os/signal: TestAtomicStop fails

[dsnet]

Just saw this failure on the darwin-amd64 builders:

--- FAIL: TestAtomicStop (1.12s)
	signal_test.go:324: iteration 0: output 
	signal_test.go:324: iteration 1: output 
	signal_test.go:324: iteration 2: output 
	signal_test.go:324: iteration 3: output 
	signal_test.go:324: iteration 4: output 
	signal_test.go:324: iteration 5: output 
	signal_test.go:324: iteration 6: output 
	signal_test.go:324: iteration 7: output lost signal on iterations: 3
	signal_test.go:331: iteration 7: lost signal
	signal_test.go:324: iteration 8: output 
	signal_test.go:324: iteration 9: output 
FAIL
FAIL	os/signal	5.514s

Full log:
https://storage.googleapis.com/go-build-log/eefc42f3/darwin-amd64-10_11_7a763c87.log

\cc @bcmills

[bcmills]

That test was added in https://golang.org/cl/46003. I assume we've only seen a failure on Darwin so far? (Would be good to know whether it's flaky everywhere or just on Darwin.)

(CC: @ianlancetaylor)

[ianlancetaylor]

Was that a trybot run? I don't see the git hash on tip. The failure could have been caused by a Darwin-specific bug in https://golang.org/cl/49590 (5500c9c), fixed last Friday by https://golang.org/cl/55032 (057a34a). Was the trybot run based on tip in between those two commits?

[dsnet]

It's a try-boy run and seems to still be happening very close to tip.

On https://build.golang.org/, you can see it happen on 17fa5a7, so it seems flaky.

[quasilyte]

Executed trybots in https://go-review.googlesource.com/#/c/go/+/63490/ and TestAtomicStop failed on Darwin again.
Second trybots run does not caused it to happen (patch set 3 passes without any code changes).

https://storage.googleapis.com/go-build-log/11750b86/darwin-amd64-10_11_83471050.log

--- FAIL: TestAtomicStop (1.09s)
	signal_test.go:324: iteration 0: output 
	signal_test.go:324: iteration 1: output lost signal on iterations: 6
	signal_test.go:331: iteration 1: lost signal
	signal_test.go:324: iteration 2: output 
	signal_test.go:324: iteration 3: output 
	signal_test.go:324: iteration 4: output 
	signal_test.go:324: iteration 5: output 
	signal_test.go:324: iteration 6: output 
	signal_test.go:324: iteration 7: output 
	signal_test.go:324: iteration 8: output 
	signal_test.go:324: iteration 9: output 

[tombergan]

One more: https://storage.googleapis.com/go-build-log/4f6f7c65/darwin-amd64-10_11_491fb13d.log

--- FAIL: TestAtomicStop (1.14s)
	signal_test.go:324: iteration 0: output 
	signal_test.go:324: iteration 1: output 
	signal_test.go:324: iteration 2: output lost signal on iterations: 6
	signal_test.go:331: iteration 2: lost signal
	signal_test.go:324: iteration 3: output 
	signal_test.go:324: iteration 4: output 
	signal_test.go:324: iteration 5: output 
	signal_test.go:324: iteration 6: output 
	signal_test.go:324: iteration 7: output 
	signal_test.go:324: iteration 8: output 
	signal_test.go:324: iteration 9: output 

[gopherbot]

Change https://golang.org/cl/79581 mentions this issue: runtime: don't block signals that will kill the program

[bradfitz]

More:
https://storage.googleapis.com/go-build-log/39f9af3b/linux-386_d50c7f23.log

[bradfitz]

Oh, Ian's CL isn't submitted.