Bug 43418

Summary:	Resolving the HiZ buffer fails when using a GLES API
Product:	Mesa	Reporter:	Neil Roberts <nroberts>
Component:	Drivers/DRI/i965	Assignee:	Chad Versace <chadversary>
Status:	RESOLVED FIXED	QA Contact:
Severity:	normal
Priority:	medium
Version:	git
Hardware:	Other
OS:	All
Whiteboard:
i915 platform:		i915 features:
Attachments:	Test case gen6_hiz: Don't bind GL_DRAW_FRAMEBUFFER on GLES

Description Neil Roberts 2011-12-01 05:47:12 UTC

When resolving the HiZ buffer Mesa uses the meta API with some internal GL calls. However when using GLES this fails because it tries to bind a renderbuffer to GL_DRAW_FRAMEBUFFER but Mesa asserts that this binding point is only used when the API is GL.

I'm attaching a patch but I'm not sure if it's the best solution because I've copied the condition for whether GL_DRAW_FRAMEBUFFER is allowed into gen6_hiz.c. This could be quite awkward to maintain. For example if Mesa ever gets support for the GL_ANGLE_framebuffer_blit extension then we'd have to remember to change this check in two places.

I wonder if the check should just be removed instead? Mesa already lets you bind to GL_DRAW_FRAMEBUFFER, just not add a renderbuffer so it's already not completely validated.

Comment 1 Neil Roberts 2011-12-01 05:50:09 UTC

Created attachment 54010 [details]
Test case

Comment 2 Neil Roberts 2011-12-01 05:53:01 UTC

Created attachment 54011 [details] [review]
gen6_hiz: Don't bind GL_DRAW_FRAMEBUFFER on GLES

When using Mesa with a GLES API, calling _mesa_FramebufferRenderbuffer
with GL_DRAW_FRAMEBUFFER will report a 'user error' because
get_framebuffer_target validates that this enum from the framebuffer
blit extension is only used on GL. To work around it this patch makes
it use the GL_FRAMEBUFFER enum instead in that case.

Comment 3 Chad Versace 2012-01-11 14:39:51 UTC

I made a modification to the patch, and posted it to mesa-dev for review.

http://article.gmane.org/gmane.comp.video.mesa3d.devel/35980

Comment 4 Chad Versace 2012-01-13 10:17:22 UTC

Mark RESOLVED/FIXED.

My change to the patch broke it because I wrongly assumed that the extension bit for EXT_framebuffer_blit was
off in an ES context, since the extension doesn't exist there. But that's not how Mesa's
extension works, which I should have known; I plumbed it before.

Neil, your original patch has been pushed to the master and 8.0 branches.
master: 9462b8447864c754252cd2580c9e1e4d36d5cc63
8.0:    399b9799de9980b8ebc0ba46304be207b28b7825

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.