(Marking as release-blocker for 1.18 because the 1-year gap in failures makes me suspect a 1.18 regression.)

Do we think this should block the beta?

The more recent failures are different from the ones a year ago. There isn't enough evidence in the test output to be sure, but I think the recent failures are caused by the test taking more than 10 seconds to run; the test makes a request with ExpectContinueTimeout set to 10 seconds and fails if it it runs over that timeout. Probably just a slow builder.

Change https://golang.org/cl/368083 mentions this issue: net/http: increase timeout in TestNoSniffExpectRequestBody_h[12]

Hmm... I'm skeptical that it's a builder slowness issue because the test itself doesn't seem to be taking anywhere near that long. From the first failure in the above logs:

--- FAIL: TestNoSniffExpectRequestBody_h2 (0.01s)
    clientserver_test.go:1341: stream error: stream ID 1; NO_ERROR; received from peer
FAIL
FAIL	net/http	13.007s

That appears to be only 10ms running the test, not 10s.

Oh, I had the timeline backwards. The old failures are from the 10s timeout being exceeded, the new failures are something different.

Friendly check-in here, this is one of the last two issues blocking the beta of 1.18. Has there been an update recently?

Looking now, but I believe this is present in 1.17.5 and does not need to block the 1.18 beta. (Should be fixed before rc1, though.)

Thanks! I removed the label. We'll consider it a release-blocker for RC1.

Checking in on this as a release-blocking issue. Are there any updates?

I believe the fix for #49645 fixed this one as well. Unless I'm misreading the builder logs, I see no instances of this failure since that fix landed.

Unless I'm misreading the builder logs, I see no instances of this failure since that fix landed.

Agreed: this does not appeared to have recurred since the backports for that fix were merged. Thanks!

greplogs --dashboard -md -l -e 'FAIL: TestNoSniffExpectRequestBody_h2' --since=2021-12-02

[no results]