Open cherrymui opened 1 year ago
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime"
(... long comment truncated ...)
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime"
(... long comment truncated ...)
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime"
(... long comment truncated ...)
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Linux doesn't use procid
anywhere quite so critical, but it is used in preemptM
-> signalM
for async preemption. preemptM
could race with an exiting M and end up sending a signal to 0. I think that should be harmless because tgkill(getpid(), 0)
returns EINVAL
.
(It looks like there could be a problematic race with semawake attempting to wake the wrong thread when racing with an extra M moving between threads. But such a race should not be possible because the semawake
M is found in the lock's waiting list. If the M is waiting for a lock it can't also be moving between threads.)
Change https://go.dev/cl/527057 mentions this issue: runtime: don't clear procid on netbsd
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Hm, perhaps not fully fixed?
That is at a commit before your CL. It's just that the builder slowly caught up. (If it were a commit newer than the CL, watchflakes would reopen the issue.)
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
Found new dashboard test flakes for:
#!watchflakes
post <- goos == "netbsd" && pkg == "runtime" && date > "2023-08-01"
On the build dashboard, the NetBSD ports start to become very flaky since CL https://go.dev/cl/526118. The failures are usually related to tests that are specific to OS threads. So the CL could be related. Filing an issue to investigate.
cc @prattmic @golang/netbsd @golang/runtime