[igt-dev] [PATCH i-g-t 8/8] i915/gem_exec_basic: Use dynamic subtests

[Petri Latvala]

On Fri, Nov 08, 2019 at 04:33:24PM +0000, Tvrtko Ursulin wrote:
> 
> On 08/11/2019 16:31, Tvrtko Ursulin wrote:
> > 
> > On 24/10/2019 12:05, Petri Latvala wrote:
> > > Instead of generating a subtest for each engine in a static list,
> > > convert to dynamic subtests, with one dynamic subtest per actually
> > > present physical engine.
> > > 
> > > Signed-off-by: Petri Latvala <petri.latvala at intel.com>
> > > ---
> > >   tests/i915/gem_exec_basic.c | 26 +++++++++++++++++++-------
> > >   1 file changed, 19 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/tests/i915/gem_exec_basic.c b/tests/i915/gem_exec_basic.c
> > > index 1287860b..b2d6ddc2 100644
> > > --- a/tests/i915/gem_exec_basic.c
> > > +++ b/tests/i915/gem_exec_basic.c
> > > @@ -159,13 +159,25 @@ igt_main
> > >       igt_subtest("gtt-all")
> > >           gtt_all(fd);
> > > -    __for_each_physical_engine(fd, e) {
> > > -        igt_subtest_f("basic-%s", e->name)
> > > -            noop(fd, e->flags);
> > > -        igt_subtest_f("readonly-%s", e->name)
> > > -            readonly(fd, e->flags);
> > > -        igt_subtest_f("gtt-%s", e->name)
> > > -            gtt(fd, e->flags);
> > > +    igt_subtest_with_dynamic_subsubtests("basic") {
> > 
> > Bikeshed on the name maybe, since "subsub" reads a bit funny.
> > 
> > igt_dynamic_subtest_group ?
> 
> Assuming at least it can both nest under igt_subtest_group and not nest ie.
> be at top-level? Do we need a nested version even?

It's a brother of igt_subtest, not a brother of
igt_subtest_group. igt_subtest_with_dynamic_subsubtests() is, like
igt_subtest(),

 * execution entry with --run-subtest
 * enumerable with --list-subtests
 * a block where you can touch the system inside it
 * nests inside an igt_subtest_group and automatically produces a
   'SKIP' if a fixture in that group triggered an igt_require

I'm all for bikeshedding on the name, but I want the closer
association to igt_subtest to show. I like none of the names discussed
so far all around the mailing list, I hate this current one the least
so far. If it's all right, I'd like to land the series and start
converting tests while the naming bikeshed (again, very welcome
bikeshed) goes on simultaneously. It's a matter of a quick sed
magickry later.


-- 
Petri Latvala