Re: Bug#23924 - crashes on next OS X release

[Alan Third]

Hi John, are you OK with the patch?

On 22 July 2016 at 22:54, Alan Third <address@hidden> wrote:

On Mon, Jul 18, 2016 at 08:03:46PM +0100, Alan Third wrote:
> On Mon, Jul 18, 2016 at 09:16:09AM -0700, John Wiegley wrote:
> > >>>>> "EZ" == Eli Zaretskii <address@hidden> writes:
> >
> > EZ> If you are confident that it won't do any trouble, and that 10.10, and
> > EZ> only that version, needs this fix, feel free to push to emacs-25.
> >
> > Before you do, can you show us the patch here so that I can have a sense of
> > what it changes?  I've been running emacs-25 crash-free on 10.11 for many
> > months now.
>
> I’m not that confident as I couldn’t find anything about the bug in
> question except what was in the bug report and comment in the code. I
> tried removing the fix and can’t reproduce the original problem on
> 10.11. It’s possible it’s actually been fixed in later releases of
> 10.10 too.

I’ve got a new patch. One of the testers of the first patch ran into
bug#18993, so it’s clearly still an issue, and further investigation
leaves me suspecting that, contrary to the bug report thread, it’s not
an OS X bug, but something to do with how Emacs is handling events.

The new patch leaves the fix in place and uses an existing GNUStep fix
to ensure that postEvent always runs in the main thread. That fixes
bug#23924.

There’s probably an argument to be made that we should fix 18993
properly, but I don’t think we have anyone who knows the NS port well
enough to do it, as it would require a substantial rewrite of the
event loop.

If this is to be applied to Emacs 25, let me know if it’s to go to
master too. I’m unsure how it’s handled.
--
Alan Third

--

Alan Third