1638305 - WebGL Rendering text using a texture atlas and per component data storage shows corruption

Status: RESOLVED WORKSFORME

(Core :: Graphics: CanvasWebGL, defect, P1)

Description

[adrian]

Attachment: webgl_testcase.zip

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/81.0.4044.138 Safari/537.36

Steps to reproduce:

Using WebGL with a glyph texture atlas that stores each character into a separate colour component with the shader only rendering one colour component at a time.

Actual results:

Reading a specific component in the shader is failing, showing the other characters underneath when it shouldn't.
https://ibb.co/jzqnGsZ firefox 77.0b6

Expected results:

The text should be legible without other unrelated characters drawn underneath
https://ibb.co/87pd312 firefox 76.0.1

Comment 1

[adrian]

This was found by a customer and so I would appreciate some urgency on this. Thanks everyone.

Comment 2

[BugBot [:suhaib / :marco/ :calixte]]

Bugbug thinks this bug should belong to this component, but please revert this change in case of error.

Comment 3

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:jgilbert, could you have a look please?

For more information, please visit auto_nag documentation.

Comment 4

[Kelsey Gilbert [:jgilbert]]

Looks like 76.0.1, 77rc1, and 78 Nightly are all unaffected on my Windows machine.

I will test on a Mac later today. It's possible this was due to a regression that was since fixed.

Comment 5

[Kelsey Gilbert [:jgilbert]]

Attachment: Error-ff-77-0b6.png

Comment 6

[Kelsey Gilbert [:jgilbert]]

Attachment: Success-ff-76-0-1.png

Comment 7

[Jessie [:jbonisteel] pls NI]

Just following up on our conversation yesterday - is this indeed S1?

Comment 8

[Kelsey Gilbert [:jgilbert]]

I can't reproduce this on 77.0.1 or Nightly 10.15.5 MacBook Pro (15-inch, 2019). (Intel(R) UHD Graphics 630)

Comment 9

[Kelsey Gilbert [:jgilbert]]

I can reproduce this on (old nightly) 77.0a1 "MacBook Pro (Retina, 15-inch, Late 2013)" 10.15.3.

Comment 10

[Kelsey Gilbert [:jgilbert]]

Letting it restart and update to 80, I can't repro. I downloaded 77.0.1 and can't repro there either.
I suspect this might be a "tired program" bug, where it won't repro right off the bat, but given enough time and usage/mem-fragmentation/something it'll start to have problems.

I'll try some workloads to see if I can't make a more reliable STR.

Does this issue happen even right after restarting the browser, Adrian?

Comment 11

[adrian]

It was happening immediately and reproducibly however I can't repro in the released version of 77 or the latest beta so possibly it was just a bad beta build. I'll ask My QA team to have a quick check to confirm then you can probably resolve this. Thanks anyway, really wanted to be sure this didn't get into the release build!

Comment 12

[Kim Moir [:kmoir] ET]

adrian: Triaging as REO for 79 - are you planning on addressing this for 79?

Comment 13

[Kelsey Gilbert [:jgilbert]]

This might be related to bug 1659225, though this bug seems to predate 1659225's regression point.

Comment 14

[Kelsey Gilbert [:jgilbert]]

Comment #11 indicates WFM