I have retested with the latest patch (from comment #15) included, the testing was done as in comment #13. As result: It all behaves nicely. Thanks for the patch, Pantxo!

Sorry, but the "mcode" approach doesn't work with non integer figure handles, and "close (gcbf ())" won't work either due to bug #55428 ... so I'll rework the patch before pushing.

I marked the bug as fixed at https://wiki.octave.org/Bug_Fix_List_-_5.0_Release#Bugs_marked_as_Crash.

There are some other crashes that are probably fixed by this such as

bug #49758: octave-cli crashes when closing fullscreen plot

I attached an updated patch that fixes the issue for me (and correctly goes through gh_manager::execute_callback). If someone can confirm that it works, I'll push it on stable soon.

(file #45926)

I have compiled and tested a current dev Octave (hg-id bf835b1d04b1) without any additional patches:

• I can reproduce the CRASH from comment #4
• I can reproduce the CRASH from comment #0
• (I have not tested the other ways of crashing)

When I then add Pantxo's patch from comment #7 and recompile:

• I canNOT provoce any crash when trying the steps from comment #4. I can click as often as I want on the x in the plot windows. Using Ctrl-C afterwards in the Octave console will properly stop the while look and Octave is still alive.
• I also canNOT provoce any crash with the steps in comment #0. The figure window just nicly closes, when clicking the x in the window.
• Also the figure windows from comment #2 just neatly closes, when clicking the x in the figure window. NO CRASH.
• I don't know if I did those steps all right (because I haven't tested this with unpatched Octave). But I was not able to provoce ANY CRASH with the steps mentioned in comment #6 either. Clicking on x closes the figure window fine. But to get my Octave console back, I still also have to Ctrl-C. But this seems fine to me (the code is an endless loop nevertheless.)

So this patch is definitly an improvement. Can anyone else confirm, that Pantxo's patch fully FIXES this issue?

I can't reproduce your suggestion with Octave from development branch. But i can crash it if i run

figure();
while 1
end

Then click on the figure "x" button. Then go to the command window and press Ctrl-C. Then octave exits with signal 11.

I tried another simpler example:
1) edit a function file tmp2.m (see below);
2) if run "tmp2", and press Ctrl+C, then nothing abnormal;
3) if run "tmp2", use mouse to close the figure ONCE (but the figure is not closed immediately in octave 4.4.0), and press Ctrl+C (until now the figure is closed in octave 4.4.0);
4) if run "tmp2", use mouse to close the figure TWICE, and press Ctrl+C, then the function "close" will report an error.
5) if run "tmp2", use mouse to close the figure THREE TIMES OR MORE, and press Ctrl+C, then the function "close" will report an error, and octave will crash with segfault.

user@debian:~$  cat tmp2.m
function tmp2
  a = imread('octave-sombrero.png');
  h = imshow(a);
  while (ishandle(h))
    a = shift(a,1);
    set(h,'cdata',a);
    drawnow;
    pause(0.01);
  end
end

user@debian:~$  octave -q
octave:1> tmp2 # do not click "X"
^C
octave:1> tmp2 # and click "X" once
^C
octave:1> tmp2 # and click "X" twice
^C
octave:1> error: close: first argument must be "all", a figure handle, or a figure name
error: parse error
error: called from
    close at line 81 column 7
octave:1> tmp2 # and click "X" three times or more
^C
octave:1> error: close: first argument must be "all", a figure handle, or a figure name
error: parse error
error: called from
    close at line 81 column 7
octave:1> fatal: caught signal Segmentation fault -- stopping myself...
Segmentation fault

user@debian:~$

When a figure is busy, closing it (e.g. click the "X" button, press Alt+F4, or Ctrl+W, etc.) does not work.
What's more, in this case, if some code tries to close it, then octave would crash with segfault.

This happens in octave 4.4.0; octave 4.2.x does not have this problem.  Is this a regression ?

My environment is Debian GNU/Linux with LXDE desktop.

Here is an example how to reproduce this bug:
1) edit a function file tmp.m (see below);
2) start "octave" or "octave --gui";
3) run "tmp" at the octave prompt and we would see an image moving in a figure;
4) NOW use mouse or shortcut key to close this figure (but this would FAIL);
5) THEN press Ctrl+C at the octave prompt to interrupt "tmp", this would run the onCleanup code which tries to close the figure (but this would CRASH octave with segfault)

user@debian:~$  cat tmp.m
function tmp
  a = imread('octave-sombrero.png');
  h = imshow(a);
  f = gcf();
  o = onCleanup(@() isfigure(f) && close(f));
  while (isfigure(f))
    a = shift(a,1);
    set(h,'cdata',a);
    drawnow;
    pause(0.01);
  end
end

user@debian:~$  octave
GNU Octave, version 4.4.0
Copyright (C) 2018 John W. Eaton and others.
This is free software; see the source code for copying conditions.
There is ABSOLUTELY NO WARRANTY; not even for MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE.  For details, type 'warranty'.

Octave was configured for "x86_64-pc-linux-gnu".

Additional information about Octave is available at https://www.octave.org.

Please contribute if you find this software useful.
For more information, visit https://www.octave.org/get-involved.html

Read https://www.octave.org/bugs.html to learn how to submit bug reports.
For information about changes from previous versions, type 'news'.

octave:1> tmp
^C
octave:1> fatal: caught signal Segmentation fault -- stopping myself...
Segmentation fault

user@debian:~$  octave --gui
octave exited with signal 11

user@debian:~$