Re: attribute warn_unused_result

[Andy Moreton]

On Thu 03 Feb 2011, Chad Brown wrote:

> On Feb 3, 2011, at 11:33 AM, Paul Eggert wrote:
>> 
>> Gnulib's ignore-value module is designed for that.  I installed the
>> following:
>
> ..and it broke the W32 and nextstep builds (at least) again.
>
> I appreciate the ideas behind the gnulib inclusion, but it seems to be
> causing a lot more broken builds than we saw before.  What can we do
> to fix this?

1) Prevail upon the developers in question to work in a more consensual
style and publicise changes that will break non-POSIX builds *before*
installing them.

2) Work on fixing the w32 build system so it is less hand-crafted and
makes more use of the POSIX build machinery. Given the number of
different environments and toolchains used to build for win32 targets,
this is tricky.

> I get the general impression via the 50+ message thread about DOS port
> compatibility that the gnulib people are mostly unwilling to make
> efforts to avoid breaking the emacs build.  Is this the case, or are
> we simply seeing bad luck?
>
> So far, the features of gnulib have not seemed to be worth the
> trouble.

As long as it does not restrict portability of emacs to platforms and
toolchains that gnulib want to support then there is a little temporary
pain and some longer term benefits.

If the two projects have divergent views on which platforms and
toolchains to support then that would indeed be a problem.

    AndyM