Closed antoine-levitt closed 2 years ago
Wow -- this seems to indicate that the X server itself segfaulted. That isn't easy to do! This implies that something, possibly a MATE component (such as the Marco window manager, like you suggest) makes the X server crash.
I need to test this on MATE 1.26, but on a preliminary test with Debian Buster and MATE 1.20.4 (that's what I'm sitting in front of right now), I couldn't reproduce the error that you're encountering. (Yes, I set the keybinding for "Switch to Workspace 2" to F2.) I tried this with and without software compositing enabled.
What kind of graphics chip are you using? Sometimes graphics drivers, the X server, and/or the window manager manage to foul things up -- i.e. often the graphics drivers are buggy and the X server doesn't know about the quirk, so the X server does something that results in a crash. I've seen that happen before, especially with NVIDIA proprietary graphics drivers. (I did and will continue to do my reproducibility tests on Intel integrated graphics hardware, since I don't have access to any fancier stuff, but will try all the various drivers -- i915
, i965
, i965
Gallium, iris
, and if possible crocus
.)
Furthermore, do you have software compositing turned on? (See the System menu -> Preferences -> Look and Feel -> Windows, under the General tab, it should be called Enable software compositing window manager or something similar.) If you toggle that checkbox, does the crash still occur when you type those magic keys?
Finally, are you sure you're actually using Marco actively? Run this terminal command just to make sure:
pgrep marco
If that doesn't return any output, you're not running Marco. In that case please specify what window manager you are running. (Beware that I've heard stories that the Compiz window manager can wreak all kinds of havoc, so if you're using that for some reason, see if switching to Marco solves the problem.)
Thanks in advance for the information.
Thanks for the advice, much appreciated! Yup it's marco, without compositing. The python traceback seems to suggest something with applets and menus, and indeed alt+f2 runs the "Run application" menu. Actually now that I try to reproduce it again, I can't... I thought it was reproducible, having reproduced it twice, but it seems not. I'm closing this until I can reproduce. In the future if I manage to trigger it, is there some sort of better log I can use to report a bug?
First off, the menus stuff that you saw, assuming it's the same messages that you posted here, was semi-unrelated: The Mint Menu crashed because the X server crashed. It doesn't seem to me that the menu caused the X server to crash; if that were so, I'd expect the errors and warnings from the menu to come before the message that the X server dumped core.
OK, so if it's not compositing-related, then I wonder what it could be? I wasn't able to reproduce it myself, with Intel integrated graphics -- all the drivers I already mentioned, plus I also tried it on the Linux system closest to Mint that I have installed: Ubuntu MATE 21.04, with MATE 1.26 all throughout. I could not make anything crash, except for Caja (that's a totally unrelated story, I promise).
Well, if even you can't reproduce it now, then I guess I'll close the issue. If it ever comes back, you can re-open this issue, but I'd also suggest you try to get help from the Linux Mint community, since I have very little background and understanding of Linux Mint -- as far as I know, their MATE desktop is incredibly tweaked beyond recognition; I've heard that their MATE desktop really looks more like Cinnamon than it does true MATE. Who knows what bugs they might be introducing? I have no clue.
Alright, thanks a lot!
My keybindings are setup so that alt+esc does the "fast" window change (alt tab but without the menu), as by default, and f2 is "switch to workspace 2". I get a segfault (back to login manager) by the following key sequence : press alt and keep it pressed, press (and release) esc, press and keep pressed f2, release alt. This is on linux mint 21 mate, marco 1.26.0. Sorry if this is not the right place, I don't know where to report it otherwise.
The following lines in syslog seem relevant: