heyas, I'm working on automatically retrying tests...
# developmentc
careful-address-89803
08/17/2023, 1:38 AMheyas, I'm working on automatically retrying tests, and I'm having trouble understanding how Pants doesn't cache failed test results. seems
class TestResult(EngineAwareReturnType)cacheableGet(TestResult, ...)careful-address-89803
08/17/2023, 2:09 AMah,
Process.cache_scopeProcessCacheScope.PER_SESSIONcareful-address-89803
08/17/2023, 2:22 AMweird that changing it to
ProcessCacheScope.SUCCESSFULcareful-address-89803
08/17/2023, 2:41 AMwait is that option only used for remote caching?
careful-address-89803
08/17/2023, 3:32 AM Copy code
fn cacheable(&self) -> bool {
match self {
NodeKey::Task(s) => s.task.cacheable,
&NodeKey::SessionValues(_) | &NodeKey::RunId(_) => false,
_ => true,
}
}b
bitter-ability-32190
08/17/2023, 11:25 AMIIRC the results cacheable is about rendering.
The process object determines if it's cached on failure (which shouldn't be default, as that should be rare)
bitter-ability-32190
08/17/2023, 11:32 AMLooks like plugins get to craft the process. So here's Pythons: https://github.com/pantsbuild/pants/blob/main/src/python/pants/backend/python/goals/pytest_runner.py#L406
bitter-ability-32190
08/17/2023, 11:34 AMAnd PER_SESSION means essentially (every run of Pants reruns), and that value is only used if the force flag is given (which makes sense since we want it to run then)
bitter-ability-32190
08/17/2023, 11:34 AMChanging it to SUCCESSFUL most certainly shouldn't cache it
bitter-ability-32190
08/17/2023, 11:41 AMOh but I think you're on to something. I think everything is cached at the session level 🥴
c
careful-address-89803
08/17/2023, 1:54 PMyeah. I saw something about output_cacheable, but didn't have the chance to look into whether that was calculated after the output was known. I also don't know if there's a way to dirty that node so it runs again
b
bitter-ability-32190
08/17/2023, 2:30 PMNo, that's a Stu question, I think. My naive brain says you'd need another field in a Process.
bitter-ability-32190
08/17/2023, 2:30 PMOr another value for the ProcessCacheScope
bitter-ability-32190
08/17/2023, 2:32 PMNew field makes sense to me, as presumably more than tests could warrant a retry (like a process that talks to the network)
3 Views