runtime: TestMemStats failure with insanely high value on openbsd-amd64-68

[bcmills]

#!watchflakes
post <- pkg == "runtime" && test == "TestMemStats" && `insanely high value`

##### GOMAXPROCS=2 runtime -cpu=1,2,4 -quick
--- FAIL: TestMemStats (0.00s)
    malloc_test.go:93: PauseTotalNs = 18446744073177999286: insanely high value (overflow?); want <= 1e+11
FAIL
FAIL	runtime	21.088s

greplogs -l -e 'FAIL: TestMemStats.*(?:\n\s+.*)*insanely high value' --since=2022-01-01
2022-09-09T01:04:57-76c94eb/openbsd-amd64-68

attn @golang/openbsd; CC @golang/runtime

[mknyszek]

... We should really update that error message.

I'm going to guess this is a product of the fact that OpenBSD has coarse-grained timers. As the error message suggests, this looks to me like a negative duration (note PauseTotalNs is a uint64, so the int64 nanotime result is cast first).

I guess the fix is to defensively check for a negative duration and clamp at zero?

[gopherbot]

Found new matching dashboard test flakes for:

#!watchflakes
post <- pkg == "runtime" && test == "TestMemStats" && `insanely high value`

2022-09-09 01:04 openbsd-amd64-68 go@76c94eb7 runtime.TestMemStats (log)

--- FAIL: TestMemStats (0.00s)
    malloc_test.go:93: PauseTotalNs = 18446744073177999286: insanely high value (overflow?); want <= 1e+11

— watchflakes