1947590 - Colorful horizontal lines in youtube video overlay background gradient due to canvas randomization

Status: NEW

(Core :: Privacy: Anti-Tracking, defect, P3)

Description

[jinra321+bugzilla]

Attachment: lines.png

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:135.0) Gecko/20100101 Firefox/135.0

Steps to reproduce:

Watch a youtube video and get the timeline overlay to show up (either by pausing or mousing over the video).

Actual results:

When the timeline/controls overlay shows up, multiple colorful horizontal lines will show up. This is especially visible with darker backgrounds and shows up more on my HDR OLED monitor vs my IPS screens (though still visible on IPS).

Expected results:

No horizontal lines when overlay is shown. Confirmed not an issue on Chrome.

Comment 1

[BugBot [:suhaib / :marco]]

The Bugbug bot thinks this bug should belong to the 'Core::Audio/Video: Playback' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[az [:az]]

Could you please share the contents of your about:support text? Click "copy raw data to clipboard", then paste into a comment here and follow the prompt to upload it as an attachment. Did this behavior occur for you with previous versions of Firefox?

Comment 3

[jinra321+bugzilla]

Attachment: raw support data

Comment 4

[jinra321+bugzilla]

(In reply to az [:az] from comment #2)

Could you please share the contents of your about:support text? Click "copy raw data to clipboard", then paste into a comment here and follow the prompt to upload it as an attachment. Did this behavior occur for you with previous versions of Firefox?

I can't recall when it first happened. I suspect it's been around for a little while but haven't really noticed until lately.

Comment 5

[Mayank Bansal]

If this used to work in an earlier version of Firefox, please do a bisection to find the exact change that caused this (https://mozilla.github.io/mozregression/)
Finding the regression range would be very useful to help diagnose this issue.

Comment 7

[Kestrel]

STR:

1. Set privacy.resistFingerprinting = true.
2. Open a youtube video.
3. Hover the cursor over the video so the controls appear.
4. Clear cookies and try again if it does not occur.

Expected:
Gradient transition behind controls from partial black transparency to full transparency.

Actual:
Randomization horizontal lines in lower half of video with no transparency gradient.

Issue occurs with FPP targets:

-AllTargets,+CanvasRandomization,+CanvasImageExtractionPrompt,+CanvasExtractionBeforeUserInputIsBlocked
-AllTargets,+CanvasRandomization,+CanvasImageExtractionPrompt
-AllTargets,+CanvasRandomization,+CanvasExtractionBeforeUserInputIsBlocked

Issue does not occur with FPP targets:

+AllTargets,-CanvasRandomization,-CanvasImageExtractionPrompt,-CanvasExtractionBeforeUserInputIsBlocked

Could not reproduce with Strict ETP which only uses +CanvasRandomization (unlike reporter).

Not a regression, reproducible back to version 110.

uBlock Origin workaround (add to "My filters" section):

youtube.com##.ytp-gradient-bottom:style(background-image: linear-gradient(rgba(0,0,0,0), rgba(0,0,0,0.4)) !important;)

Comment 8

[Kestrel]

Attachment: Horizontal lines in video overlay bottom gradient

Comment 9

[bughit]

Not a regression, reproducible back to version 110.

Bug 1948507, which you duped to this, is a regression, here's the bisection result for it:

2025-02-16T10:45:26.710000: DEBUG : Found commit message:
Bug 1942648: Support MSign with Int32 input and Double output. r=jandem

Differential Revision: https://phabricator.services.mozilla.com/D234936

2025-02-16T10:45:26.710000: DEBUG : Did not find a branch, checking all integration branches
2025-02-16T10:45:26.712000: INFO : The bisection is done.

Comment 10

[Kestrel]

Youtube seems to be doing a phased rollout so running mozregression will take you to a random unrelated change.

Comment 11

[bughit]

Can't be sure about the bisection since I only tested once for each build, but my own testing is definitely not random. Bug 1948507 consistently reproduces with 135.0.0 and is consistently fine with 134.0.2

Comment 12

[Thorin [:thorin]]

it's been going on for almost 5 years - https://github.com/arkenfox/user.js/issues/924 (back then the RFP canvas was all white)

Comment 14

[jinra321+bugzilla]

@Kestral(In reply to Kestrel from comment #10)

I'm not sure this is the same issue, I have privacy.resistantFingerprinting as false, and the issue shown is not quite the same.

Comment 15

[Tom Ritter [:tjr]]

I think this is a dupe of Bug 1724663

FWIW YT is aware of this behavior, but because RFP is an unsupported mode I'm not sure they will do anything about it.

Comment 16

[jinra321+bugzilla]

(In reply to Tom Ritter [:tjr] from comment #15)

I think this is a dupe of Bug 1724663

FWIW YT is aware of this behavior, but because RFP is an unsupported mode I'm not sure they will do anything about it.

As mentioned previous, I do not believe this is the same issue. However, since 135.0.1, this seems to be resolved.