long-standing GTK bug (was: Log out hanging after X-forwarded emacsclient)

[andres . ramirez]

Hi Adam.

>>>>> "Adam" == Adam Sjøgren via \"Emacs development discussions 
>>>>> <address@hidden> writes:

[...]

    Adam> The reason is found in src/frame.c:

Which hints leads You from xterm.c to frame.c?

[...]

    Adam> However, I can't reproduce the problem when I build Emacs
    Adam> with Lucid.  Closing all frames on a remote display doesn't
    Adam> result in a crash with Lucid - and the prompt/ssh isn't
    Adam> hanging with L2152-2153 removed.

Next time I compile emacs. I am going to remove above lines for giving
it more testing. I use lucid toolkit.

    Adam> So a step towards fixing the GTK problem is to figure out
    Adam> whether the "workaround" in src/frame.c is still needed for
    Adam> Lucid and Motif.

I think the above could just improve user experience when using lucid
toolking combined with X-forwarding.

    Adam> Ah, I've now read through the bugs referenced in the
    Adam> comment, and it sounds like the problem that was handled by
    Adam> adding this doesn't happen every time. When
    Adam> x_delete_terminal() gets called much less, the described
    Adam> crashes are less likely to happen.

I have skimmed also the bugs referenced there. A lot of info and
situations. All of that cases would need retest.


[...]


    Adam> Hm, ok. I just thought I would follow up on my question.

Nice job. This bug affects a lot of emacs users. on 2016 I digged a
little bit on this bug aka "long-standing GTK bug". I still have my
notes from 2016. I even sketched a plan (I have never fullfilled):

Those were my conclusions from 2016:
--8<---------------cut here---------------start------------->8---
compile same source code for lucid, gtk, and nextstep
remove the abort calls {after checking all bug reports}
repeat that old gtk bug on gtk and nexstep {question is nextstep is immune as 
lucid}
research how was included the lucid port {get the git hash; why gtk is 
different from lucid}
the mac port could be useful? {macport suffers the same issue as gtk when used 
with x-forwarding}
--8<---------------cut here---------------end--------------->8---

Now on 2020. I would add to that list retest all cases found on frame.c
according to your email.

Best Regards