Re: attribute warn_unused_result

[Lennart Borgman <lennart.borgman@gmail.com>]

On Fri, Feb 4, 2011 at 10:08 PM, Stefan Monnier
<monnier@iro.umontreal.ca> 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.
>> So what is wrong here, Paul? We have suggested that you make a
>> separate branch for this. Why don't you do that?
>
> AFAIK this last change introduced a bug, just like any other commit can
> (and sometimes does) introduce a bug because some simple omission that's
> easily fixed.  I.e. it's very different from the other cases where he
> knew before that his change was going to break the w32 build.
> IIUC people are reacting too strongly to this last problem because it
> just happens to follow the other ones, but really it's a completely
> different case.  I myself break the build on a regular basis as well, as
> you all know too well.

Yes, I realized that. A bit too late.

It would have been much easier to realize that with a little bit more
feedback. Mind reading is difficult face to face and even a little bit
more difficult over the net.


> BTW, I'd still like to see someone setup a build farm that can be used
> to keep track of "the last revision that bootstraps", so people who
> don't want to bump into such problems can follow that branch instead
> of trunk.

For me that would indeed be good especially when I need to test some
patch (which happens too often, actually).