On Tue, Apr 21, 2015 at 5:06 PM, Daniel Stone <daniel@xxxxxxxxxxxxx> wrote:
> On 21 April 2015 at 16:03, Micah Fedke <micah.fedke@xxxxxxxxxxxxxxx> wrote:
>> + * drm_open_any_any:
>> + *
>> + * Literally the worst-named function I've ever written.
>
> And I stand by this. This is really an RFC, partly to find out whether
> it would be better to find a new name for these functions (open a
> modeset-capable DRM node, whether it's Intel or otherwise), or whether
> adding a flag to drm_open_any and friends to specify Intel or generic,
> with the resulting mega-Cocci run, would be better.
Yeah I think a new param to restrict the driver name (i915, nouveau,
...) would be better. Plus maybe a drm_require_driver so that testcase
could open a generic driver and then only skip specific subtest.
Initial pass could be done with cocci+ follow-up patches to convert
patches from i915 specific to generic.
One thing to keep in mind is how we'll treat machines with more than 1
gpu. I do care about that at least for the prime testcases since I
have a box somewhere with i915+nouveau to run them, and that should
keep working. Aside: We should convert those tests to use
drm_open_any("nouveau") instead of the hand-rolled one. A possible
idea would be to convert the igt_main macros into a loop which would
go over all the available drm devices (well the different ones, we
don't want to separately test control/render/legacy nodes ofc). What
we definitely need is an enviroment variable to override the default
choice.
-Daniel
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
