Try to make test_tdf81457_table_merge_undo deterministic
As discussed on #libreoffice-dev: Aug 07 08:48:59 <sberg> vmiklos, the ASan+UBSan bot fails for a while now in UITest_writer_tests, e.g., <https://ci.libreoffice.org//job/lo_ubsan/995/ consoleFull#-14523843999567f988-cbcf-4519-af05-6000b834f13f>; a bit unclear when it started, as it might initially have been hidden by another (meanwhile fixed) issue that stopped the build before reaching this test; I can't reproduce it with my local ASan+UBSan build; I guess something timing dependent, the test code is at sw/qa/uitest/writer_tests/tdf79569.py:32 (i.e., the second .uno:Undo and hits the assert at sw/source/core/layout/findfrm.cxx:1702 (search for it further up in the log), after preceding warnings like "Didn't find wished box"; any idea whether anything there could be timing dependent? Aug 07 09:09:48 <vmiklos> sberg: i'm not sure if any of that is executed synchronously, actually. all of these .uno:Foo commands are dispatched async IIRC Aug 07 09:13:03 <vmiklos> sberg: the easiest i could think of (to find out if my assumption is correct) would be to insert the pyuno equivalent of Scheduler::ProcessEventsToIdle() after the dispatches and see if that helps. (we should have an UNO interface + service that exposes that) Not sure whether all of the executeCommand/executeAction would need to be followed by processEventsToIdle, but probably doesn't hurt to be conservative there (except maybe for increased run time?). Change-Id: Iac64330def636a7b7db958e83feae9ce66d0677b Reviewed-on: https://gerrit.libreoffice.org/58663 Tested-by: Jenkins Reviewed-by:Stephan Bergmann <sbergman@redhat.com>
Showing
Please
register
or
sign in
to comment