1769082 - Using the luminosity globalCompositeOperation in combination with other globalCompositeOperations causes rendering errors

Status: RESOLVED FIXED

(Core :: Graphics: Canvas2D, defect, P3)

Description

[George Broadhurst]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:101.0) Gecko/20100101 Firefox/101.0

Steps to reproduce:

I can reproduce by following these steps:

1. create a canvas element and insert an image with drawImage
2. set globalCompositeOperation to "saturation" and fill the canvas with #000
3. set globalCompositeOperation to "luminosity" and globalAlpha to 0.5 and fill the canvas with #fff

Where the desaturation layer and luminosity layer overlap, a rendering error occurs

This codepen contains a recreatable example https://codepen.io/GBro/pen/gOvwGod?editors=0010

Actual results:

On the Windows version of Firefox, it was rendered correctly.

On the Mac OSX version of Firefox (user agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:100.0) Gecko/20100101 Firefox/100.0), the resultant image has a speckling effect, with black and white spots dotted around the image where the two globalCompositerOperation effects overlap.

Expected results:

The canvas should display a desaturated and brightened version of the original image

Comment 2

[Andrew Osmond [:aosmond] (he/him)]

Same thing happens on Linux. Difference may be caused by D2D vs Skia.

If you change gfx.canvas.azure.backends to be just skia, does it reproduce on Windows? (Requires restart)

Comment 3

[George Broadhurst]

(In reply to Andrew Osmond [:aosmond] (he/him) from comment #2)

Same thing happens on Linux. Difference may be caused by D2D vs Skia.

If you change gfx.canvas.azure.backends to be just skia, does it reproduce on Windows? (Requires restart)

Yep, same issue occurs for me on Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:101.0) Gecko/20100101 Firefox/101.0 after making that config change

Comment 4

[Andrew Osmond [:aosmond] (he/him)]

Bug in Skia or DrawTargetSkia?

Comment 5

[Lee Salzman [:lsalzman]]

This bug is in Skia. It reproduces in Chromium if you disable hardware acceleration there. They fixed it in Skia's GPU pipeline, but not in the software pipeline.

Comment 6

[Lee Salzman [:lsalzman]]

I filed an upstream Skia bug: https://bugs.chromium.org/p/skia/issues/detail?id=13417

Comment 7

[Lee Salzman [:lsalzman]]

Attachment: Bug 1769082 - Avoid divide-by-zero in Skia's luminosity blend mode. r?jrmuizel

This applies a fix that is present in Skia's HW luminosity blend mode to its
CPU pipeline so that the luminosity mode no longer divides by zero, thus avoiding
infs and nans.

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/f292760857f2
Avoid divide-by-zero in Skia's luminosity blend mode. r=jrmuizel,gfx-reviewers

Comment 9

[Atila Butkovits]

Backed out for causing build bustages at SkRasterPipeline_opts.

Backout link: https://hg.mozilla.org/integration/autoland/rev/c2d1bef38ec87e21737e12a69009019dbc93b957

Push where failures started: https://treeherder.mozilla.org/jobs?repo=autoland&selectedTaskRun=OZoz5UjXSMCDwfU4UJHdew.0&resultStatus=testfailed%2Cbusted%2Cexception&revision=fe7b745e73f1a3fc88b0328ab5c4bf01d82486eeq

Failure log: https://treeherder.mozilla.org/logviewer?job_id=381498580&repo=autoland&lineNumber=14453

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/8d845c981228
Avoid divide-by-zero in Skia's luminosity blend mode. r=jrmuizel,gfx-reviewers

Comment 11

[bszekely]

https://hg.mozilla.org/mozilla-central/rev/8d845c981228

Comment 12

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

The patch landed in nightly and beta is affected.
:lsalzman, is this bug important enough to require an uplift?

• If yes, please nominate the patch for beta approval.
• If no, please set status-firefox102 to wontfix.

For more information, please visit auto_nag documentation.

Comment 13

[Lee Salzman [:lsalzman]]

Comment on attachment 9280894 [details]
Bug 1769082 - Avoid divide-by-zero in Skia's luminosity blend mode. r?jrmuizel

Beta/Release Uplift Approval Request

• User impact if declined: Visual artifacts when using the luminosity blend mode in CSS or Canvas.
• Is this code covered by automated tests?: Yes
• Has the fix been verified in Nightly?: Yes
• Needs manual test from QE?: No
• If yes, steps to reproduce:
• List of other uplifts needed: None
• Risk to taking this patch: Low
• Why is the change risky/not risky? (and alternatives if risky): Fix accepted by upstream Skia.
• String changes made/needed:
• Is Android affected?: Yes

Comment 14

[Pascal Chevrel:pascalc (PTO until April 26)]

Comment on attachment 9280894 [details]
Bug 1769082 - Avoid divide-by-zero in Skia's luminosity blend mode. r?jrmuizel

This is landing too late to take into our last beta, sorry.