1821482 - Intermittent [tier 2] TEST-UNEXPECTED-FAIL | layout/reftests/bugs/256180-6.html == layout/reftests/bugs/256180-6-ref.html | image comparison, max difference: 226, number of differing pixels: 602

Status: RESOLVED DUPLICATE of bug 1440376

(Core :: Layout, defect)

Description

[Iulian Moraru]

The failure summary does not appear in treeherder due to "Log parsing was skipped since the log exceeds the size limit."

• Push where the failure appeared
• Failure log
• Raw failure log

''' INFO '''

There are numerous areas highlighted as different. Can't really tell what is different with the naked eye.

''' INFO '''

Comment 1

[David Shin[:dshin]]

This looks like some sort of ClearType weirdness..

Comment 2

[Daniel Holbert [:dholbert]]

This looks like it was a run with bad RAM. There are zillions of reftest failures (that's why the log is so big) with all of them having small random-looking differences, e.g. a blue square with one random pixel mismatching with rgb(4,4,255) != rgb(0,0,255)

This is usually a sign of bad RAM, when it's a one-off.

Comment 3

[Daniel Holbert [:dholbert]]

Attachment: snippet of log showing some of the failures

Comment 4

[Daniel Holbert [:dholbert]]

reftest-analyzer view of the piece of the log that I just attached:
[EDIT: looks like there's a CSP that prevents reftest-analyzer from automagically working with bugzilla atachments, but you can copypaste the attachment's contents into https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml and hit "process pasted log"]

No text involved in these tests (so not a text antialiasing thing -- though when you're looking at testcases with text, that's a good default assumption).

The mismatching pixels are random pixels at the edges of shapes, with a barely mismatching color channel. Looks like bit-flips from bad RAM.

Comment 5

[Daniel Holbert [:dholbert]]

whoops, I used the wrong "$problem caused by bad memory" dupe-target. Meant to dupe to bug 1440376; fixing.