set to ‘yes’, and then use an alternate method to get the results instead of calling the macros.

A C or C++ runtime test should be portable. See Portable C and C++.

Erlang tests must exit themselves the Erlang VM by calling the halt/1 function: the given status code is used to determine the success of the test (status is 0) or its failure (status is different than 0), as explained above. It must be noted that data output through the standard output (e.g. using io:format/2) may be truncated when halting the VM. Therefore, if a test must output configuration information, it is recommended to create and to output data into the temporary file named conftest.out, using the functions of module file. The conftest.out file is automatically deleted by the AC_RUN_IFELSE macro. For instance, a simplified implementation of Autoconf's AC_ERLANG_SUBST_LIB_DIR macro is:

     AC_INIT([LibdirTest], [1.0], [bug-libdirtest@example.org])
     AC_ERLANG_NEED_ERL
     AC_LANG(Erlang)
     AC_RUN_IFELSE(
       [AC_LANG_PROGRAM([], [dnl
         file:write_file("conftest.out", code:lib_dir()),
         halt(0)])],
       [echo "code:lib_dir() returned: `cat conftest.out`"],
       [AC_MSG_FAILURE([test Erlang program execution failed])])


Next: , Previous: Runtime, Up: Writing Tests

6.7 Systemology

This section aims at presenting some systems and pointers to documentation. It may help you addressing particular problems reported by users.

Posix-conforming systems are derived from the Unix operating system.

The Rosetta Stone for Unix contains a table correlating the features of various Posix-conforming systems. Unix History is a simplified diagram of how many Unix systems were derived from each other.

The Heirloom Project provides some variants of traditional implementations of Unix utilities.

Darwin
Darwin is also known as Mac OS X. Beware that the file system can be case-preserving, but case insensitive. This can cause nasty problems, since for instance the installation attempt for a package having an INSTALL file can result in ‘make install’ report that nothing was to be done!

That's all dependent on whether the file system is a UFS (case sensitive) or HFS+ (case preserving). By default Apple wants you to install the OS on HFS+. Unfortunately, there are some pieces of software which really need to be built on UFS. We may want to rebuild Darwin to have both UFS and HFS+ available (and put the /local/build tree on the UFS).

QNX 4.25
QNX is a realtime operating system running on Intel architecture meant to be scalable from the small embedded systems to the hundred processor super-computer. It claims to be Posix certified. More information is available on the QNX home page.
Tru64
Documentation of several versions of Tru64 is available in different formats.
Unix version 7
Officially this was called the “Seventh Edition” of “the unix time-sharing system” but we use the more-common name “Unix version 7”. Documentation is available in the Unix Seventh Edition Manual. Previous versions of Unix are called “Unix version 6”, etc., but they were not as widely used.


Previous: Systemology, Up: Writing Tests

6.8 Multiple Cases

Some operations are accomplished in several possible ways, depending on the OS variant. Checking for them essentially requires a “case statement”. Autoconf does not directly provide one; however, it is easy to simulate by using a shell variable to keep track of whether a way to perform the operation has been found yet.

Here is an example that uses the shell variable fstype to keep track of whether the remaining cases need to be checked.

     AC_MSG_CHECKING([how to get file system type])
     fstype=no
     # The order of these tests is important.
     AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <sys/statvfs.h>
     #include <sys/fstyp.h>]])],
                       [AC_DEFINE([FSTYPE_STATVFS], [1],
                          [Define if statvfs exists.])
                        fstype=SVR4])
     if test $fstype = no; then
       AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <sys/statfs.h>
     #include <sys/fstyp.h>]])],
                       [AC_DEFINE([FSTYPE_USG_STATFS], [1],
                          [Define if USG statfs.])
                        fstype=SVR3])
     fi
     if test $fstype = no; then
       AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <sys/statfs.h>
     #include <sys/vmount.h>]])]),
                       [AC_DEFINE([FSTYPE_AIX_STATFS], [1],
                          [Define if AIX statfs.])
                        fstype=AIX])
     fi
     # (more cases omitted here)
     AC_MSG_RESULT([$fstype])


Next: , Previous: Writing Tests, Up: Top

7 Results of Tests

Once configure has determined whether a feature exists, what can it do to record that information? There are four sorts of things it can do: define a C preprocessor symbol, set a variable in the output files, save the result in a cache file for future configure runs, and print a message letting the user know the result of the test.


Next: , Up: Results

7.1 Defining C Preprocessor Symbols

A common action to take in response to a feature test is to define a C preprocessor symbol indicating the results of the test. That is done by calling AC_DEFINE or AC_DEFINE_UNQUOTED.

By default, AC_OUTPUT places the symbols defined by these macros into the output variable DEFS, which contains an option -Dsymbol=value for each symbol defined. Unlike in Autoconf version 1, there is no variable DEFS defined while configure is running. To check whether Autoconf macros have already defined a certain C preprocessor symbol, test the value of the appropriate cache variable, as in this example:

     AC_CHECK_FUNC([vprintf], [AC_DEFINE([HAVE_VPRINTF], [1],
                               [Define if vprintf exists.])])
     if test "$ac_cv_func_vprintf" != yes; then
       AC_CHECK_FUNC([_doprnt], [AC_DEFINE([HAVE_DOPRNT], [1],
                                 [Define if _doprnt exists.])])
     fi

If AC_CONFIG_HEADERS has been called, then instead of creating DEFS, AC_OUTPUT creates a header file by substituting the correct values into #define statements in a template file. See Configuration Headers, for more information about this kind of output.

— Macro: AC_DEFINE (variable, value, [description])
— Macro: AC_DEFINE (variable)

Define the C preprocessor variable variable to value (verbatim). value should not contain literal newlines, and if you are not using AC_CONFIG_HEADERS it should not contain any ‘#’ characters, as make tends to eat them. To use a shell variable, use AC_DEFINE_UNQUOTED instead. description is only useful if you are using AC_CONFIG_HEADERS. In this case, description is put into the generated config.h.in as the comment before the macro define. The following example defines the C preprocessor variable EQUATION to be the string constant ‘"$a > $b"’:

          AC_DEFINE([EQUATION], ["$a > $b"],
            [Equation string.])
     

If neither value nor description are given, then value defaults to 1 instead of to the empty string. This is for backwards compatibility with older versions of Autoconf, but this usage is obsolescent and may be withdrawn in future versions of Autoconf.

If the variable is a literal string, it is passed to m4_pattern_allow (see Forbidden Patterns).

— Macro: AC_DEFINE_UNQUOTED (variable, value, [description])
— Macro: AC_DEFINE_UNQUOTED (variable)

Like AC_DEFINE, but three shell expansions are performed—once—on variable and value: variable expansion (‘$’), command substitution (‘`’), and backslash escaping (‘\’). Single and double quote characters in the value have no special meaning. Use this macro instead of AC_DEFINE when variable or value is a shell variable. Examples:

          AC_DEFINE_UNQUOTED([config_machfile], ["$machfile"],
            [Configuration machine file.])
          AC_DEFINE_UNQUOTED([GETGROUPS_T], [$ac_cv_type_getgroups],
            [getgroups return type.])
          AC_DEFINE_UNQUOTED([$ac_tr_hdr], [1],
            [Translated header name.])
     

Due to a syntactical bizarreness of the Bourne shell, do not use semicolons to separate AC_DEFINE or AC_DEFINE_UNQUOTED calls from other macro calls or shell code; that can cause syntax errors in the resulting configure script. Use either blanks or newlines. That is, do this:

     AC_CHECK_HEADER([elf.h],
       [AC_DEFINE([SVR4], [1], [System V Release 4]) LIBS="-lelf $LIBS"])

or this:

     AC_CHECK_HEADER([elf.h],
       [AC_DEFINE([SVR4], [1], [System V Release 4])
        LIBS="-lelf $LIBS"])

instead of this:

     AC_CHECK_HEADER([elf.h],
       [AC_DEFINE([SVR4], [1], [System V Release 4]); LIBS="-lelf $LIBS"])


Next: , Previous: Defining Symbols, Up: Results

7.2 Setting Output Variables

Another way to record the results of tests is to set output variables, which are shell variables whose values are substituted into files that configure outputs. The two macros below create new output variables. See Preset Output Variables, for a list of output variables that are always available.

— Macro: AC_SUBST (variable, [value])

Create an output variable from a shell variable. Make AC_OUTPUT substitute the variable variable into output files (typically one or more makefiles). This means that AC_OUTPUT replaces instances of ‘@variable@’ in input files with the value that the shell variable variable has when AC_OUTPUT is called. The value can contain newlines. The substituted value is not rescanned for more output variables; occurrences of ‘@variable@’ in the value are inserted literally into the output file. (The algorithm uses the special marker |#_!!_#| internally, so the substituted value cannot contain |#_!!_#|.)

If value is given, in addition assign it to variable.

The string variable is passed to m4_pattern_allow (see Forbidden Patterns).

— Macro: AC_SUBST_FILE (variable)

Another way to create an output variable from a shell variable. Make AC_OUTPUT insert (without substitutions) the contents of the file named by shell variable variable into output files. This means that AC_OUTPUT replaces instances of ‘@variable@’ in output files (such as Makefile.in) with the contents of the file that the shell variable variable names when AC_OUTPUT is called. Set the variable to /dev/null for cases that do not have a file to insert. This substitution occurs only when the ‘@variable@’ is on a line by itself, optionally surrounded by spaces and tabs. The substitution replaces the whole line, including the spaces, tabs, and the terminating newline.

This macro is useful for inserting makefile fragments containing special dependencies or other make directives for particular host or target types into makefiles. For example, configure.ac could contain:

          AC_SUBST_FILE([host_frag])
          host_frag=$srcdir/conf/sun4.mh
     

and then a Makefile.in could contain:

          @host_frag@
     

The string variable is passed to m4_pattern_allow (see Forbidden Patterns).

Running configure in varying environments can be extremely dangerous. If for instance the user runs ‘CC=bizarre-cc ./configure’, then the cache, config.h, and many other output files depend upon bizarre-cc being the C compiler. If for some reason the user runs ./configure again, or if it is run via ‘./config.status --recheck’, (See Automatic Remaking, and see config.status Invocation), then the configuration can be inconsistent, composed of results depending upon two different compilers.

Environment variables that affect this situation, such as ‘CC’ above, are called precious variables, and can be declared as such by AC_ARG_VAR.

— Macro: AC_ARG_VAR (variable, description)

Declare variable is a precious variable, and include its description in the variable section of ‘./configure --help’.

Being precious means that


Next: , Previous: Setting Output Variables, Up: Results

7.3 Special Characters in Output Variables

Many output variables are intended to be evaluated both by make and by the shell. Some characters are expanded differently in these two contexts, so to avoid confusion these variables' values should not contain any of the following characters:

     " # $ & ' ( ) * ; < > ? [ \ ^ ` |

Also, these variables' values should neither contain newlines, nor start with ‘~’, nor contain white space or ‘:’ immediately followed by ‘~’. The values can contain nonempty sequences of white space characters like tabs and spaces, but each such sequence might arbitrarily be replaced by a single space during substitution.

These restrictions apply both to the values that configure computes, and to the values set directly by the user. For example, the following invocations of configure are problematic, since they attempt to use special characters within CPPFLAGS and white space within $(srcdir):

     CPPFLAGS='-DOUCH="&\"#$*?"' '../My Source/ouch-1.0/configure'
     
     '../My Source/ouch-1.0/configure' CPPFLAGS='-DOUCH="&\"#$*?"'


Next: , Previous: Special Chars in Variables, Up: Results

7.4 Caching Results

To avoid checking for the same features repeatedly in various configure scripts (or in repeated runs of one script), configure can optionally save the results of many checks in a cache file (see Cache Files). If a configure script runs with caching enabled and finds a cache file, it reads the results of previous runs from the cache and avoids rerunning those checks. As a result, configure can then run much faster than if it had to perform all of the checks every time.

— Macro: AC_CACHE_VAL (cache-id, commands-to-set-it)

Ensure that the results of the check identified by cache-id are available. If the results of the check were in the cache file that was read, and configure was not given the --quiet or --silent option, print a message saying that the result was cached; otherwise, run the shell commands commands-to-set-it. If the shell commands are run to determine the value, the value is saved in the cache file just before configure creates its output files. See Cache Variable Names, for how to choose the name of the cache-id variable.

The commands-to-set-it must have no side effects except for setting the variable cache-id, see below.

— Macro: AC_CACHE_CHECK (message, cache-id, commands-to-set-it)

A wrapper for AC_CACHE_VAL that takes care of printing the messages. This macro provides a convenient shorthand for the most common way to use these macros. It calls AC_MSG_CHECKING for message, then AC_CACHE_VAL with the cache-id and commands arguments, and AC_MSG_RESULT with cache-id.

The commands-to-set-it must have no side effects except for setting the variable cache-id, see below.

It is common to find buggy macros using AC_CACHE_VAL or AC_CACHE_CHECK, because people are tempted to call AC_DEFINE in the commands-to-set-it. Instead, the code that follows the call to AC_CACHE_VAL should call AC_DEFINE, by examining the value of the cache variable. For instance, the following macro is broken:

     AC_DEFUN([AC_SHELL_TRUE],
     [AC_CACHE_CHECK([whether true(1) works], [ac_cv_shell_true_works],
                     [ac_cv_shell_true_works=no
                      (true) 2>/dev/null && ac_cv_shell_true_works=yes
                      if test "$ac_cv_shell_true_works" = yes; then
                        AC_DEFINE([TRUE_WORKS], [1],
                                  [Define if `true(1)' works properly.])
                      fi])
     ])

This fails if the cache is enabled: the second time this macro is run, TRUE_WORKS will not be defined. The proper implementation is:

     AC_DEFUN([AC_SHELL_TRUE],
     [AC_CACHE_CHECK([whether true(1) works], [ac_cv_shell_true_works],
                     [ac_cv_shell_true_works=no
                      (true) 2>/dev/null && ac_cv_shell_true_works=yes])
      if test "$ac_cv_shell_true_works" = yes; then
        AC_DEFINE([TRUE_WORKS], [1],
                  [Define if `true(1)' works properly.])
      fi
     ])

Also, commands-to-set-it should not print any messages, for example with AC_MSG_CHECKING; do that before calling AC_CACHE_VAL, so the messages are printed regardless of whether the results of the check are retrieved from the cache or determined by running the shell commands.


Next: , Up: Caching Results

7.4.1 Cache Variable Names

The names of cache variables should have the following format:

     package-prefix_cv_value-type_specific-value_[additional-options]

for example, ‘ac_cv_header_stat_broken’ or ‘ac_cv_prog_gcc_traditional’. The parts of the variable name are:

package-prefix
An abbreviation for your package or organization; the same prefix you begin local Autoconf macros with, except lowercase by convention. For cache values used by the distributed Autoconf macros, this value is ‘ac’.
_cv_
Indicates that this shell variable is a cache value. This string must be present in the variable name, including the leading underscore.
value-type
A convention for classifying cache values, to produce a rational naming system. The values used in Autoconf are listed in Macro Names.
specific-value
Which member of the class of cache values this test applies to. For example, which function (‘alloca’), program (‘gcc’), or output variable (‘INSTALL’).
additional-options
Any particular behavior of the specific member that this test applies to. For example, ‘broken’ or ‘set’. This part of the name may be omitted if it does not apply.

The values assigned to cache variables may not contain newlines. Usually, their values are Boolean (‘yes’ or ‘no’) or the names of files or functions; so this is not an important restriction.


Next: , Previous: Cache Variable Names, Up: Caching Results

7.4.2 Cache Files

A cache file is a shell script that caches the results of configure tests run on one system so they can be shared between configure scripts and configure runs. It is not useful on other systems. If its contents are invalid for some reason, the user may delete or edit it.

By default, configure uses no cache file, to avoid problems caused by accidental use of stale cache files.

To enable caching, configure accepts --config-cache (or -C) to cache results in the file config.cache. Alternatively, --cache-file=file specifies that file be the cache file. The cache file is created if it does not exist already. When configure calls configure scripts in subdirectories, it uses the --cache-file argument so that they share the same cache. See Subdirectories, for information on configuring subdirectories with the AC_CONFIG_SUBDIRS macro.

config.status only pays attention to the cache file if it is given the --recheck option, which makes it rerun configure.

It is wrong to try to distribute cache files for particular system types. There is too much room for error in doing that, and too much administrative overhead in maintaining them. For any features that can't be guessed automatically, use the standard method of the canonical system type and linking files (see Manual Configuration).

The site initialization script can specify a site-wide cache file to use, instead of the usual per-program cache. In this case, the cache file gradually accumulates information whenever someone runs a new configure script. (Running configure merges the new cache results with the existing cache file.) This may cause problems, however, if the system configuration (e.g., the installed libraries or compilers) changes and the stale cache file is not deleted.


Previous: Cache Files, Up: Caching Results

7.4.3 Cache Checkpointing

If your configure script, or a macro called from configure.ac, happens to abort the configure process, it may be useful to checkpoint the cache a few times at key points using AC_CACHE_SAVE. Doing so reduces the amount of time it takes to rerun the configure script with (hopefully) the error that caused the previous abort corrected.

— Macro: AC_CACHE_LOAD

Loads values from existing cache file, or creates a new cache file if a cache file is not found. Called automatically from AC_INIT.

— Macro: AC_CACHE_SAVE

Flushes all cached values to the cache file. Called automatically from AC_OUTPUT, but it can be quite useful to call AC_CACHE_SAVE at key points in configure.ac.

For instance:

      ... AC_INIT, etc. ...
     # Checks for programs.
     AC_PROG_CC
     AC_PROG_AWK
      ... more program checks ...
     AC_CACHE_SAVE
     
     # Checks for libraries.
     AC_CHECK_LIB([nsl], [gethostbyname])
     AC_CHECK_LIB([socket], [connect])
      ... more lib checks ...
     AC_CACHE_SAVE
     
     # Might abort...
     AM_PATH_GTK([1.0.2], [], [AC_MSG_ERROR([GTK not in path])])
     AM_PATH_GTKMM([0.9.5], [], [AC_MSG_ERROR([GTK not in path])])
      ... AC_OUTPUT, etc. ...


Previous: Caching Results, Up: Results

7.5 Printing Messages

configure scripts need to give users running them several kinds of information. The following macros print messages in ways appropriate for each kind. The arguments to all of them get enclosed in shell double quotes, so the shell performs variable and back-quote substitution on them.

These macros are all wrappers around the echo shell command. They direct output to the appropriate file descriptor (see File Descriptor Macros). configure scripts should rarely need to run echo directly to print messages for the user. Using these macros makes it easy to change how and when each kind of message is printed; such changes need only be made to the macro definitions and all the callers change automatically.

To diagnose static issues, i.e., when autoconf is run, see Reporting Messages.

— Macro: AC_MSG_CHECKING (feature-description)

Notify the user that configure is checking for a particular feature. This macro prints a message that starts with ‘checking ’ and ends with ‘...’ and no newline. It must be followed by a call to AC_MSG_RESULT to print the result of the check and the newline. The feature-description should be something like ‘whether the Fortran compiler accepts C++ comments’ or ‘for c89’.

This macro prints nothing if configure is run with the --quiet or --silent option.

— Macro: AC_MSG_RESULT (result-description)

Notify the user of the results of a check. result-description is almost always the value of the cache variable for the check, typically ‘yes’, ‘no’, or a file name. This macro should follow a call to AC_MSG_CHECKING, and the result-description should be the completion of the message printed by the call to AC_MSG_CHECKING.

This macro prints nothing if configure is run with the --quiet or --silent option.

— Macro: AC_MSG_NOTICE (message)

Deliver the message to the user. It is useful mainly to print a general description of the overall purpose of a group of feature checks, e.g.,

          AC_MSG_NOTICE([checking if stack overflow is detectable])
     

This macro prints nothing if configure is run with the --quiet or --silent option.

— Macro: AC_MSG_ERROR (error-description, [exit-status])

Notify the user of an error that prevents configure from completing. This macro prints an error message to the standard error output and exits configure with exit-status (1 by default). error-description should be something like ‘invalid value $HOME for \$HOME’.

The error-description should start with a lower-case letter, and “cannot” is preferred to “can't”.

— Macro: AC_MSG_FAILURE (error-description, [exit-status])

This AC_MSG_ERROR wrapper notifies the user of an error that prevents configure from completing and that additional details are provided in config.log. This is typically used when abnormal results are found during a compilation.

— Macro: AC_MSG_WARN (problem-description)

Notify the configure user of a possible problem. This macro prints the message to the standard error output; configure continues running afterward, so macros that call AC_MSG_WARN should provide a default (back-up) behavior for the situations they warn about. problem-description should be something like ‘ln -s seems to make hard links’.


Next: , Previous: Results, Up: Top

8 Programming in M4

Autoconf is written on top of two layers: M4sugar, which provides convenient macros for pure M4 programming, and M4sh, which provides macros dedicated to shell script generation.

As of this version of Autoconf, these two layers are still experimental, and their interface might change in the future. As a matter of fact, anything that is not documented must not be used.


Next: , Up: Programming in M4

8.1 M4 Quotation

The most common problem with existing macros is an improper quotation. This section, which users of Autoconf can skip, but which macro writers must read, first justifies the quotation scheme that was chosen for Autoconf and then ends with a rule of thumb. Understanding the former helps one to follow the latter.


Next: , Up: M4 Quotation

8.1.1 Active Characters

To fully understand where proper quotation is important, you first need to know what the special characters are in Autoconf: ‘#’ introduces a comment inside which no macro expansion is performed, ‘,’ separates arguments, ‘[’ and ‘]’ are the quotes themselves, and finally ‘(’ and ‘)’ (which M4 tries to match by pairs).

In order to understand the delicate case of macro calls, we first have to present some obvious failures. Below they are “obvious-ified”, but when you find them in real life, they are usually in disguise.

Comments, introduced by a hash and running up to the newline, are opaque tokens to the top level: active characters are turned off, and there is no macro expansion:

     # define([def], ine)
     =># define([def], ine)

Each time there can be a macro expansion, there is a quotation expansion, i.e., one level of quotes is stripped:

     int tab[10];
     =>int tab10;
     [int tab[10];]
     =>int tab[10];

Without this in mind, the reader might try hopelessly to use her macro array:

     define([array], [int tab[10];])
     array
     =>int tab10;
     [array]
     =>array

How can you correctly output the intended results3?


Next: , Previous: Active Characters, Up: M4 Quotation

8.1.2 One Macro Call

Let's proceed on the interaction between active characters and macros with this small macro, which just returns its first argument:

     define([car], [$1])

The two pairs of quotes above are not part of the arguments of define; rather, they are understood by the top level when it tries to find the arguments of define. Therefore, assuming car is not already defined, it is equivalent to write:

     define(car, $1)

But, while it is acceptable for a configure.ac to avoid unnecessary quotes, it is bad practice for Autoconf macros which must both be more robust and also advocate perfect style.

At the top level, there are only two possibilities: either you quote or you don't:

     car(foo, bar, baz)
     =>foo
     [car(foo, bar, baz)]
     =>car(foo, bar, baz)

Let's pay attention to the special characters:

     car(#)
     error-->EOF in argument list

The closing parenthesis is hidden in the comment; with a hypothetical quoting, the top level understood it this way:

     car([#)]

Proper quotation, of course, fixes the problem:

     car([#])
     =>#

Here are more examples:

     car(foo, bar)
     =>foo
     car([foo, bar])
     =>foo, bar
     car((foo, bar))
     =>(foo, bar)
     car([(foo], [bar)])
     =>(foo
     define([a], [b])
     =>
     car(a)
     =>b
     car([a])
     =>b
     car([[a]])
     =>a
     car([[[a]]])
     =>[a]

With this in mind, we can explore the cases where macros invoke macros...


Next: , Previous: One Macro Call, Up: M4 Quotation

8.1.3 Quotation and Nested Macros

The examples below use the following macros:

     define([car], [$1])
     define([active], [ACT, IVE])
     define([array], [int tab[10]])

Each additional embedded macro call introduces other possible interesting quotations:

     car(active)
     =>ACT
     car([active])
     =>ACT, IVE
     car([[active]])
     =>active

In the first case, the top level looks for the arguments of car, and finds ‘active’. Because M4 evaluates its arguments before applying the macro, ‘active’ is expanded, which results in:

     car(ACT, IVE)
     =>ACT

In the second case, the top level gives ‘active’ as first and only argument of car, which results in:

     active
     =>ACT, IVE

i.e., the argument is evaluated after the macro that invokes it. In the third case, car receives ‘[active]’, which results in:

     [active]
     =>active

exactly as we already saw above.

The example above, applied to a more realistic example, gives:

     car(int tab[10];)
     =>int tab10;
     car([int tab[10];])
     =>int tab10;
     car([[int tab[10];]])
     =>int tab[10];

Huh? The first case is easily understood, but why is the second wrong, and the third right? To understand that, you must know that after M4 expands a macro, the resulting text is immediately subjected to macro expansion and quote removal. This means that the quote removal occurs twice—first before the argument is passed to the car macro, and second after the car macro expands to the first argument.

As the author of the Autoconf macro car, you then consider it to be incorrect that your users have to double-quote the arguments of car, so you “fix” your macro. Let's call it qar for quoted car:

     define([qar], [[$1]])

and check that qar is properly fixed:

     qar([int tab[10];])
     =>int tab[10];

Ahhh! That's much better.

But note what you've done: now that the arguments are literal strings, if the user wants to use the results of expansions as arguments, she has to use an unquoted macro call:

     qar(active)
     =>ACT

where she wanted to reproduce what she used to do with car:

     car([active])
     =>ACT, IVE

Worse yet: she wants to use a macro that produces a set of cpp macros:

     define([my_includes], [#include <stdio.h>])
     car([my_includes])
     =>#include <stdio.h>
     qar(my_includes)
     error-->EOF in argument list

This macro, qar, because it double quotes its arguments, forces its users to leave their macro calls unquoted, which is dangerous. Commas and other active symbols are interpreted by M4 before they are given to the macro, often not in the way the users expect. Also, because qar behaves differently from the other macros, it's an exception that should be avoided in Autoconf.


Next: , Previous: Quotation and Nested Macros, Up: M4 Quotation

8.1.4 changequote is Evil

The temptation is often high to bypass proper quotation, in particular when it's late at night. Then, many experienced Autoconf hackers finally surrender to the dark side of the force and use the ultimate weapon: changequote.

The M4 builtin changequote belongs to a set of primitives that allow one to adjust the syntax of the language to adjust it to one's needs. For instance, by default M4 uses ‘`’ and ‘'’ as quotes, but in the context of shell programming (and actually of most programming languages), that's about the worst choice one can make: because of strings and back-quoted expressions in shell code (such as ‘'this'’ and ‘`that`’), because of literal characters in usual programming languages (as in ‘'0'’), there are many unbalanced ‘`’ and ‘'’. Proper M4 quotation then becomes a nightmare, if not impossible. In order to make M4 useful in such a context, its designers have equipped it with changequote, which makes it possible to choose another pair of quotes. M4sugar, M4sh, Autoconf, and Autotest all have chosen to use ‘[’ and ‘]’. Not especially because they are unlikely characters, but because they are characters unlikely to be unbalanced.

There are other magic primitives, such as changecom to specify what syntactic forms are comments (it is common to see ‘changecom(<!--, -->)’ when M4 is used to produce HTML pages), changeword and changesyntax to change other syntactic details (such as the character to denote the nth argument, ‘$’ by default, the parenthesis around arguments, etc.).

These primitives are really meant to make M4 more useful for specific domains: they should be considered like command line options: --quotes, --comments, --words, and --syntax. Nevertheless, they are implemented as M4 builtins, as it makes M4 libraries self contained (no need for additional options).

There lies the problem...


The problem is that it is then tempting to use them in the middle of an M4 script, as opposed to its initialization. This, if not carefully thought out, can lead to disastrous effects: you are changing the language in the middle of the execution. Changing and restoring the syntax is often not enough: if you happened to invoke macros in between, these macros are lost, as the current syntax is probably not the one they were implemented with.


Next: , Previous: Changequote is Evil, Up: M4 Quotation

8.1.5 Quadrigraphs

When writing an Autoconf macro you may occasionally need to generate special characters that are difficult to express with the standard Autoconf quoting rules. For example, you may need to output the regular expression ‘[^[]’, which matches any character other than ‘[’. This expression contains unbalanced brackets so it cannot be put easily into an M4 macro.

You can work around this problem by using one of the following quadrigraphs:

@<:@
[
@:>@
]
@S|@
$
@%:@
#
@&t@
Expands to nothing.

Quadrigraphs are replaced at a late stage of the translation process, after m4 is run, so they do not get in the way of M4 quoting. For example, the string ‘^@<:@’, independently of its quotation, appears as ‘^[’ in the output.

The empty quadrigraph can be used:

The name ‘@&t@’ was suggested by Paul Eggert:

I should give some credit to the ‘@&t@’ pun. The ‘&’ is my own invention, but the ‘t’ came from the source code of the algol68c compiler, written by Steve Bourne (of Bourne shell fame), and which used ‘mt’ to denote the empty string. In C, it would have looked like something like:
          char const mt[] = "";
     

but of course the source code was written in Algol 68.

I don't know where he got ‘mt’ from: it could have been his own invention, and I suppose it could have been a common pun around the Cambridge University computer lab at the time.


Previous: Quadrigraphs, Up: M4 Quotation

8.1.6 Quotation Rule Of Thumb

To conclude, the quotation rule of thumb is:

One pair of quotes per pair of parentheses.

Never over-quote, never under-quote, in particular in the definition of macros. In the few places where the macros need to use brackets (usually in C program text or regular expressions), properly quote the arguments!

It is common to read Autoconf programs with snippets like:

     AC_TRY_LINK(
     changequote(<<, >>)dnl
     <<#include <time.h>
     #ifndef tzname /* For SGI.  */
     extern char *tzname[]; /* RS6000 and others reject char **tzname.  */
     #endif>>,
     changequote([, ])dnl
     [atoi (*tzname);], ac_cv_var_tzname=yes, ac_cv_var_tzname=no)

which is incredibly useless since AC_TRY_LINK is already double quoting, so you just need:

     AC_TRY_LINK(
     [#include <time.h>
     #ifndef tzname /* For SGI.  */
     extern char *tzname[]; /* RS6000 and others reject char **tzname.  */
     #endif],
                 [atoi (*tzname);],
                 [ac_cv_var_tzname=yes],
                 [ac_cv_var_tzname=no])

The M4-fluent reader might note that these two examples are rigorously equivalent, since M4 swallows both the ‘changequote(<<, >>)’ and ‘<<’ ‘>>’ when it collects the arguments: these quotes are not part of the arguments!

Simplified, the example above is just doing this:

     changequote(<<, >>)dnl
     <<[]>>
     changequote([, ])dnl

instead of simply:

     [[]]

With macros that do not double quote their arguments (which is the rule), double-quote the (risky) literals:

     AC_LINK_IFELSE([AC_LANG_PROGRAM(
     [[#include <time.h>
     #ifndef tzname /* For SGI.  */
     extern char *tzname[]; /* RS6000 and others reject char **tzname.  */
     #endif]],
                                     [atoi (*tzname);])],
                    [ac_cv_var_tzname=yes],
                    [ac_cv_var_tzname=no])

Please note that the macro AC_TRY_LINK is obsolete, so you really should be using AC_LINK_IFELSE instead.

See Quadrigraphs, for what to do if you run into a hopeless case where quoting does not suffice.

When you create a configure script using newly written macros, examine it carefully to check whether you need to add more quotes in your macros. If one or more words have disappeared in the M4 output, you need more quotes. When in doubt, quote.

However, it's also possible to put on too many layers of quotes. If this happens, the resulting configure script may contain unexpanded macros. The autoconf program checks for this problem by looking for the string ‘AC_’ in configure. However, this heuristic does not work in general: for example, it does not catch overquoting in AC_DEFINE descriptions.


Next: , Previous: M4 Quotation, Up: Programming in M4

8.2 Using autom4te

The Autoconf suite, including M4sugar, M4sh, and Autotest, in addition to Autoconf per se, heavily rely on M4. All these different uses revealed common needs factored into a layer over M4: autom4te4.

autom4te is a preprocessor that is like m4. It supports M4 extensions designed for use in tools like Autoconf.


Next: , Up: Using autom4te

8.2.1 Invoking autom4te

The command line arguments are modeled after M4's:

     autom4te options files

where the files are directly passed to m4. By default, GNU M4 is found during configuration, but the environment variable M4 can be set to tell autom4te where to look. In addition to the regular expansion, it handles the replacement of the quadrigraphs (see Quadrigraphs), and of ‘__oline__’, the current line in the output. It supports an extended syntax for the files:

file.m4f
This file is an M4 frozen file. Note that all the previous files are ignored. See the option --melt for the rationale.
file?
If found in the library path, the file is included for expansion, otherwise it is ignored instead of triggering a failure.

Of course, it supports the Autoconf common subset of options:
--help
-h
Print a summary of the command line options and exit.
--version
-V
Print the version number of Autoconf and exit.
--verbose
-v
Report processing steps.
--debug
-d
Don't remove the temporary files and be even more verbose.
--include=dir
-I dir
Also look for input files in dir. Multiple invocations accumulate.
--output=file
-o file
Save output (script or trace) to file. The file - stands for the standard output.

As an extension of m4, it includes the following options:
--warnings=category
-W category
Report the warnings related to category (which can actually be a comma separated list). See Reporting Messages, macro AC_DIAGNOSE, for a comprehensive list of categories. Special values include:
all
report all the warnings
none
report none
error
treats warnings as errors
no-category
disable warnings falling into category

Warnings about ‘syntax’ are enabled by default, and the environment variable WARNINGS, a comma separated list of categories, is honored. ‘autom4te -W category’ actually behaves as if you had run:

          autom4te --warnings=syntax,$WARNINGS,category
     

For example, if you want to disable defaults and WARNINGS of autom4te, but enable the warnings about obsolete constructs, you would use -W none,obsolete.

autom4te displays a back trace for errors, but not for warnings; if you want them, just pass -W error.

--melt
-M
Do not use frozen files. Any argument file.m4f is replaced by file.m4. This helps tracing the macros which are executed only when the files are frozen, typically m4_define. For instance, running:
          autom4te --melt 1.m4 2.m4f 3.m4 4.m4f input.m4
     

is roughly equivalent to running:

          m4 1.m4 2.m4 3.m4 4.m4 input.m4
     

while

          autom4te 1.m4 2.m4f 3.m4 4.m4f input.m4
     

is equivalent to:

          m4 --reload-state=4.m4f input.m4
     

--freeze
-f
Produce a frozen state file. autom4te freezing is stricter than M4's: it must produce no warnings, and no output other than empty lines (a line with white space is not empty) and comments (starting with ‘#’). Unlike m4's similarly-named option, this option takes no argument:
          autom4te 1.m4 2.m4 3.m4 --freeze --output=3.m4f
     

corresponds to

          m4 1.m4 2.m4 3.m4 --freeze-state=3.m4f
     

--mode=octal-mode
-m octal-mode
Set the mode of the non-traces output to octal-mode; by default ‘0666’.

As another additional feature over m4, autom4te caches its results. GNU M4 is able to produce a regular output and traces at the same time. Traces are heavily used in the GNU Build System: autoheader uses them to build config.h.in, autoreconf to determine what GNU Build System components are used, automake to “parse” configure.ac etc. To avoid recomputation, traces are cached while performing regular expansion, and conversely. This cache is (actually, the caches are) stored in the directory autom4te.cache. It can safely be removed at any moment (especially if for some reason autom4te considers it is trashed).
--cache=directory
-C directory
Specify the name of the directory where the result should be cached. Passing an empty value disables caching. Be sure to pass a relative file name, as for the time being, global caches are not supported.
--no-cache
Don't cache the results.
--force
-f
If a cache is used, consider it obsolete (but update it anyway).

Because traces are so important to the GNU Build System, autom4te provides high level tracing features as compared to M4, and helps exploiting the cache:
--trace=macro[:format]
-t macro[:format]
Trace the invocations of macro according to the format. Multiple --trace arguments can be used to list several macros. Multiple --trace arguments for a single macro are not cumulative; instead, you should just make format as long as needed.

The format is a regular string, with newlines if desired, and several special escape codes. It defaults to ‘$f:$l:$n:$%’. It can use the following special escapes:

$$
The character ‘$’.
$f
The file name from which macro is called.
$l
The line number from which macro is called.
$d
The depth of the macro call. This is an M4 technical detail that you probably don't want to know about.
$n
The name of the macro.
$num
The numth argument of the call to macro.
$@
$sep@
${separator}@
All the arguments passed to macro, separated by the character sep or the string separator (‘,’ by default). Each argument is quoted, i.e., enclosed in a pair of square brackets.
$*
$sep*
${separator}*
As above, but the arguments are not quoted.
$%
$sep%
${separator}%
As above, but the arguments are not quoted, all new line characters in the arguments are smashed, and the default separator is ‘:’.

The escape ‘$%’ produces single-line trace outputs (unless you put newlines in the ‘separator’), while ‘$@’ and ‘$*’ do not.

See autoconf Invocation, for examples of trace uses.

--preselect=macro
-p macro
Cache the traces of macro, but do not enable traces. This is especially important to save CPU cycles in the future. For instance, when invoked, autoconf preselects all the macros that autoheader, automake, autoreconf, etc., trace, so that running m4 is not needed to trace them: the cache suffices. This results in a huge speed-up.

Finally, autom4te introduces the concept of Autom4te libraries. They consists in a powerful yet extremely simple feature: sets of combined command line arguments:
--language=language
-l language
Use the language Autom4te library. Current languages include:
M4sugar
create M4sugar output.
M4sh
create M4sh executable shell scripts.
Autotest
create Autotest executable test suites.
Autoconf-without-aclocal-m4
create Autoconf executable configure scripts without reading aclocal.m4.
Autoconf
create Autoconf executable configure scripts. This language inherits all the characteristics of Autoconf-without-aclocal-m4 and additionally reads aclocal.m4.

--prepend-include=dir
-B dir
Prepend directory dir to the search path. This is used to include the language-specific files before any third-party macros.

As an example, if Autoconf is installed in its default location, /usr/local, the command ‘autom4te -l m4sugar foo.m4’ is strictly equivalent to the command:

     autom4te --prepend-include /usr/local/share/autoconf \
       m4sugar/m4sugar.m4f --warnings syntax foo.m4

Recursive expansion applies here: the command ‘autom4te -l m4sh foo.m4’ is the same as ‘autom4te --language M4sugar m4sugar/m4sh.m4f foo.m4’, i.e.:

     autom4te --prepend-include /usr/local/share/autoconf \
       m4sugar/m4sugar.m4f m4sugar/m4sh.m4f --mode 777 foo.m4

The definition of the languages is stored in autom4te.cfg.


Previous: autom4te Invocation, Up: Using autom4te

8.2.2 Customizing autom4te

One can customize autom4te via ~/.autom4te.cfg (i.e., as found in the user home directory), and ./.autom4te.cfg (i.e., as found in the directory from which autom4te is run). The order is first reading autom4te.cfg, then ~/.autom4te.cfg, then ./.autom4te.cfg, and finally the command line arguments.

In these text files, comments are introduced with #, and empty lines are ignored. Customization is performed on a per-language basis, wrapped in between a ‘begin-language: "language"’, ‘end-language: "language"’ pair.

Customizing a language stands for appending options (see autom4te Invocation) to the current definition of the language. Options, and more generally arguments, are introduced by ‘args: arguments’. You may use the traditional shell syntax to quote the arguments.

As an example, to disable Autoconf caches (autom4te.cache) globally, include the following lines in ~/.autom4te.cfg:

## ------------------ ##
## User Preferences.  ##
## ------------------ ##

begin-language: "Autoconf-without-aclocal-m4"
args: --no-cache
end-language: "Autoconf-without-aclocal-m4"


Next: , Previous: Using autom4te, Up: Programming in M4

8.3 Programming in M4sugar

M4 by itself provides only a small, but sufficient, set of all-purpose macros. M4sugar introduces additional generic macros. Its name was coined by Lars J. Aas: “Readability And Greater Understanding Stands 4 M4sugar”.


Next: , Up: Programming in M4sugar

8.3.1 Redefined M4 Macros

With a few exceptions, all the M4 native macros are moved in the ‘m4_’ pseudo-namespace, e.g., M4sugar renames define as m4_define etc.

Some M4 macros are redefined, and are slightly incompatible with their native equivalent.

— Macro: dnl

This macro kept its original name: no m4_dnl is defined.

— Macro: m4_defn (macro)

Unlike the M4 builtin, this macro fails if macro is not defined. See m4_undefine.

— Macro: m4_exit (exit-status)

This macro corresponds to m4exit.

— Macro: m4_if (comment)
— Macro: m4_if (string-1, string-2, equal, [not-equal])
— Macro: m4_if (string-1, string-2, equal, ...)

This macro corresponds to ifelse.

— Macro: m4_include (file)
— Macro: m4_sinclude (file)

Like the M4 builtins, but warn against multiple inclusions of file.

— Macro: m4_bpatsubst (string, regexp, [replacement])

This macro corresponds to patsubst. The name m4_patsubst is kept for future versions of M4sh, on top of GNU M4 which will provide extended regular expression syntax via epatsubst.

— Macro: m4_popdef (macro)

Unlike the M4 builtin, this macro fails if macro is not defined. See m4_undefine.

— Macro: m4_bregexp (string, regexp, [replacement])

This macro corresponds to regexp. The name m4_regexp is kept for future versions of M4sh, on top of GNU M4 which will provide extended regular expression syntax via eregexp.

— Macro: m4_wrap (text)

This macro corresponds to m4wrap.

Posix requires arguments of multiple m4wrap calls to be reprocessed at EOF in the same order as the original calls. GNU M4 versions through 1.4.x, however, reprocess them in reverse order. Your code should not depend on the order.

Also, Posix requires m4wrap to ignore its second and succeeding arguments, but GNU M4 versions through 1.4.x concatenate the arguments with intervening spaces. Your code should not pass more than one argument.

You are encouraged to end text with ‘[]’, to avoid unexpected token pasting between consecutive invocations of m4_wrap, as in:

          m4_define([foo], [bar])
          m4_define([foofoo], [OUCH])
          m4_wrap([foo])
          m4_wrap([foo])
          =>OUCH
     
— Macro: m4_undefine (macro)

Unlike the M4 builtin, this macro fails if macro is not defined. Use

          m4_ifdef([macro], [m4_undefine([macro])])
     

to recover the behavior of the builtin.


Next: , Previous: Redefined M4 Macros, Up: Programming in M4sugar

8.3.2 Looping constructs

The following macros implement loops in M4.

— Macro: m4_for (var, first, last, [step], expression)

Loop over the numeric values between first and last including bounds by increments of step. For each iteration, expand expression with the numeric value assigned to var. If step is omitted, it defaults to ‘1’ or ‘-1’ depending on the order of the limits. If given, step has to match this order.

— Macro: m4_foreach (var, list, expression)

Loop over the comma-separated M4 list list, assigning each value to var, and expand expression. The following example outputs two lines:

          m4_foreach([myvar], [[foo], [bar, baz]],
                     [echo myvar
          ])
          
     
— Macro: m4_foreach_w (var, list, expression)

Loop over the whitespace-separated list list, assigning each value to var, and expand expression.

The deprecated macro AC_FOREACH is an alias of m4_foreach_w.


Next: , Previous: Looping constructs, Up: Programming in M4sugar

8.3.3 Evaluation Macros

The following macros give some control over the order of the evaluation by adding or removing levels of quotes. They are meant for hard-core M4 programmers.

— Macro: m4_dquote (arg1, ...)

Return the arguments as a quoted list of quoted arguments.

— Macro: m4_quote (arg1, ...)

Return the arguments as a single entity, i.e., wrap them into a pair of quotes.

The following example aims at emphasizing the difference between (i), not using these macros, (ii), using m4_quote, and (iii), using m4_dquote.

     $ cat example.m4
     # Overquote, so that quotes are visible.
     m4_define([show], [$[]1 = [$1], $[]@ = [$@]])
     m4_divert(0)dnl
     show(a, b)
     show(m4_quote(a, b))
     show(m4_dquote(a, b))
     $ autom4te -l m4sugar example.m4
     $1 = a, $@ = [a],[b]
     $1 = a,b, $@ = [a,b]
     $1 = [a],[b], $@ = [[a],[b]]


Next: , Previous: Evaluation Macros, Up: Programming in M4sugar

8.3.4 Text processing Macros

The following macros may be used to manipulate strings in M4. They are not intended for casual use.

— Macro: m4_re_escape (string)

Backslash-escape all characters in string that are active in regexps.

— Macro: m4_tolower (string)
— Macro: m4_toupper (string)

Return string with letters converted to upper or lower case, respectively.

— Macro: m4_split (string, [regexp])

Split string into an M4 list of elements quoted by ‘[’ and ‘]’, while keeping white space at the beginning and at the end. If regexp is given, use it instead of ‘[\t ]+’ for splitting. If string is empty, the result is an empty list.

— Macro: m4_normalize (string)

Remove leading and trailing spaces and tabs, sequences of backslash-then-newline, and replace multiple spaces and tabs with a single space.

— Macro: m4_append (macro-name, string, [separator])
— Macro: m4_append_uniq (macro-name, string, [separator])

Redefine macro-name to its former contents with separator and string added at the end. If macro-name was undefined before (but not if it was defined but empty), then no separator is added. m4_append can be used to grow strings, and m4_append_uniq to grow strings without duplicating substrings.


Previous: Text processing Macros, Up: Programming in M4sugar

8.3.5 Forbidden Patterns

M4sugar provides a means to define suspicious patterns, patterns describing tokens which should not be found in the output. For instance, if an Autoconf configure script includes tokens such as ‘AC_DEFINE’, or ‘dnl’, then most probably something went wrong (typically a macro was not evaluated because of overquotation).

M4sugar forbids all the tokens matching ‘^m4_’ and ‘^dnl$’.

— Macro: m4_pattern_forbid (pattern)

Declare that no token matching pattern must be found in the output. Comments are not checked; this can be a problem if, for instance, you have some macro left unexpanded after an ‘#include’. No consensus is currently found in the Autoconf community, as some people consider it should be valid to name macros in comments (which doesn't make sense to the author of this documentation, as ‘#’-comments should document the output, not the input, documented by ‘dnl’ comments).

Of course, you might encounter exceptions to these generic rules, for instance you might have to refer to ‘$m4_flags’.

— Macro: m4_pattern_allow (pattern)

Any token matching pattern is allowed, including if it matches an m4_pattern_forbid pattern.


Next: , Previous: Programming in M4sugar, Up: Programming in M4

8.4 Programming in M4sh

M4sh, pronounced “mash”, is aiming at producing portable Bourne shell scripts. This name was coined by Lars J. Aas, who notes that, according to the Webster's Revised Unabridged Dictionary (1913):

Mash \Mash\, n. [Akin to G. meisch, maisch, meische, maische, mash, wash, and prob. to AS. miscian to mix. See “Mix”.]
  1. A mass of mixed ingredients reduced to a soft pulpy state by beating or pressure...
  2. A mixture of meal or bran and water fed to animals.
  3. A mess; trouble. [Obs.] –Beau. & Fl.

For the time being, it is not mature enough to be widely used.

M4sh provides portable alternatives for some common shell constructs that unfortunately are not portable in practice.

— Macro: AS_BOURNE_COMPATIBLE

Set up the shell to be more compatible with the Bourne shell as standardized by Posix, if possible. This may involve setting environment variables, or setting options, or similar implementation-specific actions.

— Macro: AS_CASE (word, [pattern1], [if-matched1], ..., [default])

Expand into a shell ‘case’ statement, where word is matched against one or more patterns. if-matched is run if the corresponding pattern matched word, else default is run.

— Macro: AS_DIRNAME (file-name)

Output the directory portion of file-name. For example, if $file is ‘/one/two/three’, the command dir=`AS_DIRNAME(["$file"])` sets dir to ‘/one/two’.

— Macro: AS_IF (test1, [run-if-true1], ..., [run-if-false])

Run shell code test1. If test1 exits with a zero status then run shell code run-if-true1, else examine further tests. If no test exits with a zero status, run shell code run-if-false, with simplifications if either run-if-true1 or run-if-false1 is empty. For example,

          AS_IF([test "$foo" = yes], [HANDLE_FOO([yes])],
                [test "$foo" != no], [HANDLE_FOO([maybe])],
                [echo foo not specified])
     

ensures any required macros of HANDLE_FOO are expanded before the first test.

— Macro: AS_MKDIR_P (file-name)

Make the directory file-name, including intervening directories as necessary. This is equivalent to ‘mkdir -p file-name’, except that it is portable to older versions of mkdir that lack support for the -p option. Also, AS_MKDIR_P succeeds if file-name is a symbolic link to an existing directory, even though Posix is unclear whether ‘mkdir -p’ should succeed in that case. If creation of file-name fails, exit the script.

Also see the AC_PROG_MKDIR_P macro (see Particular Programs).

— Macro: AS_SHELL_SANITIZE

Initialize the shell suitably for configure scripts. This has the effect of AS_BOURNE_COMPATIBLE, and sets some other environment variables for predictable results from configuration tests. For example, it sets LC_ALL to change to the default C locale. See Special Shell Variables.

— Macro: AS_TR_CPP (expression)

Transform expression into a valid right-hand side for a C #define. For example:

          # This outputs "#define HAVE_CHAR_P 1".
          type="char *"
          echo "#define AS_TR_CPP([HAVE_$type]) 1"
     
— Macro: AS_TR_SH (expression)

Transform expression into a valid shell variable name. For example:

          # This outputs "Have it!".
          header="sys/some file.h"
          AS_TR_SH([HAVE_$header])=yes
          if test "$HAVE_sys_some_file_h" = yes; then echo "Have it!"; fi
     
— Macro: AS_SET_CATFILE (var, dir, file)

Set the shell variable var to dir/file, but optimizing the common cases (dir or file is ‘.’, file is absolute, etc.).


Previous: Programming in M4sh, Up: Programming in M4

8.5 File Descriptor Macros

The following macros define file descriptors used to output messages (or input values) from configure scripts. For example:

     echo "$wombats found" >&AS_MESSAGE_LOG_FD
     echo 'Enter desired kangaroo count:' >&AS_MESSAGE_FD
     read kangaroos <&AS_ORIGINAL_STDIN_FD`

However doing so is seldom needed, because Autoconf provides higher level macros as described below.

— Macro: AS_MESSAGE_FD

The file descriptor for ‘checking for...’ messages and results. Normally this directs messages to the standard output, however when configure is run with the -q option, messages sent to AS_MESSAGE_FD are discarded.

If you want to display some messages, consider using one of the printing macros (see Printing Messages) instead. Copies of messages output via these macros are also recorded in config.log.

— Macro: AS_MESSAGE_LOG_FD

The file descriptor for messages logged to config.log. Macros that run tools, like AC_COMPILE_IFELSE (see Running the Compiler), redirect all output to this descriptor. You may want to do so if you develop such a low-level macro.

— Macro: AS_ORIGINAL_STDIN_FD

The file descriptor for the original standard input.

When configure runs, it may accidentally execute an interactive command that has the same name as the non-interactive meant to be used or checked. If the standard input was the terminal, such interactive programs would cause configure to stop, pending some user input. Therefore configure redirects its standard input from /dev/null during its initialization. This is not normally a problem, since configure normally does not need user input.

In the extreme case where your configure script really needs to obtain some values from the original standard input, you can read them explicitly from AS_ORIGINAL_STDIN_FD.


Next: , Previous: Programming in M4, Up: Top

9 Writing Autoconf Macros

When you write a feature test that could be applicable to more than one software package, the best thing to do is encapsulate it in a new macro. Here are some instructions and guidelines for writing Autoconf macros.


Next: , Up: Writing Autoconf Macros

9.1 Macro Definitions

Autoconf macros are defined using the AC_DEFUN macro, which is similar to the M4 builtin m4_define macro. In addition to defining a macro, AC_DEFUN adds to it some code that is used to constrain the order in which macros are called (see Prerequisite Macros).

An Autoconf macro definition looks like this:

     AC_DEFUN(macro-name, macro-body)

You can refer to any arguments passed to the macro as ‘$1’, ‘$2’, etc. See How to define new macros, for more complete information on writing M4 macros.

Be sure to properly quote both the macro-body and the macro-name to avoid any problems if the macro happens to have been previously defined.

Each macro should have a header comment that gives its prototype, and a brief description. When arguments have default values, display them in the prototype. For example:

     # AC_MSG_ERROR(ERROR, [EXIT-STATUS = 1])
     # --------------------------------------
     m4_define([AC_MSG_ERROR],
       [{ AS_MESSAGE([error: $1], [2])
          exit m4_default([$2], [1]); }])

Comments about the macro should be left in the header comment. Most other comments make their way into configure, so just keep using ‘#’ to introduce comments.

If you have some special comments about pure M4 code, comments that make no sense in configure and in the header comment, then use the builtin dnl: it causes M4 to discard the text through the next newline.

Keep in mind that dnl is rarely needed to introduce comments; dnl is more useful to get rid of the newlines following macros that produce no output, such as AC_REQUIRE.


Next: , Previous: Macro Definitions, Up: Writing Autoconf Macros

9.2 Macro Names

All of the Autoconf macros have all-uppercase names starting with ‘AC_’ to prevent them from accidentally conflicting with other text. All shell variables that they use for internal purposes have mostly-lowercase names starting with ‘ac_’. To ensure that your macros don't conflict with present or future Autoconf macros, you should prefix your own macro names and any shell variables they use with some other sequence. Possibilities include your initials, or an abbreviation for the name of your organization or software package.

Most of the Autoconf macros' names follow a structured naming convention that indicates the kind of feature check by the name. The macro names consist of several words, separated by underscores, going from most general to most specific. The names of their cache variables use the same convention (see Cache Variable Names, for more information on them).

The first word of the name after ‘AC_’ usually tells the category of the feature being tested. Here are the categories used in Autoconf for specific test macros, the kind of macro that you are more likely to write. They are also used for cache variables, in all-lowercase. Use them where applicable; where they're not, invent your own categories.

C
C language builtin features.
DECL
Declarations of C variables in header files.
FUNC
Functions in libraries.
GROUP
Posix group owners of files.
HEADER
Header files.
LIB
C libraries.
PATH
Absolute names of files, including programs.
PROG
The base names of programs.
MEMBER
Members of aggregates.
SYS
Operating system features.
TYPE
C builtin or declared types.
VAR
C variables in libraries.

After the category comes the name of the particular feature being tested. Any further words in the macro name indicate particular aspects of the feature. For example, AC_FUNC_FNMATCH_GNU checks whether the fnmatch function supports GNU extensions.

An internal macro should have a name that starts with an underscore; Autoconf internals should therefore start with ‘_AC_’. Additionally, a macro that is an internal subroutine of another macro should have a name that starts with an underscore and the name of that other macro, followed by one or more words saying what the internal macro does. For example, AC_PATH_X has internal macros _AC_PATH_X_XMKMF and _AC_PATH_X_DIRECT.


Next: , Previous: Macro Names, Up: Writing Autoconf Macros

9.3 Reporting Messages

When macros statically diagnose abnormal situations, benign or fatal, they should report them using these macros. For dynamic issues, i.e., when configure is run, see Printing Messages.

— Macro: AC_DIAGNOSE (category, message)

Report message as a warning (or as an error if requested by the user) if warnings of the category are turned on. You are encouraged to use standard categories, which currently include:

all
messages that don't fall into one of the following categories. Use of an empty category is equivalent.
cross
related to cross compilation issues.
obsolete
use of an obsolete construct.
syntax
dubious syntactic constructs, incorrectly ordered macro calls.

— Macro: AC_WARNING (message)

Equivalent to ‘AC_DIAGNOSE([syntax], message)’, but you are strongly encouraged to use a finer grained category.

— Macro: AC_FATAL (message)

Report a severe error message, and have autoconf die.

When the user runs ‘autoconf -W error’, warnings from AC_DIAGNOSE and AC_WARNING are reported as error, see autoconf Invocation.


Next: , Previous: Reporting Messages, Up: Writing Autoconf Macros

9.4 Dependencies Between Macros

Some Autoconf macros depend on other macros having been called first in order to work correctly. Autoconf provides a way to ensure that certain macros are called if needed and a way to warn the user if macros are called in an order that might cause incorrect operation.


Next: , Up: Dependencies Between Macros

9.4.1 Prerequisite Macros

A macro that you write might need to use values that have previously been computed by other macros. For example, AC_DECL_YYTEXT examines the output of flex or lex, so it depends on AC_PROG_LEX having been called first to set the shell variable LEX.

Rather than forcing the user of the macros to keep track of the dependencies between them, you can use the AC_REQUIRE macro to do it automatically. AC_REQUIRE can ensure that a macro is only called if it is needed, and only called once.

— Macro: AC_REQUIRE (macro-name)

If the M4 macro macro-name has not already been called, call it (without any arguments). Make sure to quote macro-name with square brackets. macro-name must have been defined using AC_DEFUN or else contain a call to AC_PROVIDE to indicate that it has been called.

AC_REQUIRE must be used inside a macro defined by AC_DEFUN; it must not be called from the top level.

AC_REQUIRE is often misunderstood. It really implements dependencies between macros in the sense that if one macro depends upon another, the latter is expanded before the body of the former. To be more precise, the required macro is expanded before the outermost defined macro in the current expansion stack. In particular, ‘AC_REQUIRE([FOO])’ is not replaced with the body of FOO. For instance, this definition of macros:

     AC_DEFUN([TRAVOLTA],
     [test "$body_temperature_in_celsius" -gt "38" &&
       dance_floor=occupied])
     AC_DEFUN([NEWTON_JOHN],
     [test "$hair_style" = "curly" &&
       dance_floor=occupied])
     
     AC_DEFUN([RESERVE_DANCE_FLOOR],
     [if date | grep '^Sat.*pm' >/dev/null 2>&1; then
       AC_REQUIRE([TRAVOLTA])
       AC_REQUIRE([NEWTON_JOHN])
     fi])

with this configure.ac

     AC_INIT([Dance Manager], [1.0], [bug-dance@example.org])
     RESERVE_DANCE_FLOOR
     if test "$dance_floor" = occupied; then
       AC_MSG_ERROR([cannot pick up here, let's move])
     fi

does not leave you with a better chance to meet a kindred soul at other times than Saturday night since it expands into:

     test "$body_temperature_in_Celsius" -gt "38" &&
       dance_floor=occupied
     test "$hair_style" = "curly" &&
       dance_floor=occupied
     fi
     if date | grep '^Sat.*pm' >/dev/null 2>&1; then
     
     
     fi

This behavior was chosen on purpose: (i) it prevents messages in required macros from interrupting the messages in the requiring macros; (ii) it avoids bad surprises when shell conditionals are used, as in:

     if ...; then
       AC_REQUIRE([SOME_CHECK])
     fi
     ...
     SOME_CHECK

The helper macros AS_IF and AS_CASE may be used to enforce expansion of required macros outside of shell conditional constructs. You are furthermore encouraged to put all AC_REQUIRE calls at the beginning of a macro. You can use dnl to avoid the empty lines they leave.


Next: , Previous: Prerequisite Macros, Up: Dependencies Between Macros

9.4.2 Suggested Ordering

Some macros should be run before another macro if both are called, but neither requires that the other be called. For example, a macro that changes the behavior of the C compiler should be called before any macros that run the C compiler. Many of these dependencies are noted in the documentation.

Autoconf provides the AC_BEFORE macro to warn users when macros with this kind of dependency appear out of order in a configure.ac file. The warning occurs when creating configure from configure.ac, not when running configure.

For example, AC_PROG_CPP checks whether the C compiler can run the C preprocessor when given the -E option. It should therefore be called after any macros that change which C compiler is being used, such as AC_PROG_CC. So AC_PROG_CC contains:

     AC_BEFORE([$0], [AC_PROG_CPP])dnl

This warns the user if a call to AC_PROG_CPP has already occurred when AC_PROG_CC is called.

— Macro: AC_BEFORE (this-macro-name, called-macro-name)

Make M4 print a warning message to the standard error output if called-macro-name has already been called. this-macro-name should be the name of the macro that is calling AC_BEFORE. The macro called-macro-name must have been defined using AC_DEFUN or else contain a call to AC_PROVIDE to indicate that it has been called.


Previous: Suggested Ordering, Up: Dependencies Between Macros

9.4.3 One-Shot Macros

Some macros should be called only once, either because calling them multiple time is unsafe, or because it is bad style. For instance Autoconf ensures that AC_CANONICAL_BUILD and cousins (see Canonicalizing) are evaluated only once, because it makes no sense to run these expensive checks more than once. Such one-shot macros can be defined using AC_DEFUN_ONCE.

— Macro: AC_DEFUN_ONCE (macro-name, macro-body)

Declare macro macro-name like AC_DEFUN would (see Macro Definitions), and emit a warning any time the macro is called more than once.

Obviously it is not sensible to evaluate a macro defined by AC_DEFUN_ONCE in a macro defined by AC_DEFUN. Most of the time you want to use AC_REQUIRE (see Prerequisite Macros).


Next: , Previous: Dependencies Between Macros, Up: Writing Autoconf Macros

9.5 Obsoleting Macros

Configuration and portability technology has evolved over the years. Often better ways of solving a particular problem are developed, or ad-hoc approaches are systematized. This process has occurred in many parts of Autoconf. One result is that some of the macros are now considered obsolete; they still work, but are no longer considered the best thing to do, hence they should be replaced with more modern macros. Ideally, autoupdate should replace the old macro calls with their modern implementation.

Autoconf provides a simple means to obsolete a macro.

— Macro: AU_DEFUN (old-macro, implementation, [message])

Define old-macro as implementation. The only difference with AC_DEFUN is that the user is warned that old-macro is now obsolete.

If she then uses autoupdate, the call to old-macro is replaced by the modern implementation. message should include information on what to do after running autoupdate; autoupdate prints it as a warning, and includes it in the updated configure.ac file.

The details of this macro are hairy: if autoconf encounters an AU_DEFUNed macro, all macros inside its second argument are expanded as usual. However, when autoupdate is run, only M4 and M4sugar macros are expanded here, while all other macros are disabled and appear literally in the updated configure.ac.

— Macro: AU_ALIAS (old-name, new-name)

Used if the old-name is to be replaced by a call to new-macro with the same parameters. This happens for example if the macro was renamed.


Previous: Obsoleting Macros, Up: Writing Autoconf Macros

9.6 Coding Style

The Autoconf macros follow a strict coding style. You are encouraged to follow this style, especially if you intend to distribute your macro, either by contributing it to Autoconf itself, or via other means.

The first requirement is to pay great attention to the quotation. For more details, see Autoconf Language, and M4 Quotation.

Do not try to invent new interfaces. It is likely that there is a macro in Autoconf that resembles the macro you are defining: try to stick to this existing interface (order of arguments, default values, etc.). We are conscious that some of these interfaces are not perfect; nevertheless, when harmless, homogeneity should be preferred over creativity.

Be careful about clashes both between M4 symbols and between shell variables.

If you stick to the suggested M4 naming scheme (see Macro Names), you are unlikely to generate conflicts. Nevertheless, when you need to set a special value, avoid using a regular macro name; rather, use an “impossible” name. For instance, up to version 2.13, the macro AC_SUBST used to remember what symbol macros were already defined by setting AC_SUBST_symbol, which is a regular macro name. But since there is a macro named AC_SUBST_FILE, it was just impossible to ‘AC_SUBST(FILE)’! In this case, AC_SUBST(symbol) or _AC_SUBST(symbol) should have been used (yes, with the parentheses).

No Autoconf macro should ever enter the user-variable name space; i.e., except for the variables that are the actual result of running the macro, all shell variables should start with ac_. In addition, small macros or any macro that is likely to be embedded in other macros should be careful not to use obvious names.

Do not use dnl to introduce comments: most of the comments you are likely to write are either header comments which are not output anyway, or comments that should make their way into configure. There are exceptional cases where you do want to comment special M4 constructs, in which case dnl is right, but keep in mind that it is unlikely.

M4 ignores the leading blanks and newlines before each argument. Use this feature to indent in such a way that arguments are (more or less) aligned with the opening parenthesis of the macro being called. For instance, instead of

     AC_CACHE_CHECK(for EMX OS/2 environment,
     ac_cv_emxos2,
     [AC_COMPILE_IFELSE([AC_LANG_PROGRAM(, [return __EMX__;])],
     [ac_cv_emxos2=yes], [ac_cv_emxos2=no])])

write

     AC_CACHE_CHECK([for EMX OS/2 environment], [ac_cv_emxos2],
     [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [return __EMX__;])],
                        [ac_cv_emxos2=yes],
                        [ac_cv_emxos2=no])])

or even

     AC_CACHE_CHECK([for EMX OS/2 environment],
                    [ac_cv_emxos2],
                    [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],
                                                        [return __EMX__;])],
                                       [ac_cv_emxos2=yes],
                                       [ac_cv_emxos2=no])])

When using AC_RUN_IFELSE or any macro that cannot work when cross-compiling, provide a pessimistic value (typically ‘no’).

Feel free to use various tricks to prevent auxiliary tools, such as syntax-highlighting editors, from behaving improperly. For instance, instead of:

     m4_bpatsubst([$1], [$"])

use

     m4_bpatsubst([$1], [$""])

so that Emacsen do not open an endless “string” at the first quote. For the same reasons, avoid:

     test $[#] != 0

and use:

     test $[@%:@] != 0

Otherwise, the closing bracket would be hidden inside a ‘#’-comment, breaking the bracket-matching highlighting from Emacsen. Note the preferred style to escape from M4: ‘$[1]’, ‘$[@]’, etc. Do not escape when it is unnecessary. Common examples of useless quotation are ‘[$]$1’ (write ‘$$1’), ‘[$]var’ (use ‘$var’), etc. If you add portability issues to the picture, you'll prefer ‘${1+"$[@]"}’ to ‘"[$]@"’, and you'll prefer do something better than hacking Autoconf :-).

When using sed, don't use -e except for indenting purposes. With the s and y commands, the preferred separator is ‘/’ unless ‘/’ itself might appear in the pattern or replacement, in which case you should use ‘|’, or optionally ‘,’ if you know the pattern and replacement cannot contain a file name. If none of these characters will do, choose a printable character that cannot appear in the pattern or replacement. Characters from the set ‘"#$&'()*;<=>?`|~’ are good choices if the pattern or replacement might contain a file name, since they have special meaning to the shell and are less likely to occur in file names.

See Macro Definitions, for details on how to define a macro. If a macro doesn't use AC_REQUIRE, is expected to never be the object of an AC_REQUIRE directive, and macros required by other macros inside arguments do not need to be expanded before this macro, then use m4_define. In case of doubt, use AC_DEFUN. All the AC_REQUIRE statements should be at the beginning of the macro, and each statement should be followed by dnl.

You should not rely on the number of arguments: instead of checking whether an argument is missing, test that it is not empty. It provides both a simpler and a more predictable interface to the user, and saves room for further arguments.

Unless the macro is short, try to leave the closing ‘])’ at the beginning of a line, followed by a comment that repeats the name of the macro being defined. This introduces an additional newline in configure; normally, that is not a problem, but if you want to remove it you can use ‘[]dnl’ on the last line. You can similarly use ‘[]dnl’ after a macro call to remove its newline. ‘[]dnl’ is recommended instead of ‘dnl’ to ensure that M4 does not interpret the ‘dnl’ as being attached to the preceding text or macro output. For example, instead of:

     AC_DEFUN([AC_PATH_X],
     [AC_MSG_CHECKING([for X])
     AC_REQUIRE_CPP()
     # ...omitted...
       AC_MSG_RESULT([libraries $x_libraries, headers $x_includes])
     fi])

you would write:

     AC_DEFUN([AC_PATH_X],
     [AC_REQUIRE_CPP()[]dnl
     AC_MSG_CHECKING([for X])
     # ...omitted...
       AC_MSG_RESULT([libraries $x_libraries, headers $x_includes])
     fi[]dnl
     ])# AC_PATH_X

If the macro is long, try to split it into logical chunks. Typically, macros that check for a bug in a function and prepare its AC_LIBOBJ replacement should have an auxiliary macro to perform this setup. Do not hesitate to introduce auxiliary macros to factor your code.

In order to highlight the recommended coding style, here is a macro written the old way:

     dnl Check for EMX on OS/2.
     dnl _AC_EMXOS2
     AC_DEFUN(_AC_EMXOS2,
     [AC_CACHE_CHECK(for EMX OS/2 environment, ac_cv_emxos2,
     [AC_COMPILE_IFELSE([AC_LANG_PROGRAM(, return __EMX__;)],
     ac_cv_emxos2=yes, ac_cv_emxos2=no)])
     test "$ac_cv_emxos2" = yes && EMXOS2=yes])

and the new way:

     # _AC_EMXOS2
     # ----------
     # Check for EMX on OS/2.
     m4_define([_AC_EMXOS2],
     [AC_CACHE_CHECK([for EMX OS/2 environment], [ac_cv_emxos2],
     [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [return __EMX__;])],
                        [ac_cv_emxos2=yes],
                        [ac_cv_emxos2=no])])
     test "$ac_cv_emxos2" = yes && EMXOS2=yes[]dnl
     ])# _AC_EMXOS2


Next: , Previous: Writing Autoconf Macros, Up: Top

10 Portable Shell Programming

When writing your own checks, there are some shell-script programming techniques you should avoid in order to make your code portable. The Bourne shell and upward-compatible shells like the Korn shell and Bash have evolved over the years, but to prevent trouble, do not take advantage of features that were added after Unix version 7, circa 1977 (see Systemology).

You should not use shell functions, aliases, negated character classes, or other features that are not found in all Bourne-compatible shells; restrict yourself to the lowest common denominator. Even unset is not supported by all shells!

Some ancient systems have quite small limits on the length of the ‘#!’ line; for instance, 32 bytes (not including the newline) on SunOS 4. A few ancient 4.2BSD based systems (such as Dynix circa 1984) required a single space between the ‘#!’ and the ‘/’. However, these ancient systems are no longer of practical concern.

The set of external programs you should run in a configure script is fairly small. See Utilities in Makefiles, for the list. This restriction allows users to start out with a fairly small set of programs and build the rest, avoiding too many interdependencies between packages.

Some of these external utilities have a portable subset of features; see Limitations of Usual Tools.

There are other sources of documentation about shells. The specification for the Posix Shell Command Language, though more generous than the restrictive shell subset described above, is fairly portable nowadays. Also please see the Shell FAQs.


Next: , Up: Portable Shell

10.1 Shellology

There are several families of shells, most prominently the Bourne family and the C shell family which are deeply incompatible. If you want to write portable shell scripts, avoid members of the C shell family. The the Shell difference FAQ includes a small history of Posix shells, and a comparison between several of them.

Below we describe some of the members of the Bourne shell family.

Ash
Ash is often used on GNU/Linux and BSD systems as a light-weight Bourne-compatible shell. Ash 0.2 has some bugs that are fixed in the 0.3.x series, but portable shell scripts should work around them, since version 0.2 is still shipped with many GNU/Linux distributions.

To be compatible with Ash 0.2:


Bash
To detect whether you are running Bash, test whether BASH_VERSION is set. To require Posix compatibility, run ‘set -o posix’. See Bash Posix Mode, for details.
Bash 2.05 and later
Versions 2.05 and later of Bash use a different format for the output of the set builtin, designed to make evaluating its output easier. However, this output is not compatible with earlier versions of Bash (or with many other shells, probably). So if you use Bash 2.05 or higher to execute configure, you'll need to use Bash 2.05 for all other build tasks as well.
Ksh
The Korn shell is compatible with the Bourne family and it mostly conforms to Posix. It has two major variants commonly called ‘ksh88’ and ‘ksh93’, named after the years of initial release. It is usually called ksh, but is called sh on some hosts if you set your path appropriately.

Solaris systems have three variants: /usr/bin/ksh is ‘ksh88’; it is standard on Solaris 2.0 and later. /usr/xpg4/bin/sh is a Posix-compliant variant of ‘ksh88’; it is standard on Solaris 9 and later. /usr/dt/bin/dtksh is ‘ksh93’. Variants that are not standard may be parts of optional packages. There is no extra charge for these packages, but they are not part of a minimal OS install and therefore some installations may not have it.

Starting with Tru64 Version 4.0, the Korn shell /usr/bin/ksh is also available as /usr/bin/posix/sh. If the environment variable BIN_SH is set to xpg4, subsidiary invocations of the standard shell conform to Posix.

Pdksh
A public-domain clone of the Korn shell called pdksh is widely available: it has most of the ‘ksh88’ features along with a few of its own. It usually sets KSH_VERSION, except if invoked as /bin/sh on OpenBSD, and similarly to Bash you can require Posix compatibility by running ‘set -o posix’. Unfortunately, with pdksh 5.2.14 (the latest stable version as of February 2006) Posix mode is buggy and causes pdksh to depart from Posix in at least one respect:
          $ echo "`echo \"hello\"`"
          hello
          $ set -o posix
          $ echo "`echo \"hello\"`"
          "hello"
     

The last line of output contains spurious quotes. This is yet another reason why portable shell code should not contain "`...\"...\"...`" constructs (see Shell Substitutions).

Zsh
To detect whether you are running zsh, test whether ZSH_VERSION is set. By default zsh is not compatible with the Bourne shell: you must execute ‘emulate sh’, and for zsh versions before 3.1.6-dev-18 you must also set NULLCMD to ‘:’. See Compatibility, for details.

The default Mac OS X sh was originally Zsh; it was changed to Bash in Mac OS X 10.2.

The following discussion between Russ Allbery and Robert Lipe is worth reading:

Russ Allbery:

The GNU assumption that /bin/sh is the one and only shell leads to a permanent deadlock. Vendors don't want to break users' existing shell scripts, and there are some corner cases in the Bourne shell that are not completely compatible with a Posix shell. Thus, vendors who have taken this route will never (OK...“never say never”) replace the Bourne shell (as /bin/sh) with a Posix shell.

Robert Lipe:

This is exactly the problem. While most (at least most System V's) do have a Bourne shell that accepts shell functions most vendor /bin/sh programs are not the Posix shell.

So while most modern systems do have a shell somewhere that meets the Posix standard, the challenge is to find it.


Next: , Previous: Shellology, Up: Portable Shell

10.2 Here-Documents

Don't rely on ‘\’ being preserved just because it has no special meaning together with the next symbol. In the native sh on OpenBSD 2.7 ‘\"’ expands to ‘"’ in here-documents with unquoted delimiter. As a general rule, if ‘\\’ expands to ‘\’ use ‘\\’ to get ‘\’.

With OpenBSD 2.7's sh

     $ cat <<EOF
     > \" \\
     > EOF
     " \

and with Bash:

     bash-2.04$ cat <<EOF
     > \" \\
     > EOF
     \" \

Some shells mishandle large here-documents: for example, Solaris 10 dtksh and the UnixWare 7.1.1 Posix shell, which are derived from Korn shell version M-12/28/93d, mishandle braced variable expansion that crosses a 1024- or 4096-byte buffer boundary within a here-document. Only the part of the variable name after the boundary is used. For example, ${variable} could be replaced by the expansion of ${ble}. If the end of the variable name is aligned with the block boundary, the shell reports an error, as if you used ${}. Instead of ${variable-default}, the shell may expand ${riable-default}, or even ${fault}. This bug can often be worked around by omitting the braces: $variable. The bug was fixed in ‘ksh93g’ (1998-04-30) but as of 2006 many operating systems were still shipping older versions with the bug.

Many older shells (including the Bourne shell) implement here-documents inefficiently. In particular, some shells can be extremely inefficient when a single statement contains many here-documents. For instance if your configure.ac includes something like:

     if <cross_compiling>; then
       assume this and that
     else
       check this
       check that
       check something else
       ...
       on and on forever
       ...
     fi

A shell parses the whole if/fi construct, creating temporary files for each here-document in it. Some shells create links for such here-documents on every fork, so that the clean-up code they had installed correctly removes them. It is creating the links that can take the shell forever.

Moving the tests out of the if/fi, or creating multiple if/fi constructs, would improve the performance significantly. Anyway, this kind of construct is not exactly the typical use of Autoconf. In fact, it's even not recommended, because M4 macros can't look into shell conditionals, so we may fail to expand a macro when it was expanded before in a conditional path, and the condition turned out to be false at runtime, and we end up not executing the macro at all.


Next: , Previous: Here-Documents, Up: Portable Shell

10.3 File Descriptors

Don't redirect the same file descriptor several times, as you are doomed to failure under Ultrix.

     ULTRIX V4.4 (Rev. 69) System #31: Thu Aug 10 19:42:23 GMT 1995
     UWS V4.4 (Rev. 11)
     $ eval 'echo matter >fullness' >void
     illegal io
     $ eval '(echo matter >fullness)' >void
     illegal io
     $ (eval '(echo matter >fullness)') >void
     Ambiguous output redirect.

In each case the expected result is of course fullness containing ‘matter’ and void being empty.

Don't try to redirect the standard error of a command substitution: it must be done inside the command substitution: when running ‘: `cd /zorglub` 2>/dev/null’ expect the error message to escape, while ‘: `cd /zorglub 2>/dev/null`’ works properly.

It is worth noting that Zsh (but not Ash nor Bash) makes it possible in assignments though: ‘foo=`cd /zorglub` 2>/dev/null’.

Most shells, if not all (including Bash, Zsh, Ash), output traces on stderr, even for subshells. This might result in undesirable content if you meant to capture the standard-error output of the inner command:

     $ ash -x -c '(eval "echo foo >&2") 2>stderr'
     $ cat stderr
     + eval echo foo >&2
     + echo foo
     foo
     $ bash -x -c '(eval "echo foo >&2") 2>stderr'
     $ cat stderr
     + eval 'echo foo >&2'
     ++ echo foo
     foo
     $ zsh -x -c '(eval "echo foo >&2") 2>stderr'
     # Traces on startup files deleted here.
     $ cat stderr
     +zsh:1> eval echo foo >&2
     +zsh:1> echo foo
     foo

You'll appreciate the various levels of detail...

One workaround is to grep out uninteresting lines, hoping not to remove good ones...

Don't try to move/delete open files, such as in ‘exec >foo; mv foo bar’; see Limitations of Builtins, mv for more details.

Don't rely on file descriptors 0, 1, and 2 remaining closed in a subsidiary program. If any of these descriptors is closed, the operating system may open an unspecified file for the descriptor in the new process image. Posix says this may be done only if the subsidiary program is set-user-ID or set-group-ID, but HP-UX 11.23 does it even for ordinary programs.

Don't rely on open file descriptors being open in child processes. In ksh, file descriptors above 2 which are opened using ‘exec n>file’ are closed by a subsequent ‘exec’ (such as that involved in the fork-and-exec which runs a program or script). Thus, using sh, we have:

     $ cat ./descrips
     #!/bin/sh -
     echo hello >&5
     $ exec 5>t
     $ ./descrips
     $ cat t
     $
     hello

But using ksh:

     $ exec 5>t
     $ ./descrips
     hello
     $ cat t
     $

Within the process which runs the ‘descrips’ script, file descriptor 5 is closed.

A few ancient systems reserved some file descriptors. By convention, file descriptor 3 was opened to /dev/tty when you logged into Eighth Edition (1985) through Tenth Edition Unix (1989). File descriptor 4 had a special use on the Stardent/Kubota Titan (circa 1990), though we don't now remember what it was. Both these systems are obsolete, so it's now safe to treat file descriptors 3 and 4 like any other file descriptors.


Next: , Previous: File Descriptors, Up: Portable Shell

10.4 File System Conventions

Autoconf uses shell-script processing extensively, so the file names that it processes should not contain characters that are special to the shell. Special characters include space, tab, newline, nul, and the following:

     " # $ & ' ( ) * ; < = > ? [ \ ` |

Also, file names should not begin with ‘~’ or ‘-’, and should contain neither ‘-’ immediately after ‘/’ nor ‘~’ immediately after ‘:’. On Posix-like platforms, directory names should not contain ‘:’, as this runs afoul of ‘:’ used as the path separator.

These restrictions apply not only to the files that you distribute, but also to the absolute file names of your source, build, and destination directories.

On some Posix-like platforms, ‘!’ and ‘^’ are special too, so they should be avoided.

Posix lets implementations treat leading // specially, but requires leading /// and beyond to be equivalent to /. Most Unix variants treat // like /. However, some treat // as a “super-root” that can provide access to files that are not otherwise reachable from /. The super-root tradition began with Apollo Domain/OS, which died out long ago, but unfortunately Cygwin has revived it.

While autoconf and friends are usually run on some Posix variety, they can be used on other systems, most notably DOS variants. This impacts several assumptions regarding file names.

For example, the following code:

     case $foo_dir in
       /*) # Absolute
          ;;
       *)
          foo_dir=$dots$foo_dir ;;
     esac

fails to properly detect absolute file names on those systems, because they can use a drivespec, and usually use a backslash as directory separator. If you want to be portable to DOS variants (at the price of rejecting valid but oddball Posix file names like a:\b), you can check for absolute file names like this:

     case $foo_dir in
       [\\/]* | ?:[\\/]* ) # Absolute
          ;;
       *)
          foo_dir=$dots$foo_dir ;;
     esac

Make sure you quote the brackets if appropriate and keep the backslash as first character (see Limitations of Builtins).

Also, because the colon is used as part of a drivespec, these systems don't use it as path separator. When creating or accessing paths, you can use the PATH_SEPARATOR output variable instead. configure sets this to the appropriate value (‘:’ or ‘;’) when it starts up.

File names need extra care as well. While DOS variants that are Posixy enough to run autoconf (such as DJGPP) are usually able to handle long file names properly, there are still limitations that can seriously break packages. Several of these issues can be easily detected by the doschk package.

A short overview follows; problems are marked with sfn/lfn to indicate where they apply: sfn means the issues are only relevant to plain DOS, not to DOS under Microsoft Windows variants, while lfn identifies problems that exist even under Microsoft Windows variants.

No multiple dots (sfn)
DOS cannot handle multiple dots in file names. This is an especially important thing to remember when building a portable configure script, as autoconf uses a .in suffix for template files.

This is perfectly OK on Posix variants:

          AC_CONFIG_HEADERS([config.h])
          AC_CONFIG_FILES([source.c foo.bar])
          AC_OUTPUT
     

but it causes problems on DOS, as it requires ‘config.h.in’, ‘source.c.in’ and ‘foo.bar.in’. To make your package more portable to DOS-based environments, you should use this instead:

          AC_CONFIG_HEADERS([config.h:config.hin])
          AC_CONFIG_FILES([source.c:source.cin foo.bar:foobar.in])
          AC_OUTPUT
     

No leading dot (sfn)
DOS cannot handle file names that start with a dot. This is usually not important for autoconf.
Case insensitivity (lfn)
DOS is case insensitive, so you cannot, for example, have both a file called ‘INSTALL’ and a directory called ‘install’. This also affects make; if there's a file called ‘INSTALL’ in the directory, ‘make install’ does nothing (unless the ‘install’ target is marked as PHONY).
The 8+3 limit (sfn)
Because the DOS file system only stores the first 8 characters of the file name and the first 3 of the extension, those must be unique. That means that foobar-part1.c, foobar-part2.c and foobar-prettybird.c all resolve to the same file name (FOOBAR-P.C). The same goes for foo.bar and foo.bartender.

The 8+3 limit is not usually a problem under Microsoft Windows, as it uses numeric tails in the short version of file names to make them unique. However, a registry setting can turn this behavior off. While this makes it possible to share file trees containing long file names between sfn and lfn environments, it also means the above problem applies there as well.

Invalid characters (lfn)
Some characters are invalid in DOS file names, and should therefore be avoided. In a lfn environment, these are ‘/’, ‘\’, ‘?’, ‘*’, ‘:’, ‘<’, ‘>’, ‘|’ and ‘"’. In a sfn environment, other characters are also invalid. These include ‘+’, ‘,’, ‘[’ and ‘]’.
Invalid names (lfn)
Some DOS file names are reserved, and cause problems if you try to use files with those names. These names include CON, AUX, COM1, COM2, COM3, COM4, LPT1, LPT2, LPT3, NUL, and PRN. File names are case insensitive, so even names like aux/config.guess are disallowed.


Next: , Previous: File System Conventions, Up: Portable Shell

10.5 Shell Substitutions

Contrary to a persistent urban legend, the Bourne shell does not systematically split variables and back-quoted expressions, in particular on the right-hand side of assignments and in the argument of case. For instance, the following code:

     case "$given_srcdir" in
     .)  top_srcdir="`echo "$dots" | sed 's,/$,,'`" ;;
     *)  top_srcdir="$dots$given_srcdir" ;;
     esac

is more readable when written as:

     case $given_srcdir in
     .)  top_srcdir=`echo "$dots" | sed 's,/$,,'` ;;
     *)  top_srcdir=$dots$given_srcdir ;;
     esac

and in fact it is even more portable: in the first case of the first attempt, the computation of top_srcdir is not portable, since not all shells properly understand "`..."..."...`". Worse yet, not all shells understand "`...\"...\"...`" the same way. There is just no portable way to use double-quoted strings inside double-quoted back-quoted expressions (pfew!).

$@
One of the most famous shell-portability issues is related to ‘"$@"’. When there are no positional arguments, Posix says that ‘"$@"’ is supposed to be equivalent to nothing, but the original Unix version 7 Bourne shell treated it as equivalent to ‘""’ instead, and this behavior survives in later implementations like Digital Unix 5.0.

The traditional way to work around this portability problem is to use ‘${1+"$@"}’. Unfortunately this method does not work with Zsh (3.x and 4.x), which is used on Mac OS X. When emulating the Bourne shell, Zsh performs word splitting on ‘${1+"$@"}’:

          zsh $ emulate sh
          zsh $ for i in "$@"; do echo $i; done
          Hello World
          !
          zsh $ for i in ${1+"$@"}; do echo $i; done
          Hello
          World
          !
     

Zsh handles plain ‘"$@"’ properly, but we can't use plain ‘"$@"’ because of the portability problems mentioned above. One workaround relies on Zsh's “global aliases” to convert ‘${1+"$@"}’ into ‘"$@"’ by itself:

          test "${ZSH_VERSION+set}" = set && alias -g '${1+"$@"}'='"$@"'
     

A more conservative workaround is to avoid ‘"$@"’ if it is possible that there may be no positional arguments. For example, instead of:

          cat conftest.c "$@"
     

you can use this instead:

          case $# in
          0) cat conftest.c;;
          *) cat conftest.c "$@";;
          esac
     

Autoconf macros often use the set command to update ‘$@’, so if you are writing shell code intended for configure you should not assume that the value of ‘$@’ persists for any length of time.

${10}
The 10th, 11th, ... positional parameters can be accessed only after a shift. The 7th Edition shell reported an error if given ${10}, and Solaris 10 /bin/sh still acts that way:
          $ set 1 2 3 4 5 6 7 8 9 10
          $ echo ${10}
          bad substitution
     

${var:-value}
Old BSD shells, including the Ultrix sh, don't accept the colon for any shell substitution, and complain and die.
${var=literal}
Be sure to quote:
          : ${var='Some words'}
     

otherwise some shells, such as on Digital Unix V 5.0, die because of a “bad substitution”.

     
     
Solaris /bin/sh has a frightening bug in its interpretation of this. Imagine you need set a variable to a string containing ‘}’. This ‘}’ character confuses Solaris /bin/sh when the affected variable was already set. This bug can be exercised by running:
          $ unset foo
          $ foo=${foo='}'}
          $ echo $foo
          }
          $ foo=${foo='}'   # no error; this hints to what the bug is
          $ echo $foo
          }
          $ foo=${foo='}'}
          $ echo $foo
          }}
           ^ ugh!
     

It seems that ‘}’ is interpreted as matching ‘${’, even though it is enclosed in single quotes. The problem doesn't happen using double quotes.

${var=expanded-value}
On Ultrix, running
          default="yu,yaa"
          : ${var="$default"}
     

sets var to ‘M-yM-uM-,M-yM-aM-a’, i.e., the 8th bit of each char is set. You don't observe the phenomenon using a simple ‘echo $var’ since apparently the shell resets the 8th bit when it expands $var. Here are two means to make this shell confess its sins:

          $ cat -v <<EOF
          $var
          EOF
     

and

          $ set | grep '^var=' | cat -v
     

One classic incarnation of this bug is:

          default="a b c"
          : ${list="$default"}
          for c in $list; do
            echo $c
          done
     

You'll get ‘a b c’ on a single line. Why? Because there are no spaces in ‘$list’: there are ‘M- ’, i.e., spaces with the 8th bit set, hence no IFS splitting is performed!!!

One piece of good news is that Ultrix works fine with ‘: ${list=$default}’; i.e., if you don't quote. The bad news is then that QNX 4.25 then sets list to the last item of default!

The portable way out consists in using a double assignment, to switch the 8th bit twice on Ultrix:

          list=${list="$default"}
     

...but beware of the ‘}’ bug from Solaris (see above). For safety, use:

          test "${var+set}" = set || var={value}
     

`commands`
Posix requires shells to trim all trailing newlines from command output before substituting it, so assignments like ‘dir=`echo "$file" | tr a A`’ do not work as expected if ‘$file’ ends in a newline.

While in general it makes no sense, do not substitute a single builtin with side effects, because Ash 0.2, trying to optimize, does not fork a subshell to perform the command.

For instance, if you wanted to check that cd is silent, do not use ‘test -z "`cd /`"’ because the following can happen:

          $ pwd
          /tmp
          $ test -z "`cd /`" && pwd
          /
     

The result of ‘foo=`exit 1`’ is left as an exercise to the reader.

The MSYS shell leaves a stray byte in the expansion of a double-quoted command substitution of a native program, if the end of the substution is not aligned with the end of the double quote. This may be worked around by inserting another pair of quotes:

          $ echo "`printf 'foo\r\n'` bar" > broken
          $ echo "`printf 'foo\r\n'`"" bar" | cmp - broken
          - broken differ: char 4, line 1
     

$(commands)
This construct is meant to replace ‘`commands`’, and it has most of the problems listed under `commands`.

This construct can be nested while this is impossible to do portably with back quotes. Unfortunately it is not yet universally supported. Most notably, even recent releases of Solaris don't support it:

          $ showrev -c /bin/sh | grep version
          Command version: SunOS 5.10 Generic 121004-01 Oct 2005
          $ echo $(echo blah)
          syntax error: `(' unexpected
     

nor does irix 6.5's Bourne shell:

          $ uname -a
          IRIX firebird-image 6.5 07151432 IP22
          $ echo $(echo blah)
          $(echo blah)
     

If you do use ‘$(commands)’, make sure that the commands do not start with a parenthesis, as that would cause confusion with a different notation ‘$((expression))’ that in modern shells is an arithmetic expression not a command. To avoid the confusion, insert a space between the two opening parentheses.

Avoid commands that contain unbalanced parentheses in here-documents, comments, or case statement patterns, as many shells mishandle them. For example, Bash 3.1, ‘ksh88’, pdksh 5.2.14, and Zsh 4.2.6 all mishandle the following valid command:

          echo $(case x in x) echo hello;; esac)
     

^
Always quote ‘^’, otherwise traditional shells such as /bin/sh on Solaris 10 treat this like ‘|’.


Next: , Previous: Shell Substitutions, Up: Portable Shell

10.6 Assignments

When setting several variables in a row, be aware that the order of the evaluation is undefined. For instance ‘foo=1 foo=2; echo $foo’ gives ‘1’ with Solaris /bin/sh, but ‘2’ with Bash. You must use ‘;’ to enforce the order: ‘foo=1; foo=2; echo $foo’.

Don't rely on the following to find subdir/program:

     PATH=subdir$PATH_SEPARATOR$PATH program

as this does not work with Zsh 3.0.6. Use something like this instead:

     (PATH=subdir$PATH_SEPARATOR$PATH; export PATH; exec program)

Don't rely on the exit status of an assignment: Ash 0.2 does not change the status and propagates that of the last statement:

     $ false || foo=bar; echo $?
     1
     $ false || foo=`:`; echo $?
     0

and to make things even worse, QNX 4.25 just sets the exit status to 0 in any case:

     $ foo=`exit 1`; echo $?
     0

To assign default values, follow this algorithm:

  1. If the default value is a literal and does not contain any closing brace, use:
              : ${var='my literal'}
         
  2. If the default value contains no closing brace, has to be expanded, and the variable being initialized is not intended to be IFS-split (i.e., it's not a list), then use:
              : ${var="$default"}
         
  3. If the default value contains no closing brace, has to be expanded, and the variable being initialized is intended to be IFS-split (i.e., it's a list), then use:
              var=${var="$default"}
         
  4. If the default value contains a closing brace, then use:
              test "${var+set}" = set || var="has a '}'"
         

In most cases ‘var=${var="$default"}’ is fine, but in case of doubt, just use the last form. See Shell Substitutions, items ‘${var:-value}’ and ‘${var=value}’ for the rationale.


Next: , Previous: Assignments, Up: Portable Shell

10.7 Parentheses in Shell Scripts

Beware of two opening parentheses in a row, as some shell implementations mishandle them. For example, ‘pdksh’ 5.2.14 misparses the following code:

     if ((true) || false); then
       echo ok
     fi

To work around this problem, insert a space between the two opening parentheses. There is a similar problem and workaround with ‘$((’; see Shell Substitutions.

Posix requires support for case patterns with opening parentheses like this:

     case $file_name in
     (*.c) echo "C source code";;
     esac

but the ( in this example is not portable to many older Bourne shell implementations. It can be omitted safely.


Next: , Previous: Parentheses, Up: Portable Shell

10.8 Slashes in Shell Scripts

Unpatched Tru64 5.1 sh omits the last slash of command-line arguments that contain two trailing slashes:

     $ echo / // /// //// .// //.
     / / // /// ./ //.
     $ x=//
     $ eval "echo \$x"
     /
     $ set -x
     $ echo abc | tr -t ab //
     + echo abc
     + tr -t ab /
     /bc

However, our understanding is that patches are available, so perhaps it's not worth worrying about working around this horrendous bug.


Next: , Previous: Slashes, Up: Portable Shell

10.9 Special Shell Variables

Some shell variables should not be used, since they can have a deep influence on the behavior of the shell. In order to recover a sane behavior from the shell, some variables should be unset, but unset is not portable (see Limitations of Builtins) and a fallback value is needed.

As a general rule, shell variable names containing a lower-case letter are safe; you can define and use these variables without worrying about their effect on the underlying system, and without worrying about whether the shell changes them unexpectedly. (The exception is the shell variable status, as described below.)

Here is a list of names that are known to cause trouble. This list is not exhaustive, but you should be safe if you avoid the name status and names containing only upper-case letters and underscores.

_
Many shells reserve ‘$_’ for various purposes, e.g., the name of the last command executed.
BIN_SH
In Tru64, if BIN_SH is set to xpg4, subsidiary invocations of the standard shell conform to Posix. Autoconf-generated scripts export this variable when they start up.
CDPATH
When this variable is set it specifies a list of directories to search when invoking cd with a relative file name that did not start with ‘./’ or ‘../’. Posix 1003.1-2001 says that if a nonempty directory name from CDPATH is used successfully, cd prints the resulting absolute file name. Unfortunately this output can break idioms like ‘abs=`cd src && pwd`’ because abs receives the name twice. Also, many shells do not conform to this part of Posix; for example, zsh prints the result only if a directory name other than . was chosen from CDPATH.

In practice the shells that have this problem also support unset, so you can work around the problem as follows:

          (unset CDPATH) >/dev/null 2>&1 && unset CDPATH
     

You can also avoid output by ensuring that your directory name is absolute or anchored at ‘./’, as in ‘abs=`cd ./src && pwd`’.

Autoconf-generated scripts automatically unset CDPATH if possible, so you need not worry about this problem in those scripts.

DUALCASE
In the MKS shell, case statements and file name generation are case-insensitive unless DUALCASE is nonzero. Autoconf-generated scripts export this variable when they start up.
ENV
MAIL
MAILPATH
PS1
PS2
PS4
These variables should not matter for shell scripts, since they are supposed to affect only interactive shells. However, at least one shell (the pre-3.0 uwin Korn shell) gets confused about whether it is interactive, which means that (for example) a PS1 with a side effect can unexpectedly modify ‘$?’. To work around this bug, Autoconf-generated scripts do something like this:
          (unset ENV) >/dev/null 2>&1 && unset ENV MAIL MAILPATH
          PS1='$ '
          PS2='> '
          PS4='+ '
     

IFS
Long ago, shell scripts inherited IFS from the environment, but this caused many problems so modern shells ignore any environment settings for IFS.

Don't set the first character of IFS to backslash. Indeed, Bourne shells use the first character (backslash) when joining the components in ‘"$@"’ and some shells then reinterpret (!) the backslash escapes, so you can end up with backspace and other strange characters.

The proper value for IFS (in regular code, not when performing splits) is ‘<SPC><TAB><RET>’. The first character is especially important, as it is used to join the arguments in ‘$*’; however, note that traditional shells, but also bash-2.04, fail to adhere to this and join with a space anyway.

LANG
LC_ALL
LC_COLLATE
LC_CTYPE
LC_MESSAGES
LC_MONETARY
LC_NUMERIC
LC_TIME
Autoconf-generated scripts normally set all these variables to ‘C’ because so much configuration code assumes the C locale and Posix requires that locale environment variables be set to ‘C’ if the C locale is desired. However, some older, nonstandard systems (notably SCO) break if locale environment variables are set to ‘C’, so when running on these systems Autoconf-generated scripts unset the variables instead.
LANGUAGE
LANGUAGE is not specified by Posix, but it is a GNU extension that overrides LC_ALL in some cases, so Autoconf-generated scripts set it too.
LC_ADDRESS
LC_IDENTIFICATION
LC_MEASUREMENT
LC_NAME
LC_PAPER
LC_TELEPHONE
These locale environment variables are GNU extensions. They are treated like their Posix brethren (LC_COLLATE, etc.) as described above.
LINENO
Most modern shells provide the current line number in LINENO. Its value is the line number of the beginning of the current command. Autoconf attempts to execute configure with a shell that supports LINENO. If no such shell is available, it attempts to implement LINENO with a Sed prepass that replaces each instance of the string $LINENO (not followed by an alphanumeric character) with the line's number.

You should not rely on LINENO within eval, as the behavior differs in practice. Also, the possibility of the Sed prepass means that you should not rely on $LINENO when quoted, when in here-documents, or when in long commands that cross line boundaries. Subshells should be OK, though. In the following example, lines 1, 6, and 9 are portable, but the other instances of LINENO are not:

          $ cat lineno
          echo 1. $LINENO
          cat <<EOF
          3. $LINENO
          4. $LINENO
          EOF
          ( echo 6. $LINENO )
          eval 'echo 7. $LINENO'
          echo 8. '$LINENO'
          echo 9. $LINENO '
          10.' $LINENO
          $ bash-2.05 lineno
          1. 1
          3. 2
          4. 2
          6. 6
          7. 1
          8. $LINENO
          9. 9
          10. 9
          $ zsh-3.0.6 lineno
          1. 1
          3. 2
          4. 2
          6. 6
          7. 7
          8. $LINENO
          9. 9
          10. 9
          $ pdksh-5.2.14 lineno
          1. 1
          3. 2
          4. 2
          6. 6
          7. 0
          8. $LINENO
          9. 9
          10. 9
          $ sed '=' <lineno |
          >   sed '
          >     N
          >     s,$,-,
          >     t loop
          >     :loop
          >     s,^\([0-9]*\)\(.*\)[$]LINENO\([^a-zA-Z0-9_]\),\1\2\1\3,
          >     t loop
          >     s,-$,,
          >     s,^[0-9]*\n,,
          >   ' |
          >   sh
          1. 1
          3. 3
          4. 4
          6. 6
          7. 7
          8. 8
          9. 9
          10. 10
     

NULLCMD
When executing the command ‘>foo’, zsh executes ‘$NULLCMD >foo’ unless it is operating in Bourne shell compatibility mode and the zsh version is newer than 3.1.6-dev-18. If you are using an older zsh and forget to set NULLCMD, your script might be suspended waiting for data on its standard input.
PATH_SEPARATOR
On DJGPP systems, the PATH_SEPARATOR environment variable can be set to either ‘:’ or ‘;’ to control the path separator Bash uses to set up certain environment variables (such as PATH). You can set this variable to ‘;’ if you want configure to use ‘;’ as a separator; this might be useful if you plan to use non-Posix shells to execute files. See File System Conventions, for more information about PATH_SEPARATOR.
PWD
Posix 1003.1-2001 requires that cd and pwd must update the PWD environment variable to point to the logical name of the current directory, but traditional shells do not support this. This can cause confusion if one shell instance maintains PWD but a subsidiary and different shell does not know about PWD and executes cd; in this case PWD points to the wrong directory. Use ‘`pwd`’ rather than ‘$PWD’.
RANDOM
Many shells provide RANDOM, a variable that returns a different integer each time it is used. Most of the time, its value does not change when it is not used, but on irix 6.5 the value changes all the time. This can be observed by using set. It is common practice to use $RANDOM as part of a file name, but code shouldn't rely on $RANDOM expanding to a nonempty string.
status
This variable is an alias to ‘$?’ for zsh (at least 3.1.6), hence read-only. Do not use it.


Next: , Previous: Special Shell Variables, Up: Portable Shell

10.10 Limitations of Shell Builtins

No, no, we are serious: some shells do have limitations! :)

You should always keep in mind that any builtin or command may support options, and therefore differ in behavior with arguments starting with a dash. For instance, the innocent ‘echo "$word"’ can give unexpected results when word starts with a dash. It is often possible to avoid this problem using ‘echo "x$word"’, taking the ‘x’ into account later in the pipe.

.
Use . only with regular files (use ‘test -f’). Bash 2.03, for instance, chokes on ‘. /dev/null’. Also, remember that . uses PATH if its argument contains no slashes, so if you want to use . on a file foo in the current directory, you must use ‘. ./foo’.
!
The Unix version 7 shell did not support negating the exit status of commands with !, and this feature is still absent from some shells (e.g., Solaris /bin/sh). Shell code like this:
          if ! cmp file1 file2 >/dev/null 2>&1; then
            echo files differ or trouble
          fi
     

is therefore not portable in practice. Typically it is easy to rewrite such code, e.g.:

          cmp file1 file2 >/dev/null 2>&1 ||
            echo files differ or trouble
     

More generally, one can always rewrite ‘! command’ as:

          if command; then (exit 1); else :; fi
     

break
The use of ‘break 2’ etc. is safe.
case
You don't need to quote the argument; no splitting is performed.

You don't need the final ‘;;’, but you should use it.

Because of a bug in its fnmatch, Bash fails to properly handle backslashes in character classes:

          bash-2.02$ case /tmp in [/\\]*) echo OK;; esac
          bash-2.02$
     

This is extremely unfortunate, since you are likely to use this code to handle Posix or ms-dos absolute file names. To work around this bug, always put the backslash first:

          bash-2.02$ case '\TMP' in [\\/]*) echo OK;; esac
          OK
          bash-2.02$ case /tmp in [\\/]*) echo OK;; esac
          OK
     

Many Bourne shells cannot handle closing brackets in character classes correctly.

Some shells also have problems with backslash escaping in case you do not want to match the backslash: both a backslash and the escaped character match this pattern. To work around this, specify the character class in a variable, so that quote removal does not apply afterwards, and the special characters don't have to be backslash-escaped:

          $ case '\' in [\<]) echo OK;; esac
          OK
          $ scanset='[<]'; case '\' in $scanset) echo OK;; esac
          $
     

Even with this, Solaris ksh matches a backslash if the set contains any of the characters ‘|’, ‘&’, ‘(’, or ‘)’.

Conversely, Tru64 ksh (circa 2003) erroneously always matches a closing parenthesis if not specified in a character class:

          $ case foo in *\)*) echo fail ;; esac
          fail
          $ case foo in *')'*) echo fail ;; esac
          fail
     

Some shells, such as Ash 0.3.8, are confused by an empty case/esac:

          ash-0.3.8 $ case foo in esac;
          error-->Syntax error: ";" unexpected (expecting ")")
     

Many shells still do not support parenthesized cases, which is a pity for those of us using tools that rely on balanced parentheses. For instance, Solaris /bin/sh:

          $ case foo in (foo) echo foo;; esac
          error-->syntax error: `(' unexpected
     

cd
Posix 1003.1-2001 requires that cd must support the -L (“logical”) and -P (“physical”) options, with -L being the default. However, traditional shells do not support these options, and their cd command has the -P behavior.

Portable scripts should assume neither option is supported, and should assume neither behavior is the default. This can be a bit tricky, since the Posix default behavior means that, for example, ‘ls ..’ and ‘cd ..’ may refer to different directories if the current logical directory is a symbolic link. It is safe to use cd dir if dir contains no .. components. Also, Autoconf-generated scripts check for this problem when computing variables like ac_top_srcdir (see Configuration Actions), so it is safe to cd to these variables.

See See Special Shell Variables, for portability problems involving cd and the CDPATH environment variable. Also please see the discussion of the pwd command.

echo
The simple echo is probably the most surprising source of portability troubles. It is not possible to use ‘echo’ portably unless both options and escape sequences are omitted. New applications which are not aiming at portability should use ‘printf’ instead of ‘echo’.

Don't expect any option. See Preset Output Variables, ECHO_N etc. for a means to simulate -n.

Do not use backslashes in the arguments, as there is no consensus on their handling. For ‘echo '\n' | wc -l’, the sh of Solaris outputs 2, but Bash and Zsh (in sh emulation mode) output 1. The problem is truly echo: all the shells understand ‘'\n'’ as the string composed of a backslash and an ‘n’.

Because of these problems, do not pass a string containing arbitrary characters to echo. For example, ‘echo "$foo"’ is safe if you know that foo's value cannot contain backslashes and cannot start with ‘-’, but otherwise you should use a here-document like this:

          cat <<EOF
          $foo
          EOF
     

eval
The eval command is useful in limited circumstances, e.g., using commands like ‘eval table_$key=\$value’ and ‘eval value=table_$key’ to simulate a hash table when the key is known to be alphanumeric. However, eval is tricky to use on arbitrary arguments, even when it is implemented correctly.

It is obviously unwise to use ‘eval $cmd’ if the string value of ‘cmd’ was derived from an untrustworthy source. But even if the string value is valid, ‘eval $cmd’ might not work as intended, since it causes field splitting and file name expansion to occur twice, once for the eval and once for the command itself. It is therefore safer to use ‘eval "$cmd"’. For example, if cmd has the value ‘cat test?.c’, ‘eval $cmd’ might expand to the equivalent of ‘cat test;.c’ if there happens to be a file named test;.c in the current directory; and this in turn mistakenly attempts to invoke cat on the file test and then execute the command .c. To avoid this problem, use ‘eval "$cmd"’ rather than ‘eval $cmd’.

However, suppose that you want to output the text of the evaluated command just before executing it. Assuming the previous example, ‘echo "Executing: $cmd"’ outputs ‘Executing: cat test?.c’, but this output doesn't show the user that ‘test;.c’ is the actual name of the copied file. Conversely, ‘eval "echo Executing: $cmd"’ works on this example, but it fails with ‘cmd='cat foo >bar'’, since it mistakenly replaces the contents of bar by the string ‘cat foo’. No simple, general, and portable solution to this problem is known.

You should also be wary of common bugs in eval implementations. In some shell implementations (e.g., older ash, OpenBSD 3.8 sh, pdksh v5.2.14 99/07/13.2, and zsh 4.2.5), the arguments of ‘eval’ are evaluated in a context where ‘$?’ is 0, so they exhibit behavior like this:

          $ false; eval 'echo $?'
          0
     

The correct behavior here is to output a nonzero value, but portable scripts should not rely on this.

You should not rely on LINENO within eval. See Special Shell Variables.

exit
The default value of exit is supposed to be $?; unfortunately, some shells, such as the DJGPP port of Bash 2.04, just perform ‘exit 0’.
          bash-2.04$ foo=`exit 1` || echo fail
          fail
          bash-2.04$ foo=`(exit 1)` || echo fail
          fail
          bash-2.04$ foo=`(exit 1); exit` || echo fail
          bash-2.04$
     

Using ‘exit $?’ restores the expected behavior.

Some shell scripts, such as those generated by autoconf, use a trap to clean up before exiting. If the last shell command exited with nonzero status, the trap also exits with nonzero status so that the invoker can tell that an error occurred.

Unfortunately, in some shells, such as Solaris /bin/sh, an exit trap ignores the exit command's argument. In these shells, a trap cannot determine whether it was invoked by plain exit or by exit 1. Instead of calling exit directly, use the AC_MSG_ERROR macro that has a workaround for this problem.

export
The builtin export dubs a shell variable environment variable. Each update of exported variables corresponds to an update of the environment variables. Conversely, each environment variable received by the shell when it is launched should be imported as a shell variable marked as exported.

Alas, many shells, such as Solaris /bin/sh, irix 6.3, irix 5.2, AIX 4.1.5, and Digital Unix 4.0, forget to export the environment variables they receive. As a result, two variables coexist: the environment variable and the shell variable. The following code demonstrates this failure:

          #!/bin/sh
          echo $FOO
          FOO=bar
          echo $FOO
          exec /bin/sh $0
     

when run with ‘FOO=foo’ in the environment, these shells print alternately ‘foo’ and ‘bar’, although they should print only ‘foo’ and then a sequence of ‘bar’s.

Therefore you should export again each environment variable that you update.

false
Don't expect false to exit with status 1: in native Solaris /bin/false exits with status 255.
for
To loop over positional arguments, use:
          for arg
          do
            echo "$arg"
          done
     

You may not leave the do on the same line as for, since some shells improperly grok:

          for arg; do
            echo "$arg"
          done
     

If you want to explicitly refer to the positional arguments, given the ‘$@’ bug (see Shell Substitutions), use:

          for arg in ${1+"$@"}; do
            echo "$arg"
          done
     

But keep in mind that Zsh, even in Bourne shell emulation mode, performs word splitting on ‘${1+"$@"}’; see Shell Substitutions, item ‘$@’, for more.

if
Using ‘!’ is not portable. Instead of:
          if ! cmp -s file file.new; then
            mv file.new file
          fi
     

use:

          if cmp -s file file.new; then :; else
            mv file.new file
          fi
     

There are shells that do not reset the exit status from an if:

          $ if (exit 42); then true; fi; echo $?
          42
     

whereas a proper shell should have printed ‘0’. This is especially bad in makefiles since it produces false failures. This is why properly written makefiles, such as Automake's, have such hairy constructs:

          if test -f "$file"; then
            install "$file" "$dest"
          else
            :
          fi
     

printf
A format string starting with a ‘-’ can cause problems. Bash (e.g., 2.05b) interprets it as an options argument and gives an error. And ‘--’ to mark the end of options is not good in the NetBSD Almquist shell (e.g., 0.4.6) which takes that literally as the format string. Putting the ‘-’ in a ‘%c’ or ‘%s’ is probably the easiest way to avoid doubt,
          printf %s -foo
     

read
Not all shells support -r (Solaris /bin/sh for example).
pwd
With modern shells, plain pwd outputs a “logical” directory name, some of whose components may be symbolic links. These directory names are in contrast to “physical” directory names, whose components are all directories.

Posix 1003.1-2001 requires that pwd must support the -L (“logical”) and -P (“physical”) options, with -L being the default. However, traditional shells do not support these options, and their pwd command has the -P behavior.

Portable scripts should assume neither option is supported, and should assume neither behavior is the default. Also, on many hosts ‘/bin/pwd’ is equivalent to ‘pwd -P’, but Posix does not require this behavior and portable scripts should not rely on it.

Typically it's best to use plain pwd. On modern hosts this outputs logical directory names, which have the following advantages:

Also please see the discussion of the cd command.

set
With the FreeBSD 6.0 shell, the set command (without any options) does not sort its output.

The set builtin faces the usual problem with arguments starting with a dash. Modern shells such as Bash or Zsh understand -- to specify the end of the options (any argument after -- is a parameter, even ‘-x’ for instance), but many traditional shells (e.g., Solaris 10 /bin/sh) simply stop option processing as soon as a non-option argument is found. Therefore, use ‘dummy’ or simply ‘x’ to end the option processing, and use shift to pop it out:

          set x $my_list; shift
     

Avoid ‘set -’, e.g., ‘set - $my_list’. Posix no longer requires support for this command, and in traditional shells ‘set - $my_list’ resets the -v and -x options, which makes scripts harder to debug.

Some nonstandard shells do not recognize more than one option (e.g., ‘set -e -x’ assigns ‘-x’ to the command line). It is better to combine them:

          set -ex
     

The BSD shell has had several problems with the -e option, partly because BSD make traditionally used -e even though this was incompatible with Posix (see Failure in Make Rules). Older versions of the BSD shell (circa 1990) mishandled ‘&&’, ‘||’, ‘if’, and ‘case’ when -e was in effect, causing the shell to exit unexpectedly in some cases. This was particularly a problem with makefiles, and led to circumlocutions like ‘sh -c 'test -f file || touch file'’, where the seemingly-unnecessary ‘sh -c '...'’ wrapper works around the bug.

Even relatively-recent versions of the BSD shell (e.g., OpenBSD 3.4) wrongly exit with -e if a command within ‘&&’ fails inside a compound statement. For example:

          #! /bin/sh
          set -e
          foo=''
          test -n "$foo" && exit 1
          echo one
          if :; then
            test -n "$foo" && exit 1
          fi
          echo two
     

does not print ‘two’. One workaround is to use ‘if test -n "$foo"; then exit 1; fi’ rather than ‘test -n "$foo" && exit 1’. Another possibility is to warn BSD users not to use ‘sh -e’.

shift
Not only is shifting a bad idea when there is nothing left to shift, but in addition it is not portable: the shell of MIPS RISC/OS 4.52 refuses to do it.

Don't use ‘shift 2’ etc.; it was not in the 7th Edition Bourne shell, and it is also absent in many pre-Posix shells.

source
This command is not portable, as Posix does not require it; use . instead.
test
The test program is the way to perform many file and string tests. It is often invoked by the alternate name ‘[’, but using that name in Autoconf code is asking for trouble since it is an M4 quote character.

If you need to make multiple checks using test, combine them with the shell operators ‘&&’ and ‘||’ instead of using the test operators -a and -o. On System V, the precedence of -a and -o is wrong relative to the unary operators; consequently, Posix does not specify them, so using them is nonportable. If you combine ‘&&’ and ‘||’ in the same statement, keep in mind that they have equal precedence.

It is safe to use ‘!’ as a test operator. For example, ‘if test ! -d foo; ...’ is portable even though ‘if ! test -d foo; ...’ is not.

test (files)
To enable configure scripts to support cross-compilation, they shouldn't do anything that tests features of the build system instead of the host system. But occasionally you may find it necessary to check whether some arbitrary file exists. To do so, use ‘test -f’ or ‘test -r’. Do not use ‘test -x’, because 4.3BSD does not have it. Do not use ‘test -e’ either, because Solaris /bin/sh lacks it. To test for symbolic links on systems that have them, use ‘test -h’ rather than ‘test -L’; either form conforms to Posix 1003.1-2001, but older shells like Solaris 8 /bin/sh support only -h.
test (strings)
Avoid ‘test "string"’, in particular if string might start with a dash, since test might interpret its argument as an option (e.g., ‘string = "-n"’).

Contrary to a common belief, ‘test -n string’ and ‘test -z stringare portable. Nevertheless many shells (such as Solaris, AIX 3.2, unicos 10.0.0.6, Digital Unix 4, etc.) have bizarre precedence and may be confused if string looks like an operator:

          $ test -n =
          test: argument expected
     

If there are risks, use ‘test "xstring" = x’ or ‘test "xstring" != x’ instead.

It is common to find variations of the following idiom:

          test -n "`echo $ac_feature | sed 's/[-a-zA-Z0-9_]//g'`" &&
            action
     

to take an action when a token matches a given pattern. Such constructs should always be avoided by using:

          echo "$ac_feature" | grep '[^-a-zA-Z0-9_]' >/dev/null 2>&1 &&
            action
     

Use case where possible since it is faster, being a shell builtin:

          case $ac_feature in
            *[!-a-zA-Z0-9_]*) action;;
          esac
     

Alas, negated character classes are probably not portable, although no shell is known to not support the Posix syntax ‘[!...]’ (when in interactive mode, zsh is confused by the ‘[!...]’ syntax and looks for an event in its history because of ‘!’). Many shells do not support the alternative syntax ‘[^...]’ (Solaris, Digital Unix, etc.).

One solution can be:

          expr "$ac_feature" : '.*[^-a-zA-Z0-9_]' >/dev/null &&
            action
     

or better yet

          expr "X$ac_feature" : '.*[^-a-zA-Z0-9_]' >/dev/null &&
            action
     

expr "Xfoo" : "Xbar"’ is more robust than ‘echo "Xfoo" | grep "^Xbar"’, because it avoids problems when ‘foo’ contains backslashes.

trap
It is safe to trap at least the signals 1, 2, 13, and 15. You can also trap 0, i.e., have the trap run when the script ends (either via an explicit exit, or the end of the script).

Posix says that ‘trap - 1 2 13 15’ resets the traps for the specified signals to their default values, but many common shells (e.g., Solaris /bin/sh) misinterpret this and attempt to execute a “command” named - when the specified conditions arise. There is no portable workaround, except for ‘trap - 0’, for which ‘trap '' 0’ is a portable substitute.

Although Posix is not absolutely clear on this point, it is widely admitted that when entering the trap ‘$?’ should be set to the exit status of the last command run before the trap. The ambiguity can be summarized as: “when the trap is launched by an exit, what is the last command run: that before exit, or exit itself?”

Bash considers exit to be the last command, while Zsh and Solaris /bin/sh consider that when the trap is run it is still in the exit, hence it is the previous exit status that the trap receives:

          $ cat trap.sh
          trap 'echo $?' 0
          (exit 42); exit 0
          $ zsh trap.sh
          42
          $ bash trap.sh
          0
     

The portable solution is then simple: when you want to ‘exit 42’, run ‘(exit 42); exit 42’, the first exit being used to set the exit status to 42 for Zsh, and the second to trigger the trap and pass 42 as exit status for Bash.

The shell in FreeBSD 4.0 has the following bug: ‘$?’ is reset to 0 by empty lines if the code is inside trap.

          $ trap 'false
          
          echo $?' 0
          $ exit
          0
     

Fortunately, this bug only affects trap.

true
Don't worry: as far as we know true is portable. Nevertheless, it's not always a builtin (e.g., Bash 1.x), and the portable shell community tends to prefer using :. This has a funny side effect: when asked whether false is more portable than true Alexandre Oliva answered:
In a sense, yes, because if it doesn't exist, the shell will produce an exit status of failure, which is correct for false, but not for true.

unset
You cannot assume the support of unset. Nevertheless, because it is extremely useful to disable embarrassing variables such as PS1, you can test for its existence and use it provided you give a neutralizing value when unset is not supported:
          if (unset FOO) >/dev/null 2>&1; then
            unset=unset
          else
            unset=false
          fi
          $unset PS1 || PS1='$ '
     

See Special Shell Variables, for some neutralizing values. Also, see Limitations of Builtins, documentation of export, for the case of environment variables.


Previous: Limitations of Builtins, Up: Portable Shell

10.11 Limitations of Usual Tools

The small set of tools you can expect to find on any machine can still include some limitations you should be aware of.

Awk
Don't leave white space before the opening parenthesis in a user function call. Posix does not allow this and GNU Awk rejects it:
          $ gawk 'function die () { print "Aaaaarg!"  }
                  BEGIN { die () }'
          gawk: cmd. line:2:         BEGIN { die () }
          gawk: cmd. line:2:                      ^ parse error
          $ gawk 'function die () { print "Aaaaarg!"  }
                  BEGIN { die() }'
          Aaaaarg!
     

If you want your program to be deterministic, don't depend on for on arrays:

          $ cat for.awk
          END {
            arr["foo"] = 1
            arr["bar"] = 1
            for (i in arr)
              print i
          }
          $ gawk -f for.awk </dev/null
          foo
          bar
          $ nawk -f for.awk </dev/null
          bar
          foo
     

Some Awk implementations, such as HP-UX 11.0's native one, mishandle anchors:

          $ echo xfoo | $AWK '/foo|^bar/ { print }'
          $ echo bar | $AWK '/foo|^bar/ { print }'
          bar
          $ echo xfoo | $AWK '/^bar|foo/ { print }'
          xfoo
          $ echo bar | $AWK '/^bar|foo/ { print }'
          bar
     

Either do not depend on such patterns (i.e., use ‘/^(.*foo|bar)/’, or use a simple test to reject such implementations.

AIX version 5.2 has an arbitrary limit of 399 on the the length of regular expressions and literal strings in an Awk program.

Traditional Awk implementations derived from Unix version 7, such as Solaris /bin/awk, have many limitations and do not conform to Posix. Nowadays AC_PROG_AWK (see Particular Programs) finds you an Awk that doesn't have these problems, but if for some reason you prefer not to use AC_PROG_AWK you may need to address them.

Traditional Awk does not support multidimensional arrays or user-defined functions.

Traditional Awk does not support the -v option. You can use assignments after the program instead, e.g., $AWK '{print v $1}' v=x; however, don't forget that such assignments are not evaluated until they are encountered (e.g., after any BEGIN action).

Traditional Awk does not support the keywords delete or do.

Traditional Awk does not support the expressions a?b:c, !a, a^b, or a^=b.

Traditional Awk does not support the predefined CONVFMT variable.

Traditional Awk supports only the predefined functions exp, int, length, log, split, sprintf, sqrt, and substr.

Traditional Awk getline is not at all compatible with Posix; avoid it.

Traditional Awk split supports only two arguments.

Traditional Awk has a limit of 99 fields in a record. You may be able to circumvent this problem by using split.

basename
Not all hosts have a working basename. You can use expr instead.
cat
Don't rely on any option.
cc
The command ‘cc -c foo.c’ traditionally produces an object file named foo.o. Most compilers allow -c to be combined with -o to specify a different object file name, but Posix does not require this combination and a few compilers lack support for it. See C Compiler, for how GNU Make tests for this feature with AC_PROG_CC_C_O.

When a compilation such as ‘cc -o foo foo.c’ fails, some compilers (such as cds on Reliant Unix) leave a foo.o.

HP-UX cc doesn't accept .S files to preprocess and assemble. ‘cc -c foo.S’ appears to succeed, but in fact does nothing.

The default executable, produced by ‘cc foo.c’, can be

The C compiler's traditional name is cc, but other names like gcc are common. Posix 1003.1-2001 specifies the name c99, but older Posix editions specified c89 and anyway these standard names are rarely used in practice. Typically the C compiler is invoked from makefiles that use ‘$(CC)’, so the value of the ‘CC’ make variable selects the compiler name.

chmod
Avoid usages like ‘chmod -w file’; use ‘chmod a-w file’ instead, for two reasons. First, plain -w does not necessarily make the file unwritable, since it does not affect mode bits that correspond to bits in the file mode creation mask. Second, Posix says that the -w might be interpreted as an implementation-specific option, not as a mode; Posix suggests using ‘chmod -- -w file’ to avoid this confusion, but unfortunately ‘--’ does not work on some older hosts.
cmp
cmp performs a raw data comparison of two files, while diff compares two text files. Therefore, if you might compare DOS files, even if only checking whether two files are different, use diff to avoid spurious differences due to differences of newline encoding.
cp
Avoid the -r option, since Posix 1003.1-2004 marks it as obsolescent and its behavior on special files is implementation-defined. Use -R instead. On GNU hosts the two options are equivalent, but on Solaris hosts (for example) cp -r reads from pipes instead of replicating them.

Some cp implementations (e.g., BSD/OS 4.2) do not allow trailing slashes at the end of nonexistent destination directories. To avoid this problem, omit the trailing slashes. For example, use ‘cp -R source /tmp/newdir’ rather than ‘cp -R source /tmp/newdir/’ if /tmp/newdir does not exist.

The ancient SunOS 4 cp does not support -f, although its mv does.

Traditionally, file timestamps had 1-second resolution, and ‘cp -p’ copied the timestamps exactly. However, many modern file systems have timestamps with 1-nanosecond resolution. Unfortunately, ‘cp -p’ implementations truncate timestamps when copying files, so this can result in the destination file appearing to be older than the source. The exact amount of truncation depends on the resolution of the system calls that cp uses; traditionally this was utime, which has 1-second resolution, but some newer cp implementations use utimes, which has 1-microsecond resolution. These newer implementations include GNU Core Utilities 5.0.91 or later, and Solaris 8 (sparc) patch 109933-02 or later. Unfortunately as of January 2006 there is still no system call to set timestamps to the full nanosecond resolution.

Bob Proulx notes that ‘cp -p’ always tries to copy ownerships. But whether it actually does copy ownerships or not is a system dependent policy decision implemented by the kernel. If the kernel allows it then it happens. If the kernel does not allow it then it does not happen. It is not something cp itself has control over.

In Unix System V any user can chown files to any other user, and System V also has a non-sticky /tmp. That probably derives from the heritage of System V in a business environment without hostile users. BSD changed this to be a more secure model where only root can chown files and a sticky /tmp is used. That undoubtedly derives from the heritage of BSD in a campus environment.

GNU/Linux and Solaris by default follow BSD, but can be configured to allow a System V style chown. On the other hand, HP-UX follows System V, but can be configured to use the modern security model and disallow chown. Since it is an administrator-configurable parameter you can't use the name of the kernel as an indicator of the behavior.

date
Some versions of date do not recognize special ‘%’ directives, and unfortunately, instead of complaining, they just pass them through, and exit with success:
          $ uname -a
          OSF1 medusa.sis.pasteur.fr V5.1 732 alpha
          $ date "+%s"
          %s
     

diff
Option -u is nonportable.

Some implementations, such as Tru64's, fail when comparing to /dev/null. Use an empty file instead.

dirname
Not all hosts have a working dirname, and you should instead use AS_DIRNAME (see Programming in M4sh). For example:
          dir=`dirname "$file"`       # This is not portable.
          dir=`AS_DIRNAME(["$file"])` # This is more portable.
     

egrep
Posix 1003.1-2001 no longer requires egrep, but many older hosts do not yet support the Posix replacement grep -E. Also, some traditional implementations do not work on long input lines. To work around these problems, invoke AC_PROG_EGREP and then use $EGREP.

Portable extended regular expressions should use ‘\’ only to escape characters in the string ‘$()*+.?[\^{|’. For example, ‘\}’ is not portable, even though it typically matches ‘}’.

The empty alternative is not portable. Use ‘?’ instead. For instance with Digital Unix v5.0:

          > printf "foo\n|foo\n" | $EGREP '^(|foo|bar)$'
          |foo
          > printf "bar\nbar|\n" | $EGREP '^(foo|bar|)$'
          bar|
          > printf "foo\nfoo|\n|bar\nbar\n" | $EGREP '^(foo||bar)$'
          foo
          |bar
     

$EGREP also suffers the limitations of grep.

expr
No expr keyword starts with ‘X’, so use ‘expr X"word" : 'Xregex'’ to keep expr from misinterpreting word.

Don't use length, substr, match and index.

expr (‘|’)
You can use ‘|’. Although Posix does require that ‘expr ''’ return the empty string, it does not specify the result when you ‘|’ together the empty string (or zero) with the empty string. For example:
          expr '' \| ''
     

Posix 1003.2-1992 returns the empty string for this case, but traditional Unix returns ‘0’ (Solaris is one such example). In Posix 1003.1-2001, the specification was changed to match traditional Unix's behavior (which is bizarre, but it's too late to fix this). Please note that the same problem does arise when the empty string results from a computation, as in:

          expr bar : foo \| foo : bar
     

Avoid this portability problem by avoiding the empty string.

expr (‘:’)
Portable expr regular expressions should use ‘\’ to escape only characters in the string ‘$()*.0123456789[\^n{}’. For example, alternation, ‘\|’, is common but Posix does not require its support, so it should be avoided in portable scripts. Similarly, ‘\+’ and ‘\?’ should be avoided.

Portable expr regular expressions should not begin with ‘^’. Patterns are automatically anchored so leading ‘^’ is not needed anyway.

The Posix standard is ambiguous as to whether ‘expr 'a' : '\(b\)'’ outputs ‘0’ or the empty string. In practice, it outputs the empty string on most platforms, but portable scripts should not assume this. For instance, the QNX 4.25 native expr returns ‘0’.

One might think that a way to get a uniform behavior would be to use the empty string as a default value:

          expr a : '\(b\)' \| ''
     

Unfortunately this behaves exactly as the original expression; see the expr (‘|’) entry for more information.

Ancient expr implementations (e.g., SunOS 4 expr and Solaris 8 /usr/ucb/expr) have a silly length limit that causes expr to fail if the matched substring is longer than 120 bytes. In this case, you might want to fall back on ‘echo|sed’ if expr fails. Nowadays this is of practical importance only for the rare installer who mistakenly puts /usr/ucb before /usr/bin in PATH.

On Mac OS X 10.4, expr mishandles the pattern ‘[^-]’ in some cases. For example, the command

          expr Xpowerpc-apple-darwin8.1.0 : 'X[^-]*-[^-]*-\(.*\)'
     

outputs ‘apple-darwin8.1.0’ rather than the correct ‘darwin8.1.0’. This particular case can be worked around by substituting ‘[^--]’ for ‘[^-]’.

Don't leave, there is some more!

The QNX 4.25 expr, in addition of preferring ‘0’ to the empty string, has a funny behavior in its exit status: it's always 1 when parentheses are used!

          $ val=`expr 'a' : 'a'`; echo "$?: $val"
          0: 1
          $ val=`expr 'a' : 'b'`; echo "$?: $val"
          1: 0
          
          $ val=`expr 'a' : '\(a\)'`; echo "?: $val"
          1: a
          $ val=`expr 'a' : '\(b\)'`; echo "?: $val"
          1: 0
     

In practice this can be a big problem if you are ready to catch failures of expr programs with some other method (such as using sed), since you may get twice the result. For instance

          $ expr 'a' : '\(a\)' || echo 'a' | sed 's/^\(a\)$/\1/'
     

outputs ‘a’ on most hosts, but ‘aa’ on QNX 4.25. A simple workaround consists of testing expr and using a variable set to expr or to false according to the result.

Tru64 expr incorrectly treats the result as a number, if it can be interpreted that way:

          $ expr 00001 : '.*\(...\)'
          1
     

fgrep
Posix 1003.1-2001 no longer requires fgrep, but many older hosts do not yet support the Posix replacement grep -F. Also, some traditional implementations do not work on long input lines. To work around these problems, invoke AC_PROG_FGREP and then use $FGREP.
find
The option -maxdepth seems to be GNU specific. Tru64 v5.1, NetBSD 1.5 and Solaris find commands do not understand it.

The replacement of ‘{}’ is guaranteed only if the argument is exactly {}, not if it's only a part of an argument. For instance on DU, and HP-UX 10.20 and HP-UX 11:

          $ touch foo
          $ find . -name foo -exec echo "{}-{}" \;
          {}-{}
     

while GNU find reports ‘./foo-./foo’.

grep
Portable scripts can rely on the grep options -c, -l, -n, and -v, but should avoid other options. For example, don't use -w, as Posix does not require it and Irix 6.5.16m's grep does not support it. Also, portable scripts should not combine -c with -l, as Posix does not allow this.

Some of the options required by Posix are not portable in practice. Don't use ‘grep -q’ to suppress output, because many grep implementations (e.g., Solaris) do not support -q. Don't use ‘grep -s’ to suppress output either, because Posix says -s does not suppress output, only some error messages; also, the -s option of traditional grep behaved like -q does in most modern implementations. Instead, redirect the standard output and standard error (in case the file doesn't exist) of grep to /dev/null. Check the exit status of grep to determine whether it found a match.

Some traditional grep implementations do not work on long input lines. On AIX the default grep silently truncates long lines on the input before matching.

Also, many implementations do not support multiple regexps with -e: they either reject -e entirely (e.g., Solaris) or honor only the last pattern (e.g., IRIX 6.5 and NeXT). To work around these problems, invoke AC_PROG_GREP and then use $GREP.

Another possible workaround for the multiple -e problem is to separate the patterns by newlines, for example:

          grep 'foo
          bar' in.txt
     

except that this fails with traditional grep implementations and with OpenBSD 3.8 grep.

Traditional grep implementations (e.g., Solaris) do not support the -E or -F options. To work around these problems, invoke AC_PROG_EGREP and then use $EGREP, and similarly for AC_PROG_FGREP and $FGREP. Even if you are willing to require support for Posix grep, your script should not use both -E and -F, since Posix does not allow this combination.

Portable grep regular expressions should use ‘\’ only to escape characters in the string ‘$()*.0123456789[\^{}’. For example, alternation, ‘\|’, is common but Posix does not require its support in basic regular expressions, so it should be avoided in portable scripts. Solaris grep does not support it. Similarly, ‘\+’ and ‘\?’ should be avoided.

join
Solaris 8 join has bugs when the second operand is standard input, and when standard input is a pipe. For example, the following shell script causes Solaris 8 join to loop forever:
          cat >file <<'EOF'
          1 x
          2 y
          EOF
          cat file | join file -
     

Use ‘join - file’ instead.

ln
Don't rely on ln having a -f option. Symbolic links are not available on old systems; use ‘$(LN_S)’ as a portable substitute.

For versions of the DJGPP before 2.04, ln emulates symbolic links to executables by generating a stub that in turn calls the real program. This feature also works with nonexistent files like in the Posix spec. So ‘ln -s file link’ generates link.exe, which attempts to call file.exe if run. But this feature only works for executables, so ‘cp -p’ is used instead for these systems. DJGPP versions 2.04 and later have full support for symbolic links.

ls
The portable options are -acdilrtu. Current practice is for -l to output both owner and group, even though ancient versions of ls omitted the group.

On ancient hosts, ‘ls foo’ sent the diagnostic ‘foo not found’ to standard output if foo did not exist. Hence a shell command like ‘sources=`ls *.c 2>/dev/null`’ did not always work, since it was equivalent to ‘sources='*.c not found'’ in the absence of ‘.c’ files. This is no longer a practical problem, since current ls implementations send diagnostics to standard error.

mkdir
No mkdir option is portable to older systems. Instead of ‘mkdir -p file-name’, you should use use AS_MKDIR_P(file-name) (see Programming in M4sh) or AC_PROG_MKDIR_P (see Particular Programs).

Posix does not clearly specify whether ‘mkdir -p foo’ should succeed when foo is a symbolic link to an already-existing directory. The GNU Core Utilities 5.1.0 mkdir succeeds, but Solaris mkdir fails.

Traditional mkdir -p implementations suffer from race conditions. For example, if you invoke mkdir -p a/b and mkdir -p a/c at the same time, both processes might detect that a is missing, one might create a, then the other might try to create a and fail with a File exists diagnostic. The GNU Core Utilities (‘fileutils’ version 4.1), FreeBSD 5.0, NetBSD 2.0.2, and OpenBSD 2.4 are known to be race-free when two processes invoke mkdir -p simultaneously, but earlier versions are vulnerable. Solaris mkdir is still vulnerable as of Solaris 10, and other traditional Unix systems are probably vulnerable too. This possible race is harmful in parallel builds when several Make rules call mkdir -p to construct directories. You may use install-sh -d as a safe replacement, provided this script is recent enough; the copy shipped with Autoconf 2.60 and Automake 1.10 is OK, but copies from older versions are vulnerable.

mktemp
Shell scripts can use temporary files safely with mktemp, but it does not exist on all systems. A portable way to create a safe temporary file name is to create a temporary directory with mode 700 and use a file inside this directory. Both methods prevent attackers from gaining control, though mktemp is far less likely to fail gratuitously under attack.

Here is sample code to create a new temporary directory safely:

          # Create a temporary directory $tmp in $TMPDIR (default /tmp).
          # Use mktemp if possible; otherwise fall back on mkdir,
          # with $RANDOM to make collisions less likely.
          : ${TMPDIR=/tmp}
          {
            tmp=`
              (umask 077 && mktemp -d "$TMPDIR/fooXXXXXX") 2>/dev/null
            ` &&
            test -n "$tmp" && test -d "$tmp"
          } || {
            tmp=$TMPDIR/foo$$-$RANDOM
            (umask 077 && mkdir "$tmp")
          } || exit $?
     

mv
The only portable options are -f and -i.

Moving individual files between file systems is portable (it was in Unix version 6), but it is not always atomic: when doing ‘mv new existing’, there's a critical section where neither the old nor the new version of existing actually exists.

On some systems moving files from /tmp can sometimes cause undesirable (but perfectly valid) warnings, even if you created these files. This is because /tmp belongs to a group that ordinary users are not members of, and files created in /tmp inherit the group of /tmp. When the file is copied, mv issues a diagnostic without failing:

          $ touch /tmp/foo
          $ mv /tmp/foo .
          error-->mv: ./foo: set owner/group (was: 100/0): Operation not permitted
          $ echo $?
          0
          $ ls foo
          foo
     

This annoying behavior conforms to Posix, unfortunately.

Moving directories across mount points is not portable, use cp and rm.

Moving/Deleting open files isn't portable. The following can't be done on DOS variants:

          exec > foo
          mv foo bar
     

nor can

          exec > foo
          rm -f foo
     

od
In Mac OS X 10.3, od does not support the standard Posix options -A, -j, -N, or -t, or the XSI option -s. The only supported Posix option is -v, and the only supported XSI options are those in -bcdox. The BSD hexdump program can be used instead.

This problem no longer exists in Mac OS X 10.4.3.

sed
Patterns should not include the separator (unless escaped), even as part of a character class. In conformance with Posix, the Cray sed rejects ‘s/[^/]*$//’: use ‘s,[^/]*$,,’.

Avoid empty patterns within parentheses (i.e., ‘\(\)’). Posix does not require support for empty patterns, and Unicos 9 sed rejects them.

Unicos 9 sed loops endlessly on patterns like ‘.*\n.*’.

Sed scripts should not use branch labels longer than 8 characters and should not contain comments. HP-UX sed has a limit of 99 commands (not counting ‘:’ commands) and 48 labels, which can not be circumvented by using more than one script file. It can execute up to 19 reads with the ‘r’ command per cycle. Solaris /usr/ucb/sed rejects usages that exceed an limit of about 6000 bytes for the internal representation of commands.

Avoid redundant ‘;’, as some sed implementations, such as NetBSD 1.4.2's, incorrectly try to interpret the second ‘;’ as a command:

          $ echo a | sed 's/x/x/;;s/x/x/'
          sed: 1: "s/x/x/;;s/x/x/": invalid command code ;
     

Input should not have unreasonably long lines, since some sed implementations have an input buffer limited to 4000 bytes.

Portable sed regular expressions should use ‘\’ only to escape characters in the string ‘$()*.0123456789[\^n{}’. For example, alternation, ‘\|’, is common but Posix does not require its support, so it should be avoided in portable scripts. Solaris sed does not support alternation; e.g., ‘sed '/a\|b/d'’ deletes only lines that contain the literal string ‘a|b’. Similarly, ‘\+’ and ‘\?’ should be avoided.

Anchors (‘^’ and ‘$’) inside groups are not portable.

Nested parenthesization in patterns (e.g., ‘\(\(a*\)b*)\)’) is quite portable to current hosts, but was not supported by some ancient sed implementations like SVR3.

Some sed implementations, e.g., Solaris, restrict the special role of the asterisk to one-character regular expressions. This may lead to unexpected behavior:

          $ echo '1*23*4' | /usr/bin/sed 's/\(.\)*/x/g'
          x2x4
          $ echo '1*23*4' | /usr/xpg4/bin/sed 's/\(.\)*/x/g'
          x
     

The -e option is portable. Some people prefer to use it:

          sed -e 'command-1' \
              -e 'command-2'
     

as opposed to the equivalent:

          sed '
            command-1
            command-2
          '
     

The following usage is sometimes equivalent:

          sed 'command-1;command-2'
     

but Posix says that this use of a semicolon has undefined effect if command-1's verb is ‘{’, ‘a’, ‘b’, ‘c’, ‘i’, ‘r’, ‘t’, ‘w’, ‘:’, or ‘#’, so you should use semicolon only with simple scripts that do not use these verbs.

Commands inside { } brackets are further restricted. Posix says that they cannot be preceded by addresses, ‘!’, or ‘;’, and that each command must be followed immediately by a newline, without any intervening blanks or semicolons. The closing bracket must be alone on a line, other than white space preceding or following it.

Contrary to yet another urban legend, you may portably use ‘&’ in the replacement part of the s command to mean “what was matched”. All descendants of Unix version 7 sed (at least; we don't have first hand experience with older sed implementations) have supported it.

Posix requires that you must not have any white space between ‘!’ and the following command. It is OK to have blanks between the address and the ‘!’. For instance, on Solaris:

          $ echo "foo" | sed -n '/bar/ ! p'
          error-->Unrecognized command: /bar/ ! p
          $ echo "foo" | sed -n '/bar/! p'
          error-->Unrecognized command: /bar/! p
          $ echo "foo" | sed -n '/bar/ !p'
          foo
     

Posix also says that you should not combine ‘!’ and ‘;’. If you use ‘!’, it is best to put it on a command that is delimited by newlines rather than ‘;’.

Also note that Posix requires that the ‘b’, ‘t’, ‘r’, and ‘w’ commands be followed by exactly one space before their argument. On the other hand, no white space is allowed between ‘:’ and the subsequent label name.

sed (‘t’)
Some old systems have sed that “forget” to reset their ‘t’ flag when starting a new cycle. For instance on MIPS RISC/OS, and on irix 5.3, if you run the following sed script (the line numbers are not actual part of the texts):
          s/keep me/kept/g  # a
          t end             # b
          s/.*/deleted/g    # c
          :end              # d
     

on

          delete me         # 1
          delete me         # 2
          keep me           # 3
          delete me         # 4
     

you get

          deleted
          delete me
          kept
          deleted
     

instead of

          deleted
          deleted
          kept
          deleted
     

Why? When processing line 1, (c) matches, therefore sets the ‘t’ flag, and the output is produced. When processing line 2, the ‘t’ flag is still set (this is the bug). Command (a) fails to match, but sed is not supposed to clear the ‘t’ flag when a substitution fails. Command (b) sees that the flag is set, therefore it clears it, and jumps to (d), hence you get ‘delete me’ instead of ‘deleted’. When processing line (3), ‘t’ is clear, (a) matches, so the flag is set, hence (b) clears the flags and jumps. Finally, since the flag is clear, line 4 is processed properly.

There are two things one should remember about ‘t’ in sed. Firstly, always remember that ‘t’ jumps if some substitution succeeded, not only the immediately preceding substitution. Therefore, always use a fake ‘t clear’ followed by a ‘:clear’ on the next line, to reset the ‘t’ flag where needed.

Secondly, you cannot rely on sed to clear the flag at each new cycle.

One portable implementation of the script above is:

          t clear
          :clear
          s/keep me/kept/g
          t end
          s/.*/deleted/g
          :end
     

touch
If you specify the desired timestamp (e.g., with the -r option), touch typically uses the utime or utimes system call, which can result in the same kind of timestamp truncation problems that ‘cp -p’ has.

On ancient BSD systems, touch or any command that results in an empty file does not update the timestamps, so use a command like echo as a workaround. Also, GNU touch 3.16r (and presumably all before that) fails to work on SunOS 4.1.3 when the empty file is on an NFS-mounted 4.2 volume. However, these problems are no longer of practical concern.


Next: , Previous: Portable Shell, Up: Top

11 Portable Make Programming

Writing portable makefiles is an art. Since a makefile's commands are executed by the shell, you must consider the shell portability issues already mentioned. However, other issues are specific to make itself.


Next: , Up: Portable Make

11.1 $< in Ordinary Make Rules

Posix says that the ‘$<’ construct in makefiles can be used only in inference rules and in the ‘.DEFAULT’ rule; its meaning in ordinary rules is unspecified. Solaris make for instance replaces it with the empty string. OpenBSD (3.0 and later) make diagnoses these uses and errors out.


Next: , Previous: $< in Ordinary Make Rules, Up: Portable Make

11.2 Failure in Make Rules

Since 1992 Posix has required that make must invoke each command with the equivalent of a ‘sh -c’ subshell. However, many make implementations, including BSD make through 2004, use ‘sh -e -c’ instead, and the -e option causes the subshell to exit immediately if a subsidiary simple-command fails. For example, the command ‘touch T; rm -f U’ always attempts to remove U with Posix make, but incompatible make implementations skip the rm if the touch fails. One way to work around this is to reword the affected simple-commands so that they always succeed, e.g., ‘touch T || :; rm -f U’. However, even this approach can run into common bugs in BSD implementations of the -e option of sh and set (see Limitations of Builtins), so if you are worried about porting to buggy BSD shells it may be simpler to migrate complicated make actions into separate scripts.


Next: , Previous: Failure in Make Rules, Up: Portable Make

11.3 Special Characters in Make Macro Names

Posix limits macro names to nonempty strings containing only ASCII letters and digits, ‘.’, and ‘_’. Many make implementations allow a wider variety of characters, but portable makefiles should avoid them. It is portable to start a name with a special character, e.g., ‘$(.FOO)’.

Some ancient make implementations don't support leading underscores in macro names. An example is NEWS-OS 4.2R.

     $ cat Makefile
     _am_include = #
     _am_quote =
     all:; @echo this is test
     $ make
     Make: Must be a separator on rules line 2.  Stop.
     $ cat Makefile2
     am_include = #
     am_quote =
     all:; @echo this is test
     $ make -f Makefile2
     this is test

However, this problem is no longer of practical concern.


Next: , Previous: Special Chars in Names, Up: Portable Make

11.4 Backslash-Newline-Newline in Make Macro Values

On some versions of HP-UX, make reads multiple newlines following a backslash, continuing to the next non-empty line. For example,

     FOO = one \
     
     BAR = two
     
     test:
             : FOO is "$(FOO)"
             : BAR is "$(BAR)"

shows FOO equal to one BAR = two. Other implementations sensibly let a backslash continue only to the immediately following line.


Next: , Previous: Backslash-Newline-Newline, Up: Portable Make

11.5 Backslash-Newline in Make Comments

According to Posix, Make comments start with # and continue until an unescaped newline is reached.

     $ cat Makefile
     # A = foo \
           bar \
           baz
     
     all:
             @echo ok
     $ make   # GNU make
     ok

However this is not always the case. Some implementations discard everything from # through the end of the line, ignoring any trailing backslash.

     $ pmake  # BSD make
     "Makefile", line 3: Need an operator
     Fatal errors encountered -- cannot continue

Therefore, if you want to comment out a multi-line definition, prefix each line with #, not only the first.

     # A = foo \
     #     bar \
     #     baz


Next: , Previous: Backslash-Newline Comments, Up: Portable Make

11.6 Long Lines in Makefiles

OSF/1 4.0d's make cannot process makefiles with lines longer than 38912 bytes. It exits with a Line too long diagnostic. A later version, Tru64 5.1's make has been reported to crash with lines around 20 kB.


Next: , Previous: Long Lines in Makefiles, Up: Portable Make

11.7 make macro=value and Submakes

A command-line variable definition such as foo=bar overrides any definition of foo in a makefile. Some make implementations (such as GNU make) propagate this override to subsidiary invocations of make. Some other implementations do not pass the substitution along to submakes.

     $ cat Makefile
     foo = foo
     one:
             @echo $(foo)
             $(MAKE) two
     two:
             @echo $(foo)
     $ make foo=bar            # GNU make 3.79.1
     bar
     make two
     make[1]: Entering directory `/home/adl'
     bar
     make[1]: Leaving directory `/home/adl'
     $ pmake foo=bar           # BSD make
     bar
     pmake two
     foo

You have a few possibilities if you do want the foo=bar override to propagate to submakes. One is to use the -e option, which causes all environment variables to have precedence over the makefile macro definitions, and declare foo as an environment variable:

     $ env foo=bar make -e

The -e option is propagated to submakes automatically, and since the environment is inherited between make invocations, the foo macro is overridden in submakes as expected.

This syntax (foo=bar make -e) is portable only when used outside of a makefile, for instance from a script or from the command line. When run inside a make rule, GNU make 3.80 and prior versions forget to propagate the -e option to submakes.

Moreover, using -e could have unexpected side effects if your environment contains some other macros usually defined by the makefile. (See also the note about make -e and SHELL below.)

Another way to propagate overrides to submakes is to do it manually, from your makefile:

     foo = foo
     one:
             @echo $(foo)
             $(MAKE) foo=$(foo) two
     two:
             @echo $(foo)

You need to foresee all macros that a user might want to override if you do that.


Next: , Previous: Macros and Submakes, Up: Portable Make

11.8 The Make Macro MAKEFLAGS

Posix requires make to use MAKEFLAGS to affect the current and recursive invocations of make, but allows implementations several formats for the variable. It is tricky to parse $MAKEFLAGS to determine whether -s for silent execution or -k for continued execution are in effect. For example, you cannot assume that the first space-separated word in $MAKEFLAGS contains single-letter options, since in the Cygwin version of GNU make it is either --unix or --win32 with the second word containing single-letter options.

     $ cat Makefile
     all:
             @echo MAKEFLAGS = $(MAKEFLAGS)
     $ make
     MAKEFLAGS = --unix
     $ make -k
     MAKEFLAGS = --unix -k


Next: , Previous: The Make Macro MAKEFLAGS, Up: Portable Make

11.9 The Make Macro SHELL

Posix-compliant make internally uses the $(SHELL) macro to spawn shell processes and execute Make rules. This is a builtin macro supplied by make, but it can be modified by a makefile or by a command-line argument.

Not all make implementations define this SHELL macro. OSF/Tru64 make is an example; this implementation always uses /bin/sh. So it's a good idea to always define SHELL in your makefiles. If you use Autoconf, do

     SHELL = @SHELL@

Do not force SHELL = /bin/sh because that is not correct everywhere. For instance DJGPP lacks /bin/sh, and when its GNU make port sees such a setting it enters a special emulation mode where features like pipes and redirections are emulated on top of DOS's command.com. Unfortunately this emulation is incomplete; for instance it does not handle command substitutions. On DJGPP SHELL should point to Bash.

Posix-compliant make should never acquire the value of $(SHELL) from the environment, even when make -e is used (otherwise, think about what would happen to your rules if SHELL=/bin/tcsh).

However not all make implementations have this exception. For instance it's not surprising that OSF/Tru64 make doesn't protect SHELL, since it doesn't use it.

     $ cat Makefile
     SHELL = /bin/sh
     FOO = foo
     all:
             @echo $(SHELL)
             @echo $(FOO)
     $ env SHELL=/bin/tcsh FOO=bar make -e   # OSF1 V4.0 Make
     /bin/tcsh
     bar
     $ env SHELL=/bin/tcsh FOO=bar gmake -e  # GNU make
     /bin/sh
     bar


Next: , Previous: The Make Macro SHELL, Up: Portable Make

11.10 Comments in Make Rules

Never put comments in a rule.

Some make treat anything starting with a tab as a command for the current rule, even if the tab is immediately followed by a #. The make from Tru64 Unix V5.1 is one of them. The following makefile runs # foo through the shell.

     all:
             # foo


Next: , Previous: Comments in Make Rules, Up: Portable Make

11.11 The obj/ Subdirectory and Make

Never name one of your subdirectories obj/ if you don't like surprises.

If an obj/ directory exists, BSD make enters it before reading the makefile. Hence the makefile in the current directory is not read.

     $ cat Makefile
     all:
             echo Hello
     $ cat obj/Makefile
     all:
             echo World
     $ make      # GNU make
     echo Hello
     Hello
     $ pmake     # BSD make
     echo World
     World


Next: , Previous: obj/ and Make, Up: Portable Make

11.12 Exit Status of make -k

Do not rely on the exit status of make -k. Some implementations reflect whether they encountered an error in their exit status; other implementations always succeed.

     $ cat Makefile
     all:
             false
     $ make -k; echo exit status: $?    # GNU make
     false
     make: *** [all] Error 1
     exit status: 2
     $ pmake -k; echo exit status: $?   # BSD make
     false
     *** Error code 1 (continuing)
     exit status: 0


Next: , Previous: make -k Status, Up: Portable Make

11.13 VPATH and Make

Posix does not specify the semantics of VPATH. Typically, make supports VPATH, but its implementation is not consistent.

Autoconf and Automake support makefiles whose usages of VPATH are portable to recent-enough popular implementations of make, but to keep the resulting makefiles portable, a package's makefile prototypes must take the following issues into account. These issues are complicated and are often poorly understood, and installers who use VPATH should expect to find many bugs in this area. If you use VPATH, the simplest way to avoid these portability bugs is to stick with GNU make, since it is the most commonly-used make among Autoconf users.

Here are some known issues with some VPATH implementations.


Next: , Up: VPATH and Make

11.13.1 VPATH and Double-colon Rules

With ancient versions of Sun make, any assignment to VPATH causes make to execute only the first set of double-colon rules. However, this problem is no longer of practical concern.


Next: , Previous: VPATH and Double-colon, Up: VPATH and Make

11.13.2 $< Not Supported in Explicit Rules

Using $< in explicit rules is not portable. The prerequisite file must be named explicitly in the rule. If you want to find the prerequisite via a VPATH search, you have to code the whole thing manually. See Build Directories.


Next: , Previous: $< in Explicit Rules, Up: VPATH and Make

11.13.3 Automatic Rule Rewriting

Some make implementations, such as Solaris make and OSF1/Tru64 make, search for prerequisites in VPATH and then rewrite each occurrence as a plain word in the rule. For instance:

     # This isn't portable to GNU make.
     VPATH = ../pkg/src
     f.c: if.c
             cp if.c f.c

executes cp ../pkg/src/if.c f.c if if.c is found in ../pkg/src.

However, this rule leads to real problems in practice. For example, if the source directory contains an ordinary file named test that is used in a dependency, Solaris make rewrites commands like ‘if test -r foo; ...’ to ‘if ../pkg/src/test -r foo; ...’, which is typically undesirable. To avoid this problem, portable makefiles should never mention a source file whose name is that of a shell keyword like until or a shell command like cat or gcc or test.

Because of these problems GNU make and many other make implementations do not rewrite commands, so portable makefiles should search VPATH manually. It is tempting to write this:

     # This isn't portable to Solaris make.
     VPATH = ../pkg/src
     f.c: if.c
             cp `test -f if.c || echo $(VPATH)/`if.c f.c

However, the “prerequisite rewriting” still applies here. So if if.c is in ../pkg/src, Solaris make and OSF1/Tru64 make executes

     cp `test -f ../pkg/src/if.c || echo ../pkg/src/`if.c f.c

which reduces to

     cp if.c f.c

and thus fails. Oops.

A simple workaround, and good practice anyway, is to use ‘$?’ and ‘$@’ when possible:

     VPATH = ../pkg/src
     f.c: if.c
             cp $? $@

but this does not generalize well to commands with multiple prerequisites. A more general workaround is to rewrite the rule so that the prerequisite if.c never appears as a plain word. For example, these three rules would be safe, assuming if.c is in ../pkg/src and the other files are in the working directory:

     VPATH = ../pkg/src
     f.c: if.c f1.c
             cat `test -f ./if.c || echo $(VPATH)/`if.c f1.c >$@
     g.c: if.c g1.c
             cat `test -f 'if.c' || echo $(VPATH)/`if.c g1.c >$@
     h.c: if.c h1.c
             cat `test -f "if.c" || echo $(VPATH)/`if.c h1.c >$@

Things get worse when your prerequisites are in a macro.

     VPATH = ../pkg/src
     HEADERS = f.h g.h h.h
     install-HEADERS: $(HEADERS)
             for i in $(HEADERS); do \
               $(INSTALL) -m 644 \
                 `test -f $$i || echo $(VPATH)/`$$i \
                 $(DESTDIR)$(includedir)/$$i; \
             done

The above install-HEADERS rule is not Solaris-proof because for i in $(HEADERS); is expanded to for i in f.h g.h h.h; where f.h and g.h are plain words and are hence subject to VPATH adjustments.

If the three files are in ../pkg/src, the rule is run as:

     for i in ../pkg/src/f.h ../pkg/src/g.h h.h; do \
       install -m 644 \
          `test -f $i || echo ../pkg/src/`$i \
          /usr/local/include/$i; \
     done

where the two first install calls fail. For instance, consider the f.h installation:

     install -m 644 \
       `test -f ../pkg/src/f.h || \
         echo ../pkg/src/ \
       `../pkg/src/f.h \
       /usr/local/include/../pkg/src/f.h;

It reduces to:

     install -m 644 \
       ../pkg/src/f.h \
       /usr/local/include/../pkg/src/f.h;

Note that the manual VPATH search did not cause any problems here; however this command installs f.h in an incorrect directory.

Trying to quote $(HEADERS) in some way, as we did for foo.c a few makefiles ago, does not help:

     install-HEADERS: $(HEADERS)
             headers='$(HEADERS)'; \
             for i in $$headers; do \
               $(INSTALL) -m 644 \
                 `test -f $$i || echo $(VPATH)/`$$i \
                 $(DESTDIR)$(includedir)/$$i; \
             done

Now, headers='$(HEADERS)' macroexpands to:

     headers='f.h g.h h.h'

but g.h is still a plain word. (As an aside, the idiom headers='$(HEADERS)'; for i in $$headers; is a good idea if $(HEADERS) can be empty, because some shells diagnose a syntax error on for i in;.)

One workaround is to strip this unwanted ../pkg/src/ prefix manually:

     VPATH = ../pkg/src
     HEADERS = f.h g.h h.h
     install-HEADERS: $(HEADERS)
             headers='$(HEADERS)'; \
             for i in $$headers; do \
               i=`expr "$$i" : '$(VPATH)/\(.*\)'`;
               $(INSTALL) -m 644 \
                 `test -f $$i || echo $(VPATH)/`$$i \
                 $(DESTDIR)$(includedir)/$$i; \
             done

Automake does something similar. However the above hack works only if the files listed in HEADERS are in the current directory or a subdirectory; they should not be in an enclosing directory. If we had HEADERS = ../f.h, the above fragment would fail in a VPATH build with OSF1/Tru64 make. The reason is that not only does OSF1/Tru64 make rewrite dependencies, but it also simplifies them. Hence ../f.h becomes ../pkg/f.h instead of ../pkg/src/../f.h. This obviously defeats any attempt to strip a leading ../pkg/src/ component.

The following example makes the behavior of OSF1/Tru64 make more apparent.

     $ cat Makefile
     VPATH = sub
     all: ../foo
             echo ../foo
     $ ls
     Makefile foo
     $ make
     echo foo
     foo

Dependency ../foo was found in sub/../foo, but OSF1/Tru64 make simplified it as foo. (Note that the sub/ directory does not even exist, this just means that the simplification occurred before the file was checked for.)

For the record here is how SunOS 4 make behaves on this example.

     $ make
     make: Fatal error: Don't know how to make target `../foo'
     $ mkdir sub
     $ make
     echo sub/../foo
     sub/../foo


Next: , Previous: Automatic Rule Rewriting, Up: VPATH and Make

11.13.4 OSF/Tru64 make Creates Prerequisite Directories Magically

When a prerequisite is a subdirectory of VPATH, Tru64 make creates it in the current directory.

     $ mkdir -p foo/bar build
     $ cd build
     $ cat >Makefile <<END
     VPATH = ..
     all: foo/bar
     END
     $ make
     mkdir foo
     mkdir foo/bar

This can yield unexpected results if a rule uses a manual VPATH search as presented before.

     VPATH = ..
     all : foo/bar
             command `test -d foo/bar || echo ../`foo/bar

The above command is run on the empty foo/bar directory that was created in the current directory.


Previous: OSF/Tru64 Directory Magic, Up: VPATH and Make

11.13.5 Make Target Lookup

GNU make uses a complex algorithm to decide when it should use files found via a VPATH search. See How Directory Searches are Performed.

If a target needs to be rebuilt, GNU make discards the file name found during the VPATH search for this target, and builds the file locally using the file name given in the makefile. If a target does not need to be rebuilt, GNU make uses the file name found during the VPATH search.

Other make implementations, like NetBSD make, are easier to describe: the file name found during the VPATH search is used whether the target needs to be rebuilt or not. Therefore new files are created locally, but existing files are updated at their VPATH location.

OpenBSD and FreeBSD make, however, never perform a VPATH search for a dependency that has an explicit rule. This is extremely annoying.

When attempting a VPATH build for an autoconfiscated package (e.g., mkdir build && cd build && ../configure), this means GNU make builds everything locally in the build directory, while BSD make builds new files locally and updates existing files in the source directory.

     $ cat Makefile
     VPATH = ..
     all: foo.x bar.x
     foo.x bar.x: newer.x
             @echo Building $@
     $ touch ../bar.x
     $ touch ../newer.x
     $ make        # GNU make
     Building foo.x
     Building bar.x
     $ pmake       # NetBSD make
     Building foo.x
     Building ../bar.x
     $ fmake       # FreeBSD make, OpenBSD make
     Building foo.x
     Building bar.x
     $ tmake       # Tru64 make
     Building foo.x
     Building bar.x
     $ touch ../bar.x
     $ make        # GNU make
     Building foo.x
     $ pmake       # NetBSD make
     Building foo.x
     $ fmake       # FreeBSD make, OpenBSD make
     Building foo.x
     Building bar.x
     $ tmake       # Tru64 make
     Building foo.x
     Building bar.x

Note how NetBSD make updates ../bar.x in its VPATH location, and how FreeBSD, OpenBSD, and Tru64 make always update bar.x, even when ../bar.x is up to date.

Another point worth mentioning is that once GNU make has decided to ignore a VPATH file name (e.g., it ignored ../bar.x in the above example) it continues to ignore it when the target occurs as a prerequisite of another rule.

The following example shows that GNU make does not look up bar.x in VPATH before performing the .x.y rule, because it ignored the VPATH result of bar.x while running the bar.x: newer.x rule.

     $ cat Makefile
     VPATH = ..
     all: bar.y
     bar.x: newer.x
             @echo Building $@
     .SUFFIXES: .x .y
     .x.y:
             cp $< $@
     $ touch ../bar.x
     $ touch ../newer.x
     $ make        # GNU make
     Building bar.x
     cp bar.x bar.y
     cp: cannot stat `bar.x': No such file or directory
     make: *** [bar.y] Error 1
     $ pmake       # NetBSD make
     Building ../bar.x
     cp ../bar.x bar.y
     $ rm bar.y
     $ fmake       # FreeBSD make, OpenBSD make
     echo Building bar.x
     cp bar.x bar.y
     cp: cannot stat `bar.x': No such file or directory
     *** Error code 1
     $ tmake       # Tru64 make
     Building bar.x
     cp: bar.x: No such file or directory
     *** Exit 1

Note that if you drop away the command from the bar.x: newer.x rule, GNU make magically starts to work: it knows that bar.x hasn't been updated, therefore it doesn't discard the result from VPATH (../bar.x) in succeeding uses. Tru64 also works, but FreeBSD and OpenBSD still don't.

     $ cat Makefile
     VPATH = ..
     all: bar.y
     bar.x: newer.x
     .SUFFIXES: .x .y
     .x.y:
             cp $< $@
     $ touch ../bar.x
     $ touch ../newer.x
     $ make        # GNU make
     cp ../bar.x bar.y
     $ rm bar.y
     $ pmake       # NetBSD make
     cp ../bar.x bar.y
     $ rm bar.y
     $ fmake       # FreeBSD make, OpenBSD make
     cp bar.x bar.y
     cp: cannot stat `bar.x': No such file or directory
     *** Error code 1
     $ tmake       # Tru64 make
     cp ../bar.x bar.y

It seems the sole solution that would please every make implementation is to never rely on VPATH searches for targets. In other words, VPATH should be reserved to unbuilt sources.


Next: , Previous: VPATH and Make, Up: Portable Make

11.14 Single Suffix Rules and Separated Dependencies

A Single Suffix Rule is basically a usual suffix (inference) rule (‘.from.to:’), but which destination suffix is empty (‘.from:’).

Separated dependencies simply refers to listing the prerequisite of a target, without defining a rule. Usually one can list on the one hand side, the rules, and on the other hand side, the dependencies.

Solaris make does not support separated dependencies for targets defined by single suffix rules:

     $ cat Makefile
     .SUFFIXES: .in
     foo: foo.in
     .in:
             cp $< $@
     $ touch foo.in
     $ make
     $ ls
     Makefile  foo.in

while GNU Make does:

     $ gmake
     cp foo.in foo
     $ ls
     Makefile  foo       foo.in

Note it works without the ‘foo: foo.in’ dependency.

     $ cat Makefile
     .SUFFIXES: .in
     .in:
             cp $< $@
     $ make foo
     cp foo.in foo

and it works with double suffix inference rules:

     $ cat Makefile
     foo.out: foo.in
     .SUFFIXES: .in .out
     .in.out:
             cp $< $@
     $ make
     cp foo.in foo.out

As a result, in such a case, you have to write target rules.


Previous: Single Suffix Rules, Up: Portable Make

11.15 Timestamp Resolution and Make

Traditionally, file timestamps had 1-second resolution, and make used those timestamps to determine whether one file was newer than the other. However, many modern file systems have timestamps with 1-nanosecond resolution. Some make implementations look at the entire timestamp; others ignore the fractional part, which can lead to incorrect results. Normally this is not a problem, but in some extreme cases you may need to use tricks like ‘sleep 1’ to work around timestamp truncation bugs.

Commands like ‘cp -p’ and ‘touch -r’ typically do not copy file timestamps to their full resolutions (see Limitations of Usual Tools). Hence you should be wary of rules like this:

     dest: src
             cp -p src dest

as dest often appears to be older than src after the timestamp is truncated, and this can cause make to do needless rework the next time it is invoked. To work around this problem, you can use a timestamp file, e.g.:

     dest-stamp: src
             cp -p src dest
             date >dest-stamp


Next: , Previous: Portable Make, Up: Top

12 Portable C and C++ Programming

C and C++ programs often use low-level features of the underlying system, and therefore are often more difficult to make portable to other platforms.

Several standards have been developed to help make your programs more portable. If you write programs with these standards in mind, you can have greater confidence that your programs work on a wide variety of systems. See Language Standards Supported by GCC, for a list of C-related standards. Many programs also assume the Posix standard.

Some old code is written to be portable to K&R C, which predates any C standard. K&R C compilers are no longer of practical interest, though, and the rest of section assumes at least C89, the first C standard.

Program portability is a huge topic, and this section can only briefly introduce common pitfalls. See Portability between System Types, for more information.


Next: , Up: Portable C and C++

12.1 Varieties of Unportability

Autoconf tests and ordinary programs often need to test what is allowed on a system, and therefore they may need to deliberately exceed the boundaries of what the standards allow, if only to see whether an optional feature is present. When you write such a program, you should keep in mind the difference between constraints, unspecified behavior, and undefined behavior.

In C, a constraint is a rule that the compiler must enforce. An example constraint is that C programs must not declare a bit-field with negative width. Tests can therefore reliably assume that programs with negative-width bit-fields are rejected by a compiler that conforms to the standard.

Unspecified behavior is valid behavior, where the standard allows multiple possibilities. For example, the order of evaluation of function arguments is unspecified. Some unspecified behavior is implementation-defined, i.e., documented by the implementation, but since Autoconf tests cannot read the documentation they cannot distinguish between implementation-defined and other unspecified behavior. It is common for Autoconf tests to probe implementations to determine otherwise-unspecified behavior.

Undefined behavior is invalid behavior, where the standard allows the implementation to do anything it pleases. For example, dereferencing a null pointer leads to undefined behavior. If possible, test programs should avoid undefined behavior, since a program with undefined behavior might succeed on a test that should fail.

The above rules apply to programs that are intended to conform to the standard. However, strictly-conforming programs are quite rare, since the standards are so limiting. A major goal of Autoconf is to support programs that use implementation features not described by the standard, and it is fairly common for test programs to violate the above rules, if the programs work well enough in practice.


Next: , Previous: Varieties of Unportability, Up: Portable C and C++

12.2 Integer Overflow

In C, signed integer overflow leads to undefined behavior. However, many programs and Autoconf tests assume that signed integer overflow after addition, subtraction, or multiplication silently wraps around modulo a power of two, using two's complement arithmetic, so long as you cast the resulting value to an integer type or store it into an integer variable. Such programs are portable to the vast majority of modern platforms. However, signed integer division is not always harmless: for example, on CPUs of the i386 family, dividing INT_MIN by -1 yields a SIGFPE signal which by default terminates the program. Worse, taking the remainder of these two values typically yields the same signal on these CPUs, even though the C standard requires INT_MIN % -1 to yield zero because the expression does not overflow.

GCC users might consider using the -ftrapv option if they are worried about porting their code to the rare platforms where signed integer overflow does not wrap around after addition, subtraction, or multiplication.

Unsigned integer overflow reliably wraps around modulo the word size. This is guaranteed by the C standard and is portable in practice.


Next: , Previous: Integer Overflow, Up: Portable C and C++

12.3 Properties of Null Pointers

Most modern hosts reliably fail when you attempt to dereference a null pointer.

On almost all modern hosts, null pointers use an all-bits-zero internal representation, so you can reliably use memset with 0 to set all the pointers in an array to null values.

If p is a null pointer to an object type, the C expression p + 0 always evaluates to p on modern hosts, even though the standard says that it has undefined behavior.


Next: , Previous: Null Pointers, Up: Portable C and C++

12.4 Buffer Overruns and Subscript Errors

Buffer overruns and subscript errors are the most common dangerous errors in C programs. They result in undefined behavior because storing outside an array typically modifies storage that is used by some other object, and most modern systems lack runtime checks to catch these errors. Programs should not rely on buffer overruns being caught.

There is one exception to the usual rule that a portable program cannot address outside an array. In C, it is valid to compute the address just past an object, e.g., &a[N] where a has N elements, so long as you do not dereference the resulting pointer. But it is not valid to compute the address just before an object, e.g., &a[-1]; nor is it valid to compute two past the end, e.g., &a[N+1]. On most platforms &a[-1] < &a[0] && &a[N] < &a[N+1], but this is not reliable in general, and it is usually easy enough to avoid the potential portability problem, e.g., by allocating an extra unused array element at the start or end.

Valgrind can catch many overruns. GCC users might also consider using the -fmudflap option to catch overruns.

Buffer overruns are usually caused by off-by-one errors, but there are more subtle ways to get them.

Using int values to index into an array or compute array sizes causes problems on typical 64-bit hosts where an array index might be 2^31 or larger. Index values of type size_t avoid this problem, but cannot be negative. Index values of type ptrdiff_t are signed, and are wide enough in practice.

If you add or multiply two numbers to calculate an array size, e.g., malloc (x * sizeof y + z), havoc ensues if the addition or multiplication overflows.

Many implementations of the alloca function silently misbehave and can generate buffer overflows if given sizes that are too large. The size limits are implementation dependent, but are at least 4000 bytes on all platforms that we know about.

The standard functions asctime, asctime_r, ctime, ctime_r, and gets are prone to buffer overflows, and portable code should not use them unless the inputs are known to be within certain limits. The time-related functions can overflow their buffers if given timestamps out of range (e.g., a year less than -999 or greater than 9999). Time-related buffer overflows cannot happen with recent-enough versions of the GNU C library, but are possible with other implementations. The gets function is the worst, since it almost invariably overflows its buffer when presented with an input line larger than the buffer.


Next: , Previous: Buffer Overruns, Up: Portable C and C++

12.5 Floating Point Portability

Almost all modern systems use IEEE-754 floating point, and it is safe to assume IEEE-754 in most portable code these days. For more information, please see David Goldberg's classic paper What Every Computer Scientist Should Know About Floating-Point Arithmetic.


Previous: Floating Point Portability, Up: Portable C and C++

12.6 Exiting Portably

A C or C++ program can exit with status N by returning N from the main function. Portable programs are supposed to exit either with status 0 or EXIT_SUCCESS to succeed, or with status EXIT_FAILURE to fail, but in practice it is portable to fail by exiting with status 1, and test programs that assume Posix can fail by exiting with status values from 1 through 255. Programs on SunOS 2.0 (1985) through 3.5.2 (1988) incorrectly exited with zero status when main returned nonzero, but ancient systems like these are no longer of practical concern.

A program can also exit with status N by passing N to the exit function, and a program can fail by calling the abort function. If a program is specialized to just some platforms, it can fail by calling functions specific to those platforms, e.g., _exit (Posix) and _Exit (C99). However, like other functions, an exit function should be declared, typically by including a header. For example, if a C program calls exit, it should include stdlib.h either directly or via the default includes (see Default Includes).

A program can fail due to undefined behavior such as dereferencing a null pointer, but this is not recommended as undefined behavior allows an implementation to do whatever it pleases and this includes exiting successfully.


Next: , Previous: Portable C and C++, Up: Top

13 Manual Configuration

A few kinds of features can't be guessed automatically by running test programs. For example, the details of the object-file format, or special options that need to be passed to the compiler or linker. You can check for such features using ad-hoc means, such as having configure check the output of the uname program, or looking for libraries that are unique to particular systems. However, Autoconf provides a uniform method for handling unguessable features.


Next: , Up: Manual Configuration

13.1 Specifying the System Type

Like other GNU configure scripts, Autoconf-generated configure scripts can make decisions based on a canonical name for the system type, which has the form: ‘cpu-vendor-os’, where os can be ‘system’ or ‘kernel-system

configure can usually guess the canonical name for the type of system it's running on. To do so it runs a script called config.guess, which infers the name using the uname command or symbols predefined by the C preprocessor.

Alternately, the user can specify the system type with command line arguments to configure. Doing so is necessary when cross-compiling. In the most complex case of cross-compiling, three system types are involved. The options to specify them are:

--build=build-type
the type of system on which the package is being configured and compiled. It defaults to the result of running config.guess.
--host=host-type
the type of system on which the package runs. By default it is the same as the build machine. Specifying it enables the cross-compilation mode.
--target=target-type
the type of system for which any compiler tools in the package produce code (rarely needed). By default, it is the same as host.

If you mean to override the result of config.guess, use --build, not --host, since the latter enables cross-compilation. For historical reasons, passing --host also changes the build type. Therefore, whenever you specify --host, be sure to specify --build too; this will be fixed in the future. So, to enter cross-compilation mode, use a command like this

     ./configure --build=i686-pc-linux-gnu --host=m68k-coff

Note that if you do not specify --host, configure fails if it can't run the code generated by the specified compiler. For example, configuring as follows fails:

     ./configure CC=m68k-coff-gcc

In the future, when cross-compiling Autoconf will not accept tools (compilers, linkers, assemblers) whose name is not prefixed with the host type. The only case when this may be useful is when you really are not cross-compiling, but only building for a least-common-denominator architecture: an example is building for i386-pc-linux-gnu while running on an i686-pc-linux-gnu architecture. In this case, some particular pairs might be similar enough to let you get away with the system compilers, but in general the compiler might make bogus assumptions on the host: if you know what you are doing, please create symbolic links from the host compiler to the build compiler.

configure recognizes short aliases for many system types; for example, ‘decstation’ can be used instead of ‘mips-dec-ultrix4.2’. configure runs a script called config.sub to canonicalize system type aliases.

This section deliberately omits the description of the obsolete interface; see Hosts and Cross-Compilation.


Next: , Previous: Specifying Names, Up: Manual Configuration

13.2 Getting the Canonical System Type

The following macros make the system type available to configure scripts.

The variables ‘build_alias’, ‘host_alias’, and ‘target_alias’ are always exactly the arguments of --build, --host, and --target; in particular, they are left empty if the user did not use them, even if the corresponding AC_CANONICAL macro was run. Any configure script may use these variables anywhere. These are the variables that should be used when in interaction with the user.

If you need to recognize some special environments based on their system type, run the following macros to get canonical system names. These variables are not set before the macro call.

If you use these macros, you must distribute config.guess and config.sub along with your source code. See Output, for information about the AC_CONFIG_AUX_DIR macro which you can use to control in which directory configure looks for those scripts.

— Macro: AC_CANONICAL_BUILD

Compute the canonical build-system type variable, build, and its three individual parts build_cpu, build_vendor, and build_os.

If --build was specified, then build is the canonicalization of build_alias by config.sub, otherwise it is determined by the shell script config.guess.

— Macro: AC_CANONICAL_HOST

Compute the canonical host-system type variable, host, and its three individual parts host_cpu, host_vendor, and host_os.

If --host was specified, then host is the canonicalization of host_alias by config.sub, otherwise it defaults to build.

— Macro: AC_CANONICAL_TARGET

Compute the canonical target-system type variable, target, and its three individual parts target_cpu, target_vendor, and target_os.

If --target was specified, then target is the canonicalization of target_alias by config.sub, otherwise it defaults to host.

Note that there can be artifacts due to the backward compatibility code. See See Hosts and Cross-Compilation, for more.


Previous: Canonicalizing, Up: Manual Configuration

13.3 Using the System Type

In configure.ac the system type is generally used by one or more case statements to select system-specifics. Shell wildcards can be used to match a group of system types.

For example, an extra assembler code object file could be chosen, giving access to a CPU cycle counter register. $(CYCLE_OBJ) in the following would be used in a makefile to add the object to a program or library.

     case $host in
       alpha*-*-*) CYCLE_OBJ=rpcc.o ;;
       i?86-*-*)   CYCLE_OBJ=rdtsc.o ;;
       *)          CYCLE_OBJ= ;;
     esac
     AC_SUBST([CYCLE_OBJ])

AC_CONFIG_LINKS (see Configuration Links) is another good way to select variant source files, for example optimized code for some CPUs. The configured CPU type doesn't always indicate exact CPU types, so some runtime capability checks may be necessary too.

     case $host in
       alpha*-*-*)   AC_CONFIG_LINKS([dither.c:alpha/dither.c]) ;;
       powerpc*-*-*) AC_CONFIG_LINKS([dither.c:powerpc/dither.c]) ;;
       *-*-*)        AC_CONFIG_LINKS([dither.c:generic/dither.c]) ;;
     esac

The host system type can also be used to find cross-compilation tools with AC_CHECK_TOOL (see Generic Programs).

The above examples all show ‘$host’, since this is where the code is going to run. Only rarely is it necessary to test ‘$build’ (which is where the build is being done).

Whenever you're tempted to use ‘$host’ it's worth considering whether some sort of probe would be better. New system types come along periodically or previously missing features are added. Well-written probes can adapt themselves to such things, but hard-coded lists of names can't. Here are some guidelines,

$target’ is for use by a package creating a compiler or similar. For ordinary packages it's meaningless and should not be used. It indicates what the created compiler should generate code for, if it can cross-compile. ‘$target’ generally selects various hard-coded CPU and system conventions, since usually the compiler or tools under construction themselves determine how the target works.


Next: , Previous: Manual Configuration, Up: Top

14 Site Configuration

configure scripts support several kinds of local configuration decisions. There are ways for users to specify where external software packages are, include or exclude optional features, install programs under modified names, and set default values for configure options.


Next: , Up: Site Configuration

14.1 Controlling Help Output

Users consult ‘configure --help’ to learn of configuration decisions specific to your package. By default, configure breaks this output into sections for each type of option; within each section, help strings appear in the order configure.ac defines them:

     Optional Features:
       ...
       --enable-bar            include bar
     
     Optional Packages:
       ...
       --with-foo              use foo
— Macro: AC_PRESERVE_HELP_ORDER

Request an alternate --help format, in which options of all types appear together, in the order defined. Call this macro before any AC_ARG_ENABLE or AC_ARG_WITH.

          Optional Features and Packages:
            ...
            --enable-bar            include bar
            --with-foo              use foo
     


Next: , Previous: Help Formatting, Up: Site Configuration

14.2 Working With External Software

Some packages require, or can optionally use, other software packages that are already installed. The user can give configure command line options to specify which such external software to use. The options have one of these forms:

     --with-package[=arg]
     --without-package

For example, --with-gnu-ld means work with the GNU linker instead of some other linker. --with-x means work with The X Window System.

The user can give an argument by following the package name with ‘=’ and the argument. Giving an argument of ‘no’ is for packages that are used by default; it says to not use the package. An argument that is neither ‘yes’ nor ‘no’ could include a name or number of a version of the other package, to specify more precisely which other package this program is supposed to work with. If no argument is given, it defaults to ‘yes’. --without-package is equivalent to --with-package=no.

configure scripts do not complain about --with-package options that they do not support. This behavior permits configuring a source tree containing multiple packages with a top-level configure script when the packages support different options, without spurious error messages about options that some of the packages support. An unfortunate side effect is that option spelling errors are not diagnosed. No better approach to this problem has been suggested so far.

For each external software package that may be used, configure.ac should call AC_ARG_WITH to detect whether the configure user asked to use it. Whether each package is used or not by default, and which arguments are valid, is up to you.

— Macro: AC_ARG_WITH (package, help-string, [action-if-given], [action-if-not-given])

If the user gave configure the option --with-package or --without-package, run shell commands action-if-given. If neither option was given, run shell commands action-if-not-given. The name package indicates another software package that this program should work with. It should consist only of alphanumeric characters and dashes.

The option's argument is available to the shell commands action-if-given in the shell variable withval, which is actually just the value of the shell variable with_package, with any - characters changed into ‘_’. You may use that variable instead, if you wish.

The argument help-string is a description of the option that looks like this:

            --with-readline         support fancy command line editing
     

help-string may be more than one line long, if more detail is needed. Just make sure the columns line up in ‘configure --help’. Avoid tabs in the help string. You'll need to enclose the help string in ‘[’ and ‘]’ in order to produce the leading blanks.

You should format your help-string with the macro AS_HELP_STRING (see Pretty Help Strings).

The following example shows how to use the AC_ARG_WITH macro in a common situation. You want to let the user decide whether to enable support for an external library (e.g., the readline library); if the user specified neither --with-readline nor --without-readline, you want to enable support for readline only if the library is available on the system.

          AC_ARG_WITH([readline],
            [AS_HELP_STRING([--with-readline],
              [support fancy command line editing @<:@default=check@:>@])],
            [],
            [with_readline=check])
          
          LIBREADLINE=
          AS_IF([test "x$with_readline" != xno],
            [AC_CHECK_LIB([readline], [main],
              [AC_SUBST([LIBREADLINE], ["-lreadline -lncurses"])
               AC_DEFINE([HAVE_LIBREADLINE], [1],
                         [Define if you have libreadline])
              ],
              [if test "x$with_readline" != xcheck; then
                 AC_MSG_FAILURE(
                   [--with-readline was given, but test for readline failed])
               fi
              ], -lncurses)])
     

The next example shows how to use AC_ARG_WITH to give the user the possibility to enable support for the readline library, in case it is still experimental and not well tested, and is therefore disabled by default.

          AC_ARG_WITH([readline],
            [AS_HELP_STRING([--with-readline],
              [enable experimental support for readline])],
            [],
            [with_readline=no])
          
          LIBREADLINE=
          AS_IF([test "x$with_readline" != xno],
            [AC_CHECK_LIB([readline], [main],
              [AC_SUBST([LIBREADLINE], ["-lreadline -lncurses"])
               AC_DEFINE([HAVE_LIBREADLINE], [1],
                         [Define if you have libreadline])
              ],
              [AC_MSG_FAILURE(
                 [--with-readline was given, but test for readline failed])],
              [-lncurses])])
     

The last example shows how to use AC_ARG_WITH to give the user the possibility to disable support for the readline library, given that it is an important feature and that it should be enabled by default.

          AC_ARG_WITH([readline],
            [AS_HELP_STRING([--without-readline],
              [disable support for readline])],
            [],
            [with_readline=yes])
          
          LIBREADLINE=
          AS_IF([test "x$with_readline" != xno],
            [AC_CHECK_LIB([readline], [main],
              [AC_SUBST([LIBREADLINE], ["-lreadline -lncurses"])
               AC_DEFINE([HAVE_LIBREADLINE], [1],
                         [Define if you have libreadline])
              ],
              [AC_MSG_FAILURE(
                 [readline test failed (--without-readline to disable)])],
              [-lncurses])])
     

These three examples can be easily adapted to the case where AC_ARG_ENABLE should be preferred to AC_ARG_WITH (see Package Options).

— Macro: AC_WITH (package, action-if-given, [action-if-not-given])

This is an obsolete version of AC_ARG_WITH that does not support providing a help string.


Next: , Previous: External Software, Up: Site Configuration

14.3 Choosing Package Options

If a software package has optional compile-time features, the user can give configure command line options to specify whether to compile them. The options have one of these forms:

     --enable-feature[=arg]
     --disable-feature

These options allow users to choose which optional features to build and install. --enable-feature options should never make a feature behave differently or cause one feature to replace another. They should only cause parts of the program to be built rather than left out.

The user can give an argument by following the feature name with ‘=’ and the argument. Giving an argument of ‘no’ requests that the feature not be made available. A feature with an argument looks like --enable-debug=stabs. If no argument is given, it defaults to ‘yes’. --disable-feature is equivalent to --enable-feature=no.

configure scripts do not complain about --enable-feature options that they do not support. This behavior permits configuring a source tree containing multiple packages with a top-level configure script when the packages support different options, without spurious error messages about options that some of the packages support. An unfortunate side effect is that option spelling errors are not diagnosed. No better approach to this problem has been suggested so far.

For each optional feature, configure.ac should call AC_ARG_ENABLE to detect whether the configure user asked to include it. Whether each feature is included or not by default, and which arguments are valid, is up to you.

— Macro: AC_ARG_ENABLE (feature, help-string, [action-if-given], [action-if-not-given])

If the user gave configure the option --enable-feature or --disable-feature, run shell commands action-if-given. If neither option was given, run shell commands action-if-not-given. The name feature indicates an optional user-level facility. It should consist only of alphanumeric characters and dashes.

The option's argument is available to the shell commands action-if-given in the shell variable enableval, which is actually just the value of the shell variable enable_feature, with any - characters changed into ‘_’. You may use that variable instead, if you wish. The help-string argument is like that of AC_ARG_WITH (see External Software).

You should format your help-string with the macro AS_HELP_STRING (see Pretty Help Strings).

See the examples suggested with the definition of AC_ARG_WITH (see External Software) to get an idea of possible applications of AC_ARG_ENABLE.

— Macro: AC_ENABLE (feature, action-if-given, [action-if-not-given])

This is an obsolete version of AC_ARG_ENABLE that does not support providing a help string.


Next: , Previous: Package Options, Up: Site Configuration

14.4 Making Your Help Strings Look Pretty

Properly formatting the ‘help strings’ which are used in AC_ARG_WITH (see External Software) and AC_ARG_ENABLE (see Package Options) can be challenging. Specifically, you want your own ‘help strings’ to line up in the appropriate columns of ‘configure --help’ just like the standard Autoconf ‘help strings’ do. This is the purpose of the AS_HELP_STRING macro.

— Macro: AS_HELP_STRING (left-hand-side, right-hand-side)

Expands into an help string that looks pretty when the user executes ‘configure --help’. It is typically used in AC_ARG_WITH (see External Software) or AC_ARG_ENABLE (see Package Options). The following example makes this clearer.

          AC_ARG_WITH([foo],
            [AS_HELP_STRING([--with-foo],
               [use foo (default is no)])],
            [use_foo=$withval],
            [use_foo=no])
     

The second argument of AS_HELP_STRING is not a literal, and should not be double quoted. See Autoconf Language, for a more detailed explanation. Then the last few lines of ‘configure --help’ appear like this:

          --enable and --with options recognized:
            --with-foo              use foo (default is no)
     

The AS_HELP_STRING macro is particularly helpful when the left-hand-side and/or right-hand-side are composed of macro arguments, as shown in the following example.

          AC_DEFUN([MY_ARG_WITH],
            [AC_ARG_WITH([$1],
               [AS_HELP_STRING([--with-$1], [use $1 (default is $2)])],
               [use_[]$1=$withval],
               [use_[]$1=$2])])
     


Next: , Previous: Pretty Help Strings, Up: Site Configuration

14.5 Configuring Site Details

Some software packages require complex site-specific information. Some examples are host names to use for certain services, company names, and email addresses to contact. Since some configuration scripts generated by Metaconfig ask for such information interactively, people sometimes wonder how to get that information in Autoconf-generated configuration scripts, which aren't interactive.

Such site configuration information should be put in a file that is edited only by users, not by programs. The location of the file can either be based on the prefix variable, or be a standard location such as the user's home directory. It could even be specified by an environment variable. The programs should examine that file at runtime, rather than at compile time. Runtime configuration is more convenient for users and makes the configuration process simpler than getting the information while configuring. See Variables for Installation Directories, for more information on where to put data files.


Next: , Previous: Site Details, Up: Site Configuration

14.6 Transforming Program Names When Installing

Autoconf supports changing the names of programs when installing them. In order to use these transformations, configure.ac must call the macro AC_ARG_PROGRAM.

— Macro: AC_ARG_PROGRAM

Place in output variable program_transform_name a sequence of sed commands for changing the names of installed programs.

If any of the options described below are given to configure, program names are transformed accordingly. Otherwise, if AC_CANONICAL_TARGET has been called and a --target value is given, the target type followed by a dash is used as a prefix. Otherwise, no program name transformation is done.


Next: , Up: Transforming Names

14.6.1 Transformation Options

You can specify name transformations by giving configure these command line options:

--program-prefix=prefix
prepend prefix to the names;
--program-suffix=suffix
append suffix to the names;
--program-transform-name=expression
perform sed substitution expression on the names.


Next: , Previous: Transformation Options, Up: Transforming Names

14.6.2 Transformation Examples

These transformations are useful with programs that can be part of a cross-compilation development environment. For example, a cross-assembler running on a Sun 4 configured with --target=i960-vxworks is normally installed as i960-vxworks-as, rather than as, which could be confused with a native Sun 4 assembler.

You can force a program name to begin with g, if you don't want GNU programs installed on your system to shadow other programs with the same name. For example, if you configure GNU diff with --program-prefix=g, then when you run ‘make install’ it is installed as /usr/local/bin/gdiff.

As a more sophisticated example, you could use

     --program-transform-name='s/^/g/; s/^gg/g/; s/^gless/less/'

to prepend ‘g’ to most of the program names in a source tree, excepting those like gdb that already have one and those like less and lesskey that aren't GNU programs. (That is assuming that you have a source tree containing those programs that is set up to use this feature.)

One way to install multiple versions of some programs simultaneously is to append a version number to the name of one or both. For example, if you want to keep Autoconf version 1 around for awhile, you can configure Autoconf version 2 using --program-suffix=2 to install the programs as /usr/local/bin/autoconf2, /usr/local/bin/autoheader2, etc. Nevertheless, pay attention that only the binaries are renamed, therefore you'd have problems with the library files which might overlap.


Previous: Transformation Examples, Up: Transforming Names

14.6.3 Transformation Rules

Here is how to use the variable program_transform_name in a Makefile.in:

     PROGRAMS = cp ls rm
     transform = @program_transform_name@
     install:
             for p in $(PROGRAMS); do \
               $(INSTALL_PROGRAM) $$p $(DESTDIR)$(bindir)/`echo $$p | \
                                                   sed '$(transform)'`; \
             done
     
     uninstall:
             for p in $(PROGRAMS); do \
               rm -f $(DESTDIR)$(bindir)/`echo $$p | sed '$(transform)'`; \
             done

It is guaranteed that program_transform_name is never empty, and that there are no useless separators. Therefore you may safely embed program_transform_name within a sed program using ‘;’:

     transform = @program_transform_name@
     transform_exe = s/$(EXEEXT)$$//;$(transform);s/$$/$(EXEEXT)/

Whether to do the transformations on documentation files (Texinfo or man) is a tricky question; there seems to be no perfect answer, due to the several reasons for name transforming. Documentation is not usually particular to a specific architecture, and Texinfo files do not conflict with system documentation. But they might conflict with earlier versions of the same files, and man pages sometimes do conflict with system documentation. As a compromise, it is probably best to do name transformations on man pages but not on Texinfo manuals.


Previous: Transforming Names, Up: Site Configuration

14.7 Setting Site Defaults

Autoconf-generated configure scripts allow your site to provide default values for some configuration values. You do this by creating site- and system-wide initialization files.

If the environment variable CONFIG_SITE is set, configure uses its value as the name of a shell script to read. Otherwise, it reads the shell script prefix/share/config.site if it exists, then prefix/etc/config.site if it exists. Thus, settings in machine-specific files override those in machine-independent ones in case of conflict.

Site files can be arbitrary shell scripts, but only certain kinds of code are really appropriate to be in them. Because configure reads any cache file after it has read any site files, a site file can define a default cache file to be shared between all Autoconf-generated configure scripts run on that system (see Cache Files). If you set a default cache file in a site file, it is a good idea to also set the output variable CC in that site file, because the cache file is only valid for a particular compiler, but many systems have several available.

You can examine or override the value set by a command line option to configure in a site file; options set shell variables that have the same names as the options, with any dashes turned into underscores. The exceptions are that --without- and --disable- options are like giving the corresponding --with- or --enable- option and the value ‘no’. Thus, --cache-file=localcache sets the variable cache_file to the value ‘localcache’; --enable-warnings=no or --disable-warnings sets the variable enable_warnings to the value ‘no’; --prefix=/usr sets the variable prefix to the value ‘/usr’; etc.

Site files are also good places to set default values for other output variables, such as CFLAGS, if you need to give them non-default values: anything you would normally do, repetitively, on the command line. If you use non-default values for prefix or exec_prefix (wherever you locate the site file), you can set them in the site file if you specify it with the CONFIG_SITE environment variable.

You can set some cache values in the site file itself. Doing this is useful if you are cross-compiling, where it is impossible to check features that require running a test program. You could “prime the cache” by setting those values correctly for that system in prefix/etc/config.site. To find out the names of the cache variables you need to set, look for shell variables with ‘_cv_’ in their names in the affected configure scripts, or in the Autoconf M4 source code for those macros.

The cache file is careful to not override any variables set in the site files. Similarly, you should not override command-line options in the site files. Your code should check that variables such as prefix and cache_file have their default values (as set near the top of configure) before changing them.

Here is a sample file /usr/share/local/gnu/share/config.site. The command ‘configure --prefix=/usr/share/local/gnu’ would read this file (if CONFIG_SITE is not set to a different file).

     # config.site for configure
     #
     # Change some defaults.
     test "$prefix" = NONE && prefix=/usr/share/local/gnu
     test "$exec_prefix" = NONE && exec_prefix=/usr/local/gnu
     test "$sharedstatedir" = '$prefix/com' && sharedstatedir=/var
     test "$localstatedir" = '$prefix/var' && localstatedir=/var
     
     # Give Autoconf 2.x generated configure scripts a shared default
     # cache file for feature test results, architecture-specific.
     if test "$cache_file" = /dev/null; then
       cache_file="$prefix/var/config.cache"
       # A cache file is only valid for one C compiler.
       CC=gcc
     fi


Next: , Previous: Site Configuration, Up: Top

15 Running configure Scripts

Below are instructions on how to configure a package that uses a configure script, suitable for inclusion as an INSTALL file in the package. A plain-text version of INSTALL which you may use comes with Autoconf.


Next: , Up: Running configure Scripts

15.1 Basic Installation

Briefly, the shell commands ‘./configure; make; make install’ should configure, build, and install this package. The following more-detailed instructions are generic; see the README file for instructions specific to this package.

The configure shell script attempts to guess correct values for various system-dependent variables used during compilation. It uses those values to create a Makefile in each directory of the package. It may also create one or more .h files containing system-dependent definitions. Finally, it creates a shell script config.status that you can run in the future to recreate the current configuration, and a file config.log containing compiler output (useful mainly for debugging configure).

It can also use an optional file (typically called config.cache and enabled with --cache-file=config.cache or simply -C) that saves the results of its tests to speed up reconfiguring. Caching is disabled by default to prevent problems with accidental use of stale cache files.

If you need to do unusual things to compile the package, please try to figure out how configure could check whether to do them, and mail diffs or instructions to the address given in the README so they can be considered for the next release. If you are using the cache, and at some point config.cache contains results you don't want to keep, you may remove or edit it.

The file configure.ac (or configure.in) is used to create configure by a program called autoconf. You need configure.ac if you want to change it or regenerate configure using a newer version of autoconf.

The simplest way to compile this package is:

  1. cd to the directory containing the package's source code and type ‘./configure’ to configure the package for your system.

    Running configure might take a while. While running, it prints some messages telling which features it is checking for.

  2. Type ‘make’ to compile the package.
  3. Optionally, type ‘make check’ to run any self-tests that come with the package.
  4. Type ‘make install’ to install the programs and any data files and documentation.
  5. You can remove the program binaries and object files from the source code directory by typing ‘make clean’. To also remove the files that configure created (so you can compile the package for a different kind of computer), type ‘make distclean’. There is also a ‘make maintainer-clean’ target, but that is intended mainly for the package's developers. If you use it, you may have to get all sorts of other programs in order to regenerate files that came with the distribution.


Next: , Previous: Basic Installation, Up: Running configure Scripts

15.2 Compilers and Options

Some systems require unusual options for compilation or linking that the configure script does not know about. Run ‘./configure --help’ for details on some of the pertinent environment variables.

You can give configure initial values for configuration parameters by setting variables in the command line or in the environment. Here is an example:

     ./configure CC=c99 CFLAGS=-g LIBS=-lposix

See Defining Variables, for more details.


Next: , Previous: Compilers and Options, Up: Running configure Scripts

15.3 Compiling For Multiple Architectures

You can compile the package for more than one kind of computer at the same time, by placing the object files for each architecture in their own directory. To do this, you can use GNU make. cd to the directory where you want the object files and executables to go and run the configure script. configure automatically checks for the source code in the directory that configure is in and in ...

With a non-GNU make, it is safer to compile the package for one architecture at a time in the source code directory. After you have installed the package for one architecture, use ‘make distclean’ before reconfiguring for another architecture.


Next: , Previous: Multiple Architectures, Up: Running configure Scripts

15.4 Installation Names

By default, ‘make install’ installs the package's commands under /usr/local/bin, include files under /usr/local/include, etc. You can specify an installation prefix other than /usr/local by giving configure the option --prefix=prefix.

You can specify separate installation prefixes for architecture-specific files and architecture-independent files. If you pass the option --exec-prefix=prefix to configure, the package uses prefix as the prefix for installing programs and libraries. Documentation and other data files still use the regular prefix.

In addition, if you use an unusual directory layout you can give options like --bindir=dir to specify different values for particular kinds of files. Run ‘configure --help’ for a list of the directories you can set and what kinds of files go in them.

If the package supports it, you can cause programs to be installed with an extra prefix or suffix on their names by giving configure the option --program-prefix=PREFIX or --program-suffix=SUFFIX.


Next: , Previous: Installation Names, Up: Running configure Scripts

15.5 Optional Features

Some packages pay attention to --enable-feature options to configure, where feature indicates an optional part of the package. They may also pay attention to --with-package options, where package is something like ‘gnu-as’ or ‘x’ (for the X Window System). The README should mention any --enable- and --with- options that the package recognizes.

For packages that use the X Window System, configure can usually find the X include and library files automatically, but if it doesn't, you can use the configure options --x-includes=dir and --x-libraries=dir to specify their locations.


Next: , Previous: Optional Features, Up: Running configure Scripts

15.6 Specifying the System Type

There may be some features configure cannot figure out automatically, but needs to determine by the type of machine the package will run on. Usually, assuming the package is built to be run on the same architectures, configure can figure that out, but if it prints a message saying it cannot guess the machine type, give it the --build=type option. type can either be a short name for the system type, such as ‘sun4’, or a canonical name which has the form:

     cpu-company-system

where system can have one of these forms:

     os kernel-os

See the file config.sub for the possible values of each field. If config.sub isn't included in this package, then this package doesn't need to know the machine type.

If you are building compiler tools for cross-compiling, you should use the option --target=type to select the type of system they will produce code for.

If you want to use a cross compiler, that generates code for a platform different from the build platform, you should specify the host platform (i.e., that on which the generated programs will eventually be run) with --host=type.


Next: , Previous: System Type, Up: Running configure Scripts

15.7 Sharing Defaults

If you want to set default values for configure scripts to share, you can create a site shell script called config.site that gives default values for variables like CC, cache_file, and prefix. configure looks for prefix/share/config.site if it exists, then prefix/etc/config.site if it exists. Or, you can set the CONFIG_SITE environment variable to the location of the site script. A warning: not all configure scripts look for a site script.


Next: , Previous: Sharing Defaults, Up: Running configure Scripts

15.8 Defining Variables

Variables not defined in a site shell script can be set in the environment passed to configure. However, some packages may run configure again during the build, and the customized values of these variables may be lost. In order to avoid this problem, you should set them in the configure command line, using ‘VAR=value’. For example:

     ./configure CC=/usr/local2/bin/gcc

causes the specified gcc to be used as the C compiler (unless it is overridden in the site shell script).

Unfortunately, this technique does not work for CONFIG_SHELL due to an Autoconf bug. Until the bug is fixed you can use this workaround:

     CONFIG_SHELL=/bin/bash /bin/bash ./configure CONFIG_SHELL=/bin/bash


Previous: Defining Variables, Up: Running configure Scripts

15.9 configure Invocation

configure recognizes the following options to control how it operates.

--help
-h
Print a summary of the options to configure, and exit.
--version
-V
Print the version of Autoconf used to generate the configure script, and exit.
--cache-file=file
Enable the cache: use and save the results of the tests in file, traditionally config.cache. file defaults to /dev/null to disable caching.
--config-cache
-C
Alias for --cache-file=config.cache.
--quiet
--silent
-q
Do not print messages saying which checks are being made. To suppress all normal output, redirect it to /dev/null (any error messages will still be shown).
--srcdir=dir
Look for the package's source code in directory dir. Usually configure can determine that directory automatically.

configure also accepts some other, not widely useful, options. Run ‘configure --help’ for more details.


Next: , Previous: Running configure Scripts, Up: Top

16 Recreating a Configuration

The configure script creates a file named config.status, which actually configures, instantiates, the template files. It also records the configuration options that were specified when the package was last configured in case reconfiguring is needed.

Synopsis:

     ./config.status option... [file...]

It configures the files; if none are specified, all the templates are instantiated. The files must be specified without their dependencies, as in

     ./config.status foobar

not

     ./config.status foobar:foo.in:bar.in

The supported options are:

--help
-h
Print a summary of the command line options, the list of the template files, and exit.
--version
-V
Print the version number of Autoconf and exit.
--silent
--quiet
-q
Do not print progress messages.
--debug
-d
Don't remove the temporary files.
--file=file[:template]
Require that file be instantiated as if ‘AC_CONFIG_FILES(file:template)’ was used. Both file and template may be ‘-’ in which case the standard output and/or standard input, respectively, is used. If a template file name is relative, it is first looked for in the build tree, and then in the source tree. See Configuration Actions, for more details.

This option and the following ones provide one way for separately distributed packages to share the values computed by configure. Doing so can be useful if some of the packages need a superset of the features that one of them, perhaps a common library, does. These options allow a config.status file to create files other than the ones that its configure.ac specifies, so it can be used for a different package.

--header=file[:template]
Same as --file above, but with ‘AC_CONFIG_HEADERS’.
--recheck
Ask config.status to update itself and exit (no instantiation). This option is useful if you change configure, so that the results of some tests might be different from the previous run. The --recheck option reruns configure with the same arguments you used before, plus the --no-create option, which prevents configure from running config.status and creating Makefile and other files, and the --no-recursion option, which prevents configure from running other configure scripts in subdirectories. (This is so other Make rules can run config.status when it changes; see Automatic Remaking, for an example).

config.status checks several optional environment variables that can alter its behavior:

— Variable: CONFIG_SHELL

The shell with which to run configure for the --recheck option. It must be Bourne-compatible. The default is a shell that supports LINENO if available, and /bin/sh otherwise. Invoking configure by hand bypasses this setting, so you may need to use a command like ‘CONFIG_SHELL=/bin/bash /bin/bash ./configure’ to insure that the same shell is used everywhere. The absolute name of the shell should be passed.

— Variable: CONFIG_STATUS

The file name to use for the shell script that records the configuration. The default is ./config.status. This variable is useful when one package uses parts of another and the configure scripts shouldn't be merged because they are maintained separately.

You can use ./config.status in your makefiles. For example, in the dependencies given above (see Automatic Remaking), config.status is run twice when configure.ac has changed. If that bothers you, you can make each run only regenerate the files for that rule:

     config.h: stamp-h
     stamp-h: config.h.in config.status
             ./config.status config.h
             echo > stamp-h
     
     Makefile: Makefile.in config.status
             ./config.status Makefile

The calling convention of config.status has changed; see Obsolete config.status Use, for details.


Next: , Previous: config.status Invocation, Up: Top

17 Obsolete Constructs

Autoconf changes, and throughout the years some constructs have been obsoleted. Most of the changes involve the macros, but in some cases the tools themselves, or even some concepts, are now considered obsolete.

You may completely skip this chapter if you are new to Autoconf. Its intention is mainly to help maintainers updating their packages by understanding how to move to more modern constructs.


Next: , Up: Obsolete Constructs

17.1 Obsolete config.status Invocation

config.status now supports arguments to specify the files to instantiate; see config.status Invocation, for more details. Before, environment variables had to be used.

— Variable: CONFIG_COMMANDS

The tags of the commands to execute. The default is the arguments given to AC_OUTPUT and AC_CONFIG_COMMANDS in configure.ac.

— Variable: CONFIG_FILES

The files in which to perform ‘@variable@’ substitutions. The default is the arguments given to AC_OUTPUT and AC_CONFIG_FILES in configure.ac.

— Variable: CONFIG_HEADERS

The files in which to substitute C #define statements. The default is the arguments given to AC_CONFIG_HEADERS; if that macro was not called, config.status ignores this variable.

— Variable: CONFIG_LINKS

The symbolic links to establish. The default is the arguments given to AC_CONFIG_LINKS; if that macro was not called, config.status ignores this variable.

In config.status Invocation, using this old interface, the example would be:

     config.h: stamp-h
     stamp-h: config.h.in config.status
             CONFIG_COMMANDS= CONFIG_LINKS= CONFIG_FILES= \
               CONFIG_HEADERS=config.h ./config.status
             echo > stamp-h
     
     Makefile: Makefile.in config.status
             CONFIG_COMMANDS= CONFIG_LINKS= CONFIG_HEADERS= \
               CONFIG_FILES=Makefile ./config.status

(If configure.ac does not call AC_CONFIG_HEADERS, there is no need to set CONFIG_HEADERS in the make rules. Equally for CONFIG_COMMANDS, etc.)


Next: , Previous: Obsolete config.status Use, Up: Obsolete Constructs

17.2 acconfig.h

In order to produce config.h.in, autoheader needs to build or to find templates for each symbol. Modern releases of Autoconf use AH_VERBATIM and AH_TEMPLATE (see Autoheader Macros), but in older releases a file, acconfig.h, contained the list of needed templates. autoheader copied comments and #define and #undef statements from acconfig.h in the current directory, if present. This file used to be mandatory if you AC_DEFINE any additional symbols.

Modern releases of Autoconf also provide AH_TOP and AH_BOTTOM if you need to prepend/append some information to config.h.in. Ancient versions of Autoconf had a similar feature: if ./acconfig.h contains the string ‘@TOP@’, autoheader copies the lines before the line containing ‘@TOP@’ into the top of the file that it generates. Similarly, if ./acconfig.h contains the string ‘@BOTTOM@’, autoheader copies the lines after that line to the end of the file it generates. Either or both of those strings may be omitted. An even older alternate way to produce the same effect in ancient versions of Autoconf is to create the files file.top (typically config.h.top) and/or file.bot in the current directory. If they exist, autoheader copies them to the beginning and end, respectively, of its output.

In former versions of Autoconf, the files used in preparing a software package for distribution were:

     configure.ac --.   .------> autoconf* -----> configure
                    +---+
     [aclocal.m4] --+   `---.
     [acsite.m4] ---'       |
                            +--> [autoheader*] -> [config.h.in]
     [acconfig.h] ----.     |
                      +-----'
     [config.h.top] --+
     [config.h.bot] --'

Using only the AH_ macros, configure.ac should be self-contained, and should not depend upon acconfig.h etc.


Next: , Previous: acconfig.h, Up: Obsolete Constructs

17.3 Using autoupdate to Modernize configure.ac

The autoupdate program updates a configure.ac file that calls Autoconf macros by their old names to use the current macro names. In version 2 of Autoconf, most of the macros were renamed to use a more uniform and descriptive naming scheme. See Macro Names, for a description of the new scheme. Although the old names still work (see Obsolete Macros, for a list of the old macros and the corresponding new names), you can make your configure.ac files more readable and make it easier to use the current Autoconf documentation if you update them to use the new macro names.

If given no arguments, autoupdate updates configure.ac, backing up the original version with the suffix ~ (or the value of the environment variable SIMPLE_BACKUP_SUFFIX, if that is set). If you give autoupdate an argument, it reads that file instead of configure.ac and writes the updated file to the standard output.

autoupdate accepts the following options:

--help
-h
Print a summary of the command line options and exit.
--version
-V
Print the version number of Autoconf and exit.
--verbose
-v
Report processing steps.
--debug
-d
Don't remove the temporary files.
--force
-f
Force the update even if the file has not changed. Disregard the cache.
--include=dir
-I dir
Also look for input files in dir. Multiple invocations accumulate. Directories are browsed from last to first.


Next: , Previous: autoupdate Invocation, Up: Obsolete Constructs

17.4 Obsolete Macros

Several macros are obsoleted in Autoconf, for various reasons (typically they failed to quote properly, couldn't be extended for more recent issues, etc.). They are still supported, but deprecated: their use should be avoided.

During the jump from Autoconf version 1 to version 2, most of the macros were renamed to use a more uniform and descriptive naming scheme, but their signature did not change. See Macro Names, for a description of the new naming scheme. Below, if there is just the mapping from old names to new names for these macros, the reader is invited to refer to the definition of the new macro for the signature and the description.

— Macro: AC_ALLOCA

AC_FUNC_ALLOCA

— Macro: AC_ARG_ARRAY

removed because of limited usefulness

— Macro: AC_C_CROSS

This macro is obsolete; it does nothing.

— Macro: AC_C_LONG_DOUBLE

If the C compiler supports a working long double type with more range or precision than the double type, define HAVE_LONG_DOUBLE.

You should use AC_TYPE_LONG_DOUBLE or AC_TYPE_LONG_DOUBLE_WIDER instead. See Particular Types.

— Macro: AC_CANONICAL_SYSTEM

Determine the system type and set output variables to the names of the canonical system types. See Canonicalizing, for details about the variables this macro sets.

The user is encouraged to use either AC_CANONICAL_BUILD, or AC_CANONICAL_HOST, or AC_CANONICAL_TARGET, depending on the needs. Using AC_CANONICAL_TARGET is enough to run the two other macros.

— Macro: AC_CHAR_UNSIGNED

AC_C_CHAR_UNSIGNED

— Macro: AC_CHECK_TYPE (type, default)

Autoconf, up to 2.13, used to provide this version of AC_CHECK_TYPE, deprecated because of its flaws. First, although it is a member of the CHECK clan, it does more than just checking. Secondly, missing types are defined using #define, not typedef, and this can lead to problems in the case of pointer types.

This use of AC_CHECK_TYPE is obsolete and discouraged; see Generic Types, for the description of the current macro.

If the type type is not defined, define it to be the C (or C++) builtin type default, e.g., ‘short int’ or ‘unsigned int’.

This macro is equivalent to:

          AC_CHECK_TYPE([type], [],
            [AC_DEFINE_UNQUOTED([type], [default],
               [Define to `default'
                if <sys/types.h> does not define.])])
     

In order to keep backward compatibility, the two versions of AC_CHECK_TYPE are implemented, selected by a simple heuristics:

  1. If there are three or four arguments, the modern version is used.
  2. If the second argument appears to be a C or C++ type, then the obsolete version is used. This happens if the argument is a C or C++ builtin type or a C identifier ending in ‘_t’, optionally followed by one of ‘[(* ’ and then by a string of zero or more characters taken from the set ‘[]()* _a-zA-Z0-9’.
  3. If the second argument is spelled with the alphabet of valid C and C++ types, the user is warned and the modern version is used.
  4. Otherwise, the modern version is used.

You are encouraged either to use a valid builtin type, or to use the equivalent modern code (see above), or better yet, to use AC_CHECK_TYPES together with

          #if !HAVE_LOFF_T
          typedef loff_t off_t;
          #endif
     
— Macro: AC_CHECKING (feature-description)

Same as ‘AC_MSG_NOTICE([checking feature-description...]’.

— Macro: AC_COMPILE_CHECK (echo-text, includes, function-body, action-if-true, [action-if-false])

This is an obsolete version of AC_TRY_COMPILE itself replaced by AC_COMPILE_IFELSE (see Running the Compiler), with the addition that it prints ‘checking for echo-text’ to the standard output first, if echo-text is non-empty. Use AC_MSG_CHECKING and AC_MSG_RESULT instead to print messages (see Printing Messages).

— Macro: AC_CONST

AC_C_CONST

— Macro: AC_CROSS_CHECK

Same as AC_C_CROSS, which is obsolete too, and does nothing :-).

— Macro: AC_CYGWIN

Check for the Cygwin environment in which case the shell variable CYGWIN is set to ‘yes’. Don't use this macro, the dignified means to check the nature of the host is using AC_CANONICAL_HOST. As a matter of fact this macro is defined as:

          AC_REQUIRE([AC_CANONICAL_HOST])[]dnl
          case $host_os in
            *cygwin* ) CYGWIN=yes;;
                   * ) CYGWIN=no;;
          esac
     

Beware that the variable CYGWIN has a special meaning when running Cygwin, and should not be changed. That's yet another reason not to use this macro.

— Macro: AC_DECL_SYS_SIGLIST

Same as:

          AC_CHECK_DECLS([sys_siglist], [], [],
          [#include <signal.h>
          /* NetBSD declares sys_siglist in unistd.h.  */
          #if HAVE_UNISTD_H
          # include <unistd.h>
          #endif
          ])
     
— Macro: AC_DECL_YYTEXT

Does nothing, now integrated in AC_PROG_LEX.

— Macro: AC_DIR_HEADER

Like calling AC_FUNC_CLOSEDIR_VOID andAC_HEADER_DIRENT, but defines a different set of C preprocessor macros to indicate which header file is found:

Header Old Symbol New Symbol
dirent.h DIRENT HAVE_DIRENT_H
sys/ndir.h SYSNDIR HAVE_SYS_NDIR_H
sys/dir.h SYSDIR HAVE_SYS_DIR_H
ndir.h NDIR HAVE_NDIR_H

— Macro: AC_DYNIX_SEQ

If on DYNIX/ptx, add -lseq to output variable LIBS. This macro used to be defined as

          AC_CHECK_LIB([seq], [getmntent], [LIBS="-lseq $LIBS"])
     

now it is just AC_FUNC_GETMNTENT.

— Macro: AC_EXEEXT

Defined the output variable EXEEXT based on the output of the compiler, which is now done automatically. Typically set to empty string if Posix and ‘.exe’ if a DOS variant.

— Macro: AC_EMXOS2

Similar to AC_CYGWIN but checks for the EMX environment on OS/2 and sets EMXOS2.

— Macro: AC_ERROR

AC_MSG_ERROR

— Macro: AC_FIND_X

AC_PATH_X

— Macro: AC_FIND_XTRA

AC_PATH_XTRA

— Macro: AC_FOREACH

m4_foreach_w

— Macro: AC_FUNC_CHECK

AC_CHECK_FUNC

— Macro: AC_FUNC_WAIT3

If wait3 is found and fills in the contents of its third argument (a ‘struct rusage *’), which HP-UX does not do, define HAVE_WAIT3.

These days portable programs should use waitpid, not wait3, as wait3 has been removed from Posix.

— Macro: AC_GCC_TRADITIONAL

AC_PROG_GCC_TRADITIONAL

— Macro: AC_GETGROUPS_T

AC_TYPE_GETGROUPS

— Macro: AC_GETLOADAVG

AC_FUNC_GETLOADAVG

— Macro: AC_HAVE_FUNCS

AC_CHECK_FUNCS

— Macro: AC_HAVE_HEADERS

AC_CHECK_HEADERS

— Macro: AC_HAVE_LIBRARY (library, [action-if-found], [action-if-not-found], [other-libraries])

This macro is equivalent to calling AC_CHECK_LIB with a function argument of main. In addition, library can be written as any of ‘foo’, -lfoo, or ‘libfoo.a’. In all of those cases, the compiler is passed -lfoo. However, library cannot be a shell variable; it must be a literal name.

— Macro: AC_HAVE_POUNDBANG

AC_SYS_INTERPRETER (different calling convention)

— Macro: AC_HEADER_CHECK

AC_CHECK_HEADER

— Macro: AC_HEADER_EGREP

AC_EGREP_HEADER

— Macro: AC_HELP_STRING

AS_HELP_STRING

— Macro: AC_INIT (unique-file-in-source-dir)

Formerly AC_INIT used to have a single argument, and was equivalent to:

          AC_INIT
          AC_CONFIG_SRCDIR(unique-file-in-source-dir)
     
— Macro: AC_INLINE

AC_C_INLINE

— Macro: AC_INT_16_BITS

If the C type int is 16 bits wide, define INT_16_BITS. Use ‘AC_CHECK_SIZEOF(int)’ instead.

— Macro: AC_IRIX_SUN

If on irix (Silicon Graphics Unix), add -lsun to output LIBS. If you were using it to get getmntent, use AC_FUNC_GETMNTENT instead. If you used it for the NIS versions of the password and group functions, use ‘AC_CHECK_LIB(sun, getpwnam)’. Up to Autoconf 2.13, it used to be

          AC_CHECK_LIB([sun], [getmntent], [LIBS="-lsun $LIBS"])
     

now it is defined as

          AC_FUNC_GETMNTENT
          AC_CHECK_LIB([sun], [getpwnam])
     
— Macro: AC_LANG_C

Same as ‘AC_LANG([C])’.

— Macro: AC_LANG_CPLUSPLUS

Same as ‘AC_LANG([C++])’.

— Macro: AC_LANG_FORTRAN77

Same as ‘AC_LANG([Fortran 77])’.

— Macro: AC_LANG_RESTORE

Select the language that is saved on the top of the stack, as set by AC_LANG_SAVE, remove it from the stack, and call AC_LANG(language).

— Macro: AC_LANG_SAVE

Remember the current language (as set by AC_LANG) on a stack. The current language does not change. AC_LANG_PUSH is preferred.

— Macro: AC_LINK_FILES (source..., dest...)

This is an obsolete version of AC_CONFIG_LINKS. An updated version of:

          AC_LINK_FILES(config/$machine.h config/$obj_format.h,
                        host.h            object.h)
     

is:

          AC_CONFIG_LINKS([host.h:config/$machine.h
                          object.h:config/$obj_format.h])
     
— Macro: AC_LN_S

AC_PROG_LN_S

— Macro: AC_LONG_64_BITS

Define LONG_64_BITS if the C type long int is 64 bits wide. Use the generic macro ‘AC_CHECK_SIZEOF([long int])’ instead.

— Macro: AC_LONG_DOUBLE

If the C compiler supports a working long double type with more range or precision than the double type, define HAVE_LONG_DOUBLE.

You should use AC_TYPE_LONG_DOUBLE or AC_TYPE_LONG_DOUBLE_WIDER instead. See Particular Types.

— Macro: AC_LONG_FILE_NAMES

AC_SYS_LONG_FILE_NAMES

— Macro: AC_MAJOR_HEADER

AC_HEADER_MAJOR

— Macro: AC_MEMORY_H

Used to define NEED_MEMORY_H if the mem functions were defined in memory.h. Today it is equivalent to ‘AC_CHECK_HEADERS([memory.h])’. Adjust your code to depend upon HAVE_MEMORY_H, not NEED_MEMORY_H; see Standard Symbols.

— Macro: AC_MINGW32

Similar to AC_CYGWIN but checks for the MinGW compiler environment and sets MINGW32.

— Macro: AC_MINUS_C_MINUS_O

AC_PROG_CC_C_O

— Macro: AC_MMAP

AC_FUNC_MMAP

— Macro: AC_MODE_T

AC_TYPE_MODE_T

— Macro: AC_OBJEXT

Defined the output variable OBJEXT based on the output of the compiler, after .c files have been excluded. Typically set to ‘o’ if Posix, ‘obj’ if a DOS variant. Now the compiler checking macros handle this automatically.

— Macro: AC_OBSOLETE (this-macro-name, [suggestion])

Make M4 print a message to the standard error output warning that this-macro-name is obsolete, and giving the file and line number where it was called. this-macro-name should be the name of the macro that is calling AC_OBSOLETE. If suggestion is given, it is printed at the end of the warning message; for example, it can be a suggestion for what to use instead of this-macro-name.

For instance

          AC_OBSOLETE([$0], [; use AC_CHECK_HEADERS(unistd.h) instead])dnl
     

You are encouraged to use AU_DEFUN instead, since it gives better services to the user.

— Macro: AC_OFF_T

AC_TYPE_OFF_T

— Macro: AC_OUTPUT ([file]..., [extra-cmds], [init-cmds])

The use of AC_OUTPUT with argument is deprecated. This obsoleted interface is equivalent to:

          AC_CONFIG_FILES(file...)
          AC_CONFIG_COMMANDS([default],
                             extra-cmds, init-cmds)
          AC_OUTPUT
     
— Macro: AC_OUTPUT_COMMANDS (extra-cmds, [init-cmds])

Specify additional shell commands to run at the end of config.status, and shell commands to initialize any variables from configure. This macro may be called multiple times. It is obsolete, replaced by AC_CONFIG_COMMANDS.

Here is an unrealistic example:

          fubar=27
          AC_OUTPUT_COMMANDS([echo this is extra $fubar, and so on.],
                             [fubar=$fubar])
          AC_OUTPUT_COMMANDS([echo this is another, extra, bit],
                             [echo init bit])
     

Aside from the fact that AC_CONFIG_COMMANDS requires an additional key, an important difference is that AC_OUTPUT_COMMANDS is quoting its arguments twice, unlike AC_CONFIG_COMMANDS. This means that AC_CONFIG_COMMANDS can safely be given macro calls as arguments:

          AC_CONFIG_COMMANDS(foo, [my_FOO()])
     

Conversely, where one level of quoting was enough for literal strings with AC_OUTPUT_COMMANDS, you need two with AC_CONFIG_COMMANDS. The following lines are equivalent:

          AC_OUTPUT_COMMANDS([echo "Square brackets: []"])
          AC_CONFIG_COMMANDS([default], [[echo "Square brackets: []"]])
     
— Macro: AC_PID_T

AC_TYPE_PID_T

— Macro: AC_PREFIX

AC_PREFIX_PROGRAM

— Macro: AC_PROGRAMS_CHECK

AC_CHECK_PROGS

— Macro: AC_PROGRAMS_PATH

AC_PATH_PROGS

— Macro: AC_PROGRAM_CHECK

AC_CHECK_PROG

— Macro: AC_PROGRAM_EGREP

AC_EGREP_CPP

— Macro: AC_PROGRAM_PATH

AC_PATH_PROG

— Macro: AC_REMOTE_TAPE

removed because of limited usefulness

— Macro: AC_RESTARTABLE_SYSCALLS

AC_SYS_RESTARTABLE_SYSCALLS

— Macro: AC_RETSIGTYPE

AC_TYPE_SIGNAL

— Macro: AC_RSH

removed because of limited usefulness

— Macro: AC_SCO_INTL

If on SCO Unix, add -lintl to output variable LIBS. This macro used to do this:

          AC_CHECK_LIB([intl], [strftime], [LIBS="-lintl $LIBS"])
     

Now it just calls AC_FUNC_STRFTIME instead.

— Macro: AC_SETVBUF_REVERSED

AC_FUNC_SETVBUF_REVERSED

— Macro: AC_SET_MAKE

AC_PROG_MAKE_SET

— Macro: AC_SIZEOF_TYPE

AC_CHECK_SIZEOF

— Macro: AC_SIZE_T

AC_TYPE_SIZE_T

— Macro: AC_STAT_MACROS_BROKEN

AC_HEADER_STAT

— Macro: AC_STDC_HEADERS

AC_HEADER_STDC

— Macro: AC_STRCOLL

AC_FUNC_STRCOLL

— Macro: AC_ST_BLKSIZE

AC_CHECK_MEMBERS

— Macro: AC_ST_BLOCKS

AC_STRUCT_ST_BLOCKS

— Macro: AC_ST_RDEV

AC_CHECK_MEMBERS

— Macro: AC_SYS_RESTARTABLE_SYSCALLS

If the system automatically restarts a system call that is interrupted by a signal, define HAVE_RESTARTABLE_SYSCALLS. This macro does not check whether system calls are restarted in general—it checks whether a signal handler installed with signal (but not sigaction) causes system calls to be restarted. It does not check whether system calls can be restarted when interrupted by signals that have no handler.

These days portable programs should use sigaction with SA_RESTART if they want restartable system calls. They should not rely on HAVE_RESTARTABLE_SYSCALLS, since nowadays whether a system call is restartable is a dynamic issue, not a configuration-time issue.

— Macro: AC_SYS_SIGLIST_DECLARED

AC_DECL_SYS_SIGLIST

— Macro: AC_TEST_CPP

AC_TRY_CPP, replaced by AC_PREPROC_IFELSE.

— Macro: AC_TEST_PROGRAM

AC_TRY_RUN, replaced by AC_RUN_IFELSE.

— Macro: AC_TIMEZONE

AC_STRUCT_TIMEZONE

— Macro: AC_TIME_WITH_SYS_TIME

AC_HEADER_TIME

— Macro: AC_TRY_COMPILE (includes, function-body, [action-if-true], [action-if-false])

Same as:

          AC_COMPILE_IFELSE(
            [AC_LANG_PROGRAM([[includes]],
               [[function-body]])],
            [action-if-true],
            [action-if-false])
     

See Running the Compiler.

This macro double quotes both includes and function-body.

For C and C++, includes is any #include statements needed by the code in function-body (includes is ignored if the currently selected language is Fortran or Fortran 77). The compiler and compilation flags are determined by the current language (see Language Choice).

— Macro: AC_TRY_CPP (input, [action-if-true], [action-if-false])

Same as:

          AC_PREPROC_IFELSE(
            [AC_LANG_SOURCE([[input]])],
            [action-if-true],
            [action-if-false])
     

See Running the Preprocessor.

This macro double quotes the input.

— Macro: AC_TRY_LINK (includes, function-body, [action-if-true], [action-if-false])

Same as:

          AC_LINK_IFELSE(
            [AC_LANG_PROGRAM([[includes]],
               [[function-body]])],
            [action-if-true],
            [action-if-false])
     

See Running the Compiler.

This macro double quotes both includes and function-body.

Depending on the current language (see Language Choice), create a test program to see whether a function whose body consists of function-body can be compiled and linked. If the file compiles and links successfully, run shell commands action-if-found, otherwise run action-if-not-found.

This macro double quotes both includes and function-body.

For C and C++, includes is any #include statements needed by the code in function-body (includes is ignored if the currently selected language is Fortran or Fortran 77). The compiler and compilation flags are determined by the current language (see Language Choice), and in addition LDFLAGS and LIBS are used for linking.

— Macro: AC_TRY_LINK_FUNC (function, [action-if-found], [action-if-not-found])

This macro is equivalent to ‘AC_LINK_IFELSE([AC_LANG_CALL([], [function])], [action-if-found], [action-if-not-found])’.

— Macro: AC_TRY_RUN (program, [action-if-true], [action-if-false], [action-if-cross-compiling])

Same as:

          AC_RUN_IFELSE(
            [AC_LANG_SOURCE([[program]])],
            [action-if-true],
            [action-if-false],
            [action-if-cross-compiling])
     

See Runtime.

— Macro: AC_UID_T

AC_TYPE_UID_T

— Macro: AC_UNISTD_H

Same as ‘AC_CHECK_HEADERS([unistd.h])’.

— Macro: AC_USG

Define USG if the BSD string functions are defined in strings.h. You should no longer depend upon USG, but on HAVE_STRING_H; see Standard Symbols.

— Macro: AC_UTIME_NULL

AC_FUNC_UTIME_NULL

— Macro: AC_VALIDATE_CACHED_SYSTEM_TUPLE ([cmd])

If the cache file is inconsistent with the current host, target and build system types, it used to execute cmd or print a default error message. This is now handled by default.

— Macro: AC_VERBOSE (result-description)

AC_MSG_RESULT.

— Macro: AC_VFORK

AC_FUNC_VFORK

— Macro: AC_VPRINTF

AC_FUNC_VPRINTF

— Macro: AC_WAIT3

AC_FUNC_WAIT3

— Macro: AC_WARN

AC_MSG_WARN

— Macro: AC_WORDS_BIGENDIAN

AC_C_BIGENDIAN

— Macro: AC_XENIX_DIR

This macro used to add -lx to output variable LIBS if on Xenix. Also, if dirent.h is being checked for, added -ldir to LIBS. Now it is merely an alias of AC_HEADER_DIRENT instead, plus some code to detect whether running xenix on which you should not depend:

          AC_MSG_CHECKING([for Xenix])
          AC_EGREP_CPP([yes],
          [#if defined M_XENIX && !defined M_UNIX
            yes
          #endif],
                       [AC_MSG_RESULT([yes]); XENIX=yes],
                       [AC_MSG_RESULT([no]); XENIX=])
     
— Macro: AC_YYTEXT_POINTER

AC_DECL_YYTEXT


Next: , Previous: Obsolete Macros, Up: Obsolete Constructs

17.5 Upgrading From Version 1

Autoconf version 2 is mostly backward compatible with version 1. However, it introduces better ways to do some things, and doesn't support some of the ugly things in version 1. So, depending on how sophisticated your configure.ac files are, you might have to do some manual work in order to upgrade to version 2. This chapter points out some problems to watch for when upgrading. Also, perhaps your configure scripts could benefit from some of the new features in version 2; the changes are summarized in the file NEWS in the Autoconf distribution.


Next: , Up: Autoconf 1

17.5.1 Changed File Names

If you have an aclocal.m4 installed with Autoconf (as opposed to in a particular package's source directory), you must rename it to acsite.m4. See autoconf Invocation.

If you distribute install.sh with your package, rename it to install-sh so make builtin rules don't inadvertently create a file called install from it. AC_PROG_INSTALL looks for the script under both names, but it is best to use the new name.

If you were using config.h.top, config.h.bot, or acconfig.h, you still can, but you have less clutter if you use the AH_ macros. See Autoheader Macros.


Next: , Previous: Changed File Names, Up: Autoconf 1

17.5.2 Changed Makefiles

Add ‘@CFLAGS@’, ‘@CPPFLAGS@’, and ‘@LDFLAGS@’ in your Makefile.in files, so they can take advantage of the values of those variables in the environment when configure is run. Doing this isn't necessary, but it's a convenience for users.

Also add ‘@configure_input@’ in a comment to each input file for AC_OUTPUT, so that the output files contain a comment saying they were produced by configure. Automatically selecting the right comment syntax for all the kinds of files that people call AC_OUTPUT on became too much work.

Add config.log and config.cache to the list of files you remove in distclean targets.

If you have the following in Makefile.in:

     prefix = /usr/local
     exec_prefix = $(prefix)

you must change it to:

     prefix = @prefix@
     exec_prefix = @exec_prefix@

The old behavior of replacing those variables without ‘@’ characters around them has been removed.


Next: , Previous: Changed Makefiles, Up: Autoconf 1

17.5.3 Changed Macros

Many of the macros were renamed in Autoconf version 2. You can still use the old names, but the new ones are clearer, and it's easier to find the documentation for them. See Obsolete Macros, for a table showing the new names for the old macros. Use the autoupdate program to convert your configure.ac to using the new macro names. See autoupdate Invocation.

Some macros have been superseded by similar ones that do the job better, but are not call-compatible. If you get warnings about calling obsolete macros while running autoconf, you may safely ignore them, but your configure script generally works better if you follow the advice that is printed about what to replace the obsolete macros with. In particular, the mechanism for reporting the results of tests has changed. If you were using echo or AC_VERBOSE (perhaps via AC_COMPILE_CHECK), your configure script's output looks better if you switch to AC_MSG_CHECKING and AC_MSG_RESULT. See Printing Messages. Those macros work best in conjunction with cache variables. See Caching Results.


Next: , Previous: Changed Macros, Up: Autoconf 1

17.5.4 Changed Results

If you were checking the results of previous tests by examining the shell variable DEFS, you need to switch to checking the values of the cache variables for those tests. DEFS no longer exists while configure is running; it is only created when generating output files. This difference from version 1 is because properly quoting the contents of that variable turned out to be too cumbersome and inefficient to do every time AC_DEFINE is called. See Cache Variable Names.

For example, here is a configure.ac fragment written for Autoconf version 1:

     AC_HAVE_FUNCS(syslog)
     case "$DEFS" in
     *-DHAVE_SYSLOG*) ;;
     *) # syslog is not in the default libraries.  See if it's in some other.
       saved_LIBS="$LIBS"
       for lib in bsd socket inet; do
         AC_CHECKING(for syslog in -l$lib)
         LIBS="-l$lib $saved_LIBS"
         AC_HAVE_FUNCS(syslog)
         case "$DEFS" in
         *-DHAVE_SYSLOG*) break ;;
         *) ;;
         esac
         LIBS="$saved_LIBS"
       done ;;
     esac

Here is a way to write it for version 2:

     AC_CHECK_FUNCS([syslog])
     if test $ac_cv_func_syslog = no; then
       # syslog is not in the default libraries.  See if it's in some other.
       for lib in bsd socket inet; do
         AC_CHECK_LIB([$lib], [syslog], [AC_DEFINE([HAVE_SYSLOG])
           LIBS="-l$lib $LIBS"; break])
       done
     fi

If you were working around bugs in AC_DEFINE_UNQUOTED by adding backslashes before quotes, you need to remove them. It now works predictably, and does not treat quotes (except back quotes) specially. See Setting Output Variables.

All of the Boolean shell variables set by Autoconf macros now use ‘yes’ for the true value. Most of them use ‘no’ for false, though for backward compatibility some use the empty string instead. If you were relying on a shell variable being set to something like 1 or ‘t’ for true, you need to change your tests.


Previous: Changed Results, Up: Autoconf 1

17.5.5 Changed Macro Writing

When defining your own macros, you should now use AC_DEFUN instead of define. AC_DEFUN automatically calls AC_PROVIDE and ensures that macros called via AC_REQUIRE do not interrupt other macros, to prevent nested ‘checking...’ messages on the screen. There's no actual harm in continuing to use the older way, but it's less convenient and attractive. See Macro Definitions.

You probably looked at the macros that came with Autoconf as a guide for how to do things. It would be a good idea to take a look at the new versions of them, as the style is somewhat improved and they take advantage of some new features.

If you were doing tricky things with undocumented Autoconf internals (macros, variables, diversions), check whether you need to change anything to account for changes that have been made. Perhaps you can even use an officially supported technique in version 2 instead of kludging. Or perhaps not.

To speed up your locally written feature tests, add caching to them. See whether any of your tests are of general enough usefulness to encapsulate them into macros that you can share.


Previous: Autoconf 1, Up: Obsolete Constructs

17.6 Upgrading From Version 2.13

The introduction of the previous section (see Autoconf 1) perfectly suits this section...

Autoconf version 2.50 is mostly backward compatible with version 2.13. However, it introduces better ways to do some things, and doesn't support some of the ugly things in version 2.13. So, depending on how sophisticated your configure.ac files are, you might have to do some manual work in order to upgrade to version 2.50. This chapter points out some problems to watch for when upgrading. Also, perhaps your configure scripts could benefit from some of the new features in version 2.50; the changes are summarized in the file NEWS in the Autoconf distribution.


Next: , Up: Autoconf 2.13

17.6.1 Changed Quotation

The most important changes are invisible to you: the implementation of most macros have completely changed. This allowed more factorization of the code, better error messages, a higher uniformity of the user's interface etc. Unfortunately, as a side effect, some construct which used to (miraculously) work might break starting with Autoconf 2.50. The most common culprit is bad quotation.

For instance, in the following example, the message is not properly quoted:

     AC_INIT
     AC_CHECK_HEADERS(foo.h, ,
       AC_MSG_ERROR(cannot find foo.h, bailing out))
     AC_OUTPUT

Autoconf 2.13 simply ignores it:

     $ autoconf-2.13; ./configure --silent
     creating cache ./config.cache
     configure: error: cannot find foo.h
     $

while Autoconf 2.50 produces a broken configure:

     $ autoconf-2.50; ./configure --silent
     configure: error: cannot find foo.h
     ./configure: exit: bad non-numeric arg `bailing'
     ./configure: exit: bad non-numeric arg `bailing'
     $

The message needs to be quoted, and the AC_MSG_ERROR invocation too!

     AC_INIT([Example], [1.0], [bug-example@example.org])
     AC_CHECK_HEADERS([foo.h], [],
       [AC_MSG_ERROR([cannot find foo.h, bailing out])])
     AC_OUTPUT

Many many (and many more) Autoconf macros were lacking proper quotation, including no less than... AC_DEFUN itself!

     $ cat configure.in
     AC_DEFUN([AC_PROG_INSTALL],
     [# My own much better version
     ])
     AC_INIT
     AC_PROG_INSTALL
     AC_OUTPUT
     $ autoconf-2.13
     autoconf: Undefined macros:
     ***BUG in Autoconf--please report*** AC_FD_MSG
     ***BUG in Autoconf--please report*** AC_EPI
     configure.in:1:AC_DEFUN([AC_PROG_INSTALL],
     configure.in:5:AC_PROG_INSTALL
     $ autoconf-2.50
     $


Next: , Previous: Changed Quotation, Up: Autoconf 2.13

17.6.2 New Macros

While Autoconf was relatively dormant in the late 1990s, Automake provided Autoconf-like macros for a while. Starting with Autoconf 2.50 in 2001, Autoconf provided versions of these macros, integrated in the AC_ namespace, instead of AM_. But in order to ease the upgrading via autoupdate, bindings to such AM_ macros are provided.

Unfortunately older versions of Automake (e.g., Automake 1.4) did not quote the names of these macros. Therefore, when m4 finds something like ‘AC_DEFUN(AM_TYPE_PTRDIFF_T, ...)’ in aclocal.m4, AM_TYPE_PTRDIFF_T is expanded, replaced with its Autoconf definition.

Fortunately Autoconf catches pre-AC_INIT expansions, and complains, in its own words:

     $ cat configure.ac
     AC_INIT([Example], [1.0], [bug-example@example.org])
     AM_TYPE_PTRDIFF_T
     $ aclocal-1.4
     $ autoconf
     aclocal.m4:17: error: m4_defn: undefined macro: _m4_divert_diversion
     aclocal.m4:17: the top level
     autom4te: m4 failed with exit status: 1
     $

Modern versions of Automake no longer define most of these macros, and properly quote the names of the remaining macros. If you must use an old Automake, do not depend upon macros from Automake as it is simply not its job to provide macros (but the one it requires itself):

     $ cat configure.ac
     AC_INIT([Example], [1.0], [bug-example@example.org])
     AM_TYPE_PTRDIFF_T
     $ rm aclocal.m4
     $ autoupdate
     autoupdate: `configure.ac' is updated
     $ cat configure.ac
     AC_INIT([Example], [1.0], [bug-example@example.org])
     AC_CHECK_TYPES([ptrdiff_t])
     $ aclocal-1.4
     $ autoconf
     $


Next: , Previous: New Macros, Up: Autoconf 2.13

17.6.3 Hosts and Cross-Compilation

Based on the experience of compiler writers, and after long public debates, many aspects of the cross-compilation chain have changed:


The relationship between build, host, and target have been cleaned up: the chain of default is now simply: target defaults to host, host to build, and build to the result of config.guess. Nevertheless, in order to ease the transition from 2.13 to 2.50, the following transition scheme is implemented. Do not rely on it, as it will be completely disabled in a couple of releases (we cannot keep it, as it proves to cause more problems than it cures).

They all default to the result of running config.guess, unless you specify either --build or --host. In this case, the default becomes the system type you specified. If you specify both, and they're different, configure enters cross compilation mode, so it doesn't run any tests that require execution.

Hint: if you mean to override the result of config.guess, prefer --build over --host. In the future, --host will not override the name of the build system type. Whenever you specify --host, be sure to specify --build too.


For backward compatibility, configure accepts a system type as an option by itself. Such an option overrides the defaults for build, host, and target system types. The following configure statement configures a cross toolchain that runs on NetBSD/alpha but generates code for GNU Hurd/sparc, which is also the build platform.
     ./configure --host=alpha-netbsd sparc-gnu

In Autoconf 2.13 and before, the variables build, host, and target had a different semantics before and after the invocation of AC_CANONICAL_BUILD etc. Now, the argument of --build is strictly copied into build_alias, and is left empty otherwise. After the AC_CANONICAL_BUILD, build is set to the canonicalized build type. To ease the transition, before, its contents is the same as that of build_alias. Do not rely on this broken feature.

For consistency with the backward compatibility scheme exposed above, when --host is specified but --build isn't, the build system is assumed to be the same as --host, and ‘build_alias’ is set to that value. Eventually, this historically incorrect behavior will go away.


The former scheme to enable cross-compilation proved to cause more harm than good, in particular, it used to be triggered too easily, leaving regular end users puzzled in front of cryptic error messages. configure could even enter cross-compilation mode only because the compiler was not functional. This is mainly because configure used to try to detect cross-compilation, instead of waiting for an explicit flag from the user.

Now, configure enters cross-compilation mode if and only if --host is passed.

That's the short documentation. To ease the transition between 2.13 and its successors, a more complicated scheme is implemented. Do not rely on the following, as it will be removed in the near future.

If you specify --host, but not --build, when configure performs the first compiler test it tries to run an executable produced by the compiler. If the execution fails, it enters cross-compilation mode. This is fragile. Moreover, by the time the compiler test is performed, it may be too late to modify the build-system type: other tests may have already been performed. Therefore, whenever you specify --host, be sure to specify --build too.

     ./configure --build=i686-pc-linux-gnu --host=m68k-coff

enters cross-compilation mode. The former interface, which consisted in setting the compiler to a cross-compiler without informing configure is obsolete. For instance, configure fails if it can't run the code generated by the specified compiler if you configure as follows:

     ./configure CC=m68k-coff-gcc


Next: , Previous: Hosts and Cross-Compilation, Up: Autoconf 2.13

17.6.4 AC_LIBOBJ vs. LIBOBJS

Up to Autoconf 2.13, the replacement of functions was triggered via the variable LIBOBJS. Since Autoconf 2.50, the macro AC_LIBOBJ should be used instead (see Generic Functions). Starting at Autoconf 2.53, the use of LIBOBJS is an error.

This change is mandated by the unification of the GNU Build System components. In particular, the various fragile techniques used to parse a configure.ac are all replaced with the use of traces. As a consequence, any action must be traceable, which obsoletes critical variable assignments. Fortunately, LIBOBJS was the only problem, and it can even be handled gracefully (read, “without your having to change something”).

There were two typical uses of LIBOBJS: asking for a replacement function, and adjusting LIBOBJS for Automake and/or Libtool.


As for function replacement, the fix is immediate: use AC_LIBOBJ. For instance:
     LIBOBJS="$LIBOBJS fnmatch.o"
     LIBOBJS="$LIBOBJS malloc.$ac_objext"

should be replaced with:

     AC_LIBOBJ([fnmatch])
     AC_LIBOBJ([malloc])

When used with Automake 1.10 or newer, a suitable value for LIBOBJDIR is set so that the LIBOBJS and LTLIBOBJS can be referenced from any Makefile.am. Even without Automake, arranging for LIBOBJDIR to be set correctly enables referencing LIBOBJS and LTLIBOBJS in another directory. The LIBOJBDIR feature is experimental.


Previous: AC_LIBOBJ vs LIBOBJS, Up: Autoconf 2.13

17.6.5 AC_FOO_IFELSE vs. AC_TRY_FOO

Since Autoconf 2.50, internal codes uses AC_PREPROC_IFELSE, AC_COMPILE_IFELSE, AC_LINK_IFELSE, and AC_RUN_IFELSE on one hand and AC_LANG_SOURCES, and AC_LANG_PROGRAM on the other hand instead of the deprecated AC_TRY_CPP, AC_TRY_COMPILE, AC_TRY_LINK, and AC_TRY_RUN. The motivations where:

In addition to the change of syntax, the philosophy has changed too: while emphasis was put on speed at the expense of accuracy, today's Autoconf promotes accuracy of the testing framework at, ahem..., the expense of speed.

As a perfect example of what is not to be done, here is how to find out whether a header file contains a particular declaration, such as a typedef, a structure, a structure member, or a function. Use AC_EGREP_HEADER instead of running grep directly on the header file; on some systems the symbol might be defined in another header file that the file you are checking includes.

As a (bad) example, here is how you should not check for C preprocessor symbols, either defined by header files or predefined by the C preprocessor: using AC_EGREP_CPP:

     AC_EGREP_CPP(yes,
     [#ifdef _AIX
       yes
     #endif
     ], is_aix=yes, is_aix=no)

The above example, properly written would (i) use AC_LANG_PROGRAM, and (ii) run the compiler:

     AC_COMPILE_IFELSE([AC_LANG_PROGRAM(
     [[#if !defined _AIX
      error: This isn't AIX!
     #endif
     ]])],
                        [is_aix=yes],
                        [is_aix=no])


Next: , Previous: Obsolete Constructs, Up: Top

18 Generating Test Suites with Autotest

     N.B.: This section describes an experimental feature which will
     be part of Autoconf in a forthcoming release.  Although we believe
     Autotest is stabilizing, this documentation describes an interface which
     might change in the future: do not depend upon Autotest without
     subscribing to the Autoconf mailing lists.

It is paradoxical that portable projects depend on nonportable tools to run their test suite. Autoconf by itself is the paragon of this problem: although it aims at perfectly portability, up to 2.13 its test suite was using DejaGNU, a rich and complex testing framework, but which is far from being standard on Posix systems. Worse yet, it was likely to be missing on the most fragile platforms, the very platforms that are most likely to torture Autoconf and exhibit deficiencies.

To circumvent this problem, many package maintainers have developed their own testing framework, based on simple shell scripts whose sole outputs are exit status values describing whether the test succeeded. Most of these tests share common patterns, and this can result in lots of duplicated code and tedious maintenance.

Following exactly the same reasoning that yielded to the inception of Autoconf, Autotest provides a test suite generation framework, based on M4 macros building a portable shell script. The suite itself is equipped with automatic logging and tracing facilities which greatly diminish the interaction with bug reporters, and simple timing reports.

Autoconf itself has been using Autotest for years, and we do attest that it has considerably improved the strength of the test suite and the quality of bug reports. Other projects are known to use some generation of Autotest, such as Bison, Free Recode, Free Wdiff, GNU Tar, each of them with different needs, and this usage has validated Autotest as a general testing framework.

Nonetheless, compared to DejaGNU, Autotest is inadequate for interactive tool testing, which is probably its main limitation.


Next: , Up: Using Autotest

18.1 Using an Autotest Test Suite


Next: , Up: Using an Autotest Test Suite

18.1.1 testsuite Scripts

Generating testing or validation suites using Autotest is rather easy. The whole validation suite is held in a file to be processed through autom4te, itself using GNU M4 under the scene, to produce a stand-alone Bourne shell script which then gets distributed. Neither autom4te nor GNU M4 are needed at the installer's end.

Each test of the validation suite should be part of some test group. A test group is a sequence of interwoven tests that ought to be executed together, usually because one test in the group creates data files than a later test in the same group needs to read. Complex test groups make later debugging more tedious. It is much better to keep only a few tests per test group. Ideally there is only one test per test group.

For all but the simplest packages, some file such as testsuite.at does not fully hold all test sources, as these are often easier to maintain in separate files. Each of these separate files holds a single test group, or a sequence of test groups all addressing some common functionality in the package. In such cases, testsuite.at merely initializes the validation suite, and sometimes does elementary health checking, before listing include statements for all other test files. The special file package.m4, containing the identification of the package, is automatically included if found.

A convenient alternative consists in moving all the global issues (local Autotest macros, elementary health checking, and AT_INIT invocation) into the file local.at, and making testsuite.at be a simple list of m4_include of sub test suites. In such case, generating the whole test suite or pieces of it is only a matter of choosing the autom4te command line arguments.

The validation scripts that Autotest produces are by convention called testsuite. When run, testsuite executes each test group in turn, producing only one summary line per test to say if that particular test succeeded or failed. At end of all tests, summarizing counters get printed. One debugging directory is left for each test group which failed, if any: such directories are named testsuite.dir/nn, where nn is the sequence number of the test group, and they include:

In the ideal situation, none of the tests fail, and consequently no debugging directory is left behind for validation.

It often happens in practice that individual tests in the validation suite need to get information coming out of the configuration process. Some of this information, common for all validation suites, is provided through the file atconfig, automatically created by AC_CONFIG_TESTDIR. For configuration informations which your testing environment specifically needs, you might prepare an optional file named atlocal.in, instantiated by AC_CONFIG_FILES. The configuration process produces atconfig and atlocal out of these two input files, and these two produced files are automatically read by the testsuite script.

Here is a diagram showing the relationship between files.

Files used in preparing a software package for distribution:

                     [package.m4] -->.
                                      \
     subfile-1.at ->.  [local.at] ---->+
         ...         \                  \
     subfile-i.at ---->-- testsuite.at -->-- autom4te* -->testsuite
         ...         /
     subfile-n.at ->'

Files used in configuring a software package:

                                          .--> atconfig
                                         /
     [atlocal.in] -->  config.status* --<
                                         \
                                          `--> [atlocal]

Files created during the test suite execution:

     atconfig -->.                    .--> testsuite.log
                  \                  /
                   >-- testsuite* --<
                  /                  \
     [atlocal] ->'                    `--> [testsuite.dir]


Previous: testsuite Scripts, Up: Using an Autotest Test Suite

18.1.2 Autotest Logs

When run, the test suite creates a log file named after itself, e.g., a test suite named testsuite creates testsuite.log. It contains a lot of information, usually more than maintainers actually need, but therefore most of the time it contains all that is needed:

command line arguments
A bad but unfortunately widespread habit consists of setting environment variables before the command, such as in ‘CC=my-home-grown-cc ./testsuite’. The test suite does not know this change, hence (i) it cannot report it to you, and (ii) it cannot preserve the value of CC for subsequent runs. Autoconf faced exactly the same problem, and solved it by asking users to pass the variable definitions as command line arguments. Autotest requires this rule, too, but has no means to enforce it; the log then contains a trace of the variables that were changed by the user.
ChangeLog excerpts
The topmost lines of all the ChangeLog files found in the source hierarchy. This is especially useful when bugs are reported against development versions of the package, since the version string does not provide sufficient information to know the exact state of the sources the user compiled. Of course, this relies on the use of a ChangeLog.
build machine
Running a test suite in a cross-compile environment is not an easy task, since it would mean having the test suite run on a machine build, while running programs on a machine host. It is much simpler to run both the test suite and the programs on host, but then, from the point of view of the test suite, there remains a single environment, host = build. The log contains relevant information on the state of the build machine, including some important environment variables.
tested programs
The absolute file name and answers to --version of the tested programs (see Writing testsuite.at, AT_TESTED).
configuration log
The contents of config.log, as created by configure, are appended. It contains the configuration flags and a detailed report on the configuration itself.


Next: , Previous: Using an Autotest Test Suite, Up: Using Autotest

18.2 Writing testsuite.at

The testsuite.at is a Bourne shell script making use of special Autotest M4 macros. It often contains a call to AT_INIT near its beginning followed by one call to m4_include per source file for tests. Each such included file, or the remainder of testsuite.at if include files are not used, contain a sequence of test groups. Each test group begins with a call to AT_SETUP, then an arbitrary number of shell commands or calls to AT_CHECK, and then completes with a call to AT_CLEANUP.

— Macro: AT_INIT ([name])

Initialize Autotest. Giving a name to the test suite is encouraged if your package includes several test suites. In any case, the test suite always displays the package name and version. It also inherits the package bug report address.

— Macro: AT_COPYRIGHT (copyright-notice)

State that, in addition to the Free Software Foundation's copyright on the Autotest macros, parts of your test suite are covered by copyright-notice.

The copyright-notice shows up in both the head of testsuite and in ‘testsuite --version’.

— Macro: AT_TESTED (executables)

Log the file name and answer to --version of each program in space-separated list executables. Several invocations register new executables, in other words, don't fear registering one program several times.

Autotest test suites rely on PATH to find the tested program. This avoids the need to generate absolute names of the various tools, and makes it possible to test installed programs. Therefore, knowing which programs are being exercised is crucial to understanding problems in the test suite itself, or its occasional misuses. It is a good idea to also subscribe foreign programs you depend upon, to avoid incompatible diagnostics.


— Macro: AT_SETUP (test-group-name)

This macro starts a group of related tests, all to be executed in the same subshell. It accepts a single argument, which holds a few words (no more than about 30 or 40 characters) quickly describing the purpose of the test group being started.

— Macro: AT_KEYWORDS (keywords)

Associate the space-separated list of keywords to the enclosing test group. This makes it possible to run “slices” of the test suite. For instance, if some of your test groups exercise some ‘foo’ feature, then using ‘AT_KEYWORDS(foo)’ lets you run ‘./testsuite -k foo’ to run exclusively these test groups. The title of the test group is automatically recorded to AT_KEYWORDS.

Several invocations within a test group accumulate new keywords. In other words, don't fear registering the same keyword several times in a test group.

— Macro: AT_CAPTURE_FILE (file)

If the current test group fails, log the contents of file. Several identical calls within one test group have no additional effect.

— Macro: AT_XFAIL_IF (shell-condition)

Determine whether the test is expected to fail because it is a known bug (for unsupported features, you should skip the test). shell-condition is a shell expression such as a test command; you can instantiate this macro many times from within the same test group, and one of the conditions is enough to turn the test into an expected failure.

— Macro: AT_CLEANUP

End the current test group.


— Macro: AT_DATA (file, contents)

Initialize an input data file with given contents. Of course, the contents have to be properly quoted between square brackets to protect against included commas or spurious M4 expansion. The contents ought to end with an end of line.

— Macro: AT_CHECK (commands, [status = ‘0], [stdout = ‘], [stderr = ‘], [run-if-fail], [run-if-pass])

Execute a test by performing given shell commands. These commands should normally exit with status, while producing expected stdout and stderr contents. If commands exit with status 77, then the whole test group is skipped. Otherwise, if this test fails, run shell commands run-if-fail or, if this test passes, run shell commands run-if-pass.

The commands must not redirect the standard output, nor the standard error.

If status, or stdout, or stderr is ‘ignore’, then the corresponding value is not checked.

The special value ‘expout’ for stdout means the expected output of the commands is the content of the file expout. If stdout is ‘stdout’, then the standard output of the commands is available for further tests in the file stdout. Similarly for stderr with ‘expout’ and ‘stderr’.


Next: , Previous: Writing testsuite.at, Up: Using Autotest

18.3 Running testsuite Scripts

Autotest test suites support the following arguments:

--help
-h
Display the list of options and exit successfully.
--version
-V
Display the version of the test suite and exit successfully.
--clean
-c
Remove all the files the test suite might have created and exit. Meant for clean Make targets.
--list
-l
List all the tests (or only the selection), including their possible keywords.

By default all tests are performed (or described with --list) in the default environment first silently, then verbosely, but the environment, set of tests, and verbosity level can be tuned:
variable=value
Set the environment variable to value. Use this rather than ‘FOO=foo ./testsuite’ as debugging scripts would then run in a different environment.

The variable AUTOTEST_PATH specifies the testing path to prepend to PATH. Relative directory names (not starting with ‘/’) are considered to be relative to the top level of the package being built. All directories are made absolute, first starting from the top level build tree, then from the source tree. For instance ‘./testsuite AUTOTEST_PATH=tests:bin’ for a /src/foo-1.0 source package built in /tmp/foo results in ‘/tmp/foo/tests:/tmp/foo/bin’ and then ‘/src/foo-1.0/tests:/src/foo-1.0/bin’ being prepended to PATH.

number
number-number
number-
-number
Add the corresponding test groups, with obvious semantics, to the selection.
--keywords=keywords
-k keywords
Add to the selection the test groups with title or keywords (arguments to AT_SETUP or AT_KEYWORDS) that match all keywords of the comma separated list keywords, case-insensitively. Use ‘!’ immediately before the keyword to invert the selection for this keyword. By default, the keywords match whole words; enclose them in ‘.*’ to also match parts of words.

For example, running

          ./testsuite -k 'autoupdate,.*FUNC.*'
     

selects all tests tagged ‘autoupdateand with tags containing ‘FUNC’ (as in ‘AC_CHECK_FUNC’, ‘AC_FUNC_FNMATCH’, etc.), while

          ./testsuite -k '!autoupdate' -k '.*FUNC.*'
     

selects all tests not tagged ‘autoupdateor with tags containing ‘FUNC’.

--errexit
-e
If any test fails, immediately abort testing. It implies --debug: post test group clean up, and top-level logging are inhibited. This option is meant for the full test suite, it is not really useful for generated debugging scripts.
--verbose
-v
Force more verbosity in the detailed output of what is being done. This is the default for debugging scripts.
--debug
-d
Do not remove the files after a test group was performed —but they are still removed before, therefore using this option is sane when running several test groups. Create debugging scripts. Do not overwrite the top-level log (in order to preserve supposedly existing full log file). This is the default for debugging scripts, but it can also be useful to debug the testsuite itself.
--trace
-x
Trigger shell tracing of the test groups.


Previous: testsuite Invocation, Up: Using Autotest

18.4 Making testsuite Scripts

For putting Autotest into movement, you need some configuration and makefile machinery. We recommend, at least if your package uses deep or shallow hierarchies, that you use tests/ as the name of the directory holding all your tests and their makefile. Here is a check list of things to do.

With Automake, here is a minimal example about how to link ‘make check’ with a validation suite.

     EXTRA_DIST = testsuite.at $(TESTSUITE) atlocal.in
     TESTSUITE = $(srcdir)/testsuite
     
     check-local: atconfig atlocal $(TESTSUITE)
             $(SHELL) '$(TESTSUITE)' $(TESTSUITEFLAGS)
     
     installcheck-local: atconfig atlocal $(TESTSUITE)
             $(SHELL) '$(TESTSUITE)' AUTOTEST_PATH='$(bindir)' \
               $(TESTSUITEFLAGS)
     
     clean-local:
             test ! -f '$(TESTSUITE)' || \
              $(SHELL) '$(TESTSUITE)' --clean
     
     AUTOTEST = $(AUTOM4TE) --language=autotest
     $(TESTSUITE): $(srcdir)/testsuite.at
             $(AUTOTEST) -I '$(srcdir)' -o $@.tmp $@.at
             mv $@.tmp $@

You might want to list explicitly the dependencies, i.e., the list of the files testsuite.at includes.

With strict Autoconf, you might need to add lines inspired from the following:

     subdir = tests
     
     atconfig: $(top_builddir)/config.status
             cd $(top_builddir) && \
                $(SHELL) ./config.status $(subdir)/$@
     
     atlocal: $(srcdir)/atlocal.in $(top_builddir)/config.status
             cd $(top_builddir) && \
                $(SHELL) ./config.status $(subdir)/$@

and manage to have atconfig.in and $(EXTRA_DIST) distributed.

With all this in place, and if you have not initialized ‘TESTSUITEFLAGS’ within your makefile, you can fine-tune test suite execution with this variable, for example:

     make check TESTSUITEFLAGS='-v -d -x 75 -k AC_PROG_CC CFLAGS=-g'


Next: , Previous: Using Autotest, Up: Top

19 Frequent Autoconf Questions, with answers

Several questions about Autoconf come up occasionally. Here some of them are addressed.


Next: , Up: FAQ

19.1 Distributing configure Scripts

     What are the restrictions on distributing configure
     scripts that Autoconf generates?  How does that affect my
     programs that use them?

There are no restrictions on how the configuration scripts that Autoconf produces may be distributed or used. In Autoconf version 1, they were covered by the GNU General Public License. We still encourage software authors to distribute their work under terms like those of the GPL, but doing so is not required to use Autoconf.

Of the other files that might be used with configure, config.h.in is under whatever copyright you use for your configure.ac. config.sub and config.guess have an exception to the GPL when they are used with an Autoconf-generated configure script, which permits you to distribute them under the same terms as the rest of your package. install-sh is from the X Consortium and is not copyrighted.


Next: , Previous: Distributing, Up: FAQ

19.2 Why Require GNU M4?

     Why does Autoconf require GNU M4?

Many M4 implementations have hard-coded limitations on the size and number of macros that Autoconf exceeds. They also lack several builtin macros that it would be difficult to get along without in a sophisticated application like Autoconf, including:

     m4_builtin
     m4_indir
     m4_bpatsubst
     __file__
     __line__

Autoconf requires version 1.4.4 or later of GNU M4.

Since only software maintainers need to use Autoconf, and since GNU M4 is simple to configure and install, it seems reasonable to require GNU M4 to be installed also. Many maintainers of GNU and other free software already have most of the GNU utilities installed, since they prefer them.


Next: , Previous: Why GNU M4, Up: FAQ

19.3 How Can I Bootstrap?

     If Autoconf requires GNU M4 and GNU M4 has an Autoconf
     configure script, how do I bootstrap?  It seems like a chicken
     and egg problem!

This is a misunderstanding. Although GNU M4 does come with a configure script produced by Autoconf, Autoconf is not required in order to run the script and install GNU M4. Autoconf is only required if you want to change the M4 configure script, which few people have to do (mainly its maintainer).


Next: , Previous: Bootstrapping, Up: FAQ

19.4 Why Not Imake?

     Why not use Imake instead of configure scripts?

Several people have written addressing this question, so I include adaptations of their explanations here.

The following answer is based on one written by Richard Pixley:

Autoconf generated scripts frequently work on machines that it has never been set up to handle before. That is, it does a good job of inferring a configuration for a new system. Imake cannot do this.

Imake uses a common database of host specific data. For X11, this makes sense because the distribution is made as a collection of tools, by one central authority who has control over the database.

GNU tools are not released this way. Each GNU tool has a maintainer; these maintainers are scattered across the world. Using a common database would be a maintenance nightmare. Autoconf may appear to be this kind of database, but in fact it is not. Instead of listing host dependencies, it lists program requirements.

If you view the GNU suite as a collection of native tools, then the problems are similar. But the GNU development tools can be configured as cross tools in almost any host+target permutation. All of these configurations can be installed concurrently. They can even be configured to share host independent files across hosts. Imake doesn't address these issues.

Imake templates are a form of standardization. The GNU coding standards address the same issues without necessarily imposing the same restrictions.

Here is some further explanation, written by Per Bothner:

One of the advantages of Imake is that it easy to generate large makefiles using the ‘#include’ and macro mechanisms of cpp. However, cpp is not programmable: it has limited conditional facilities, and no looping. And cpp cannot inspect its environment.

All of these problems are solved by using sh instead of cpp. The shell is fully programmable, has macro substitution, can execute (or source) other shell scripts, and can inspect its environment.

Paul Eggert elaborates more:

With Autoconf, installers need not assume that Imake itself is already installed and working well. This may not seem like much of an advantage to people who are accustomed to Imake. But on many hosts Imake is not installed or the default installation is not working well, and requiring Imake to install a package hinders the acceptance of that package on those hosts. For example, the Imake template and configuration files might not be installed properly on a host, or the Imake build procedure might wrongly assume that all source files are in one big directory tree, or the Imake configuration might assume one compiler whereas the package or the installer needs to use another, or there might be a version mismatch between the Imake expected by the package and the Imake supported by the host. These problems are much rarer with Autoconf, where each package comes with its own independent configuration processor.

Also, Imake often suffers from unexpected interactions between make and the installer's C preprocessor. The fundamental problem here is that the C preprocessor was designed to preprocess C programs, not makefiles. This is much less of a problem with Autoconf, which uses the general-purpose preprocessor M4, and where the package's author (rather than the installer) does the preprocessing in a standard way.

Finally, Mark Eichin notes:

Imake isn't all that extensible, either. In order to add new features to Imake, you need to provide your own project template, and duplicate most of the features of the existing one. This means that for a sophisticated project, using the vendor-provided Imake templates fails to provide any leverage—since they don't cover anything that your own project needs (unless it is an X11 program).

On the other side, though:

The one advantage that Imake has over configure: Imakefile files tend to be much shorter (likewise, less redundant) than Makefile.in files. There is a fix to this, however—at least for the Kerberos V5 tree, we've modified things to call in common post.in and pre.in makefile fragments for the entire tree. This means that a lot of common things don't have to be duplicated, even though they normally are in configure setups.


Next: , Previous: Why Not Imake, Up: FAQ

19.5 How Do I #define Installation Directories?

     My program needs library files, installed in datadir and
     similar.  If I use
     
          AC_DEFINE_UNQUOTED([DATADIR], [$datadir],
            [Define to the read-only architecture-independent
             data directory.])
     

I get

          #define DATADIR "${prefix}/share"
     

As already explained, this behavior is on purpose, mandated by the GNU Coding Standards, see Installation Directory Variables. There are several means to achieve a similar goal:


Next: , Previous: Defining Directories, Up: FAQ

19.6 What is autom4te.cache?

     What is this directory autom4te.cache?  Can I safely remove it?

In the GNU Build System, configure.ac plays a central role and is read by many tools: autoconf to create configure, autoheader to create config.h.in, automake to create Makefile.in, autoscan to check the completeness of configure.ac, autoreconf to check the GNU Build System components that are used. To “read configure.ac” actually means to compile it with M4, which can be a long process for complex configure.ac.

This is why all these tools, instead of running directly M4, invoke autom4te (see autom4te Invocation) which, while answering to a specific demand, stores additional information in autom4te.cache for future runs. For instance, if you run autoconf, behind the scenes, autom4te also stores information for the other tools, so that when you invoke autoheader or automake etc., reprocessing configure.ac is not needed. The speed up is frequently of 30%, and is increasing with the size of configure.ac.

But it is and remains being simply a cache: you can safely remove it.


     Can I permanently get rid of it?

The creation of this cache can be disabled from ~/.autom4te.cfg, see Customizing autom4te, for more details. You should be aware that disabling the cache slows down the Autoconf test suite by 40%. The more GNU Build System components are used, the more the cache is useful; for instance running ‘autoreconf -f’ on the Core Utilities is twice slower without the cache although --force implies that the cache is not fully exploited, and eight times slower than without --force.


Previous: autom4te.cache, Up: FAQ

19.7 Header Present But Cannot Be Compiled

The most important guideline to bear in mind when checking for features is to mimic as much as possible the intended use. Unfortunately, old versions of AC_CHECK_HEADER and AC_CHECK_HEADERS failed to follow this idea, and called the preprocessor, instead of the compiler, to check for headers. As a result, incompatibilities between headers went unnoticed during configuration, and maintainers finally had to deal with this issue elsewhere.

As of Autoconf 2.56 both checks are performed, and configure complains loudly if the compiler and the preprocessor do not agree. For the time being the result used is that of the preprocessor, to give maintainers time to adjust their configure.ac, but in the future, only the compiler will be considered.

Consider the following example:

     $ cat number.h
     typedef int number;
     $ cat pi.h
     const number pi = 3;
     $ cat configure.ac
     AC_INIT([Example], [1.0], [bug-example@example.org])
     AC_CHECK_HEADERS([pi.h])
     $ autoconf -Wall
     $ ./configure
     checking for gcc... gcc
     checking for C compiler default output file name... a.out
     checking whether the C compiler works... yes
     checking whether we are cross compiling... no
     checking for suffix of executables...
     checking for suffix of object files... o
     checking whether we are using the GNU C compiler... yes
     checking whether gcc accepts -g... yes
     checking for gcc option to accept ISO C89... none needed
     checking how to run the C preprocessor... gcc -E
     checking for grep that handles long lines and -e... grep
     checking for egrep... grep -E
     checking for ANSI C header files... yes
     checking for sys/types.h... yes
     checking for sys/stat.h... yes
     checking for stdlib.h... yes
     checking for string.h... yes
     checking for memory.h... yes
     checking for strings.h... yes
     checking for inttypes.h... yes
     checking for stdint.h... yes
     checking for unistd.h... yes
     checking pi.h usability... no
     checking pi.h presence... yes
     configure: WARNING: pi.h: present but cannot be compiled
     configure: WARNING: pi.h:     check for missing prerequisite headers?
     configure: WARNING: pi.h: see the Autoconf documentation
     configure: WARNING: pi.h:     section "Present But Cannot Be Compiled"
     configure: WARNING: pi.h: proceeding with the preprocessor's result
     configure: WARNING: pi.h: in the future, the compiler will take precedence
     configure: WARNING:     ## -------------------------------------- ##
     configure: WARNING:     ## Report this to bug-example@example.org ##
     configure: WARNING:     ## -------------------------------------- ##
     checking for pi.h... yes

The proper way the handle this case is using the fourth argument (see Generic Headers):

     $ cat configure.ac
     AC_INIT([Example], [1.0], [bug-example@example.org])
     AC_CHECK_HEADERS([number.h pi.h], [], [],
     [[#if HAVE_NUMBER_H
     # include <number.h>
     #endif
     ]])
     $ autoconf -Wall
     $ ./configure
     checking for gcc... gcc
     checking for C compiler default output... a.out
     checking whether the C compiler works... yes
     checking whether we are cross compiling... no
     checking for suffix of executables...
     checking for suffix of object files... o
     checking whether we are using the GNU C compiler... yes
     checking whether gcc accepts -g... yes
     checking for gcc option to accept ANSI C... none needed
     checking for number.h... yes
     checking for pi.h... yes

See Particular Headers, for a list of headers with their prerequisite.


Next: , Previous: FAQ, Up: Top

20 History of Autoconf

You may be wondering, Why was Autoconf originally written? How did it get into its present form? (Why does it look like gorilla spit?) If you're not wondering, then this chapter contains no information useful to you, and you might as well skip it. If you are wondering, then let there be light...


Next: , Up: History

20.1 Genesis

In June 1991 I was maintaining many of the GNU utilities for the Free Software Foundation. As they were ported to more platforms and more programs were added, the number of -D options that users had to select in the makefile (around 20) became burdensome. Especially for me—I had to test each new release on a bunch of different systems. So I wrote a little shell script to guess some of the correct settings for the fileutils package, and released it as part of fileutils 2.0. That configure script worked well enough that the next month I adapted it (by hand) to create similar configure scripts for several other GNU utilities packages. Brian Berliner also adapted one of my scripts for his CVS revision control system.

Later that summer, I learned that Richard Stallman and Richard Pixley were developing similar scripts to use in the GNU compiler tools; so I adapted my configure scripts to support their evolving interface: using the file name Makefile.in as the templates; adding ‘+srcdir’, the first option (of many); and creating config.status files.


Next: , Previous: Genesis, Up: History

20.2 Exodus

As I got feedback from users, I incorporated many improvements, using Emacs to search and replace, cut and paste, similar changes in each of the scripts. As I adapted more GNU utilities packages to use configure scripts, updating them all by hand became impractical. Rich Murphey, the maintainer of the GNU graphics utilities, sent me mail saying that the configure scripts were great, and asking if I had a tool for generating them that I could send him