bug#63589: [PATCH] 29.0.91; crash after creating graphical frames via emacsclient when compiled with cairo-xcb

[Eli Zaretskii <eliz@gnu.org>]

> From: Po Lu <luangruo@yahoo.com>
> Cc: tmdmelo@gmail.com,  63589@debbugs.gnu.org
> Date: Tue, 23 May 2023 21:18:04 +0800
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Then I don't understand your objections to the proposed patch.  Please
> > elaborate.
> 
> Please see the other thread(s), where I explained in detail the two
> problems with `gtk_init_check': upon success, it leaves two display
> connections open, which is quite fragile, and upon failure, it makes
> creating another display impossible, even if a display then becomes
> available.

I've read all those discussions in real time, and I still don't see
the obvious connection.  So please humor me with a more detailed and
complete explanation of why the last suggested patch somehow causes
extra connections open.  And let me remind you that your objection,
which is what caused my question, was

> We want closing displays (think x-delete-terminal) to still work on such
> builds if the user uses it.

Which seems to be about _closing_ connections, not about opening too
many of them.  It's the leap between the extra connections on the one
hand and closing display not working OTOH that I cannot make.  Please
help me fill the dots.

TIA