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

Issue 794035 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner: ----
Closed: Jan 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Mac
Pri: 2
Type: Bug-Regression



Sign in to add a comment

Incorrect output on some hardware when gl.clearColor arguments identically zero

Reported by ri...@standardcyborg.com, Dec 12 2017

Issue description

UserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_11_6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/63.0.3239.84 Safari/537.36

Steps to reproduce the problem:
1. create a webgl context
2. create a floating point framebuffer using oes_texture_float extension
3. clear the framebuffer with color [0, 0, 0, 0]
4. draw a triangle to the framebuffer with color [1, 0, 0, 0.5] and blending (gl.ONE + gl.ONE_MINUS_SRC_ALPHA)
5. draw the framebuffer to the screen using blending (gl.ONE + gl.ONE_MINUS_SRC_ALPHA)

What is the expected behavior?
A 50% opacity red triangle is drawn to the screen.

What went wrong?
Output is a white screen with no triangle, but only on some computers.

A workaround is to set any of the components in the color in step 3 to not-identically-zero. For example, clearing the floating point framebuffer in step 3 to [0, 0, 0, 1e-30] fixes the problem on affected computers.

Changing the 50% opacity a bit, it seems to be merely completely blown out (saturated at rgb = [1, 1, 1]). It could be some color space issue or some other optimization, but the output seems exceptionally washed out and displays a sharp cliff between identically zero and almost-zero clearcolors.

My computer has specs:

MacBook Pro (Retina, 13-inch, Early 2015)
Processor: 2.7 GHz Intel Core i5
Memory: 8 GB 1867 MHz DDR3
Graphics: Intel Iris Graphics 6100 1536 MB

My nearly-identical macbook pro (Late 2013, Retina, Intel Iris 1536 MB) works fine. Every other laptop in the office of ten works fine. However, we're aware of some client's computers that also display the same bug.

Did this work before? Yes 59.0.3071.86

Does this work in other browsers? N/A

Chrome version: 63.0.3239.84  Channel: stable
OS Version: OS X 10.11.6
Flash Version: 

Chrome version <= 59.0 displays the expected result
Chrome version > 59 displays a white screen
Firefox latest displays the expected result

This is also visible in the plotly.js library with this test mock: https://rreusser.github.io/plotly-mock-viewer/#gl3d_bunny

Firefox and Chrome <= 59.0 display the expected plot. Newer chrome displays a nearly-white screen. Results attached above. Changing the clearColor to not-identically-zero fixes this problem also. That bug report is here: https://github.com/gl-vis/gl-plot3d/issues/9
 
index.html
4.3 KB View Download
plotly-bunny-incorrect-output.png
78.1 KB View Download
plotly-bunny-correct-output-firefox.png
136 KB View Download
correct-output-firefox.png
38.3 KB View Download
incorrect-output-chrome-63.png
36.2 KB View Download
Oh, and for simplicity, you may find a codepen-ified example here: https://codepen.io/rsreusser/pen/MrgNda

And to the extent that you're familiar with regl (https://github.com/regl-project/regl), a significantly easier to follow regl example here: https://codepen.io/rsreusser/pen/LePvJj
One last note is that we create thumbnails on the client so we know for certain that this is occurring on at least two separate computers. I don't know details about the client's computer. Otherwise we were inclined to chalk it up to a glitchy gpu, but it seems a bit unlikely it would fail in exactly the same way if just a bad gpu.
Labels: Needs-Bisect

Comment 4 by gov...@chromium.org, Dec 12 2017

Cc: pbomm...@chromium.org kbr@chromium.org kainino@chromium.org
Could this be related to  bug 792966 ?
Unable to reproduce the issue on reported chrome version 63.0.3239.84 using MacBook Air Sierra 10.12.6, we have also checked it on MacBook retina 10.12.6 and MacBook retina 10.13.1, still unable to reproduce the issue with steps mentioned below:

Steps followed to reproduce the issue:
1) Launched chrome and navigated to URL: https://codepen.io/rsreusser/pen/MrgNda
2) Able to see red triangle on the screen as expecetd

@Reporter:
Can you please try to test the issue by creating a new person on your Mac 10.11.6, if the issue still persists can you please upgrade your Mac to 10.12.6 Sierra or 10.13.1 High Sierra and try to reproduce the issue

Thanks!
794035.png
222 KB View Download
Cc: viswatej...@techmahindra.com sc00335...@techmahindra.com
Labels: Needs-Triage-M63 Triaged-ET Needs-Feedback
It's not related to 792966.

Reporter, your machine is on an outdated macOS 10.11.6. Have you seen other
machines in your office with 10.11 which don't see this problem? I suspect
it may have been fixed in 10.12.
I didn't expect the OS upgrade to fix this since it remained dependent upon the Chrome version, but upgrading to os x 10.13.2 leads to the expected behavior. I'd failed to re-upgrade after wiping my work computer. We'll retain our failsafe since it's best if it doesn't fail for users.

I'm content to consider fixed, close, and hope it serves as a sign post in case someone else encounters it. Thanks.
Project Member

Comment 9 by sheriffbot@chromium.org, Dec 14 2017

Labels: -Needs-Feedback
Thank you for providing more feedback. Adding requester "sc00335628@techmahindra.com" to the cc list and removing "Needs-Feedback" label.

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

Comment 10 by kbr@chromium.org, Dec 14 2017

Submitter: we don't have any 10.11 machines in house any more so can't verify this, but if you run this variant of your pen:
https://codepen.io/anon/pen/xpbBNB

what is printed to the console?

If we can detect this bug via gl.readPixels then we can write a conformance test for it. I'm not 100% sure if just the on-screen result is wrong, or if the actual WebGL rendering result is.

@Reporter:
Can you please respond to Comment#10, hence adding Needs-Feedback label

Thanks!
Labels: Needs-Feedback
Status: WontFix (was: Unconfirmed)
Mac triage: marking this WontFix. Reporter in #8 remarked that they are happy to have this issue closed (and have upgraded their machine) so followup is unlikely.

Sign in to add a comment