[Piglit] [PATCH 7/7] Add occlusion-query tests for various meta operations.

[Carl Worth]

Eric Anholt <eric at anholt.net> writes:
> concurrent

Thanks for the review, Eric. I'll take care of little things like
"concurrent" etc. throughout, as you mention.

> It would be nice to verify that the correct number of fragments is
> actually generated, not just != 0.  It's not obvious to me without
> reading the spec what the right answer would be for glBitmap(), for
> example.

Yes, that would be a nice improvement to the tests. I don't know the
actual correct numbers myself, (though I guess I can just sample the
current implementation and see if the numbers make sense).

It's interesting that you should mention glBitmap. My reading of the
spec. suggests it should yield a non-zero result, but the current
implementation on master, (without any of my patches), is yielding a
value of 0 for this operation. I haven't looked closer to see what's
going on there, but I'd be glad if anyone has any ideas.

>> +		glBindFramebuffer(GL_READ_FRAMEBUFFER, fb);
>> +		glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
>> +		glBlitFramebuffer(0, 0, 2, 2,
>> +				  2, 2, 20, 20,
>> +				  GL_COLOR_BUFFER_BIT, GL_NEAREST);
>
> Ah, it's done this way to specifically hit the meta path?  That might
> deserve mention, since I was about to suggest copying from the window to
> the window to get the job done.

No, you're giving me too much credit. You're mostly just seeing me
thrashing around trying to learn how to use GL. I saw "BlitFramebuffer"
and likely just mistakenly assumed a needed to do a bunch of
"BindFramebuffer" stuff to set it up.

So I'll see if I can't simplify things here.

>> +	/* Paint the copied texture just ensure it worked. */
>> +	glEnable(GL_TEXTURE_2D);
>> +	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
>> +	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
>> +	piglit_draw_rect_tex(22, 2, 18, 18, 0, 0, 1, 1);
>
> If we're not probing the results, it's not much of an ensuring it worked
> :P

Right. The test isn't verifying that. I meant more that I wanted to put
the results on the framebuffer so that I could manually verify that the
test was doing what I expected it to. Let me know if you think any of
this should be removed. (I don't think adding any probing would improve
the test, but it does seem like making the results visible could help
anyone trying to decipher the test in the future. Perhaps I should just
update the comment along those lines.)

Anyway, thanks for all the review. I'll update the series according to
all your comments and post that in the morning, (and then probably just
push it unless there's more immediate feedback).

-Carl
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/piglit/attachments/20130110/9e5a678a/attachment-0001.pgp>