[Piglit] [PATCH 1/2] egl: Add sanity test for EGL_EXT_device_query

[Adam Jackson]

On Thu, 2016-07-21 at 12:37 -0700, Eric Anholt wrote:
> Adam Jackson <ajax at redhat.com> writes:
> > +
> > +/**
> > + * \file
> > + *
> > + * Tests for EGL_EXT_device_query
> > + */
> 
> I don't think "\file" works without a file mentioned.  Maybe just drop
> this comment entirely, since it seems pretty empty.

It might not! I started by cloning egl_khr_get_all_proc_addresses.c
(hence the Intel copyright) which also has an empty \file.

> > +	if ((devstring = queryDeviceString(dpy, EGL_EXTENSIONS)) == NULL) {
> > +		printf("Empty device extension string\n");
> > +		piglit_report_result(PIGLIT_WARN);
> > +	}
> 
> Style nitpick: I'd prefer moving the assignment out of the conditional.

Sure.

> Possible additional tests:
> 
> - Make sure that EGL_NOT_INTIIALIZED is errored from
>   queryDisplayAttrib before eglInitialize().
> 
> - Make sure that EGL_NOT_INTIIALIZED is errored from
>   queryDisplayAttrib(dpy, 0x0d0d0d0)

EGL_BAD_ATTRIBUTE I think you mean?

> - Make sure that EGL_BAD_DEVICE_EXT is returned from
>   queryDeviceString(EGL_NO_DEVICE_EXT, EGL_EXTENSIONS)

I think this is sane? I'm trying to imagine something like the client
extension trick for display extensions and I don't think that really
makes sense. Can always fix the test after if EGL_NO_DEVICE suddenly
becomes meaningful I guess.

> - Make sure that EGL_BAD_PARAMETER is returned from
>   queryDeviceString(queryDeviceString(device, 0xd0d0d0d0)
> 
> - Make sure that eglQueryDeviceAttribEXT is also
>   eglGetProcAddress()able.
> 
> - Make sure that queryDeviceAttrib(device, 0xd0d0d0d0) throws
>   EGL_BAD_ATTRIBUTE
> 
> (We can't test EGL_BAD_DEVICE from that function because we don't have a
> valid attrib to use)

Yeah, plan is to add separate tests for for EGL_EXT_device_drm. I might
or might not end up writing a spec for EGL_EXT_device_glx, remains to
be seen whether that's useful.

- ajax