On 02/02/2013 09:09 AM, Michał Górny wrote: > The major issue we've been having is difference in inline semantics > between gnu89 and c99. It hit us at least with libgcrypt [1,2], > pkg-config [3,4], e2fsprogs [5]. OK, this is a different matter than what AC_PROG_CC will do. AC_PROG_CC will enable extensions if available. But the inline semantics issue is something where, if I understand things correctly, a package explicitly wants the old GNU inline semantics for 'inline', and doesn't want the C99 semantics for 'inline' that is supported by clang and by newer versions of GCC. That's not an extension: it's a change to the language. One possibility is that Autoconf could add a macro AC_CC_OLD_GNU_INLINE, or something like that, which would cause 'configure' to arrange for the old GNU inline semantics and to fail if this can't be arranged. Is that a macro that you could write? But maybe we don't need to have such a macro. Aren't the packages in question intended to be used with newer compilers? If so, they need to be updated to use C99 semantics for 'inline', since we can't expect every new compiler to support the old GCC 'inline' semantics. And then AC_PROG_CC will already do what we want, since it'll (in the next version of Autoconf, anyway), ask newer GCCs to use the C99-compatible 'inline' semantics. By "newer GCCs" I mean GCC 4.3 or later, if I'm reading the release notes correctly. _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx https://lists.gnu.org/mailman/listinfo/autoconf
