Title: Windows buildbots hang after fatal exit
Type: Stage: patch review
Components: Tests, Windows Versions: Python 3.11, Python 3.10, Python 3.9
Status: open Resolution:
Dependencies: Superseder:
Assigned To: Nosy List: jkloth, paul.moore, steve.dower, tim.golden, zach.ware
Priority: normal Keywords: patch

Created on 2021-06-07 17:07 by jkloth, last changed 2021-06-10 01:34 by jkloth.

Pull Requests
URL Status Linked Edit
PR 26578 open jkloth, 2021-06-07 17:16
Messages (4)
msg395268 - (view) Author: Jeremy Kloth (jkloth) * Date: 2021-06-07 17:07
Currently, a stack overflow is causing the debug build Windows buildbots to abort (bpo-11105).  Once the regrtest process is terminated, the buildbot test process hangs indefinitely waiting for handles to be closed (see msg350191 from bpo-37531 for some details).
msg395269 - (view) Author: Jeremy Kloth (jkloth) * Date: 2021-06-07 17:24
To verify the PR, can someone please add the test-with-buildbots label on GH?
msg395274 - (view) Author: Jeremy Kloth (jkloth) * Date: 2021-06-07 18:46
The PR has been successfully run on the buildbots.



With these changes, at least now aborted runs can be seen as direct failures of Python's tests instead of just builder exceptions.
msg395496 - (view) Author: Jeremy Kloth (jkloth) * Date: 2021-06-10 01:34
While now not as immediately beneficial, I believe that the linked PR would be good for the long run.  The ramifications of bpo-11105 meant that the Windows buildbots were basically unusable for 5 days.

Realistically, any commit that triggers aborts in the Windows test runs will exhibit this problematic behavior.  I'm confident that the PR is in ready-to-go form.
Date User Action Args
2021-06-10 01:34:56jklothsetmessages: + msg395496
2021-06-07 18:46:14jklothsetmessages: + msg395274
2021-06-07 17:24:36jklothsetmessages: + msg395269
2021-06-07 17:16:40jklothsetkeywords: + patch
stage: patch review
pull_requests: + pull_request25166
2021-06-07 17:07:52jklothcreate