[Intel-gfx] [PATCH i-g-t] tests/gem_exec_basic: don't use gem_require_ring to check ring availability

Chris Wilson chris at chris-wilson.co.uk
Fri Jan 29 04:11:41 PST 2016

On Fri, Jan 29, 2016 at 11:58:14AM +0000, Daniele Ceraolo Spurio wrote:
> On 29/01/16 11:35, Chris Wilson wrote:
> >On Fri, Jan 29, 2016 at 11:16:37AM +0000, Daniele Ceraolo Spurio wrote:
> >>
> >>On 29/01/16 10:58, Chris Wilson wrote:
> >>>On Fri, Jan 29, 2016 at 09:21:33AM +0000, daniele.ceraolospurio at intel.com wrote:
> >>>>From: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
> >>>>
> >>>>gem_require_ring will submit an execbuf using the provided flags and
> >>>>skip the test if the ioctl fails. This test is however designed to catch
> >>>>issues with the ioctl, so it should fail if the ioctl fails on a ring
> >>>>that the HW possesses.
> >>>>
> >>>>Instead of using gem_require_ring we can use the getparam ioctl. The new
> >>>>checker has been added to the test file and not to the commmon library
> >>>>because this test is the only special case where we want to not use
> >>>>gem_has_ring
> >>>That would be gem_exec_param.
> >>>-Chris
> >>I don't understand what you mean, can you elaborate a bit?
> >For the purposes of checking that the kernel honours the ABI, the tests
> >belong in gem_exec_params.
> >
> >For the purposes of CI, a testing going from PASS -> SKIP is just as
> >indicative of a problem as test going from PASS -> FAIL or any other
> >state.
> The difference would be that the CI system still reports that BAT
> succeeded if one or more tests go from PASS to SKIP (e.g. http://lists.freedesktop.org/archives/intel-gfx/2016-January/086586.html).

Fortunately, the results are sometimes even read.
> >>What I wanted to fix here is the fact that the logic to skip the
> >>test and the test itself are identical, which means that this test
> >>can't fail. As far as I can tell gem_exec_param is trying to catch
> >>errors in the handling of invalid flags, while in this test we check
> >>for errors in the handling of valid flags instead.
> >Basically the logic is repeated, that is not an issue for its purpose.
> >-Chris
> This patch can be dropped then.

But can be refactored for gem_exec_param!

Chris Wilson, Intel Open Source Technology Centre

More information about the Intel-gfx mailing list