Re: master 4ac4cec652f: Prevent freezes on macOS (bug#69561)

Re: master 4ac4cec652f: Prevent freezes on macOS (bug#69561)

[Po Lu]

Gerd Moellmann <gerd@gnu.org> writes:

>    if (NSApp == nil
>        || ![NSThread isMainThread]
>        || (timeout && timeout->tv_sec == 0 && timeout->tv_nsec == 0))
> -    return thread_select (pselect, nfds, readfds, writefds,
> -			  exceptfds, timeout, sigmask);
> +    thread_select (pselect, nfds, readfds, writefds,
> +		   exceptfds, timeout, sigmask);

Far be it from me to offer advice on code I know very little about, but
on the surface this appears to force non-main threads to enter [NSApp
run] whenever they enter select.  Is this really the desired behavior?

I should think that this function, being provided by the toolkit for the
purpose of running its event loop, must be called by strictly one thread
at a time.

Re: master 4ac4cec652f: Prevent freezes on macOS (bug#69561)

[Gerd Möllmann]

Po Lu <luangruo@yahoo.com> writes:

> Gerd Moellmann <gerd@gnu.org> writes:
>
>>    if (NSApp == nil
>>        || ![NSThread isMainThread]
>>        || (timeout && timeout->tv_sec == 0 && timeout->tv_nsec == 0))
>> -    return thread_select (pselect, nfds, readfds, writefds,
>> -			  exceptfds, timeout, sigmask);
>> +    thread_select (pselect, nfds, readfds, writefds,
>> +		   exceptfds, timeout, sigmask);
>
> Far be it from me to offer advice on code I know very little about, but
> on the surface this appears to force non-main threads to enter [NSApp
> run] whenever they enter select.  Is this really the desired behavior?
>
> I should think that this function, being provided by the toolkit for the
> purpose of running its event loop, must be called by strictly one thread
> at a time.

Well, let me first make the usual disclaimer that I don't know much
about about Cocoa. (And don't really want to, I must admit ;-)).

I couldn't find much advice which I'd call definite on using Cocoa from
multiple threads, except

  https://developer.apple.com/library/archive/documentation/Cocoa/Conceptual/Multithreading/ThreadSafetySummary/ThreadSafetySummary.html#:~:text=Guidelines%20for%20using%20Cocoa%20from,are%20generally%20not%20thread%2Dsafe.

I interpret this as meaning that it's possible to do this, considering
that Emacs' threads run one at a time only, AFAIU. If that's no longer
the case, we have a problem, I agree.

Re: master 4ac4cec652f: Prevent freezes on macOS (bug#69561)

[Po Lu]

Gerd Möllmann <gerd.moellmann@gmail.com> writes:

> I interpret this as meaning that it's possible to do this, considering
> that Emacs' threads run one at a time only, AFAIU.  If that's no
> longer the case, we have a problem, I agree.

My mistake.  I assumed thread_select was the function calling ns_select,
not the other way around.  Nevertheless, we've received reports of the
toolkit aborting when invoked from a non-main thread, so would you
please test this change in an environment where several threads are
active?

Re: master 4ac4cec652f: Prevent freezes on macOS (bug#69561)

[Gerd Möllmann]

Po Lu <luangruo@yahoo.com> writes:

> Gerd Möllmann <gerd.moellmann@gmail.com> writes:
>
>> I interpret this as meaning that it's possible to do this, considering
>> that Emacs' threads run one at a time only, AFAIU.  If that's no
>> longer the case, we have a problem, I agree.
>
> My mistake.  I assumed thread_select was the function calling ns_select,
> not the other way around.  Nevertheless, we've received reports of the
> toolkit aborting when invoked from a non-main thread, so would you
> please test this change in an environment where several threads are
> active?

I'll put that on my todo list, but it can take me some time.

I think the change is necessary if or if not what you describe can
happen, because without it Emacs freezes for me somewhere between 0s and
a few hours after starting it. With the fix, it didn't freeze at all.