On 10/18/2013 09:54 AM, Vladimir 'φ-coder/phcoder' Serbinenko wrote: > On 18.10.2013 17:40, Eric Blake wrote: >> On 10/18/2013 09:30 AM, Vladimir 'φ-coder/phcoder' Serbinenko wrote: >>> When testing with old gcc, I found out that gnulib uses >>> __GNUC_GNU_INLINE__ without checking whether it's defined first. I >>> propose following fix: >>> >>> === modified file 'm4/extern-inline.m4' >>> --- m4/extern-inline.m4 2013-04-11 19:12:46 +0000 >>> +++ m4/extern-inline.m4 2013-10-18 14:55:19 +0000 >>> @@ -32,7 +32,7 @@ >>> # define _GL_INLINE inline >>> # define _GL_EXTERN_INLINE extern inline >>> #elif 2 < __GNUC__ + (7 <= __GNUC_MINOR__) && !defined __APPLE__ >>> -# if __GNUC_GNU_INLINE__ >>> +# if defined (__GNUC_GNU_INLINE__) && __GNUC_GNU_INLINE__
Our style eschews unneeded () inside #if defined. >> >> This has no semantic change, unless you are compiling with -Wundef; and >> we generally state that use of -Wundef for gnulib .c files is >> unsupported. Or is your complaint that this snippet of code gets >> included into header files that get included in your files, and not just >> gnulib.c files, and where we should try to honor -Wundef? >> > Yes, this code ends up in config.h (called config-util.h in our project) > which is not restricted to gnulib. > We compile gnulib with -Wno-undef Mentioning that would have helped. Also, posting the patch as a full git commit, with a ChangeLog entry, would have helped, too :) At any rate, the patch is now pushed. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
