net/http/pprof: occasional failures in TestDeltaProfile #38544
Comments
bcmills
changed the title
|
any objection to disabling this test in all platforms other than in linux? |
|
Is there a reason to suspect that it is not flaky on Linux? (In my experience, the unusual builders are often just more timing-sensitive than the mainline builders — tests that are flaky on more than one of those platforms also tend to be flaky on |
|
Or, maybe we completely delete this test? |
|
Change https://golang.org/cl/229498 mentions this issue: |
In some slow environment, the goroutine for mutexHog2 may not run within 1secs. So, try with increasing seconds parameters, and declare failure if it still fails with the longest duration parameter (32sec). Also, relax the test condition - previously we expected the profile's duration is within 0.5~2sec. But obviously, in some slow environment, that's not even guaranteed. Just check we get non-zero duration in the result. Update #38544 Change-Id: Ia9b0d51429a2093e6c9eb92cf463ff6952ef3e10 Reviewed-on: https://go-review.googlesource.com/c/go/+/229498 Run-TryBot: Hyang-Ah Hana Kim <hyangah@gmail.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
This comment was marked as off-topic.
|
I don't see any more test failures after the cl. Closing it. Thanks! |
|
Still appears to be flaky, but removing the 2020-07-13T17:48:55-e88ea87/openbsd-arm-jsing |
|
Hasn’t been flaky for the past 6 months, so punting to Go1.17. |
|
No new failures, calling this fixed. |
|
Apparently still not fixed, although the failure rate does still seem to be lower than it was. |
TestDeltaProfile, added in CL 147598 for #23401, seems to be flaky on some of thearmbuilders.(Perhaps it is relying on some property of scheduler latency that does not hold on the very slowest builders?)
2020-04-18T20:01:46-f5291cf/plan9-arm
2020-04-18T18:07:52-2a20f5c/android-arm-corellium
CC @hyangah @odeke-em @bradfitz
Marking as release-blocker for 1.15 since the test appears to be new for that release.
The text was updated successfully, but these errors were encountered: