#general
- l
limited-terabyte-34878
09/22/2015, 10:29 PMIt’s because we want every test/test-suite to have a timeout, i couldn’t figure out how to enforce that and allow teams to configure their own timeouts at the individual test level, so I put it at the test target level. - l
limited-terabyte-34878
09/22/2015, 10:30 PMI actually don’t know twitter politics at all - l
limited-terabyte-34878
09/22/2015, 10:30 PMi’ve only been here 3 weeks - e
enough-analyst-54434
09/22/2015, 10:30 PMk - well pytest-timeout allows them to do this already - e
enough-analyst-54434
09/22/2015, 10:30 PMas does junit @Test(timeout=) - e
enough-analyst-54434
09/22/2015, 10:31 PMI think you may find the politics /accounting I describe is waht motivated others to ask you to do this! - l
limited-terabyte-34878
09/22/2015, 10:31 PMcan we enforce that if they don’t put in a timeout using either mechanism, some other timeout is used? - l
limited-terabyte-34878
09/22/2015, 10:31 PMyeah that politics reason does make sense - e
enough-analyst-54434
09/22/2015, 10:32 PMread up on pytest-timeout - it looks like yes - globally, not at the target level - e
enough-analyst-54434
09/22/2015, 10:32 PMfor junit, pants owns the runner - e
enough-analyst-54434
09/22/2015, 10:32 PMso a similar global timeout could be plumbed if its not already - e
enough-analyst-54434
09/22/2015, 10:33 PMPer-target would need something like what you've done, but the utility of that would need open debate - e
enough-analyst-54434
09/22/2015, 10:33 PMSince we already removed that capability from the codebase for python_tests some time back and would be revisting that with a proposed reversal - e
enough-analyst-54434
09/22/2015, 10:35 PMIt looks like the junit runner does not have global timeout capabilities today looking at the flags here: https://github.com/pantsbuild/pants/blob/master/src/java/org/pantsbuild/tools/junit/impl/ConsoleRunnerImpl.java - h
happy-kitchen-89482
09/22/2015, 10:38 PMIt does seem like the code, not the BUILD file, is the right place for a timeout. - h
happy-kitchen-89482
09/22/2015, 10:38 PMThe timeout would be set based on what the test code does. So it would make sense for it to live as near to the code as possible. - h
happy-kitchen-89482
09/22/2015, 10:38 PMSameer, sorry if you're currently caught between the hammer of Pants politics and the anvil of Twitter politics... - l
limited-terabyte-34878
09/22/2015, 10:38 PMYeah I agree that is a better place. And then if we can override it in the test-runner (or provide an upper bound) then that will accomplish what we need. - h
happy-kitchen-89482
09/22/2015, 10:39 PMGreat. Pants community would greatly prefer that solution, as it doesn't require adding any complexity to pants... simple smile - l
limited-terabyte-34878
09/22/2015, 10:39 PMwell wouldn’t we have to pass the timeout parameter through to pytest, and also add the global timeout to the junit runner? - e
enough-analyst-54434
09/22/2015, 10:40 PMGlobal timeouts yes - w
witty-crayon-22786
09/22/2015, 10:40 PMyea. still the case that you need a default. - l
limited-terabyte-34878
09/22/2015, 10:41 PMso there would still be some added pants complexity, but less, I would hope. (don’t know pants well enough to know if this is more or less complex) - e
enough-analyst-54434
09/22/2015, 10:41 PMMy apologies for putting you in the jaws @limited-terabyte-34878 , I just think Twitter often falls short educating new contributors about things like this - e
enough-analyst-54434
09/22/2015, 10:41 PMI appreciate your position and help - l
limited-terabyte-34878
09/22/2015, 10:42 PMno problem. - l
limited-terabyte-34878
09/22/2015, 10:43 PMI believe I should discard this RB and post to the pants list with my proposal to modify pants to enable passing the timeout parameter through to pytest-timeout and to modify the junit runner to have a global timeout default as well? - w
witty-crayon-22786
09/22/2015, 10:45 PMwell, if sirois and benjy are good with it, you can probably skip the posting to the list step - w
witty-crayon-22786
09/22/2015, 10:45 PM(original/core contributors) - f
fancy-queen-20734
09/22/2015, 11:04 PMQuestion, if a task has a flag that includes another option scope as a prefix, should that flag be referenceable in a scoped context, or should the global always override?