bug#46881: 28.0.50; pdumper dumping causes way too many syscalls

[Pip Cet]

On Wed, Mar 3, 2021 at 7:58 PM Alan Third <alan@idiocy.org> wrote:
> On Wed, Mar 03, 2021 at 07:10:28AM +0000, Pip Cet wrote:
> > > My quick test on macOS by doing:
> > >
> > > rm src/*.pdmp
> > > time make
> > >
> > > sees it going from ~26s without patch to ~10s with patch, so a
> > > considerable improvement.
> >
> > Thanks for testing!
> >
> > > > > Patch will be attached once this has a bug number.
> > > >
> > > > And here's the patch. Testing would be very appreciated.
> > >
> > > It appears to work fine here, but I don't know if there's anything
> > > specific to test other than just running Emacs.
> >
> > I suspect there may be problems on systems with very little memory. Do
> > you have an easy way to determine maximum resident size (on Debian
> > GNU/Linux, /bin/time works)? It would be interesting to see if that's
> > actually different.
>
> I tried using time -l the same as above and both came  out with
> roughly the same values, but I suspect that's probably just me getting
> the values for running make.

Thanks.

> Is there a better way of testing dumping?

I guess you could run "time  ./temacs --batch  -l loadup
--temacs=pbootstrap" directly (in src/)...

I realize that's not an answer to your question, but IME, dumping bugs
will often lead to immediate failures to bootstrap, whereas GC bugs
are tricky and don't. Since this bug doesn't affect GC in any way, I
think we've done enough initial testing.

Pip