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

[Alan Third]

On Tue, Nov 02, 2021 at 10:15:29PM +0200, Eli Zaretskii wrote:
> > Date: Tue, 2 Nov 2021 19:58:10 +0000
> > From: Alan Third <alan@idiocy.org>
> > Cc: Philipp Stephani <p.stephani2@gmail.com>
> > 
> > When we put in the fix to use vfork on macOS back in Emacs 26 the
> > measured difference between fork and vfork was something like 15x. Now
> > Apple have turned vfork into fork I guess we're back to the same
> > situation. Packages that make heavy use of external processes, such as
> > Magit, are badly hit by this slow down.
> > 
> > Is there anything we need to do to get scratch/posix-spawn-no-gnulib
> > merged into master (or even emacs-28)?
> 
> Does it solve the problem?  And which builds does it affect?

As I understand it Yamamoto Mitsuharu has already merged it into the
Mac port and I've seen it reported that the problems are solved with
it.

I don't have a Mac with the latest macOS on to test it myself.

This is really only a problem on macOS, so we could limit the use of
posix_spawn to Darwin builds.

I am concerned that there could be hidden issues with the branch since
it doesn't seem to have had much testing, so it's probably best going
into master, but I expect we'll see an increasing number of complaints
about performance as more people switch to macOS 12. Perhaps we should
keep it in mind as a possible update for 28.2, whenever that may be.

-- 
Alan Third