Collin Funk wrote: > On Haiku I see the following warning: > > test-strsignal.c:24:18: warning: initialization of 'char * (*)(int)' from > incompatible pointer type 'const char * (*)(int)' > [-Wincompatible-pointer-types] > 24 | SIGNATURE_CHECK (strsignal, char *, (int)); > | ^~~~~~~~~ > signature.h:46:57: note: in definition of macro 'SIGNATURE_CHECK2' > 46 | _GL_UNUSED static ret (*signature_check ## id) args = fn > | ^~ > signature.h:39:3: note: in expansion of macro 'SIGNATURE_CHECK1' > 39 | SIGNATURE_CHECK1 (fn, ret, args, __LINE__) > | ^~~~~~~~~~~~~~~~ > test-strsignal.c:24:1: note: in expansion of macro 'SIGNATURE_CHECK' > 24 | SIGNATURE_CHECK (strsignal, char *, (int)); > | ^~~~~~~~~~~~~~~ > > I don't think the missing const is worth a workaround, so I have > disabled the signature check and documented the bug. > > Reported to Haiku in case they want to comply with POSIX [1].
Thanks for both. In fact, since POSIX [1] says "The application shall not modify the string returned.", Haiku's non-standard prototype makes sense. It would be useful if POSIX at some point changed the return type to 'const char *', and likewise for the *gettext() functions (cf. [2]). Bruno [1] https://pubs.opengroup.org/onlinepubs/9799919799/functions/strsignal.html [2] https://lists.gnu.org/archive/html/bug-gettext/2025-04/msg00006.html
