1824158 - Perma Android layout/reftests/svg/svg-integration/clip-path/clip-path-polygon-013.html == layout/reftests/svg/svg-integration/clip-path/clip-path-stripes-003-ref.html | image comparison, max difference: 255 when Gecko 113 merges to Beta on 2023-04-10

Status: VERIFIED FIXED

(Core :: SVG, defect)

Description

[Serban Stanca [:SerbanS]]

Central-as-beta-simulation

How to run these simulations

Failure log

The green stripe from the bar in the middle in the test image doesn't have same length as in the reference image

Comment 1

[Jamie Nicol [:jnicol]]

I changed vClipMaskUv in prim_shared.glsl to mediump in bug 1823411, and perhaps incorrectly added a fuzzy annotation. The amount of fuzz is different on the beta simulation, hence the unexpected pass.

I can reproduce the failure on a variety of adreno devices, but not on Mali. I think technically mediump should be correct for this variable, but for now at least probably safer to revert to highp.

Comment 2

[Jamie Nicol [:jnicol]]

Attachment: Bug 1824158 - Make vClipMaskUv highp again. r?#gfx-reviewers

In bug 1823411 prim_shared.glsl's vClipMaskUv was made mediump,
assuming it was safe to do so as it is an unnormalized texture
coordinate. This is, however, causing fuzziness in a test on Adreno
devices, so we are now reverting it to highp.

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/9dd24249fabe
Make vClipMaskUv highp again. r=gfx-reviewers,lsalzman

Comment 4

[Cristina Horotan [:chorotan]]

https://hg.mozilla.org/mozilla-central/rev/9dd24249fabe

Comment 5

[Cristian Tuns]

This seems to be working in today’s beta-simulation: https://treeherder.mozilla.org/jobs?repo=try&group_state=expanded&revision=d1aa3fed8a0dda2d0552fc7ae96a24e156542033&searchStr=android%2C13.0%2Cpixel5%2Caarch64%2Cshippable%2Copt%2Creftests%2Cwithout%2Cfission%2Cenabled%2Ctest-android-hw-p5-13-0-android-aarch64-shippable-qr%2Fopt-geckoview-reftest-qr-nofis%2Cr