Segfault on adding new location to calendar #75
Labels
Comments
|
With new gtk 3.24.11 (plus commits from that point to Sep 12 for proper meson build versioning) and glib 2.62.0 I got what might be a more useful backtrace, looks like problems in glib or gtk itself but not sure: |
|
It doesn't crash here in f30 with glib2-2.60.7-1.fc30.x86_64 and gtk3-3.24.11-1.fc30.x86_64. |
|
I can't reproduce this issue with |
|
I'm not sure this is anywhere in mate-panel itself given the backtrace. With gtk3.24.11 and glib 2.62 the segfault was on gtk_widget_get_request_mode called somewhere after gtk_widget_get_preferred_height_and_baseline_for_width, and that does not appear in our code, only in GTK itself. The whole mess seems to come out of gtk_main , posssibly from mate-panel-applet.c as I don't see any other cases of it that might be affected from fiddling with the clock. Other instances of gtk_main() in parents of this would be in main.c (unlikely it would seem here), or maybe panel.shell.c but all other instances of it are in things adding a location to the clock ought not to invoke or modify at all.
I've tested both with and without mate-desktop/mate-panel#1016 and gotten the same results, getting this segfault almost but not quite all of the time. |
|
New test with: mate-panel is built from released 1.23.1 tarball with support for X11, wayland and out-of-process-applets (all applets). |
|
Same result with matel-panel built from master and --with-in-process-applets=all |
|
I have no idea in the world what is different, what exact glib and GTK versions are you using? Stacktrace does not look like the segfault is from within our code but I'm not skilled enough in reading them to be sure. I do not have gtk and glib debugging symbols installed for space reasons. |
|
I was able reproduce the issue today when adding a location to clock-applet in fedora 32 or 30. |
|
My backtrace is different, here the clock-applet crashes with libsoup, which is a dependency from libmateweather. So, i assume it is more an issue with libmateweather. |
|
Reverting 4cad2cf seems to fix the issue. |
|
Any plans to make new release? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Expected behaviour
Adding location to calendar works normally
Actual behaviour
Panel segfaults on clicking "OK" button in dialog to add location
Steps to reproduce the behaviour
Open the calendar popup , it will come up with "locations" selected
click "edit" and then "add" buttons
Select any location
click "OK" and that's when the panel segfaults. When it restarts the location will have been sucessfully added.
Got this backtrace in gdb
MATE general version
1.23 from git master
Package version
local build from git master as of 9-11-2019
Linux Distribution
Debian Unstable last updated Sep 6
Link to downstream report of your Distribution
None as this is a local build
The text was updated successfully, but these errors were encountered: