02:53 <dnkl> sewn: foot should exit, but not segfault

02:54 <sewn> well it occurs when I want to halt my system, and a signal is sent to all processes to suicide which also kills wayland and its sockets

02:55 <dnkl> if you can get a backtrace from a debug build, I'll take a look

02:56 <sewn> ok I'll see

04:41 <dnkl> sewn: which signal is being sent to all processes? SIGTERM? Might be the signal that triggers the segfault, rather than the compositor dying.

06:02 <dnkl> fwiw, I can't reproduce. Neither by killing the compositor, nor by sending various signals to the foot process

10:43 <llyyr> I can reproduce it by sending SIGTERM to the foot process, but it only happens if the cursor is hovering on the foot window when you send the SIGTERM

10:43 <llyyr> I'll get a backtrace

10:43 <llyyr> dnkl: https://0x0.st/HjDF.txt

10:50 <dnkl> llyyr: it doesn't reproduce on River. Which compositor are you running?

10:50 <llyyr> sway with the scene graph PR

10:50 <llyyr> I can try with master branch as well

10:52 <llyyr> Oh, can't reproduce on master

10:52 <llyyr> I guess it's a bug with the PR?

10:54 <dnkl> maybe, but not necessarily. Only way to tell is to try to figure out why it crashes...

11:05 <dnkl> the chain of events is: term_destroy() -> wayl_win_destroy() -> keyboard_leave() -> term_xcursor_update_for_seat() -> term_mouse_grabbed().

11:06 <dnkl> there we crash, because term_destroy() had already freed the key bindings associated with the terminal instance

11:07 <dnkl> we should be able to move the destruction of the key bindings until after the window has been destroyed

11:15 <llyyr> trace without optimized out just in case it's helpful https://0x0.st/HjkH.txt

11:24 <dnkl> llyyr: thanks, that shows my guess was right: 'bindings' is null in the top most frame

11:27 <dnkl> I'd consider this a bug in foot. One could question wether it makes sense for the compositor to send a keyboard_leave event after the client explicitly unmapped the window. But doesn't really matter... this should be easy to handle in foot

11:34 <llyyr> thanks

14:17 <dnkl> llyyr: any chance you can test https://codeberg.org/dnkl/foot/pulls/1412?

15:23 <llyyr> yep I'll give it a try in a sec

15:27 <llyyr> it fixes the crash, thanks for the quick fix

17:42 <Arnavion> (Also doesn't repro in sway 1.8.1)