[Intel-gfx] [PATCH i-g-t v4 2/3] lib/igt_core: Unify handling of slow/combinatorial tests

David Weinehall david.weinehall at linux.intel.com
Thu Feb 11 13:40:55 UTC 2016


On Thu, Feb 11, 2016 at 01:04:14PM +0000, Chris Wilson wrote:
> On Thu, Feb 11, 2016 at 01:09:33PM +0200, David Weinehall wrote:
> > +enum {
> > +	/* The set of tests run if nothing else is specified */
> > +	SUBTEST_TYPE_NORMAL = 1 << 0,
> > +	/* Basic Acceptance Testing set */
> > +	SUBTEST_TYPE_BASIC = 1 << 1,
> > +	/* Tests that are very slow */
> > +	SUBTEST_TYPE_SLOW = 1 << 2,
> 
> I still feel that slow isn't a useful descriminant for tests. Off the
> opt of my head, I would like
> 
> HANG,
> SWAP,
> STRESS (this is too also undefined)
> 
> that should cover the majority of the GEM cases. Looking at kms_flip
> should also offer a few categories.

As I said the first time I posted this patch, I'm open to both
alternative names and help with categorising the subtests.

> As for gem_concurrent_blit, I basically expect to
> have h->subtest_flags | p->subtest_flags | subtest_flags.
> 
> I also expect QA to run the full gem_concurrent_blit anyway since it is
> showing up regressions from over a year ago.

Yeah, I did a partial run of the full set yesterday and spotted about
150 failed tests within an hour of so. After that I had to abort the
test to continue with other work on that machine.

But it's fairly obvious that the limited set is a lot better tested.


Kind regards, David


More information about the Intel-gfx mailing list