bug#48629: 28.0.50; GUI emacsclient frames stop accepting keyboard input around recv

["Basil L. Contovounesios" <contovob@tcd.ie>]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: "Basil L. Contovounesios" <contovob@tcd.ie>
>> Cc: Eli Zaretskii <eliz@gnu.org>,  48629@debbugs.gnu.org
>> Date: Wed, 26 May 2021 16:58:17 +0100
>> 
>> Thread 1 (Thread 0x7fb2f78e9040 (LWP 46955) "emacs"):
>> #0 0x00007fb2fa7ae9c6 in __pselect (nfds=24, readfds=0x7ffc492342e0,
>> writefds=0x7ffc49234360, exceptfds=0x0, timeout=<optimized out>,
>> sigmask=0x7ffc49234130) at ../sysdeps/unix/sysv/linux/pselect.c:48
>>         resultvar = 18446744073709551102
>>         sc_cancel_oldtype = 0
>>         tval = {tv_sec = 14, tv_nsec = 461936780}
>>         data = {ss = 0, ss_len = 8}
>> #1  0x000055dfc63615b6 in really_call_select ()
>> #2  0x000055dfc6362320 in thread_select ()
>> #3  0x000055dfc637ef58 in xg_select ()
>> #4  0x000055dfc633f89d in wait_reading_process_output ()
>> #5  0x000055dfc62843e8 in read_char ()
>> #6  0x000055dfc6286822 in read_key_sequence ()
>> #7  0x000055dfc628822c in command_loop_1 ()
>
> This is slightly more interesting, because it shows that timeout for
> pselect was 14 sec plus change.  So the interesting question is what
> happens after the timeout expires?
>
> IOW, when you find Emacs stuck in pselect, type "finish" and let Emacs
> exit the pselect call.

I did this, and it returned zero after a while:

Attaching to process 46955
[New LWP 46956]
[New LWP 46957]
[New LWP 46958]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
0x00007fb2fa7ae9c6 in __pselect (nfds=24, readfds=0x7ffc492342e0, writefds=0x7ffc49234360, exceptfds=0x0, 
    timeout=<optimized out>, sigmask=0x7ffc49234130) at ../sysdeps/unix/sysv/linux/pselect.c:48
48      ../sysdeps/unix/sysv/linux/pselect.c: No such file or directory.
(gdb) finish
Run till exit from #0  0x00007fb2fa7ae9c6 in __pselect (nfds=24, readfds=0x7ffc492342e0, writefds=0x7ffc49234360, 
    exceptfds=0x0, timeout=<optimized out>, sigmask=0x7ffc49234130) at ../sysdeps/unix/sysv/linux/pselect.c:48
0x000055dfc63615b6 in really_call_select ()
Value returned is $1 = 0

> Then step through the code and see what it does and why it doesn't
> respond to keyboard input.

I guess I need debugging symbols for this, because at this point
step/next progressively and slowly climb frames, without any indication
of the surrounding code or values (GDB maintains that it's in pselect.c
the whole time).

> Is the keyboard descriptor in the set of descriptors pselect waits
> for?

How do I determine what the keyboard descriptor is?

Thanks,

-- 
Basil