Open bcmills opened 11 months ago
Sorry, but there were parse errors in the watch flakes script. The script I found was:
#!watchflakes
post <- pkg == "cmd/go" && test == "TestScript" && `cover_.*.txt:\d+: .*link.* unexpected success
And the problems were:
script:2.52: newline in backquoted regexp
See https://go.dev/wiki/Watchflakes for details.
Found new dashboard test flakes for:
#!watchflakes
post <- pkg == "cmd/go" && test == "TestScript" && `cover_.*.txt:\d+: .*link.* unexpected success`
(attn @thanm)
Seems pretty clear from the failing logs that for some reason the Go command decided that the the cache entry was stale and the action needed to be redone... which is a puzzle to me: seems like all the action inputs should be deterministic. Not sure how we could have flakiness in such a situation.
Hmm. I wonder if this is related to #49267?
That does look related, thanks
Found new dashboard test flakes for:
#!watchflakes
post <- pkg == "cmd/go" && test == "TestScript" && `cover_.*.txt:\d+: .*link.* unexpected success`
Found new dashboard test flakes for:
#!watchflakes
post <- pkg == "cmd/go" && test == "TestScript" && `cover_.*.txt:\d+: .*link.* unexpected success`
Found new dashboard test flakes for:
#!watchflakes
post <- pkg == "cmd/go" && test == "TestScript" && `cover_.*.txt:\d+: .*link.* unexpected success`
Found new dashboard test flakes for:
#!watchflakes
post <- pkg == "cmd/go" && test == "TestScript" && `cover_.*.txt:\d+: .*link.* unexpected success`
Found new dashboard test flakes for:
#!watchflakes
post <- pkg == "cmd/go" && test == "TestScript" && `cover_.*.txt:\d+: .*link.* unexpected success`
It looks like I can reproduce the problem with this script:
I still don't have a good idea as to why caching fails in the first place, however. I think given that the test is sensitive to caching behavior and cache invalidation, it is probably a good idea to run the test with its own local cache. I will send a CL for that.
Change https://go.dev/cl/545235 mentions this issue: cmd/go: use local go cache for cover_statements script test
https://go.dev/cl/545235 should in theory fix this bug. Leaving open for now just in case there is more than one underlying problem.
CL 545235 should stop the failures, but it doesn't fix the underlying caching bug. I think we should leave this open until/unless that bug is understood (or we have reason to believe it may have been fixed).
Indeed. And it's probably useful to have a way to locally reproduce the caching bug (by reverting this change and then using the script from https://github.com/golang/go/issues/64014#issuecomment-1828291943