Closed Bug 2025750 Opened 7 months ago Closed 6 months ago

[Late Beta] Perma win reftest dom/media/test/reftest/image-10bits-rendering-90-video.html == dom/media/test/reftest/image-10bits-rendering-90-ref.html - when Gecko 151 switches to beta on 2026-05-01

Categories

(Core :: Audio/Video: Playback, defect, P2)

defect

Tracking

()

RESOLVED FIXED
151 Branch
Tracking Status
firefox-esr115 --- unaffected
firefox-esr140 --- unaffected
firefox149 --- unaffected
firefox150 + wontfix
firefox151 + fixed

People

(Reporter: amarc, Unassigned)

References

(Regression)

Details

(Keywords: regression)

Summary: [Late Beta] Perma dom/media/test/reftest/image-10bits-rendering-90-video.html == dom/media/test/reftest/image-10bits-rendering-90-ref.html - when Gecko 151 switches to beta on 2026-05-01 → [Late Beta] Perma win reftest dom/media/test/reftest/image-10bits-rendering-90-video.html == dom/media/test/reftest/image-10bits-rendering-90-ref.html - when Gecko 151 switches to beta on 2026-05-01

:jmoss, since you are the author of the regressor, bug 2024599, could you take a look? Also, could you set the severity field?

For more information, please visit BugBot documentation.

Flags: needinfo?(jmoss)
Flags: needinfo?(jmoss)

The regressor landed during Fx150 nightly, so Fx150 is affected

The bug is marked as tracked for firefox150 (beta) and tracked for firefox151 (nightly). However, the bug still isn't assigned.

:jimm, could you please find an assignee for this tracked bug? Given that it is a regression and we know the cause, we could also simply backout the regressor. If you disagree with the tracking decision, please talk with the release managers.

For more information, please visit BugBot documentation.

Flags: needinfo?(jmathies)

This is primarily a CSS filter rendering on video test. It should probably be over in Layout. But it sounds like the fix is straight forward. Right now the team is a bit busy but I'll add this to our triage queue.

  • Summary: [Late Beta] Perma win reftest dom/media/test/reftest/image-10bits-rendering-90-video.html == dom/media/test/reftest/image-10bits-rendering-90-ref.html - when Gecko 151 switches to beta on 2026-05-01
  • Reporter: amarc@mozilla.com (Mozilla employee)
  • Status: NEW
  • Product: Core
  • Component: Audio/Video: Playback
  • Created: 2026-03-24 (2 days ago)
  • Severity / Priority: -- / --
  • Security Sensitive: No
  • Tracked: Firefox 150 (beta) affected, Firefox 151 (nightly) tracked
  • Pending needinfo: :jimm (jmathies@mozilla.com) — asked to find an assignee or backout regressor

Research Summary and Key Findings

This is a test-only CI regression with no user-facing impact. The permanently failing reftest is image-10bits-rendering-90-video.html == image-10bits-rendering-90-ref.html on Windows. It was caught via a Late Beta simulation that predicts failures that will materialize when Gecko 151 switches to beta on 2026-05-01.

The identified regressor is Bug 2024599 (RESOLVED FIXED, authored by :jmoss). That bug fixed color comparison failures caused by the Win11 25H2 worker pool migration (Bug 2024195 — "Migrate reftest to win11 25h2 worker pools"). The Win11 25H2 workers produce slightly different rendered output for 10-bit video reftests, requiring relaxed fuzzy-if(winWidget,...) annotations.

The fix in Bug 2024599 was incomplete. Looking at dom/media/test/reftest/reftest.list:

Line Test winWidget annotation
6 image-10bits-rendering-video.html fuzzy-if(winWidget,0-240,0-769689) ← added by Bug 2024599
7 image-10bits-rendering-90-video.html none — MISSING ← this is the failing test
9 image-10bits-rendering-720-video.html fuzzy-if(winWidget,0-255,0-576000) ← added by Bug 2024599
10 image-10bits-rendering-720-90-video.html none — potentially also affected

The 90-degree rotated variants (lines 7 and 10) were skipped when the fix was applied. Line 7 currently only has fuzzy(0-10,0-778536) fuzzy-if(appleSilicon,0-37,0-699614) with no Windows-specific tolerance, so the Windows rendering difference causes a permanent failure.

Candidate fix: Add a fuzzy-if(winWidget,...) annotation to line 7 (and possibly line 10) with appropriate max-difference and pixel-count values. The exact values need to come from the Windows failure log (Treeherder link in comment 1). The non-rotated counterpart uses 0-240,0-769689, which is a reasonable starting estimate for the same video at 90-degree rotation (pixel count may differ due to the rotation flipping width/height).

Regression Timeline

  • 2026-03-18: Bug 2024195 filed — migration to Win11 25H2 reftest worker pools begins.
  • 2026-03-19: Bug 2024599 filed — intermittent color comparison failures appear. :jmoss authors a patch ("Relax color_quads and image-10bits fuzzy annotations for 25H2 winWidget") and lands it the same day.
  • 2026-03-24: Bug 2025750 filed — Late Beta simulation reveals the image-10bits-rendering-90 test was not covered by the Bug 2024599 fix and will permanently fail when 151 goes to beta on 2026-05-01.
  • 2026-03-24: :dmeehan confirms Firefox 150 is also affected (the regressor landed in Fx150 nightly).
  • 2026-03-26: Release management bot needinfos :jimm to find an assignee.

Assessment

  • Suggested Severity: S3
  • Suggested Priority: P2
Blocks: media-triage
Severity: -- → S3
Flags: needinfo?(jmathies)
Priority: -- → P2

Is there someone that needs to be assigned here?

Flags: needinfo?(jmathies)
Flags: needinfo?(jmathies) → needinfo?(alwu)

looks like it's fixed

Status: NEW → RESOLVED
Closed: 6 months ago
Resolution: --- → FIXED
Target Milestone: --- → 151 Branch
Flags: needinfo?(alwu)
No longer blocks: media-triage
You need to log in before you can comment on or make changes to this bug.