stress: skip.UnderStress() doesn't #84312
Labels
A-build-system
C-bug
Code not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior.
T-dev-inf
Describe the problem
skip.UnderStress()
,skip.UnderStressRace()
, andskip.NightlyStress()
all reference theCOCKROACH_NIGHTLY_STRESS
environment variable to check if we're running under stress. This variable isn't set bymake stress
or./dev test --stress
, though, it's set by shell scripts that call those commands. So runningdev test --stress
locally won't skip those tests.I haven't done the forensics to figure out whether this is intentional--it's plausibly a good idea. But it is definitely confusing, because some people will use skip.UnderStress if they have a test that fails under stress (due to a timeout or something) and thus effectively break
./dev test --stress
for their package outside nightly test runs.To Reproduce
Create a test that goes
and run it with
./dev test --stress
.Expected behavior
Maybe rename it to
skip.UnderNightlyStress()
and call it a day?My preference though would I think be to have the ability to actually skip tests under stress. For example, right after I file this issue I'm going to go add
skip.UnderStressRace()
to TestChangefeedRestartMultiNode because it reliably hangs under stressrace, I think because it's callingTestCluster.WaitForFullReplication()
and under stressrace we just never get there. I'd rather not have that be a trap for anyone running tests locally.So maybe --stress could set a different flag that's also visible to
skip
, and then we could have an API likeJira issue: CRDB-17584
The text was updated successfully, but these errors were encountered: