Re: [PATCH] Cygwin: signal.cc: Include <unistd.h>

[Corinna Vinschen <vinschen@redhat.com>]

On May 12 08:48, Brian Inglis wrote:
> On 2022-05-12 02:46, Corinna Vinschen wrote:
> > On May 12 10:22, Sebastian Huber wrote:
> > > On 12/05/2022 10:14, Corinna Vinschen wrote:
> > > > >    sleep (unsigned int seconds)
> > > > >    {
> > > > >      struct timespec req, rem;
> > > > > @@ -174,7 +175,7 @@ sleep (unsigned int seconds)
> > > > >      return 0;
> > > > >    }
> > > > > -extern "C" unsigned int
> > > > > +int
> > > > >    usleep (useconds_t useconds)
> > > > Good catch, otherwise, thank you!  I'd prefer if you push just the
> > > > type change and keep the extern "C" as a marker for exported functions.
> > > 
> > > Ok, what about the unistd.h include?
> > 
> > Let's keep it.  It might help to uncover other issues in future.
> 
> Shouldn't these Cygwin winsup patches be posted separately to
> cygwin-patches rather than newlib, once any newlib patches they depend on
> have been accepted.

Not really.  If the patch changes something generic which Cygwin uses,
then the patch should ideally contain the required Cygwin patch, too.
Otherwise you'd have a deliberately build breakage in git history.  We
have a few of them in the past, they are a PITA.


Corinna