https://bugs.winehq.org/show_bug.cgi?id=54975
Bug ID: 54975 Summary: win32u:win32u exits prematurely, leaves child process behind, interferes with windows.globalization:globalization Product: Wine Version: unspecified Hardware: x86-64 OS: Linux Status: NEW Severity: normal Priority: P2 Component: win32u Assignee: wine-bugs@winehq.org Reporter: fgouget@codeweavers.com Distribution: ---
win32u:win32u exits prematurely, leaves child process behind:
win32u:win32u start dlls/win32u/tests/win32u.c [...] win32u.c:1037: Test marked todo: res = 4 win32u.c:1040: Test marked todo: buf[1] = 0 win32u:win32u:09f0 done (0) in 0s 1295B win32u.c:1045: Test failed: res = 0 win32u.c:1046: Test failed: buf = d win32u.c:1049: Test failed: res = 0 win32u.c:1059: Test failed: res = 0 09e0:win32u: 14 tests executed (2 marked as todo, 0 as flaky, 4 failures), 0 skipped.
See https://test.winehq.org/data/patterns.html#win32u:win32u
There are two things to notice in the above log: * The main win32u:win32u process is 09e0 but it did not issue a 'tests executed' summary. This means it exited prematurely. * 09e0 is a child process that continues running and printing messages which interferes with the next test, specifically windows.globalization:globalization.
Also on the TestBot one can find the corresponding Wine log which contains these lines that seem strongly correlated to this issue:
0480:err:seh:KiUserCallbackDispatcher ignoring exception 0480:err:seh:KiUserCallbackDispatcher ignoring exception wine client error:480: partial write 57344 Running: win32u:win32u (694 of 774) Running tests - 6 failures
This issue only happens on Linux and happens systematically but only on three configurations: fg-deb64-t32 but not for the wow32 and wow64 builds on the same machine fgtb-debian11-wow64 but not for the wow32 build on the same machine debian11b-wow64 but but not for the wow32 build on the same machine (and none of the 32-bit builds on debian11 is impacted)
And I have been unable to reproduce this failure on my machine (fg-deb64-t32). So it's unclear what causes this failure but it clearly started on 2023-05-15 so it's probably some commit on that day.