1740922 - High frequency [tier 2] layout/reftests/bugs/1738700-1.html == layout/reftests/bugs/1738700-1-ref.html | image comparison, max difference: 3, number of differing pixels: 1707 Gecko 96 merges to Beta on 2021-12-06

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[Cristian Tuns]

[Tracking Requested - why for this release]:

Central-as-late beta simulation:
How to run these simulations:

Failure log:
Reftest analyzer link:

''' INFO '''

Many of the arabic characters are highlighted as different from the reference image.

''' INFO '''

[task 2021-11-12T13:09:54.710Z] 13:09:54     INFO -  REFTEST TEST-START | layout/reftests/bugs/1738700-1.html == layout/reftests/bugs/1738700-1-ref.html
[task 2021-11-12T13:10:05.830Z] 13:09:54     INFO -  REFTEST TEST-LOAD | http://10.7.205.220:8854/tests/layout/reftests/bugs/1738700-1.html | 67 / 2169 (3%)
[task 2021-11-12T13:10:05.830Z] 13:10:05     INFO -  REFTEST TEST-LOAD | http://10.7.205.220:8854/tests/layout/reftests/bugs/1738700-1-ref.html | 67 / 2169 (3%)
[task 2021-11-12T13:10:05.830Z] 13:10:05     INFO -  REFTEST INFO | REFTEST fuzzy test (0, 0) <= (3, 1707) <= (3, 1328)
[task 2021-11-12T13:10:05.830Z] 13:10:05  WARNING -  REFTEST TEST-UNEXPECTED-FAIL | layout/reftests/bugs/1738700-1.html == layout/reftests/bugs/1738700-1-ref.html | image comparison, max difference: 3, number of differing pixels: 1707

Comment 1

[Cristian Tuns]

Hi Jonathan, your latest patch from bug 1740413 doesn't seem to work. This is still happening on beta sim, we got 1 green from 10. Can you please take a look? Thanks.

Comment 2

[Jonathan Kew [:jfkthame]]

(In reply to Cristian Tuns from comment #1)

Hi Jonathan, your latest patch from bug 1740413 doesn't seem to work. This is still happening on beta sim, we got 1 green from 10. Can you please take a look? Thanks.

Looks like the fuzz limit needs to be raised further. The glyphs are all rendering in the right places, there's just some discrepancy in the antialiased edges.

(It's frustrating that there seems to be so much variability between runs; do we know why this is? Are we running these Android test jobs on a variety of slightly different configurations picked at random from among a pool of machines? Otherwise, I don't see why we should see such inconsistency in the number of fuzzy antialiasing pixels.)

Comment 3

[Jonathan Kew [:jfkthame]]

Attachment: Bug 1740922 - Increase antialiasing fuzz again to allow for varying results on Android. r?lsalzman

Comment 4

[Jonathan Kew [:jfkthame]]

Seems like we get a more-or-less arbitrary number of fuzz-pixels, but as long as the max-difference is only 3, there's no real problem here; an actual misplaced glyph would result in a much greater difference and would be clearly visible as red in the screenshot.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/9f6de6f4ead4
Increase antialiasing fuzz again to allow for varying results on Android. r=lsalzman

Comment 6

[Noemi Erli[:noemi_erli]]

https://hg.mozilla.org/mozilla-central/rev/9f6de6f4ead4

Comment 7

[Intermittent Failures Robot]

1 failures in 2723 pushes (0.0 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• try: 1

Platform and build breakdown:

• android-hw-p2-8-0-android-aarch64-shippable-qr: 1
  • opt: 1

For more details, see:
https://treeherder.mozilla.org/intermittent-failures/bugdetails?bug=1740922&startday=2021-11-08&endday=2021-11-14&tree=all

Comment 8

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

No failure observed in latest central-as-late-beta simulation.