IRC, freenode, #hurd, 2012-11-08
<pinotree> tschwinge: i'm not totally sure, but it seems the pie options for gcc/ld are causing issues <pinotree> namely, producing executables that sigsegv straight away <tschwinge> pinotree: OK, I do remember some issues about these, too. <tschwinge> Also for -pg. <tschwinge> These have in common that they use different crt*.o files for linking. <tschwinge> Might well be there's some bugs there. <pinotree> one way is to try the w3m debian build: the current build configuration enables also pie, which in turns makes an helper executable (mktable) sigsegv when invoked <pinotree> if «,-pie» is appended to the DEB_BUILD_MAINT_OPTIONS variable in debian/rules, pie is not added and the resulting mktable runs correctly
IRC, OFTC, #debian-hurd, 2012-11-09
<pinotree> youpi: ah, as i noted to tschwinge earlier, it seems -fPIE -pie miscompile stuff <youpi> uh <pinotree> this causes the w3m build failure and (indirectly, due to elinks built with -pie) aptitude
IRC, freenode, #hurd, 2013-01-19
<gnu_srs> pinotree: I can confirm that -fPIE -pie fails and only -fPIE works for mktable in w3m. Still have to check with elinks. What's up doc?
Profiling has been fixed (GCC specs; commit
9b2eb5d3268cf674f9a6964479f20428e0b43500); PIE unknown.
PIE_SPEC comments in gcc.