bpo-35189, bpo-35316: Make test_eintr less strict #10782
Conversation
test_eintr no longer fails if the signal handler has not been called.
|
I recently added a check in test_eintr to see if the signal handler has been called at least once in each test method. Sadly the test fails randomly on macOS. I don't want to bother to check wait, I propose to simply remove the check. I wasn't sure that it was a good idea to add such check in the first place. test_eintr is already causing enough troubles: |
There was a problem hiding this comment.
LGTM
I still would want to investigate what's going on with SIGALRM in MacOs (https://bugs.python.org/issue35316) because we could be missing something,
but I don't mind removing flaky extra checks :)
|
@pablogsal: My intent for this PR is to repair the CIs, to make them more reliable. It's now clear to me that the check fails sometimes. If you would like to investigate: please, go ahead ;-) ... Have a good luck. Writing reliable test on EINTR is painful, and it seems like it's even more funny on macOS! Once I wanted to port test_eintr to Windows, but I gave up on this idea. |
|
Thanks @vstinner for the PR 🌮🎉.. I'm working now to backport this PR to: 3.6, 3.7. |
test_eintr no longer fails if the signal handler has not been called. (cherry picked from commit 2956bff) Co-authored-by: Victor Stinner <vstinner@redhat.com>
|
GH-10784 is a backport of this pull request to the 3.7 branch. |
|
GH-10785 is a backport of this pull request to the 3.6 branch. |
test_eintr no longer fails if the signal handler has not been called. (cherry picked from commit 2956bff) Co-authored-by: Victor Stinner <vstinner@redhat.com>
test_eintr no longer fails if the signal handler has not been called.
https://bugs.python.org/issue35189