[Bug 53458] New: windows.media.speech:speech - The 64-bit test_Recognition() gets unexpected ref count on Windows after ISpeechContinuousRecognitionSession_Release()
https://bugs.winehq.org/show_bug.cgi?id=53458
Bug ID: 53458 Summary: windows.media.speech:speech - The 64-bit test_Recognition() gets unexpected ref count on Windows after ISpeechContinuousRecognitionSession_Release() Product: Wine Version: unspecified Hardware: x86-64 OS: Windows Status: NEW Severity: normal Priority: P2 Component: -unknown Assignee: wine-bugs@winehq.org Reporter: fgouget@codeweavers.com
windows.media.speech:speech - The 64-bit test_Recognition() gets unexpected ref count on Windows after ISpeechContinuousRecognitionSession_Release():
w1064v1507, w1064v1607 and w1064-adm (21H2): speech.c:1781: Test failed: Got unexpected ref 6.
or on w1064v1709: speech.c:1781: Test failed: Got unexpected ref 6. speech.c:1784: Test failed: Got unexpected ref 3. speech.c:1787: Test failed: Got unexpected ref 2. speech.c:1790: Test failed: Got unexpected ref 1.
https://test.winehq.org/data/patterns.html#windows.media.speech:speech
This is specific to the 64-bit test and only happens on w1064v1507, w1064v1709 and w1064-adm (21H2).
Also these failures started as soon as test_Recognition() was introduced:
commit 0a4b2bbd55a015dc603d0e073f7d000a574ab1b4 Author: Bernhard Kölbl besentv@gmail.com Date: Mon Apr 25 23:18:06 2022 +0200
windows.media.speech/tests: Add some recognition tests.
Signed-off-by: Bernhard Kölbl besentv@gmail.com Signed-off-by: Alexandre Julliard julliard@winehq.org
https://bugs.winehq.org/show_bug.cgi?id=53458
François Gouget fgouget@codeweavers.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |regression, source, | |testcase Regression SHA1| |0a4b2bbd55a015dc603d0e073f7 | |d000a574ab1b4
https://bugs.winehq.org/show_bug.cgi?id=53458
--- Comment #1 from François Gouget fgouget@codeweavers.com --- We can actually get all invalid ref values up to 7.
https://bugs.winehq.org/show_bug.cgi?id=53458
Bernhard Kölbl besentv@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |besentv@gmail.com
--- Comment #2 from Bernhard Kölbl besentv@gmail.com --- Yeah, those ref checks are pretty flaky because the recognition session might outlive the test function, I wonder if I should just remove them all.
https://bugs.winehq.org/show_bug.cgi?id=53458
--- Comment #3 from François Gouget fgouget@codeweavers.com --- That's up to you:
1. Either these tests are very important and then it's worth spending time to make sure they are reliable.
2. Or they are not that important and removing them is the right thing to do.
Assuming the tests are important, how long can the recognition session outlive the test function? Is it a matter of waiting a few milliseconds or is this something that could stay around for seconds?
Another point to take into account is that this seems very specific to some Windows test configurations. One could initially think this is specific to old Windows 10 versions (1507, 1607, 1709) but w1064 (21H2) invalidates that theory since it's new and only the -adm snapshot is impacted, not the regular and -tsign ones. So there is something else at play here. If these tests are important then it may be worth figuring out what.
https://bugs.winehq.org/show_bug.cgi?id=53458
--- Comment #4 from Bernhard Kölbl besentv@gmail.com --- I plan to just remove the tests for the simple reason, that we already have a ref check for the recognition session at another place. There it's not expected to outlive the test, because it hasn't been started, yet. Starting it, seems to spawn a background thread or so
The -adm VM is only impacted in later Win10 versions, because Windows.Media.Speech just refuses to run with elevated privileges. The other VMs skip these tests.
https://bugs.winehq.org/show_bug.cgi?id=53458
François Gouget fgouget@codeweavers.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Resolution|--- |FIXED Status|NEW |RESOLVED Fixed by SHA1| |6dc873bcaeed77dfaabc6f9d51c | |dd9fd56cd1a32
--- Comment #5 from François Gouget fgouget@codeweavers.com --- This got fixed by:
commit 6dc873bcaeed77dfaabc6f9d51cdd9fd56cd1a32 Author: Bernhard Kölbl besentv@gmail.com Date: Tue Aug 23 13:33:06 2022 +0200
windows.media.speech/tests: Remove some flaky refcount checks.
They can fail, when the recognition session outlives the test function, which is expected to happen, as the session is concurrent.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=53458
https://bugs.winehq.org/show_bug.cgi?id=53458
Alexandre Julliard julliard@winehq.org changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #6 from Alexandre Julliard julliard@winehq.org --- Closing bugs fixed in 7.18.
-
WineHQ Bugzilla