Re: master af152ffab1: Don't freeze if the compositing manager crashes

Re: master af152ffab1: Don't freeze if the compositing manager crashes

[Eli Zaretskii]

There are a few places in Emacs that wait for similar reasons, so my
suggestion is to do this like they do, i.e. by calling
wait_reading_process_output.  The advantage is that we let other code
and other threads to run while we wait.

Re: master af152ffab1: Don't freeze if the compositing manager crashes

[Po Lu]

Eli Zaretskii <eliz@gnu.org> writes:

> There are a few places in Emacs that wait for similar reasons, so my
> suggestion is to do this like they do, i.e. by calling
> wait_reading_process_output.  The advantage is that we let other code
> and other threads to run while we wait.

The problem here is Emacs doesn't just "wait" for an event, it has to
use X(Check)IfEvent to remove that single event from the event queue, in
effect processing it "out-of-order".

So unfortunately wait_reading_process_output doesn't quite fit the bill
here.

Re: master af152ffab1: Don't freeze if the compositing manager crashes

[Eli Zaretskii]

> From: Po Lu <luangruo@yahoo.com>
> Cc: emacs-devel@gnu.org
> Date: Sat, 30 Jul 2022 15:02:32 +0800
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > There are a few places in Emacs that wait for similar reasons, so my
> > suggestion is to do this like they do, i.e. by calling
> > wait_reading_process_output.  The advantage is that we let other code
> > and other threads to run while we wait.
> 
> The problem here is Emacs doesn't just "wait" for an event, it has to
> use X(Check)IfEvent to remove that single event from the event queue, in
> effect processing it "out-of-order".
> 
> So unfortunately wait_reading_process_output doesn't quite fit the bill
> here.

Why not?

Re: master af152ffab1: Don't freeze if the compositing manager crashes

[Po Lu]

Eli Zaretskii <eliz@gnu.org> writes:

> Why not?

Because wait_reading_process_output will call read_socket_hook to read
events, and at that point it's too late to extract the event we want out
of the queue.

Re: master af152ffab1: Don't freeze if the compositing manager crashes

[Eli Zaretskii]

> From: Po Lu <luangruo@yahoo.com>
> Cc: emacs-devel@gnu.org
> Date: Sat, 30 Jul 2022 16:37:37 +0800
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Why not?
> 
> Because wait_reading_process_output will call read_socket_hook to read
> events, and at that point it's too late to extract the event we want out
> of the queue.

Why too late?

Anyway, your change freezes Emacs for a whole second, which doesn't
seem to be a good idea.

Re: master af152ffab1: Don't freeze if the compositing manager crashes

[Po Lu]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Po Lu <luangruo@yahoo.com>
>> Cc: emacs-devel@gnu.org
>> Date: Sat, 30 Jul 2022 16:37:37 +0800
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>> 
>> > Why not?
>> 
>> Because wait_reading_process_output will call read_socket_hook to read
>> events, and at that point it's too late to extract the event we want out
>> of the queue.
>
> Why too late?

Because read_socket_hook has already removed the event (which is
probably not the event we want) from the queue.

> Anyway, your change freezes Emacs for a whole second, which doesn't
> seem to be a good idea.

It only freezes in a rare situation, which is when the compositing
manager crashes after the user has explictly enabled frame
synchronization.