Compile error with GCC 4.9

Marc Glisse marc.glisse at inria.fr
Fri May 2 16:26:08 UTC 2014


On Fri, 2 May 2014, Torbjörn Granlund wrote:

> Jaak Ristioja <jaak.ristioja at cyber.ee> writes:
>
>  Even http://gcc.gnu.org/gcc-4.9/porting_to.html bashes GMP for this.
>  Probably related to __need_size_t?
> 
> Then the GCC project deserves to be chastised.
>
> They urge people into using a 4 year old GMP release.  This is done in
> many places, as well as enforced by the 'prerequisites' script.
>
> Then they find an incompatibility with their latest and greatest
> (pre-)release and bashes GMP.  They do in lieu of reporting the
> perceived problem to us.

Those release notes are not bashing GMP at all. They explain in clear 
technical terms (no bashing) that old versions (they do specify the 
versions) have an issue, and how people can work around it. Thanks to 
them, we were supposed to avoid receiving messages from confused users: 
they can google for the error message and find that explanation. This 
issue was listed because when preparing the 4.9 release (gmp-6 was not 
released yet) this was one of the common issues encountered while 
rebuilding fedora / debian with the new compiler.

I had already fixed this bug for a while when they noticed, and they did 
contact me to discuss this message in the release notes (and ask when 
gmp-6 was due).

> (Of course, our reply would have been to stop using long outdated GMP
> releases.)

5.1.3 is not "long outdated", and this issue is about C++11 so it is kind 
of orthogonal to the build dependency of gcc on gmp (which is mostly an 
issue for people not using GMP for anything else).


If there is an issue with these release notes, please blame me, maybe I 
should have brought it up with other GMP developers, but GCC people acted 
perfectly well in this case.

-- 
Marc Glisse


More information about the gmp-bugs mailing list