[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5 Option Definitions

AutoOpts uses an AutoGen definitions file for the definitions of the program options and overall configuration attributes. The complete list of program and option attributes is quite extensive, so if you are reading to understand how to use AutoOpts, I recommend reading the "Quick Start" section (see section Quick Start) and paying attention to the following:

  1. prog-name, prog-title, and argument, program attributes, See section Program Description Attributes.
  2. name and descrip option attributes, See section Required Attributes.
  3. value (flag character) and min (occurrence counts) option attributes, See section Common Option Attributes.
  4. arg-type from the option argument specification section, See section Option Argument Specification.
  5. Read the overall how to, See section Using AutoOpts.
  6. Highly recommended, but not required, are the several "man" and "info" documentation attributes, See section Man and Info doc Attributes.

Keep in mind that the majority are rarely used and can be safely ignored. However, when you have special option processing requirements, the flexibility is there.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.1 Program Description Attributes

The following global definitions are used to define attributes of the entire program. These generally alter the configuration or global behavior of the AutoOpts option parser. The first two are required of every program. The third is required if there are to be any left over arguments (operands) after option processing. The rest have been grouped below. Except as noted, there may be only one copy of each of these definitions:


This attribute is required. Variable names derived from this name are derived using string->c_name! (see section string->c-name!’ - map non-name chars to underscore).


This attribute is required and may be any descriptive text.


This attribute is required if your program uses operand arguments. It specifies the syntax of the arguments that follow the options. It may not be empty, but if it is not supplied, then option processing must consume all the arguments. If it is supplied and starts with an open bracket ([), then there is no requirement on the presence or absence of command line arguments following the options. Lastly, if it is supplied and does not start with an open bracket, then option processing must not consume all of the command line arguments.


If your build has a configuration header, it must be included before anything else. Specifying the configuration header file name with this attribute will cause that to happen.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Usage and Version Info Display

These will affect the way usage is seen and whether or not version information gets displayed.


If this attribute is provided, it may specify the full length usage text, or a variable name assignable to a char const * pointer, or it may be empty. The meanings are determined by the length.

This string should be readily translatable. Provision will be made to translate it if this is provided, if the source code is compiled with ENABLE_NLS defined, and no-xlate has not been set to the value anything. The untranslated text will be handed to dgettext("libopts", txt) and then gettext(txt) for translation, one paragraph at a time.

To facilitate the creation and maintenance of this text, you can force the string to be ignored and recomputed by specifying


in the environment and requesting help or usage information. See See section Developer and User Notes.


If this attribute is provided, it is used to specify an abbreviated version of the usage text. This text is constructed in the same way as the full-usage, described above.


AutoOpts normaly displays usage text in a format that provides more information than the standard GNU layout, but that also means it is not the standard GNU layout. This attribute changes the default to GNU layout, with the AUTOOPTS_USAGE environment variable used to request autoopts layout. See See section Developer and User Notes.


I apologize for too many confusing usages of usage. This attribute specifies that ‘--usage’ and/or ‘-u’ be supported. The help (usage) text displayed will be abbreviated when compared to the default help text.


When there is a command line syntax error, by default AutoOpts will display the abbreviated usage text, rather than just a one line “you goofed it, ask for usage” message. You can change the default behavior for your program by supplying this attribute. The user may override this choice, again, with the AUTOOPTS_USAGE environment variable. See See section Developer and User Notes.


The version text in the ‘getopt.tpl’ template will include this text in parentheses after the program name, when this attribute is specified. For example:

mumble (stumble) 1.0

says that the ‘mumble’ program is version 1.0 and is part of the ‘stumble’ group of programs.


If your program has some cleanup work that must be done before exiting on usage mode issues, or if you have to customize the usage message in some way, specify this procedure and it will be called instead of the default optionUsage() function. For example, if a program is using the curses library and needs to invoke the usage display, then you must arrange to call endwin() before invoking the library function optionUsage(). This can be handled by specifying your own usage function, thus:

my_usage(tOptions * opts, int ex)
    if (curses_window_active)
    optionUsage(opts, ex);

Specifies the program version and activates the VERSION option, See section Automatically Supported Options.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Program Configuration

Programs may be “pre-configured” before normal command line options are processed (See see section Immediate Action Attributes). How configuration files and environment variables are handled get specified with these attributes.


Indicates that the command line usage of ‘--load-opts’ and/or ‘--save-opts’ are disallowed.


Indicates looking in the environment for values of variables named, PROGRAM_OPTNAME or PROGRAM, where PROGRAM is the upper cased C-name of the program and ‘OPTNAME’ is the upper cased C-name of a specific option. The contents of the PROGRAM variable, if found, are tokenized and processed. The contents of PROGRAM_OPTNAME environment variables are taken as the option argument to the option nameed ‘--optname’.


Specifies that option settings may be loaded from and stored into configuration files. Each instance of this attribute is either a directory or a file using a specific path, a path based on an environment variable or a path relative to installation directories. The method used depends on the name. If the one entry is empty, it enables the loading and storing of settings, but no specific files are searched for. Otherwise, a series of configuration files are hunted down and, if found, loaded.

If the first character of the ‘homerc’ value is not the dollar character ($), then it is presumed to be a path name based on the current directory. Otherwise, the method depends on the second character:


The path is relative to the directory where the executable was found.


The path is relative to the package data directory, e.g. ‘/usr/local/share/autogen’.


The path is derived from the named environment variable.

Use as many as you like. The presence of this attribute activates the ‘--save-opts’ and ‘--load-opts’ options. However, saving into a file may be disabled with the ‘disable-save’. See section configuration file presets. See the optionMakePath(3AGEN) man page for excruciating details.


Specifies the configuration file name. This is only useful if you have provided at least one homerc attribute.

default: .<prog-name>rc

This option implements the ‘-W’ vendor option command line option.

For POSIX specified utilities, the options are constrained to the options that are specified by POSIX. Extensions should be handled with ‘-W’ command line options, the short flag form. Long option name processing must be disabled. In fact, the long-opts attribute must not be provided, and some options must be specified without flag values.

The ‘-W long-name’ is processed by looking up the long option name that follows it. It cannot be a short flag because that would conflict with the POSIX flag name space. It will be processed as if long options were accepted and ‘--long-name’ were found on the command line.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Programming Details

These attributes affect some of the ways that the option data are used and made available to the program.


The contents of this attribute should be just the name of the configuration file. A "#include" naming this file will be inserted at the top of the generated header.


These values should be defined as indexed values, thus:

exit-name[0] = success;
exit-desc[0] = 'Successful program execution.';
exit-name[1] = failure;
exit-desc[1] = 'The operation failed or command syntax was not valid.';

By default, all programs have these effectively defined for them. They may be overridden by explicitly defining any or all of these values. Additional names and descriptions may be defined. They will cause an enumeration to be emitted, like this one for getdefs:

typedef enum {
} getdefs_exit_code_t;

which will be augmented by any exit-name definitions beyond ‘1’.

Some of the generated code will exit non-zero if there is an allocation error. This exit will always be code ‘1’, unless there is an exit named ‘no_mem’ or ‘nomem’. In that case, that value will be used. Additionally, if there is such a value, and if die-code is specified, then a function nomem_err(size_t len, char const * what) will be emitted as an inline function for reporting out-of-memory conditions.


This attribute will cause two procedures to be added to the code file: usage_message() and vusage_message(), with any applicable prefix (see prefix, below). They are declared in the generated header, thus:

extern void vusage_message(char const * fmt, va_list ap);
extern void usage_message(char const * fmt, ...);

These functions print the message to ‘stderr’ and invoke the usage function with the exit code set to 1 (EXIT_FAILURE).


This tells AutoOpts templates to emit code for vdie(), die(), fserr(), and, possibly the nomem_err() functions. The latter is emitted if an exit name of ‘no-mem’ or ‘nomem’ is specified. If the die-code is assigned a text value, then that code will be inserted in the vdie function immediately before it prints the death rattle message.

The profiles for these functions are:

extern void vdie( int exit_code, char const * fmt, va_list);
extern void die(  int exit_code, char const * fmt, ...);
extern void fserr(int exit_code, char const * op, char const * fname);
static inline void
nomem_err(size_t sz, char const * what) {...}

If usage-message or die-code are specified, you may also specify that the generated functions are marked as “noreturn” with this attribute. If this attribute is not empty, then the specified string is used instead of “noreturn”. If “noreturn” has not been defined before these functions are declared, then it will be “#define”-d to the empty string. No such protection is made for any non-default value. These functions will be declared “extern noreturn void”.


This string is inserted into the .h interface file. Generally used for global variables or #include directives required by flag-code text and shared with other program text. Do not specify your configuration header (‘config.h’) in this attribute or the include attribute, however. Instead, use config-header, above.


AutoOpts generates macros that presume that there are no cpp macros with the same name as the option name. For example, if you have an option named, ‘--debug’, then you must not use #ifdef DEBUG in your code. If you specify this attribute, every option name will be guarded. If the name is #define-d, then a warning will be issued and the name undefined. If you do not specify this and there is a conflict, you will get strange error messages.

This attribute may be set to any of four recognized states:


This string is inserted into the .c file. Generally used for global variables required only by flag-code program text.


If you are going to handle your option processing with the ‘getopt.tpl’ template instead of using libopts, then specify this attribute. It will suppress mention of ‘--more-help’ in the generated documentation. (getopt_long does not support ‘--more-help’.)


This value is inserted into all global names. This will disambiguate them if more than one set of options are to be compiled into a single program.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] User Presentation Attributes

Attributes that affect the user’s experience.


The presence of this attribute indicates ignoring any command line option errors. This may also be turned on and off by invoking the macros ERRSKIP_OPTERR and ERRSTOP_OPTERR from the generated interface file.


Presence indicates GNU-standard long option processing. Partial name matches are accepted, if they are at least two characters long and the partial match is unique. The matching is not case sensitive, and the underscore, hyphen and carat characters are all equivalent (they match).

If any options do not have an option value (flag character) specified, and least one does specify such a value, then you must specify long-opts. If none of your options specify an option value (flag character) and you do not specify long-opts, then command line arguments are processed in "named option mode". This means that:


Modifies when or whether option names get translated. If provided, it must be assigned one of these values:


to suppress option name translation for configuration file and and environment variable processing.


to suppress option name translation completely. The usage text will always be translated if ENABLE_NLS is defined and you have translations for that text.


Specifies disabling all internationalization support for option code, completely.

See also the various XLAT interface entries in the AutoOpts Programmatic Interface section (see section Programmatic Interface).


Normally, POSIX compliant commands do not allow for options to be interleaved with operands. If this is necessary for historical reasons, there are two approaches available:


Specifies that the ‘--reset-option’ command line option is to be supported. This makes it possible to suppress any setting that might be found in a configuration file or environment variable.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.2 Options for Library Code

Some libraries provide their own code for processing command line options, and this may be used by programs that utilize AutoOpts. You may also wish to write a library that gets configured with AutoOpts options and config files. Such a library may either supply its own configury routine and process its own options, or it may export its option descriptions to programs that also use AutoOpts. This section will describe how to do all of these different things.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] AutoOpt-ed Library for AutoOpt-ed Program

The library source code must provide an option definition file that consists of only the attribute library and flag entries. The library attribute does not need any associated value, so it will generally appeary by itself on a line folowed by a semi-colon. The first flag entry must contain the following attributes:


This name is used in the construction of a global pointer of type tOptDesc const*. It is always required.


It tells AutoOpts that this option serves no normal purpose. It will be used to add usage clarity and to locate option descriptors in the library code.


This is a string that is inserted in the extended usage display before the options specific to the current library. It is always required.


This should match the name of the library. This string is also used in the construction of the option descriptor pointer name. In the end, it looks like this:

extern tOptDesc const* <<lib-name>>_<<name>>_optDesc_p;

and is used in the macros generated for the library’s ‘.h’ file.

In order to compile this AutoOpts using library, you must create a special header that is not used by the client program. This is accomplished by creating an option definition file that contains essentially exactly the following:

AutoGen definitions options;
prog-name  = does-not-matter;  // but is always required
prog-title = 'also does not matter';  // also required
config-header = 'config.h'; // optional, but common
#include library-options-only.def

and nothing else. AutoGen will produce only the ‘.h’ file. You may now compile your library, referencing just this ‘.h’ file. The macros it creates will utilize a global variable that will be defined by the AutoOpts-using client program. That program will need to have the following #include in its option definition file:

#include library-options-only.def

All the right things will magically happen so that the global variables named <<lib-name>>_<<name>>_optDesc_p are initialized correctly. For an example, please see the AutoOpts test script: ‘autoopts/test/library.test’.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] AutoOpt-ed Library for Regular Program

In this case, your library must provide an option processing function to a calling program. This is accomplished by setting the allow-errors global option attribute. Each time your option handling function is called, you must determine where your scan is to resume and tell the AutoOpts library by invoking:


and then invoke not_opt_index = optionProcess(...). The not_opt_index value can be used to set optind, if that is the global being used to scan the program argument array.

In this method, do NOT utilize the global library attribute. Your library must specify its options as if it were a complete program. You may choose to specify an alternate usage() function so that usage for other parts of the option interface may be displayed as well. See “Program Information Attributes” (see section Program Information Attributes).

At the moment, there is no method for calling optionUsage() telling it to produce just the information about the options and not the program as a whole. Some later revision after somebody asks.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] AutoOpt-ed Program Calls Regular Library

As with providing an AutoOpt-ed library to a non-AutoOpt-ed program, you must write the option description file as if you were writing all the options for the program, but you should specify the allow-errors global option attribute and you will likely want an alternate usage() function (see “Program Information Attributes” see section Program Information Attributes). In this case, though, when optionProcess() returns, you need to test to see if there might be library options. If there might be, then call the library’s exported routine for handling command line options, set the next-option-to-process with the RESTART_OPT() macro, and recall optionProcess(). Repeat until done.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.3 Program Information Attributes

These attributes are used to define how and what information is displayed to the user of the program.


The copyright is a structured value containing three to five values. If copyright is used, then the first three are required.

  1. date’ - the list of applicable dates for the copyright.
  2. owner’ - the name of the copyright holder.
  3. type’ - specifies the type of distribution license. AutoOpts/AutoGen supports the text of the GNU Public License (‘gpl’), the GNU Lesser General Public License with Library extensions (‘lgpl’), the Modified Free BSD license (‘mbsd’) and a few others. Other licenses may be specified, but you must provide your own license file. The list of license files provided by AutoOpts may be seen by typing:
    ls $(autoopts-config pkgdatadir)/*.lic
  4. text’ - the text of the copyright notice. This must be provided if ‘type’ is set to ‘NOTE’.
  5. author’ - in case the author name is to appear in the documentation and is different from the copyright owner.
  6. eaddr’ - email address for receiving praises and complaints. Typically that of the author or copyright holder.

An example of this might be:

copyright = {
    date  = "1992-2015";
    owner = "Bruce Korb";
    eaddr = 'bkorb@gnu.org';
    type  = GPL;

This string is added to the usage output when the HELP option is selected.


Gives additional information whenever the usage routine is invoked.


The name of the package the program belongs to. This will appear parenthetically after the program name in the version and usage output, e.g.: autogen (GNU autogen) - The Automated Program Generator.


This attribute will not change anything except appearance. Normally, the option names are all documented in lower case. However, if you specify this attribute, then they will display in the case used in their specification. Command line options will still be matched without case sensitivity. This is useful for specifying option names in camel-case.

prog-desc and

These define global pointer variables that point to the program descriptor and the first option descriptor for a library option. This is intended for use by certain libraries that need command line and/or initialization file option processing. These definitions have no effect on the option template output, but are used for creating a library interface file. Normally, the first "option" for a library will be a documentation option that cannot be specified on the command line, but is marked as settable. The library client program will invoke the SET_OPTION macro which will invoke a handler function that will finally set these global variables.


Optionally names the usage procedure, if the library routine optionUsage() does not work for you. If you specify my_usage as the value of this attribute, for example, you will use a procedure by that name for displaying usage. Of course, you will need to provide that procedure and it must conform to this profile:

void my_usage( tOptions* pOptions, int exitCode )

Normally, the default format produced by the optionUsage procedure is AutoOpts Standard. By specifying this attribute, the default format will be GNU-ish style. Either default may be overridden by the user with the AUTOOPTS_USAGE environment variable. If it is set to gnu or autoopts, it will alter the style appropriately. This attribute will conflict with the usage attribute.


Some applications traditionally require that the command operands be intermixed with the command options. In order to handle that, the arguments must be reordered. If you are writing such an application, specify this global option. All of the options (and any associated option arguments) will be brought to the beginning of the argument list. New applications should not use this feature, if at all possible. This feature is disabled if POSIXLY_CORRECT is defined in the environment.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.4 Generating main procedures

When AutoOpts generates the code to parse the command line options, it has the ability to produce any of several types of main() procedures. This is done by specifying a global structured value for main. The values that it contains are dependent on the value set for the one value it must have: main-type.

The recognized values for main-type are guile, shell-process, shell-parser, main, include, invoke, and for-each.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] guile: main and inner_main procedures

When the main-type is specified to be guile, a main() procedure is generated that calls gh_enter(), providing it with a generated inner_main() to invoke. If you must perform certain tasks before calling gh_enter(), you may specify such code in the value for the before-guile-boot attribute.

The inner_main() procedure itself will process the command line arguments (by calling optionProcess(), see section optionProcess), and then either invoke the code specified with the guile-main attribute, or else export the parsed options to Guile symbols and invoke the scm_shell() function from the Guile library. This latter will render the program nearly identical to the stock guile(1) program.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] shell-process: emit Bourne shell results

This will produce a main() procedure that parses the command line options and emits to ‘stdout’ Bourne shell commands that puts the option state into environment variables. This can be used within a shell script as follows:

eval "`opt_parser \"$@\"`"
test ${OPTION_CT} -gt 0 && shift ${OPTION_CT}

If the option parsing code detects an error or a request for usage or version, it will emit a command to exit with an appropriate exit code to ‘stdout’. This form of main will cause all messages, including requested usage and version information, to be emitted to ‘stderr’. Otherwise, a numeric value for OPTION_CT is guaranteed to be emitted, along with assignments for all the options parsed, something along the lines of the following will be written to ‘stdout’ for evaluation:

export OPTION_CT

If the arguments are to be reordered, however, then the resulting set of operands will be emitted and OPTION_CT will be set to zero. For example, the following would be appended to the above:

set -- 'operand1' 'operand2' 'operand3'

OPTION_CT is set to zero since it is not necessary to shift off any options.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] shell-parser: emit Bourne shell script

This will produce a main() procedure that emits a shell script that will parse the command line options. That script can be emitted to ‘stdout’ or inserted or substituted into a pre-existing shell script file. Improbable markers are used to identify previously inserted parsing text:

# # # # # # # # # # -- do not modify this marker --

The program is also pretty insistent upon starting its parsing script on the second line.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] main: user supplied main procedure

You must supply a value for the main-text attribute. You may also supply a value for option-code. If you do, then the optionProcess invocation will not be emitted into the code. AutoOpts will wrap the main-text inside of:

main( int argc, char** argv )
    int res = <<success-exit-code>>;
    { // replaced by option-code, if that exists
        int ct = optionProcess( &<<prog-name>>Options, argc, argv);
        argc -= ct;
        argv += ct;
    return res;

so you can most conveniently set the value with a here string (see section A Here String):

code = <<- _EndOfMainProc_
	<<your text goes here>>

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] include: code emitted from included template

You must write a template to produce your main procedure. You specify the name of the template with the tpl attribute and it will be incorporated at the point where AutoOpts is ready to emit the main() procedure.

This can be very useful if, in your working environment, you have many programs with highly similar main() procedures. All you need to do is parameterize the variations and specify which variant is needed within the main AutoOpts specification. Since you are coding the template for this, the attributes needed for this variation would be dictated by your template.

Here is an example of an include variation:

main = {
  main-type = include;
  tpl       = "main-template.tpl";

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] invoke: code emitted from AutoGen macro

You must write a template to produce your main procedure. That template must contain a definition for the function specified with the func attribute to this main() procedure specification. This variation operates in much the same way as include (see section include: code emitted from included template) method.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] for-each: perform function on each operand

This produces a main procedure that invokes a procedure once for each operand on the command line (non-option arguments), OR once for each non-blank, non-comment stdin input line. Leading and trailing white space is trimmed from the input line and comment lines are lines that are empty or begin with a comment character, defaulting to a hash (’#’) character.

NB: The argument program attribute (see section Program Description Attributes) must begin with the [ character, to indicate that there are command operands, but that they are optional.

For an example of the produced main procedure, in the ‘autoopts/test’ build directory, type the following command and look at ‘main.c’:

make verbose TESTS=main.test

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

procedure to handle each argument

The handler-proc attribute is required. It is used to name the procedure to call. That procedure is presumed to be external, but if you provide the code for it, then the procedure is emitted as a static procedure in the generated code.

This procedure should return 0 on success, a cumulative error code on warning and exit without returning on an unrecoverable error. As the cumulative warning codes are or-ed together, the codes should be some sort of bit mask in order to be ultimately decipherable (if you need to do that).

If the called procedure needs to cause a fail-exit, it is expected to call exit(3) directly. If you want to cause a warning exit code, then this handler function should return a non-zero status. That value will be OR-ed into a result integer for computing the final exit code. E.g., here is part of the emitted code:

  int res = 0;
  if (argc > 0) {
     do  {
         res |= my_handler( *(argv++) );
     } while (--argc > 0);
  } else { ...

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

handler procedure type

If you do not supply the handler-type attribute, your handler procedure must be the default type. The profile of the procedure must be:

int my_handler(char const * pz_entry);

However, if you do supply this attribute, you may set the value to any of four alternate flavors:


This is essentially the same as the default handler type, except that before your procedure is invoked, the generated code has verified that the string names an existing file. The profile is unchanged.


Before calling your procedure, the file is f-opened according to the X, where X may be any of the legal modes for fopen(3C). In this case, the profile for your procedure must be:

int my_handler(char const * pz_fname, FILE * entry_fp);

When processing inputs as file pointer stream files, there are several ways of treating standard input. It may be an ordinary input file, or it may contain a list of files to operate on.

If the file handler type is more specifically set to ‘file-r’ and a command line operand consists of a single hyphen, then my_handler will be called with entry_fp set to stdin and the pz_fname set to the translatable string, "standard input". Consequently, in this case, if the input list is being read from stdin, a line containing a hyphen by itself will be ignored.


This attribute specifies that standard input is a data input file. By default, for-each main procedures will read standard input for operands if no operands appear on the command line. If there are operands after the command line options, then standard input is typically ignored. It can always be processed as an input data file, however, if a single bare hyphen is put on the command line.


Before calling your procedure, the contents of the file are read or mapped into memory. (Excessively large files may cause problems.) The ‘some-text-of-file’ disallows empty files. Both require regular files. In this case, the profile for your procedure must be:

my_handler(char const * fname, char * file_text,
           size_t text_size);

Note that though the file_text is not const, any changes made to it are not written back to the original file. It is merely a memory image of the file contents. Also, the memory allocated to hold the text is text_size + 1 bytes long and the final byte is always NUL. The file contents need not be text, as the data are read with the read(2) system call.

file_text is automatically freed, unless you specify a handler-frees attribute. Then your code must free(3) the text.

If you select one of these file type handlers, then on access or usage errors the PROGRAM_EXIT_FAILURE exit code will, by default, be or-ed into the final exit code. This can be changed by specifying the global file-fail-code attribute and naming a different value. That is, something other than failure. You may choose success, in which case file access issues will not affect the exit code and the error message will not be printed.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

code for handler procedure

With the MYHANDLER-code attribute, you provide the code for your handler procedure in the option definition file. Note that the spelling of this attribute depends on the name provided with the handler-proc attribute, so we represent it here with MYHANDLER as a place holder. As an example, your main() procedure specification might look something like this:

main = {
  main-type    = for-each;
  handler-proc = MYHANDLER;
  MYHANDLER-code = <<- EndOfMyCode
	/* whatever you want to do */

and instead of an emitted external reference, a procedure will be emitted that looks like this:

static int
MYHANDLER( char const* pz_entry )
    int res = 0;
    <<MYHANDLER-code goes here>>
    return res;

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

for-each main procedure options

These attributes affect the main procedure and how it processes each argument or input line.


If this attribute is specified, then options and operands may be interleaved. Arguments or input lines beginning with a hyphen will cause it to be passed through to an option processing function and will take effect for the remainder of the operands (or input lines) processed.


This is code that gets inserted after the options have been processed, but before the handler procs get invoked.


This is code that gets inserted after all the entries have been processed, just before returning from main().


When reading operands from standard input, if you wish comment lines to start with a character other than a hash (#) character, then specify one character with this attribute. If string value is empty, then only blank lines will be considered comments.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.5 Option Attributes

For each option you wish to specify, you must have a block macro named flag defined. There are two required attributes: name and descrip. If any options do not have a value (traditional flag character) attribute, then the long-opts program attribute must also be defined. As a special exception, if no options have a value and long-opts is not defined and argument is not defined, then all arguments to the program are named options. In this case, the ‘-’ and ‘--’ command line option markers are optional.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Required Attributes

Every option must have exactly one copy of both of these attributes.


Long name for the option. Even if you are not accepting long options and are only accepting flags, it must be provided. AutoOpts generates private, named storage that requires this name. This name also causes a #define-d name to be emitted. It must not conflict with any other names you may be using in your program.

For example, if your option name is, debug or munged-up, you must not use the #define names DEBUG (or MUNGED_UP) in your program for non-AutoOpts related purposes. They are now used by AutoOpts.

Sometimes (most especially under Windows), you may get a surprise. For example, INTERFACE is apparently a user space name that one should be free to use. Windows usurps this name. To solve this, you must do one of the following:

  1. Change the name of your option
  2. add the program attribute (see section Program Description Attributes):
    export = '#undef INTERFACE';
  3. add the program attribute:

Except for documentation options, a very brief description of the option. About 40 characters on one line, maximum, not counting any texinfo markups. Texinfo markups are stripped before printing in the usage text. It appears on the usage() output next to the option name.

If, however, the option is a documentation option, it will appear on one or more lines by itself. It is thus used to visually separate and comment upon groups of options in the usage text.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Common Option Attributes

These option attributes are optional. Any that do appear in the definition of a flag, may appear only once.


The flag character to specify for traditional option flags, e.g., ‘-L’.


Maximum occurrence count (invalid if disable present). The default maximum is 1. NOLIMIT can be used for the value, otherwise it must be a number or a #define that evaluates to a number.


Minimum occurrence count. If present, then the option must appear on the command line. Do not define it with the value zero (0).


If an option must be specified, but it need not be specified on the command line, then specify this attribute for the option.


There are two effects to this attribute: the usage text will not show the option, and the generated documentation will mark it with: NOTE: THIS OPTION IS DEPRECATED.


Prefix for disabling (inverting sense of) the option. Only useful if long option names are being processed. When an option has this attribute, the test ENABLED_OPT(OPTNAME) is false when either of the following is true:

To detect that the option has been specified with the disabling prefix, you must use:


Long-name prefix for enabling the option (invalid if disable not present). Only useful if long option names are being processed.


If default is for option being enabled. (Otherwise, the OPTST_DISABLED bit is set at compile time.) Only useful if the option can be disabled.


If an option is relevant on certain platforms or when certain features are enabled or disabled, you can specify the compile time flag used to indicate when the option should be compiled in or out. For example, if you have a configurable feature, mumble that is indicated with the compile time define, WITH_MUMBLING, then add:


Take care when using these. There are several caveats:


This option specifies that the option is not allowed on the command line. Such an option may not take a value (flag character) attribute. The program must have the homerc (see section Program Description Attributes) option set.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Immediate Action Attributes

Certain options may need to be processed early. For example, in order to suppress the processing of configuration files, it is necessary to process the command line option ‘--no-load-optsbefore the config files are processed. To accommodate this, certain options may have their enabled or disabled forms marked for immediate processing. The consequence of this is that they are processed ahead of all other options in the reverse of normal order.

Normally, the first options processed are the options specified in the first homerc file, followed by then next homerc file through to the end of config file processing. Next, environment variables are processed and finally, the command line options. The later options override settings processed earlier. That actually gives them higher priority. Command line immediate action options actually have the lowest priority of all. They would be used only if they are to have an effect on the processing of subsequent options.


Use this option attribute to specify that the enabled form of the option is to be processed immediately. The help and more-help options are so specified. They will also call exit() upon completion, so they do have an effect on the processing of the remaining options :-).


Use this option attribute to specify that the disabled form of the option is to be processed immediately. The load-opts option is so specified. The ‘--no-load-opts’ command line option will suppress the processing of config files and environment variables. Contrariwise, the ‘--load-opts’ command line option is processed normally. That means that the options specified in that file will be processed after all the homerc files and, in fact, after options that precede it on the command line.


If either the immediate or the immed-disable attributes are set to the string, also, then the option will actually be processed twice: first at the immediate processing phase and again at the normal time.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Option Conflict Attributes

These attributes may be used as many times as you need. They are used at the end of the option processing to verify that the context within which each option is found does not conflict with the presence or absence of other options.

This is not a complete cover of all possible conflicts and requirements, but it simple to implement and covers the more common situations.


one entry for every option that must be present when this option is present


one entry for every option that cannot be present when this option is present

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Program may set option

If the option can be set outside of option processing, specify settable. If this attribute is defined, special macros for setting this particular option will be inserted into the interface file. For example, TEMPL_DIRS is a settable option for AutoGen, so a macro named SET_OPT_TEMPL_DIRS(a) appears in the interface file. This attribute interacts with the documentation attribute.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Option cannot be pre-configured

If presetting this option is not allowed, specify no-preset. (Thus, environment variables and values set in configuration files will be ignored.)

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Option Equivalence Class

Generally, when several options are mutually exclusive and basically serve the purpose of selecting one of several processing modes, specify the equivalence attribute. These options will be considered an equivalence class. Sometimes, it is just easier to deal with them as such. All members of the equivalence class must contain the same equivalenced-to option, including the equivalenced-to option itself. Thus, it must be a class member.

For an option equivalence class, there is a single occurrence counter for the class. It can be referenced with the interface macro, COUNT_OPT(BASE_OPTION), where BASE_OPTION is the equivalenced-to option name.

Also, please take careful note: since the options are mapped to the equivalenced-to option descriptor, any option argument values are mapped to that descriptor also. Be sure you know which “equivalent option” was selected before getting an option argument value!

During the presetting phase of option processing (see section Configuring your program), equivalenced options may be specified. However, if different equivalenced members are specified, only the last instance will be recognized and the others will be discarded. A conflict error is indicated only when multiple different members appear on the command line itself.

As an example of where equivalenced options might be useful, cpio(1) has three options ‘-o’, ‘-i’, and ‘-p’ that define the operational mode of the program (create, extract and pass-through, respectively). They form an equivalence class from which one and only one member must appear on the command line. If cpio were an AutoOpt-ed program, then each of these option definitions would contain:

equivalence = create;

and the program would be able to determine the operating mode with code that worked something like this:

case INDEX_OPT_CREATE:       ...
case INDEX_OPT_EXTRACT:      ...
default:    /* cannot happen */

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Option Aliasing

Sometimes, for backwards compatibility or tradition or just plain convenience, it works better to define one option as a pure alias for another option. For such situations, provide the following pieces of information:

flag = {
   name  = aliasing-option-name;
   value = aliasing-flag-char; // optional !
   aliases = aliased-to-option;

Do not provide anything else. The usage text for such an option will be:

   This is an alias for aliased-to-option

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Default Option

If your program processes its arguments in named option mode (See long-opts in Program Description Attributes), then you may select one of your options to be the default option. Do so by using attribute default with one of the options. The option so specified must have an arg-type (see section Option Argument Specification) specified, but not the arg-optional (see section Option Argument Optional) attribute. That is to say, the option argument must be required.

If you have done this, then any arguments that do not match an option name and do not contain an equal sign (=) will be interpreted as an option argument to the default option.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Option Sectioning Comment

This attribute means the option exists for the purpose of separating option description text in the usage output and texi documentation. Without this attribute, every option is a separate node in the texi docs. With this attribute, the documentation options become texi doc nodes and the options are collected under them. Choose the name attribute carefully because it will appear in the texi documentation.

Libraries may also choose to make it settable so that the library can determine which command line option is the first one that pertains to the library.

If the ‘documentation’ attribute is present, then all other attributes are disabled except settable, call-proc and flag-code. settable must be and is only specified if call-proc, extract-code or flag-code has been specified. When present, the descrip attribute will be displayed only when the ‘--help’ option has been specified. It will be displayed flush to the left hand margin and may consist of one or more lines of text, filled to 72 columns.

The name of the option will not be printed in the help text. It will, however, be printed as section headers in the texi documentation. If the attribute is given a non-empty value, this text will be reproduced in the man page and texi doc immediately after the descrip text.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Translator Notes

If you need to give the translators a special note about a particular option, please use the translators attribute. The attribute text will be emitted into the generated .c text where the option related strings get defined. To make a general comment about all of the option code, add comments to an include attribute (see section Program Description Attributes). Do not use this attribute globally, or it will get emitted into every option definition block.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.6 Option Argument Specification

Command line options come in three flavors: options that do not take arguments, those that do and those that may. Without an "arg-type" attribute, AutoOpts will not process an argument to an option. If "arg-type" is specified and "arg-optional" is also specified, then the next command line token will be taken to be an argument, unless it looks like the name of another option.

If the argument type is specified to be anything other than "str[ing]", then AutoOpts will specify a callback procedure to handle the argument. Some of these procedures will be created and inserted into the generated ‘.c’ file, and others are already built into the ‘libopts’ library. Therefore, if you write your own callback procedure (see section Option Argument Handling), then you must either not specify an "arg-type" attribute, or else specify it to be of type "str[ing]". Your callback function will be able to place its own restrictions on what that string may contain or represent.

Option argument handling attributes depend upon the value set for the arg-type attribute. It specifies the type of argument the option will take. If not present, the option cannot take an argument. If present, it must be an entry in the following table. The first three letters is sufficient.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type String

arg-type = string;

The argument may be any arbitrary string, though your program or option callback procedure may place additional constraints upon it.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type Number

arg-type = number;

The argument must be a correctly formed integer, without any trailing U’s or L’s. AutoOpts contains a library procedure to convert the string to a number. If you specify range checking with arg-range (see below), then AutoOpts produces a special purpose procedure for this option.


scaled marks the option so that suffixes of ‘k’, ‘K’, ‘m’, ‘M’, ‘g’, ‘G’, ‘t’, and ‘T’ will multiply the given number by a power of 1000 or 1024. Lower case letters scale by a power of 1000 and upper case scale by a power of 1024.


arg-range is used to create a callback procedure for validating the range of the option argument. It must match one of the range entries. Each arg-range should consist of either an integer by itself or an integer range. The integer range is specified by one or two integers separated by the two character sequence, ->. Be sure to quote the entire range string. The definitions parser will not accept the range syntax as a single string token.

The generated procedure imposes the range constraints as follows:

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type Boolean

arg-type = boolean;

The argument will be interpreted and always yield either AG_TRUE or AG_FALSE. False values are the empty string, the number zero, or a string that starts with f, F, n or N (representing False or No). Anything else will be interpreted as True.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type Keyword

arg-type = keyword;

The argument must match a specified list of strings (see section Keyword list). Assuming you have named the option, optn-name, the strings will be converted into an enumeration of type te_Optn_Name with the values OPTN_NAME_KEYWORD.* If you have not specified a default value, the value OPTN_NAME_UNDEFINED will be inserted with the value zero. The option will be initialized to that value. You may now use this in your code as follows:

te_Optn_Name opt = OPT_VALUE_OPTN_NAME;
switch (opt) {
case OPTN_NAME_UNDEFINED:  /* undefined things */ break;
case OPTN_NAME_KEYWORD:    /* `keyword' things */ break;
default: /* utterly impossible */ ;

AutoOpts produces a special purpose procedure for this option. You may not specify an alternate handling procedure.

If you have need for the string name of the selected keyword, you may obtain this with the macro, OPT_OPTN_NAME_VAL2STR(val). The value you pass would normally be OPT_VALUE_OPTN_NAME, but anything with numeric value that is legal for te_Optn_Name may be passed. Anything out of range will result in the string, ‘"*INVALID*"’ being returned. The strings are read only. It may be used as in:

te_Optn_Name opt = OPT_VALUE_OPTN_NAME;
printf( "you selected the %s keyword\n",
        OPT_OPTN_NAME_VAL2STR(opt) );

* Note: you may replace the OPTN_NAME enumeration prefix with another prefix by specifying a prefix-enum attribute.

Finally, users may specify the argument either by name or by number. Since the numeric equivalents change by having new entries inserted into the keyword list, this would not be a recommended practice. However, either -1 or ~0 will always be equivalent to specifying the last keyword.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type Set Membership

arg-type = set;

The argument must be a list of names each of which must match the strings “all”, “none” or one of the keywords (see section Keyword list) specified for this option. all will turn on all membership bits and none will turn them all off. Specifying one of the keywords will set the corresponding set membership bit on (or off, if negated) . Literal numbers may also be used and may, thereby, set or clear more than one bit.

The membership result starts with the previous (or initialized) result. To clear previous results, either start the membership string with ‘none +’ or with the equals character (‘=’). To invert (bit flip) the final result (regardless of whether the previous result is carried over or not), start the string with a carat character (‘^’). If you wish to invert the result and start without a carried over value, use one of the following: =^ or ^none+. These are equivalent.

The list of names or numbers must be separated by one of the following characters: ‘+-|!,’ or whitespace. The comma is equivalent to whitespace, except that only one may appear between two entries and it may not appear in conjunction with the or bar (‘|’). The ‘+|’ leading characters or unadorned name signify adding the next named bit to the mask, and the ‘-!’ leading characters indicate removing it.

The number of keywords allowed is constrained by the number of bits in a pointer, as the bit set is kept in a void * pointer.

If, for example, you specified first in your list of keywords, then you can use the following code to test to see if either first or all was specified:

uintptr_t opt = OPT_VALUE_OPTN_NAME;
if (opt & OPTN_NAME_FIRST)
    /* OPTN_NAME_FIRST bit was set */ ;

AutoOpts produces a special purpose procedure for this option. To set multiple bits as the default (initial) value, you must specify an initial numeric value (which might become inaccurate over time), or else specify arg-default multiple times. Do not specify a series of names conjoined with + symbols as the value for any of the arg-default attributes. That works for option parsing, but not for the option code generation.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type Hierarchical

arg-type = hierarchy;
arg-type = nested;

This denotes an option with a structure-valued argument, a.k.a. subopts in getopts terminology. The argument is parsed and the values made available to the program via the find and find next calls (See section optionFindValue, See section optionGetValue, and see section optionFindNextValue).

tOptionValue * val = optionGetValue(VALUE_OPT_OPTN_NAME, "name");
while (val != NULL) {
  val = optionNextValue(VALUE_OPT_OPTN_NAME, val);
  if (wrong_name(val, "name"))

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type File Name

arg-type = file;

This argument type will have some validations on the argument and, optionally, actually open the file. You must specify several additonal attributes for the option:


If not specified or empty, then the directory portion of the name is checked. The directory must exist or the argument is rejected and the usage procedure is invoked.

Otherwise, both the directory as above and the full name is tested for existence. If the value begins with the two letters no, then the file must not pre-exist. Otherwise, the file is expected to exist.


If not specified or empty, the file is left alone. If the value begins with the four letters desc[riptor], then open(2) is used and optArg.argFd is set. Otherwise, the file is opened with fopen and optArg.argFp is set.


If open-file is set and not empty, then you must specify the open mode. Set the value to the flag bits or mode string as appropriate for the open type.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type Time Duration

arg-type = time-duration;

The argument will be converted into a number of seconds. It may be a multi-part number with different parts being multiplied into a seconds value and added into the final result. Valid forms are in the table below. Upper cased letters represent numbers that must be used in the expressions.


HH is multiplied by 3600 and MM multiplied by 60 before they are added to SS. This time specification may not be followed by any other time specs. HH and MM are both optional, though HH cannot be specified without MM.


DAYS is multiplied by the number of seconds in a day. This value may be followed by (and added to) values specified by HH:MM:SS or the suffixed values below. If present, it must always be first.


HRS is multiplied by the number of seconds in an hour. This value may be followed by (and added to) values specified by MM:SS or the suffixed values below.


MINS is multiplied by the number of seconds in a minute. This value may be followed by (and added to) a count of seconds.


This value can only be the last value in a time specification. The s suffix is optional.

   5 d 1:10:05    ==> 5 days + 1 hour 10 minutes and 5 seconds
   5 d 1 h 10 m 5 ==> yields: 436205 seconds
   5d1h10m5s      ==> same result -- spaces are optional.

When saved into a config file, the value will be stored as a simple count of seconds. There are actually more (many) accepted time duration strings. The full documentation can be found with ISO-8601 documentation and the more extedded documentation when parse_duration() becomes more widely available.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Arg Type Time and Date

arg-type = time-date;

The argument will be converted into the number of seconds since the epoch. The conversion rules are very complicated, please see the ‘getdate_r(3GNU)’ man page. There are some additional restrictions:

  1. Your project must be compiled with PKGDATADIR defined and naming a valid directory.
  2. The DATEMSK environment variable will be set to the ‘datemsk’ file within that directory.

If that file is not accessible for any reason, the string will be parsed as a time duration (see section Arg Type Time Duration) instead of a specific date and time.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Keyword list

If the arg-type is keyword (see section Arg Type Keyword) or set-membership (see section Arg Type Set Membership), then you must specify the list of keywords by a series of keyword entries. The interface file will contain values for <OPTN_NAME>_<KEYWORD> for each keyword entry. keyword option types will have an enumeration and set-membership option types will have a set of unsigned bits #define-d.

If the arg-type is specifically keyword, you may also add special handling code with a extra-code attribute. After optionEnumerationVal has converted the input string into an enumeration, you may insert code to process this enumeration value (pOptDesc->optArg.argEnum).

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Option Argument Optional

The arg-optional attribute indicates that the argument to the option is optional (need not be specified on the command line). This is only valid if the arg-type is string (see section Arg Type String) or keyword (see section Arg Type Keyword). If it is keyword, then this attribute may also specify the default keyword to assume when the argument is not supplied. If left empty, arg-default (see section Default Option Argument Value) or the zero-valued keyword will be used.

The syntax rules for identifying the option argument are:

This is overridden and the options are required if the libopts library gets configured with ‘--disable-optional-args’.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Default Option Argument Value

This specifies the default option argument value to be used when the option is not specified or preset. You may specify multiple arg-default values if the argument type is set membership.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.7 Option Argument Handling

AutoOpts will either specify or automatically generate callback procedures for options that take specialized arguments. The only option argument types that are not specialized are plain string arguments and no argument at all. For options that fall into one of those two categories, you may specify your own callback function, as specified below. If you do this and if you specify that options are resettable (see section Automatically Supported Options), then your option handling code must look for the ‘OPTST_RESET’ bit in the fOptState field of the option descriptor.

If the option takes a string argument, then the stack-arg attribute can be used to specify that the option is to be handled by the libopts stackOptArg() and unstackOptArg() library procedures (see below). In this case, you may not provide option handling code.

Finally, ‘documentation’ options (see section Option Sectioning Comment) may also be marked as ‘settable’ (see section Program may set option) and have special callback functions (either ‘flag-code’, ‘extract-code’, or ‘call-proc’).


statements to execute when the option is encountered. This may be used in conjunction with option argument types that cause AutoOpts to emit handler code. If you do this, the ‘flag-code’ with index zero (0) is emitted into the handler code before the argument is handled, and the entry with index one (1) is handled afterward.

The generated procedure will be laid out something like this:

static void
doOpt<name>(tOptions* pOptions, tOptDesc* pOptDesc)
<AutoOpts defined handler code>

Only certain fields within the tOptions and tOptDesc structures may be accessed. See section Data for Option Processing. When writing this code, you must be very careful with the pOptions pointer. The handler code is called with this pointer set to special values for handling special situations. Your code must handle them. As an example, look at optionEnumerationVal in ‘enum.c’.


This is effectively identical to flag-code, except that the source is kept in the output file instead of the definitions file and you cannot use this in conjunction with options with arguments, other than string arguments.

A long comment is used to demarcate the code. You must not modify that marker. Before regenerating the option code file, the old file is renamed from MUMBLE.c to MUMBLE.c.save. The template will be looking there for the text to copy into the new output file.


external procedure to call when option is encountered. The calling sequence must conform to the sequence defined above for the generated procedure, doOpt<name>. It has the same restrictions regarding the fields within the structures passed in as arguments. See section Data for Option Processing.


Name of another option whose flag-code can be executed when this option is encountered.


Call a special library routine to stack the option’s arguments. Special macros in the interface file are provided for determining how many of the options were found (STACKCT_OPT(NAME)) and to obtain a pointer to a list of pointers to the argument values (STACKLST_OPT(NAME)). Obviously, for a stackable argument, the max attribute (see section Common Option Attributes) needs to be set higher than 1.

If this stacked argument option has a disablement prefix, then the entire stack of arguments will be cleared by specifying the option with that disablement prefix.


Call a special library routine to remove (unstack) strings from a stack-arg option stack. This attribute must name the option that is to be unstacked. Neither this option nor the stacked argument option it references may be equivalenced to another option.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.8 Internationalizing Options

Normally, AutoOpts produces usage text that is difficult to translate. It is pieced together on the fly using words and phrases scattered around here and there, piecing together toe document. This does not translate well.

Incorporated into this package are some ways around the problem. First, you should specify the full-usage and short-usage program attributes (see section Program Description Attributes). This will enable your translators to translate the usage text as a whole.

Your translators will also be able to translate long option names. The option name translations will then become the names searched for both on the command line and in configuration files. However, it will not affect the names of environment variable names used to configure your program.

If it is considered desireable to keep configuration files in the C locale, then several macros are available to suppress or delay the translations of option names at run time. These are all disabled if ENABLE_NLS is not defined at compile time or if no-xlate has been set to the value anything. These macros must be invoked before the first invocation of optionProcess.


Disable (or enable) the translations of option names for configuration files. If you enable translation for config files, then they will be translated for command line options.


Disable (or enable) the translations of option names for command line processing. If you disable the translation for command line processing, you will also disable it for configuration file processing. Once translated, the option names will remain translated.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.9 Man and Info doc Attributes

AutoOpts includes AutoGen templates for producing abbreviated man pages and for producing the invoking section of an info document. To take advantage of these templates, you must add several attributes to your option definitions.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Per option documentation attributes

These attributes are sub-attributes (sub-stanzas) of the flag stanzas.


If an option has an argument, the argument should have a name for documentation purposes. It will default to arg-type, but it will likely be clearer with something else like, file-name instead of string (the type).


First, every flag definition other than documentation definitions, must have a doc attribute defined. If the option takes an argument, then it will need an arg-name attribute as well. The doc text should be in plain sentences with minimal formatting. The Texinfo commands @code, and @var will have its enclosed text made into \fB entries in the man page, and the @file text will be made into \fI entries. The arg-name attribute is used to display the option’s argument in the man page.

Options marked with the documentation attribute are for documenting the usage text. All other options should have the doc attribute in order to document the usage of the option in the generated man pages.

Since these blocks of text are inserted into all output forms, any markup text included in these blocks must be massaged for each output format. By default, it is presumed to be ‘texi’ format.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ] Global documentation attributes


If your command is a game or a system management command, specify this attribute with the value 5 or 8, respectively. The default is a user command (section 1).


This attribute is used to add a very short explanation about what a program is used for when the title attribute is insufficient. If there is no doc-section stanza of type DESCRIPTION, then this text is used for the man page DESCRIPTION section, too.


This attribute tells the template that the generated code should be surrounded with the following doxygen comments:

/** @file <header-or-code-file-name>
 *  @addtogroup <value-of-addtogroup>
 *  @{


/** @} */

Specify the default markup style for the doc stanzas. By default, it is texi, but man and mdoc may also be selected. There are nine converter programs that do a partial job of converting one form of markup into another. texi2texi, man2man and mdoc2mdoc work pretty well.

You may also post process the document by using doc-sub stanzas, see below.


This text will be inserted as a lead-in paragraph in the OPTIONS section of the generated man page.


This is a compound attribute that requires three subattributes:


This describes the format of the associated ds-text section. man, mdoc and texi formats are supported. Regardless of the chosen format, the formatting tags in the output text will be converted to man macros for man pages, mdoc macros for mdoc pages, and texi macros for texinfo pages.


This is the descriptive text, written according to the rules for ds-format documents.


This describes the section type. Basically, the title of the section that will be added to all output documentation. There may be only one doc-section for any given ds-type. If there are duplicates, the results are undefined (it might work, it might not).

There are five categories of ds-type sections. They are those that the documentation templates would otherwise:

  1. always create itself, ignoring any ds-types by this name. These are marked, below, as ao-only.
  2. create, if none was provided. These are marked, alternate.
  3. create, but augment if the doc-section was provided. These are marked, augments.
  4. do nothing, but inserts them into the output in a prescribed order. These are marked, known
  5. knows nothing about them. They will be alphabetized and inserted after the list of leading sections and before the list of trailing sections. These are not marked because I don’t know their names.

Some of these are emitted by the documentation templates only if certain conditions are met. If there are conditions, they are explained below. If there are no conditions, then you will always see the named section in the output.

The output sections will appear in this order:










ao-only, if environment presets or configuration file processing has been specified.


At this point, the unknown, alphabetized sections are inserted.




augments, if environment presets have been specified.


augments, if configuration file processing has been specified.
















alternate, if the copyright stanza has either an author or an owner attribute.


alternate, if there is a copyright stanza.


augments, if the copyright stanza has an eaddr attribute.



Here is an example of a doc-section for a SEE ALSO type.

doc-section = {
  ds-type   = 'SEE ALSO'; // or anything else
  ds-format = 'man';      // or texi or mdoc format
  ds-text   = <<-_EOText_
	text relevant to this section type,
	in the chosen format

This attribute will cause the resulting documentation to be post-processed. This is normally with sed, see doc-sub-cmd below. This attribute has several sub-attributes:


This is the name of an autogen text definition value, like prog-name or version. In the sub-text field, occurrences of this name preceded by two less than characters and followed by two greater than characters will be replaced by the text value of the definition, e.g. ‘<<prog-name>>’.


The text that gets added to the command file for the post processing program.


If this command only applies to certain types of output, specify this with a regular expression that will match one of the valid output format types, e.g. ‘man|mdoc’ will match those two kinds, but not texi output. If omitted, it will always apply.

For example, if you want to reference the program name in the doc text for an option common to two programs, put ‘#PROG#’ into the text. The following will replace all occrrences of ‘#PROG#’ with the current value for prog:

doc-sub = {
  sub-name = prog-name;
  sub-text = 's/#PROG#/<<prog-name>>/g';

A formatting string for constructing the post-processing command. The first parameter is the name of the file with editing commands in it, and the second is the file containing the unprocessed document. The default value is:

sed -f %s %s

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.10 Automatically Supported Options

AutoOpts provides automated support for several options. help and more-help are always provided. The others are conditional upon various global program attributes being defined See section Program Description Attributes.

Below are the option names and default flag values. The flags are activated if and only if at least one user-defined option also uses a flag value. The long names are supported as option names if long-opts has been specified. These option flags may be deleted or changed to characters of your choosing by specifying xxx-value = "y";, where xxx is one of the option names below and y is either empty or the character of your choice. For example, to change the help flag from ? to h, specify help-value = "h";; and to require that save-opts be specified only with its long option name, specify save-opts-value = "";.

Additionally, the procedure that prints out the program version may be replaced by specifying version-proc. This procedure must be defined to be of external scope (non-static). By default, the AutoOpts library provides optionPrintVersion and it will be the specified callback function in the option definition structure.

With the exception of the load-opts option, none of these automatically supported options will be recognized in configuration files or environment variables.

help -?

This option will immediately invoke the USAGE() procedure and display the usage line, a description of each option with its description and option usage information. This is followed by the contents of the definition of the detail text macro.

more-help -!

This option is identical to the help option, except that the output is passed through a pager program. (more by default, or the program identified by the PAGER environment variable.)

usage -u

This option must be requested by specifying, usage-opt in the option definition file. It will produce abbreviated help text to ‘stdout’ and exit with zero status (EXIT_SUCCESS).

version -v

This will print the program name, title and version. If it is followed by the letter c and a value for copyright and owner have been provided, then the copyright will be printed, too. If it is followed by the letter n, then the full copyright notice (if available) will be printed. The version attribute must be specified in the option definition file.

load-opts -<

This option will load options from the named file. They will be treated exactly as if they were loaded from the normally found configuration files, but will not be loaded until the option is actually processed. This can also be used within another configuration file, causing them to nest. This is the only automatically supported option that can be activated inside of config files or with environment variables.

Specifying the negated form of the option (‘--no-load-opts’) will suppress the processing of configuration files and environment variables.

This option is activated by specifying one or more homerc attributes.

save-opts ->

This option will cause the option state to be printed in the configuration file format when option processing is done but not yet verified for consistency. The program will terminate successfully without running when this has completed. Note that for most shells you will have to quote or escape the flag character to restrict special meanings to the shell.

The output file will be the configuration file name (default or provided by rcfile) in the last directory named in a homerc definition.

This option may be set from within your program by invoking the "SET_OPT_SAVE_OPTS(filename)" macro (see section SET_OPT_name - Force an option to be set). Invoking this macro will set the file name for saving the option processing state, but the state will not actually be saved. You must call optionSaveFile to do that (see section optionSaveFile). CAVEAT: if, after invoking this macro, you call optionProcess, the option processing state will be saved to this file and optionProcess will not return. You may wish to invoke CLEAR_OPT( SAVE_OPTS ) (see section CLEAR_OPT( <NAME> ) - Clear Option Markings) beforehand if you do need to reinvoke optionProcess.

This option is activated by specifying one or more homerc attributes.

reset-option -R

This option takes the name of an option for the current program and resets its state such that it is set back to its original, compile-time initialized value. If the option state is subsequently stored (via ‘--save-opts’), the named option will not appear in that file.

This option is activated by specifying the resettable attribute.

BEWARE: If the resettable attribute is specified, all option callbacks must look for the OPTST_RESET bit in the fOptState field of the option descriptor. If set, the optCookie and optArg fields will be unchanged from their last setting. When the callback returns, these fields will be set to their original values. If you use this feature and you have allocated data hanging off of the cookie, you need to deallocate it.

[ < ] [ > ]   [ << ] [ Up ] [ >> ]         [Top] [Contents] [Index] [ ? ]

7.5.11 Library of Standard Options

AutoOpts has developed a set of standardized options. You may incorporate these options in your program simply by first adding a #define for the options you want, and then the line,

#include stdoptions.def

in your option definitions. The supported options are specified thus:

#define DEBUG
#define DRY_RUN
#define INPUT
#define OUTPUT
#define WARN

#define SILENT
#define QUIET
#define BRIEF
#define VERBOSE

By default, only the long form of the option will be available. To specify the short (flag) form, suffix these names with _FLAG. e.g.,

#define DEBUG_FLAG

--silent’, ‘--quiet’, ‘--brief’ and ‘--verbose’ are related in that they all indicate some level of diagnostic output. These options are all designed to conflict with each other. Instead of four different options, however, several levels can be incorporated by #define-ing VERBOSE_ENUM. In conjunction with VERBOSE, it incorporates the notion of 5 levels in an enumeration: silent, quiet, brief, informative and verbose; with the default being brief.

Here is an example program that uses the following set of definitions:

AutoGen Definitions options;

prog-name  = default-test;
prog-title = 'Default Option Example';
homerc     = '$$/../share/default-test', '$HOME', '.';
version    = '1.0';
main = {
  main-type = shell-process;
#define DEBUG_FLAG
#define WARN_FLAG
#define WARN_LEVEL
#define DRY_RUN_FLAG
#define INPUT_FLAG
#include stdoptions.def

Running a few simple commands on that definition file:

autogen default-test.def
copts="-DTEST_DEFAULT_TEST_OPTS `autoopts-config cflags`"
lopts="`autoopts-config ldflags`"
cc -o default-test ${copts} default-test.c ${lopts}

Yields a program which, when run with ‘--help’, prints out:

exit 0

[ < ] [ > ]   [ << ] [ Up ] [ >> ]

This document was generated by Bruce Korb on August 21, 2015 using texi2html 1.82.