Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "cmd/cgo/internal/test" && test == "TestThreadLock"

scatter = 0x56e1a3
--- FAIL: TestThreadLock (0.00s)
    cgo_thread_lock.go:50: cgo has not locked OS thread

— watchflakes

Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "cmd/cgo/internal/test" && test == "TestThreadLock"

--- FAIL: TestThreadLock (0.00s)
    cgo_thread_lock.go:50: cgo has not locked OS thread

— watchflakes

@aclements points out we might just be running this test under conditions we previously weren't, because they moved this test out of it's original location. It's also possible we were having this failure before under misc/cgo/test, but there don't seem to be any other failures there.

Added release-blocker until we can figure out more details. Probably won't be long-term.

This test has been in the tree for years, but never ran before CL 493603, which I submitted 2 days ago.

Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "cmd/cgo/internal/test" && test == "TestThreadLock"

--- FAIL: TestThreadLock (0.00s)
    cgo_thread_lock.go:50: cgo has not locked OS thread

— watchflakes

Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "cmd/cgo/internal/test" && test == "TestThreadLock"

--- FAIL: TestThreadLock (0.00s)
    cgo_thread_lock.go:50: cgo has not locked OS thread

— watchflakes

I think the test is inherently flaky. The goroutine running testThreadLock could get preempted and moved to a different thread in between the call to syscall.Gettid and the call to C.Ctid.

It was added in https://go.dev/cl/18882.

I suggest that we either delete it (since it never ran before anyway) or change it to make both of the gettid calls from within the C function (before and after the Gosched).

Change https://go.dev/cl/496182 mentions this issue: cmd/cgo/internal/test: fix TestThreadLock

Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "cmd/cgo/internal/test" && test == "TestThreadLock"

scatter = 0xe87733c0
sqrt is: 0
fatal error: m changed unexpectedly in cgocallbackghello from C


cmd/cgo/internal/test.TestThreadLock(0xbf93d4a0)
	/workdir/go/src/cmd/cgo/internal/test/cgo_test.go:105 +0x37 fp=0xbf8de774 sp=0xbf8de770 pc=0xe873da47
testing.tRunner(0xbf93d4a0, 0xe87f05f8)

— watchflakes

The above is a different failure mode — looks like a runtime bug rather than a bug in the test.