Re: Opportunistic GC

[Pip Cet <pipcet@gmail.com>]

On Thu, Mar 11, 2021 at 9:34 AM Eli Zaretskii <eliz@gnu.org> wrote:
> > From: Pip Cet <pipcet@gmail.com>
> > Date: Thu, 11 Mar 2021 09:03:23 +0000
> > Cc: Stefan Monnier <monnier@iro.umontreal.ca>, Andrea Corallo <akrl@sdf.org>, emacs-devel@gnu.org
> >
> > > See thread_select.
> >
> > I have no intention of calling thread_select.
>
> If you call wait_reading_process_output, or any of its callers (such
> as accept-process-output or sleep-for), Emacs _will_ call
> thread_select.

Why would I do any of those things in the child process?

Note that they'll continue working fine in the parent process, which
is unaffected by the fork.

> > > We don't call pselect directly.
> >
> > In this case, we should. I don't see a problem with that, as the child
> > Emacs does not need to perform other I/O.
>
> If you want to do that for any of the above-mentioned calls, you will
> have to modify the code of wait_reading_process_output.

The child process's sole job is to inspect the heap, perform GC, and
return collectable objects to the parent via a pipe. It does not need
to call Lisp. It certainly shouldn't call thread_select, or
wait_reading_process_output.

What am I missing?

Pip