shaver pointed out in bug 586418 that removing files after we're done a test run will probably be much faster than doing the cleanup ahead of time, since the OS cache will have most of the test files in cache. it's still a good idea to do cleanup before the test run, just in case.

In the unittest case, you can do the rm in the background while you send the results, and then wait for it to finish, if you want to squeeze the time down even more. But I suspect that time-to-cleanup will drop to "not a huge deal" once it's at the tail anyway.

patch for adding the cleanup step after the end of the run. Still performs a cleanup before as well.

Would it make sense to add "addCleanupSteps" inside of "addTearDownSteps" since it is part of the tearing down of the test run?

I agree that making the cleanup part of the teardown makes sense. I've also removed the second cleanup from the the talos runs, since when testing it on staging, I observed little-to-no difference in the amount of time taken to cleanup before/after (usually just a few seconds anyway), so adding it afterwards would probably just be adding a second or so of filesystem churn.

Comment on attachment 472616 [details] [diff] [review] addCleanupStep inside addTeardownSteps, remove from TalosFactory changeset: 953:6f1dbbcf0aee