runtime: TestGcLastTime failures

[gopherbot]

#!watchflakes
default <- test == "TestGcLastTime"

Issue created automatically to collect these failures.

Example (log):

=== RUN   TestGcLastTime
    gc_test.go:131: bad last GC time: got 1742950790082949000, want [1742950790141094000, 1742950790083385000]
--- FAIL: TestGcLastTime (0.00s)

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
default <- pkg == "runtime:cpu2" && test == "TestGcLastTime"

2025-03-25 18:36 gotip-darwin-amd64_14 go@936bb82d runtime:cpu2.TestGcLastTime (log)

=== RUN   TestGcLastTime
    gc_test.go:131: bad last GC time: got 1742950790082949000, want [1742950790141094000, 1742950790083385000]
--- FAIL: TestGcLastTime (0.00s)

— watchflakes

[randall77]

Something went wrong with UnixNano. The window listed has negative width. That corresponds to (in the test):

	t0 := time.Now().UnixNano()
	runtime.GC()
	t1 := time.Now().UnixNano()

Getting t1 < t0. Pretty clearly a bug in either UnixNano or the os itself. (Skewed clocks across processes?)

[mknyszek]

In triage, we might want to wait for more failures. I agree with @randall77.

[mknyszek]

Maybe the test could be slightly more robust by calling LockOSThread? That seems like a hack though. This sort of thing should just work, but time is hard.