New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 658623 link

Starred by 1 user

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Mac
Pri: 3
Type: Bug



Sign in to add a comment

Multisampled depth renderbuffers of certain widths are corrupted

Project Member Reported by jie.a.c...@intel.com, Oct 24 2016

Issue description

[Radar 28901861]
On macOS machines with an Intel Haswell GPU, when drawing to an FBO with a multisampled depth renderbuffer attached, some incorrect depth data will be written into the buffer if the buffer width falls into the ranges of [8N+4, 8N+8) where N>=0. 
This bug causes failures of fbomultisample.*_samples.html’s tests in the WebGL 2.0 conformance suite.
This bug does not occur on other GPU types on macOS, nor on other operating systems.


Steps to reproduce:

1. Download Chrome Canary and run from the Terminal:
/Applications/Google\ Chrome\ Canary.app/Contents/MacOS/Google\ Chrome\ Canary --enable-unsafe-es3-apis

2. Check out the KhronosGroup/WebGL repository:
git clone https://github.com/KhronosGroup/WebGL.git

3. Launch an HTTP server against that repository:
cd WebGL
python -m SimpleHTTPServer

4. Navigate the browser to the test cases:
http://localhost:8000/sdk/tests/deqp/functional/gles3/fbomultisample.8_samples.html?webglVersion=2&quiet=0
http://localhost:8000/sdk/tests/deqp/functional/gles3/fbomultisample.4_samples.html?webglVersion=2&quiet=0
http://localhost:8000/sdk/tests/deqp/functional/gles3/fbomultisample.2_samples.html?webglVersion=2&quiet=0

These tests are also available online at:
http://www.khronos.org/registry/webgl/sdk/tests/deqp/functional/gles3/fbomultisample.8_samples.html?webglVersion=2&quiet=0
http://www.khronos.org/registry/webgl/sdk/tests/deqp/functional/gles3/fbomultisample.4_samples.html?webglVersion=2&quiet=0
http://www.khronos.org/registry/webgl/sdk/tests/deqp/functional/gles3/fbomultisample.2_samples.html?webglVersion=2&quiet=0


Expected Results:

Expect all tests to pass, as they do on other operating systems and on other GPU types on macOS.


Actual Results:

The test cases involving renderbuffers of any depth format fail.


OS X Build/Version:

macOS Sierra (10.12)
MacBook Pro-Mid 2015-Intel Iris Pro Graphics 5200 

Notes:

NOT reproducible on MacBook-Early 2016-Intel HD Graphics 515

 
Cc: kbr@chromium.org zmo@chromium.org yang...@intel.com qiankun....@intel.com yunchao...@intel.com
Components: Blink>WebGL
Status: Available (was: Untriaged)
Project Member

Comment 4 by sheriffbot@chromium.org, Apr 11 2018

Labels: Hotlist-Recharge-Cold
Status: Untriaged (was: Available)
This issue has been Available for over a year. If it's no longer important or seems unlikely to be fixed, please consider closing it out. If it is important, please re-triage the issue.

Sorry for the inconvenience if the bug really should have been left as Available.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Comment 5 by kbr@chromium.org, Apr 11 2018

Cc: -qiankun....@intel.com
Owner: jie.a.c...@intel.com
Status: Assigned (was: Untriaged)
Has this been fixed in 10.13?

Still reproducible.

Chrome version	Chrome/67.0.3395.0
Operating system	Mac OS X 10.13.4
GPU1	VENDOR = 0x8086, DEVICE= 0x0d26 *ACTIVE*
Driver version	10.32.48
GL_RENDERER	Intel Iris Pro OpenGL Engine

Comment 7 by kbr@chromium.org, Apr 13 2018

Owner: yizhou.j...@intel.com
Thanks for verifying. That's a bit disappointing. May I assign this to you for tracking?

Comment 8 by kbr@chromium.org, Apr 13 2018

I filed Radar 39402487 about this since I didn't have access to the old Radar 28901861.

Comment 9 by kbr@chromium.org, Apr 13 2018

Also contacted Apple about the new Radar.

Comment 10 by kbr@chromium.org, Apr 13 2018

Labels: -Hotlist-Recharge-Cold
Cc: yizhou.j...@intel.com
Owner: jie.a.c...@intel.com
Ken, Jie will track the status of this issue. Actually it is Jie that verified the test on latest Mac OSX, but he used yizhou's account by mistake. 

Sign in to add a comment