bug#4970: 23.1; Emacs Gtk running nuts

[Jan Djärv]

Eli Zaretskii skrev:
> > Date: Fri, 20 Nov 2009 11:37:01 +0100
> > From: Jan =?UTF-8?Q?Dj=C3=A4rv?= <jan.h.d@swipnet.se>
> > Cc: 4970@emacsbugs.donarmstrong.com
> >
> > Dan Nicolaescu skrev:
> > > Jan Djärv <jan.h.d@swipnet.se> writes:
> > >
> > >   > What happens is that reading from the terminal fails and Emacs tries
> > >   > to remove that terminal, but in term.c:
> > >   > 
> > >   >   if (last_terminal)
> > >   >       error ("Attempt to delete the sole terminal device with live frames");
> > >   > 
> > >   > 
> > >   > which goes back to the command loop, tries to read agan, fails, and
> > >   > tries to delete the terminal again, and so on.
> > >   > 
> > >   > If you remove this check, Emacs exits.  But I suppose it is there for
> > >   > a reason, but I don't know what.  Anybody?
> > >
> > > It's there so that if you do:
> > > emacs -Q -nw
> > > C-x 5 0
> > > does not exit emacs.
> > Well, the check in term.c isn't preventing that.  It is the check in frame.c 
> > delete_frame that does that:
> >
> >    if (NILP (force) && !other_visible_frames (f))
> >      error ("Attempt to delete the sole visible or iconified frame");
>
> What about delete-terminal?

Rhat check in term.c prevents delete-terminal from working when the FORCE 
argument is t.  Deleting an X11 terminal woth FORCE set to t makes a core dump...

> And btw, are there any live frames when the test in term.c is made, in
> the recipe to reproduce the original bug?  If not, maybe it needs to
> check for live frames explicitly.

Yes there are. When read_socket_hook returns -2 and this is the last terminal, 
a SIGHUP is sent to ourselves (why not just call shutdown_emacs and exit?) and 
Fdelete_terminal is called.  So frames has not been removed yet.

        Jan D.