GNOME Bugzilla – Bug 702370
a11y's /child/add-remove/GtkEntry test reads invalid memory => fails
Last modified: 2013-08-08 14:49:47 UTC
Hi, /child/add-remove/GtkEntry was failing in Debian/ia64, see http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=711107 Stephan Schreiber analyzed the failure and found that parent_data is never set in the GtkEntry test but is tested in an if. Indeed running valgrind on x86_64 I get: /child/add-remove/GtkEntry: ==10919== Conditional jump or move depends on uninitialised value(s) ==10919== at 0x40205D: test_add_remove (children.c:200) ==10919== by 0x7FD09BC: g_test_run_suite_internal (gtestutils.c:1714) ==10919== by 0x7FD0B35: g_test_run_suite_internal (gtestutils.c:1778) ==10919== by 0x7FD0B35: g_test_run_suite_internal (gtestutils.c:1778) ==10919== by 0x7FD0E9A: g_test_run_suite (gtestutils.c:1823) ==10919== by 0x4019D8: main (children.c:267) See the full analysis at: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=711107#42 I'm attaching his proposed patch. I have had a look but the test case has grown to be quite complex. I wonder if special cases such as the GtkEntry one shouldn't be split out into separate functions.
Created attachment 246911 [details] [review] Fix invalid read in /child/add-remove/GtkEntry Patch from Stephan Schreiber <info@fs-driver.org>
For the record, another fix was applied in: https://git.gnome.org/browse/gtk+/commit?id=da208f89d5c2722df372cd7f91d6046092eb9e45