On Tue, 19 Jan 2010, karl williamson wrote: > But as I prepare this for submittal, I wonder if it takes too much CPU > time. It runs 360K tests and takes 3.5 minutes on my almost-a-year-old > (hence pretty fast) PC. > > One solution would be to run a random selection of tests each time with > an option to run all. About 20 tests are run currently for each code > point, so we would run at least one per code point in a reasonable > amount of time. That would have found the bug that spurred this. I don't like the idea of a *random* selection being run; it should be the same set every time so you get reproducible results. In general I agree we'll eventually need some way to keep the time required to run the testsuite at some reasonable level, but don't know what the best way to do so would be. So I would run all tests by default, and only run the reduced set if e.g. PERL_RUN_SLOW_TESTS is defined. That way developers can set an environment variable for quicker turnaround of patch/test/commit cycles but the smokers and default builds would still get the full test coverage. Cheers, -JanThread Previous | Thread Next