Re: long-standing GTK bug

[Robert Pluim]

>>>>> On Mon, 10 May 2021 11:00:08 -0400, Stefan Monnier 
>>>>> <monnier@iro.umontreal.ca> said:

    Stefan> Robert Pluim [2021-05-10 16:45:05] wrote:
    >>>>>>> On Mon, 10 May 2021 10:40:20 -0400, Stefan Monnier 
<monnier@iro.umontreal.ca> said:
    Stefan> IIUC the difference (with the patch) is between stopping ssh via 
`C-c` or
    Stefan> via `kill`.  I wonder why this matters, and I think it's not at all
    Stefan> obvious if an "abrupt" network disconnection would be more like 
`C-c` or
    Stefan> more like `kill`.
    >> In the 'C-c' case Emacs has had a chance to clean up its internal
    >> state (with the patch applied).

    Stefan> Why?  What does ssh do that is gentler?

I assume itʼs just shutting down the X forwarding connection
gracefully. Since that doesnʼt result in Emacs losing its existing
local display connection, thereʼs no crash.

    >> With 'kill' weʼre pulling the rug out from under GTK (see the
    >> backtrace several messages up-thread).

    Stefan> I suspect it then also depends on the `kill` signal.

    Stefan> In my WM (ctwm), I can also distinguish between `f.delete` and
    Stefan> `f.destroy` where the first sends something like a WM_DELETE 
request to
    Stefan> the application (which Emacs treats via `handle-delete-frame`, IIRC)
    Stefan> whereas `f.destroy` asks the X server to cut the connection AFAIK.

    Stefan> Would `f.destroy` be more like the `C-c` or the `kill` case of ssh?

I suspect like 'kill'.

Robert
--