Re: optional compat?

[Rich Felker <dalias@aerifal.cx>]

On Tue, Dec 31, 2013 at 09:23:37AM +0000, Justin Cormack wrote:
> On Sat, Dec 28, 2013 at 1:32 AM, Rich Felker <dalias@aerifal.cx> wrote:
> > On Fri, Dec 27, 2013 at 11:11:45PM +0000, Justin Cormack wrote:
> >> The include files sys/errno.h and sys/signal.h, to work around broken
> >> code are a bit inconsistent, one has a #warning and the other doesnt.
> >> I would actually rather that they were only installed if requested, as
> >> (in an unusual situation) they are actually causing me issues. It
> >> would seem nicer to have a make installbroken as well...
> >
> > Could you elaborate on how they're causing a problem? It's preferred
> > to avoid creating multiple configuration variants which would then
> > need to have their own application compatibility studies...
> 
> I worked out what the issue was - the #warning breaks the make depend
> step as it can't parse the output. So removing the warning fixes it,
> and so does removing the file as the dependency seems to be bogus
> anyway. Will still try to fix it upstream, but coming round to the
> view that if we have these files they shouldn't have #warnings in.

Is this script mixing stderr with stdout for the purpose of parsing?
Warnings should not affect parsing, and in fact other warnings could
potentially happen, e.g. with custom CFLAGS or newer compiler versions
that add new default warnings...

Rich