Serious stability problem in current pretest of Emacs 27.1

Serious stability problem in current pretest of Emacs 27.1

[Eli Zaretskii]

See bug#41321 for the details.  Did someone see anything similar?

Needless to say, this is a serious problem, so I'd like to ask
everyone to please run the latest pretest under a debugger and report
any similar problems with all the details they can provide.  We must
hunt down this bug and fix it before the release.

TIA

Re: Serious stability problem in current pretest of Emacs 27.1

[Dmitry Gutov]

On 16.05.2020 13:48, Eli Zaretskii wrote:
> See bug#41321 for the details.  Did someone see anything similar?

FWIW, emacs-27 has been very stable here in the recent months 
(GNU/Linux). No crashes. And I do use global-auto-revert-mode. CC 
Mode--not so much, only occasionally.

Some minor font rendering artefacts in the echo area are the main issue 
I've been considering reporting, but they're very sporadic, have no 
stable repro, and have been here for a while.

Re: Serious stability problem in current pretest of Emacs 27.1

[Eli Zaretskii]

> From: Dmitry Gutov <dgutov@yandex.ru>
> Date: Sat, 16 May 2020 21:25:48 +0300
> 
> On 16.05.2020 13:48, Eli Zaretskii wrote:
> > See bug#41321 for the details.  Did someone see anything similar?
> 
> FWIW, emacs-27 has been very stable here in the recent months 
> (GNU/Linux). No crashes. And I do use global-auto-revert-mode. CC 
> Mode--not so much, only occasionally.

Thanks for the feedback.  I don't use auto-revert, so if the problem
only happens when the changes are large (for some value of "large"),
auto-revert might prevent you from seeing that.

OTOH, if I'm the only one who sees this, then the problem is not so
serious.

Re: Serious stability problem in current pretest of Emacs 27.1

[Alan Mackenzie]

Hello, Eli.

On Sat, May 16, 2020 at 21:35:20 +0300, Eli Zaretskii wrote:
> > From: Dmitry Gutov <dgutov@yandex.ru>
> > Date: Sat, 16 May 2020 21:25:48 +0300

> > On 16.05.2020 13:48, Eli Zaretskii wrote:
> > > See bug#41321 for the details.  Did someone see anything similar?

> > FWIW, emacs-27 has been very stable here in the recent months 
> > (GNU/Linux). No crashes. And I do use global-auto-revert-mode. CC 
> > Mode--not so much, only occasionally.

> Thanks for the feedback.  I don't use auto-revert, so if the problem
> only happens when the changes are large (for some value of "large"),
> auto-revert might prevent you from seeing that.

> OTOH, if I'm the only one who sees this, then the problem is not so
> serious.

I think that would be being very optimistic.  There's a definite problem
here.  Maybe it only hits 32-bit systems, or only Windows systems.  But
...

It screams out "Finsert_file_contents" to me.  Maybe you've thought of
this yourself already, but the next time you need to revert a file
because of this, why not put a breakpoint at the start of that function
and eyeball all the clever manipulation it does with markers.  Or
something like that.

But there haven't been any recent changes to Finsert_file_contents in
the emacs-27 branch.  Maybe there's been a recent change to something it
depends on.

I'm not sure how helpful this will be, but I thought I'd say it anyhow.

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: Serious stability problem in current pretest of Emacs 27.1

[Dmitry Gutov]

On 16.05.2020 21:35, Eli Zaretskii wrote:
> I don't use auto-revert, so if the problem
> only happens when the changes are large (for some value of "large"),
> auto-revert might prevent you from seeing that.

I don't know: the largest changes usually come from 'git checkout', and 
then the order of their magnitude is probably the same whether you 
revert manually or automatically.

OTOH, the size of the files I work with is probably smaller than yours 
on average.

Re: Serious stability problem in current pretest of Emacs 27.1

[Eli Zaretskii]

> Date: Sat, 16 May 2020 19:25:31 +0000
> Cc: Dmitry Gutov <dgutov@yandex.ru>, emacs-devel@gnu.org
> From: Alan Mackenzie <acm@muc.de>
> 
> But there haven't been any recent changes to Finsert_file_contents in
> the emacs-27 branch.

Exactly.

> Maybe there's been a recent change to something it depends on.

Concrete ideas about the changes that could matter are welcome.