[Piglit] [PATCH] egl_ext_device_drm: don't fail the test if open("/dev/dri/cardX") errors

Emil Velikov emil.l.velikov at gmail.com
Wed Oct 10 13:22:16 UTC 2018


On Tue, 9 Oct 2018 at 16:50, Dylan Baker <dylan at pnwbakers.com> wrote:
>
> Quoting Emil Velikov (2018-10-09 05:35:39)
> > On Fri, 5 Oct 2018 at 19:13, Dylan Baker <dylan at pnwbakers.com> wrote:
> > >
> > > In other cases like this we return skip instead of warn.
> > Can you be more specific about those "other cases like these"? It's
> > the core part of your argument so w/o a quote/reference it's
> > impossible to make a decision.
>
> For example, if a test binary doesn't exist or isn't executable.
>
I fear you're confusing things. There's extension with it's requirements.
There are parts which are optional and the extension interacts with,
when available.

> > > I think that it should
> > > be skip because:
> > >
> > > 1) warn doesn't have any real meaning
> > Not a usual black-and-white one, but it does.
> > WARN is something that should be addressed (looked at), but isn't
> > considered an blocking issue.
>
> Piglit's warn is a bad particularly because it doesn't have a concise meaning,
> it means "something might have happened". Tests results should not be ambiguous
> like that. pass, fail, crash, and skip all have clear concise meanings.
>
I'm not sure that this is the best place to discuss WARN.
Please start a separate thread, instead.

> > > 2) we couldn't actually run the test so there isn't a result
> > >
> > There is no clear "one test" here, but a collection of small subtests
> > - more or less one for each piglit_report_result.
> > You don't want to skip the whole thing, because a section is no applicable.
>
> In that case, I think a more correct solution would be to use
> piglit_report_subtest_result in the test so that each subtest returns an
> individual result.
>
$git grep piglit_report_subtest_result -- tests/egl/ | wc -l
0

So I'm consistent ... despite how good or bad that is.
If you want to go ahead and fix that up, please be my guest.

Personally, I've wanted to fix multiple things in piglit for ages,
sadly I haven't had time to do much apart from the odd test.

Thanks
Emil


More information about the Piglit mailing list