Bug 30993

Created attachment 39558 [details] [review]
proposed fix/patch

Can you test the attached patch?

Mesa is correct in generating GL_INVALID_ENUM when the object type is not one of GL_TEXTURE or GL_RENDERBUFFER.  The GL_ARB_framebuffer_objects spec says:

            "If the value of FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE is
            RENDERBUFFER, then

                ...

            If the value of FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE is
            TEXTURE, then

                ...

            Otherwise, an INVALID_ENUM error is generated."

I looked at the GL_ARB_fbo spec too, but the GL 3.3 spec says (page 276):

"""If the value of FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE is NONE, no
framebuffer is bound to target. In this case querying pname FRAMEBUFFER_-
ATTACHMENT_OBJECT_NAME will return zero, and all other queries will generate
an INVALID_OPERATION error.
"""

Though I guess since we don't yet support GL 3.3 the code is OK as-is.

(In reply to comment #3)
> I looked at the GL_ARB_fbo spec too, but the GL 3.3 spec says (page 276):
> 
> """If the value of FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE is NONE, no
> framebuffer is bound to target. In this case querying pname FRAMEBUFFER_-
> ATTACHMENT_OBJECT_NAME will return zero, and all other queries will generate
> an INVALID_OPERATION error.
> """
> 
> Though I guess since we don't yet support GL 3.3 the code is OK as-is.

ARG!  The same language is in the 3.0 spec, and GL_ARB_framebuffer_objects is supposed to match 3.0.  It seems we've found a spec bug.  I've submitted a bug against the GL_ARB_framebuffer_objects spec.

I've also added piglit test fbo-getframebufferattachmentparameter-01 the reproduces this bug and another similar bug.

(In reply to comment #1)
> Created an attachment (id=39558) [details]
> proposed fix/patch
> 
> Can you test the attached patch?

Tested, the fix works for me.

I've committed the previously attached patch.

The fbo-getframebufferattachmentparameter-01 test does not pass yet but Mesa fails in the same way as NVIDIA's driver.

Closing because the fix has been committed.