bug#22385: 24.5; Build failure with modules on OpenBSD

[Gabriele Svelto]

 Hello Paul,
thanks for getting in touch about this. There's a way to fix this issue
in my macro and it's by running the test twice with and without the
attribute and comparing the compiler error output instead of simply
assuming that any error output means that the attribute is not
supported. It's something I had considered doing in the beginning but
deemed unnecessarily complicated at the time. Turns out I was wrong;
I'll try to push a fix into autoconf-archive in the next few days.

 Gabriele

On 17/01/2016 23:30, Paul Eggert wrote:
> Thanks for reporting the problem. The OpenBSD compiler warnings about
> strcpy etc. are false alarms, and these false alarms confuse the Emacs
> build procedure and therefore introduce bugs into the Emacs module code.
> We're lucky that a run-time assertion caught this.
> 
> I installed the attached patch into the emacs-25 branch. It removes the
> offending test in 'configure', and so should fix the bug so I am boldly
> marking it as done. Please give the fix a try; if it doesn't work I can
> unmark the bug.
> 
> Is there some way to tell the OpenBSD C compiler to not chatter about
> uses of strcpy, strcat, and sprintf? We don't want to use the
> OpenBSD-suggested replacements in Emacs, as that would be more likely to
> introduce bugs than to fix them. If we can suggest (or get 'configure'
> to use) a build-time option to suppress the false alarms that would save
> everybody some work and would help make Emacs more reliable.
> 
> Although the OpenBSD porting problem with AX_GCC_VAR_ATTRIBUTE no longer
> affects Emacs, it might affect other programs that use that macro, so
> I'll CC: this to the macro's author Gabriele Svelto. Gabriele, the bug
> report is here:
> 
> http://bugs.gnu.org/22385