Closed marianosimone closed 1 year ago
Thanks for reporting. Yes I think this particular test is flakey even in our internal CI. We haven't really gotten the time to sit down and debug this.
If you like you are welcome to send us your local patch and we can get it merged.
I have a proposed fix in [watchman/pr/1103] 🐛 Fix flaky test_sinceRelativeRoot
I'm trying to build from source, and I keep getting flaky results on https://github.com/facebook/watchman/blob/0b238b069f54a4de27c35a3e2352c37bccc12d60/watchman/integration/test_since.py#L90
While debugging, I added some log lines, and it immediately stopped flaking (Heisenbug flag!)
I haven't fully looked into the implementation to understand this 100%, but could it be that:
https://github.com/facebook/watchman/blob/0b238b069f54a4de27c35a3e2352c37bccc12d60/watchman/integration/test_since.py#L78-L89
can flake if not enough time passes between the assignment to
clock
and thequery
?I've locally patched this to mock the
clock
with an older time, and it stopped flaking