You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The test ran a for loop without preemption points. The loop checked a
condition that would only become true after another goroutine had been
scheduled and carried out its job.
If, with only few cores (four in my case) GC kicked in before that other
goroutine got scheduled, that loop would just run hot forever until the
test timed out, and the resulting stack dump looked quite unhelpful.
Add a small sleep so the runtime can preempt the goroutine.
The issue was harder to run into when stressing only the test, since there
was less garbage available at that point. Adding some print statements,
I accidentally made it much more likely.
Previously flaked (got stuck) within <500iters, now ran past 1.5k without
problems.
Fixescockroachdb#19397.
Fixescockroachdb#19388.
Touches cockroachdb#19367.
Fixescockroachdb#18554.
The following tests appear to have failed:
#383473:
The text was updated successfully, but these errors were encountered: