Re: master 2c79a8f 2/2: Use posix_spawn if possible.

[Philipp Stephani]

Am Mi., 30. Dez. 2020 um 04:33 Uhr schrieb Eli Zaretskii <eliz@gnu.org>:
>
> > From: Philipp Stephani <p.stephani2@gmail.com>
> > Date: Tue, 29 Dec 2020 22:36:40 +0100
> > Cc: Emacs developers <emacs-devel@gnu.org>
> >
> > >     CC       getdtablesize.o
> > >   getdtablesize.c:58:1: warning: no previous prototype for 
> > > 'getdtablesize' [-Wmissing-prototypes]
> > >      58 | getdtablesize (void)
> > >         | ^~~~~~~~~~~~~
> >
> > Ah, I guess this is because the MinGW build excludes Gnulib's unistd.h
> > etc. So for now it looks like we do have to exclude the spawn modules
> > from the Windows build (unless we can include the POSIX headers from
> > Gnulib).
>
> The Gnulib unistd.h caused trouble in the MinGW build, so we disabled
> it long ago.  On the trunk, we don't compile getdtablesize.c, either,
> and I don't see why we would need to.
>
> (It's not that we always exclude headers from Gnulib on Windows, we
> only do that with headers that cause trouble.)

I've now created the branch scratch/posix-spawn-no-gnulib that calls
the libc posix_spawn function/syscall directly. I've realized that for
now that's the better approach, as the Gnulib replacement doesn't let
us call setsid() yet (that's a new functionality that is available on
GNU/Linux and macOS, but not yet in Gnulib). This should also lower
the impact on the Windows build.