Re: [Bug-gnubg] bad __attribute__ handling breaks w/newer standards, newer gcc, & glibc and causes infinite loop - FIX NEEDS TESTING

[Mike Frysinger]

On 07 Oct 2015 18:46, Michael Petch wrote:
> I have put away a fix for this issue and did some other cleanup that
> involved files that may be using __attribute__. I took the approach of
> moving the __attribute__ check to configure.ac using
> ax_c___attribute__.m4 . common.h now includes config.h .
> This fix could use some testing, so please let me know if i have caused
> any regressions or something isn't fixed.
> 
> The Changelog is here:
> 
> 2015-10-07  Michael Petch  <address@hidden>
> 
>     * common.h, configure.ac, copying.awk, credits.sh,
>       output.h, lib/simd.h, m4/Makefile.am, non-src/copying.c,
>       non-src/credits.c, non-src/gnubg-stock-pixbufs.h,
>       +m4/ax_c___attribute__.m4, external_l.l, external_y.y,
>       sgf_l.l, sgf_y.y, non-src/external_l.c, non-src/external_y.c,
>       non-src/external_y.h, non-src/sgf_l.c, non-src/sgf_y.c,
>       non-src/sgf_y.h, backgammon.h : Fix problem identified by
>     Mike Frysinger where __attribute__ and  _GNU_SOURCE may not
>     be properly defined in some situations. This may result in
>     an infinite loop on newer compilers.

my (additional) 2 cents (after reading some of the other notes posted):
redefining __attribute__ is generally dangerous and i'm not sure trying
to mitigate it is worth the effort.  the way other projects do it seems
much safer to me -- namely, creating your own FOO_ATTRIBUTE(xxx) macro
which in turn would expand into __attribute__ or nothing.  that way even
if the configure or header test goes wrong, you won't inadvertently break
system headers ... only your own :).
-mike

signature.asc
Description: Digital signature