On 05/07/17 23:53, Lyude Paul wrote:
NAK. You're right that these don't actually give us any advantage over just using CRCs and are just slower, however I left these tests in here moreso just so we had something to actually test the frame dumping functions so that we could avoid regressing them by accident since we're the only users of those functions right now. If I recall properly, isn't there a list of tests in igt's source that they use for determining which tests to run on the CI? I think a better solution would be to just disable this for CI runs, and maybe add some comments pointing out that this test is only really useful for developers making changes to the chamelium library.
The general direction of IGT is to make sure that all tests are being executed by default, very much alike piglit.
I however do understand what you mean and I would then propose that we start defining naming conventions that allow filtering tests based on how level directive. For instance, we could have every test that need it export a slow and fast version, so as to easily be able to filter stuff out. I guess we need to start a separate thread for that.
As for this particular test, I would say that it could be simplified to prevent downloading so many images (all the exported resolutions * all the connected connectors). We could instead test one resolution per connector and call it good-enough. By better testing stuff, we just increase the likeliness of the test being completely blacklisted due to its excessive execution time. In this case, the better is the enemy of good.
_______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
