I really like the thought of having this functionality in i-g-t, especially if combined with my patches that allows for categorising subtests (I'll submit a new version of those patches soon, since they never got merged last time around). I'll refrain from bikeshedding on the syntax. In the longer run I think a rewrite of the subtest logic to allow not only specifying what tests to run and/or exclude, but also to specify what *order* to run the tests in (if reordering is possible) and to randomise the order would be useful. Since some testcases contain so many subtests that having those testcases be a part of the CI/Piglit testruns isn't possible, we need some way to have those tests run anyway. My current train of thought is to either traverse through the tests (1..20, 21..40, 41..60, etc.; which would require piglit to keep state, which it currently doesn't) or to randomise a set of subtests that would take something like 30s. This might mean that things could slip under the radar for a long time, but compared to today's situation where these tests aren't run at all it'd still be an improvement. Just my 2¢. Kind regards, David Weinehall _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
