New occurrence of this GTK3 crash, this time in seahorse 41.0-2 on Ubuntu
22.04.5 (jammy), with a core that confirms the parent widget was already
finalized.

Environment
  Ubuntu 22.04.5 LTS, X11 session
  seahorse 41.0-2
  libgtk-3-0 3.24.33-1ubuntu2.2
    build-id 90b2d0ad0e5bb5d7d9acd575efe769320ee34f75
  kernel 6.8.0-149-generic
  command: /usr/bin/seahorse --gapplication-service

Crash
  2026-11-18 09:44:40Z, PID 51599, signal 11 (SEGV)
  #0  widget_needs_widget_path (widget=0x5e7728168e10)
        at gtk/gtkcsswidgetnode.c:171
  #1  gtk_css_widget_node_init_matcher ...:190
  #2  gtk_css_node_init_matcher ...:1416
  #3  gtk_css_node_create_style ...:370
  #4  gtk_css_node_real_update_style ...:425
  #5  gtk_css_node_ensure_style ...:1007
  #17 gtk_css_node_get_style ...:1033
  #18 gtk_style_context_lookup_style ...:3077
  #19 _gtk_style_context_get_pango_attributes ...:3079
  #20 gtk_entry_create_layout ...:6351
  #22 gtk_entry_get_cursor_locations ...:6889
  #23 update_im_cursor_location ...:6255
  #25 gtk_entry_preedit_changed_cb ...:6099
  #30 im-ibus.so
  #35 libibus-1.0.so.5
  #36 g_object_run_dispose
  #41 libgio-2.0.so.0

The faulting instruction is the same one as in the reports above:
mov 0x380(%rax),%rax with rax == 0.

Core analysis with libgtk-3-0-dbgsym
  In widget_needs_widget_path, parent == 0x5e7728161170.
  That parent object has g_class == NULL and ref_count == 0: it has
  already been finalized.
  The crashing child widget (0x5e7728168e10) is still alive (ref_count 1),
  but its priv->parent still points at the finalized parent. The
  GTK_CONTAINER_GET_CLASS(parent)->get_path_for_child access therefore
  dereferences NULL.

  The call originates from libibus disposing an input context
  (g_object_run_dispose) during widget teardown. The dispose emits a final
  preedit-changed, which reaches gtk_entry_preedit_changed_cb and re-enters
  GTK CSS style resolution on the half-destroyed widget tree.

Seahorse logged this 28 minutes before the crash:
  seahorse_key_manager_on_sidebar_selected_rows_changed:
  assertion 'place != NULL' failed
That is consistent with the key-manager UI being torn down while an
entry/IM context was still alive.

This is not nautilus-specific: any GTK3 application with an entry and an
active IBus context can reach the same path. The gtk+3.0 (Ubuntu Jammy) task
still needs a fix.

The raw core is not attached because a seahorse process can contain keyring
material. I can provide a redacted gdb extract (stack, registers, and the
freed-parent analysis above) if that helps.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1726129

Title:
  nautilus crashed with SIGSEGV in widget_needs_widget_path

To manage notifications about this bug go to:
https://bugs.launchpad.net/nautilus/+bug/1726129/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to