Allowed GLES query target migration between binary occlusion types

Allowed GLES query target migration between binary occlusion types

BUG=429053

[bajones, 2015-06-24 20:43:01]

bajones@chromium.org changed reviewers:
+ piman@chromium.org, zmo@chromium.org

[bajones, 2015-06-24 20:43:02]

PTAL. Behavior is described in the OpenGL ES3 spec section 2.13, or more
readably here: http://docs.gl/es3/glBeginQuery

[piman, 2015-06-24 21:41:31]

I don't think we're reading the spec the same way.

What the spec says is that GL_ANY_SAMPLES_PASSED and
GL_ANY_SAMPLES_PASSED_CONSERVATIVE are mutually exclusive (effectively the
current query is the same state for both), so that you can't begin a
GL_ANY_SAMPLES_PASSED query if you already have begun (and not ended) a
GL_ANY_SAMPLES_PASSED_CONSERVATIVE query, and vice versa.

There's no "upgrading" of the target.

[bajones, 2015-06-24 21:53:38]

Hm... re-reading the spec I agree that it doesn't explicitly state that. The man
page (https://www.khronos.org/opengles/sdk/docs/man31/html/glBeginQuery.xhtml)
goes further, though:

"If target is GL_ANY_SAMPLES_PASSED, id must be an unused name, or the name of
an existing boolean occlusion query object" ...
"If target is GL_ANY_SAMPLES_PASSED_CONSERVATIVE, id must be an unused name, or
the name of an existing boolean occlusion query object." ...

Both of those types are considered boolean occlusion query objects (which is why
only one can be active at a time). This text suggests they're interchangeable.

How canonical do we consider the man pages to be?

[piman, 2015-06-24 23:14:38]

On Wed, Jun 24, 2015 at 2:53 PM, <bajones@chromium.org> wrote:

> Hm... re-reading the spec I agree that it doesn't explicitly state that.
> The man
> page (
> https://www.khronos.org/opengles/sdk/docs/man31/html/glBeginQuery.xhtml)
> goes further, though:
>
> "If target is GL_ANY_SAMPLES_PASSED, id must be an unused name, or the
> name of
> an existing boolean occlusion query object" ...
> "If target is GL_ANY_SAMPLES_PASSED_CONSERVATIVE, id must be an unused
> name, or
> the name of an existing boolean occlusion query object." ...
>
> Both of those types are considered boolean occlusion query objects (which
> is why
> only one can be active at a time). This text suggests they're
> interchangeable.
>

I don't think that part is inconsistent. In any case, this man page also
explicitly says:
"GL_INVALID_OPERATION is generated if glBeginQuery is executed while a
query object of the same target is already active. Note:
GL_ANY_SAMPLES_PASSED and GL_ANY_SAMPLES_PASSED_CONSERVATIVE alias to the
same target for the purposes of this error."


> How canonical do we consider the man pages to be?
>

The spec is what's authoritative.

From the spec (under BeginQuery):
"An INVALID_OPERATION error is generated if the active query object
name for target is non-zero (for the targets ANY_SAMPLES_PASSED and
ANY_SAMPLES_PASSED_CONSERVATIVE, if the active query for either target
is non-zero)".


First principles also suggest that it's not possible to
"upgrade" GL_ANY_SAMPLES_PASSED_CONSERVATIVE to GL_ANY_SAMPLES_PASSED
anyway.


> https://codereview.chromium.org/1199313008/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[bajones, 2015-06-24 23:23:56]

Okay then. Closing this CL and I'll update the corresponding code in Blink and
the conformance tests.

[piman, 2015-06-24 23:37:01]

On Wed, Jun 24, 2015 at 4:23 PM, <bajones@chromium.org> wrote:

> Okay then. Closing this CL and I'll update the corresponding code in Blink
> and
> the conformance tests.
>

Note, I believe it is true that we're not properly failing
ANY_SAMPLES_PASSED when ANY_SAMPLES_PASSED_CONSERVATIVE is on (neither on
the client side nor the service side), and that's possibly something we
want to do and test.


> https://codereview.chromium.org/1199313008/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.