[Piglit] [PATCH 3/3 V2] GL 3.2: Test layered framebuffers clear the depth attachment properly.

[Chad Versace]

On 09/17/2013 10:59 AM, Paul Berry wrote:
> On 14 September 2013 09:25, Henri Verbeet <hverbeet at gmail.com <mailto:hverbeet at gmail.com>> wrote:
>
>     On 13 September 2013 23:40, Paul Berry <stereotype441 at gmail.com <mailto:stereotype441 at gmail.com>> wrote:
>      > Looking at the specs yesterday, I *thought* the call is necessary in order
>      > to ensure framebuffer completeness, since there's no color attachment to the
>      > framebuffer.  In GL 3.2 (section 4.4.4 Framebuffer Completeness, under the
>      > "Whole Framebuffer Completeness" section), the following is listed as a
>      > criterion for framebuffer completeness:
>      >
>      > The value of FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE must not be
>      > NONE for any color attachment point(s) named by DRAW_BUFFERi.
>      >
>      > { FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER }
>      >
>      >
>      > However, maybe I'm interpreting the spec wrong, because neither Mesa nor the
>      > NVIDIA proprietary driver requires the glDrawBuffer() call in order for the
>      > framebuffer to be complete.
>      >
>     ARB_ES2_compatibility / GL 4.1 remove those completeness requirements,
>     and make the behaviour as if the READ/DRAW buffer is NONE for missing
>     attachments.
>
>
> Ah, ok.  In that case I think we should go ahead and keep the glDrawBuffer() call, just to make sure that the test works
> on implementations that use the pre-GL-4.1 completeness rules.

Hmm... I wonder how many of Piglit tests fail to due to this oversight on non-Mesa drivers.
In particular, I'm considering Mac OS pre-Mavericks, which supports 3.2, but neither
4.1 nor ARB_ES2_compatibility.