[PATCH v5 10/10] egl: EGL image trace support

[José Fonseca]

On Thu, Aug 2, 2012 at 7:16 PM, Imre Deak <imre.deak at intel.com> wrote:
> Hi,
>
> On Thu, 2012-08-02 at 18:25 +0100, José Fonseca wrote:
>> Imre,
>>
>> I've refactored this change so that instead of lumping image info and
>> data into a blob, we simply emit a fake glTexImage2D.
>>
>> I've been using fake calls in other cases (most notably for user guest
>> drivers). I hope to use this trick for GLX_texture_from_pixmap, and
>> alike. I also hope one day to make it easier to distinguish fake vs
>> real, but's its not urgent.
>
> I discussed this possibility with Pauli earlier and I think he was right
> saying that retracing should show the fact that EGL images are involved.
> That should reveal a lot about the underlying problem we are tracing.

Oh I agree, and no EGL call is dropped with my follow on changes.
Simply an additional fake call is emitted after the real
glEGLImageTargetTexture2DOES.

My plan for the future is to flag fake calls so that they are not
normally visible -- but are still executed by glretrace.

>> I tested that this works using mesa demos
>> src/egl/opengles1/texture_from_pixmap.c . But I had to tweak the image
>> to be a power-of-two -- it looks like EGLImageKHR don't need to be a
>> power-of-two size, but regular textures do. A bit confusing.
>> Eitherway, I believe this is a problem with your original solution
>> too.
>
> Hm, I'm not sure what's the reason for this. But my version did a
> texture from EGL image too, so the same restrictions apply there.

> I will run our test case with your version and get back to you with the
> result. I'm also in the process of going through your earlier comments
> about the remaining parts of the patchset.

Thanks

Jose