Terry J. Reedy <tjre...@udel.edu> added the comment:
My position is that tests that are known to occasionally fail independently of any changes in particular PRs should not be allowed to block merging. In the context of CI, the failure is a lie. One solution, and the easiest, is to disable the test. That is all I could do here. If the module/test author wants to keep running the test, another solution is to change the test so that bogus results are no longer seen as a failure by the test framework. How depends on the situation. Perhaps turn particular exceptions into a warning message? In this case, there are multiple exceptions listed in the log: OSError: [WinError 6] The handle is invalid x 2 OSError: [WinError 995] The I/O operation has been aborted because of either a thread exit or an application request, leading to ConnectionResetError: [WinError 995] ... OSError: [WinError 64] The specified network name is no longer available, leading to another ConnectionResetError. I don't see am a little surprised at and don't understand the vague end failure message. So I cannot suggest anything more specific. The third solution I am thinking of is somehow change the test framework so we can somehow mark 'heisentests' as such and not count know false positives as merge-blocking failures. ---------- _______________________________________ Python tracker <rep...@bugs.python.org> <https://bugs.python.org/issue35620> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
