1210966 - Intermittent emoji-03.html | image comparison (!=), max difference: 90, number of differing pixels: 1

Status: RESOLVED INVALID

(Core :: Layout, defect)

Description

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

http://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=http://ftp.mozilla.org/pub/mozilla.org/mobile/tinderbox-builds/mozilla-inbound-android-api-11-debug/1443810336/mozilla-inbound_panda_android-debug_test-plain-reftest-8-bm89-tests1-panda-build796.txt.gz&only_show_unexpected=1

Comment 1

[Daniel Holbert [:dholbert]]

tl;dr: this looks like bad memory (cosmic ray?) on the machine.

The single mismatching pixel is nowhere near the actual test content -- it's hundreds of pixels to the bottom-right. And it's just a single pixel which is rgb(255,165,255), a light purple color, instead of white.

For the record, the failure log itself is:
http://ftp.mozilla.org/pub/mozilla.org/mobile/tinderbox-builds/mozilla-inbound-android-api-11-debug/1443810336/mozilla-inbound_panda_android-debug_test-plain-reftest-8-bm89-tests1-panda-build796.txt.gz


Making things a bit more fun: this test is actually marked as an "UNEXPECTED-PASS", because these files are *intended* to render differently on this platform (android), but they don't normally, due to a bug [I guess], which is why this line is marked as "fails" in the manifest.  But on this one run, the bad memory made this one pixel render in a bogus way, which made them actually render differently from each other and "pass" the != comparison.  But it wasn't a real pass -- it was just a single bogus pixel.

Comment 2

[Daniel Holbert [:dholbert]]

Attachment: gzipped log, for the record

Here's the log, for the record. The relevant lines, with data URIs omitted:

>12:08:59     INFO -  REFTEST TEST-START | http://10.26.132.20:30481/tests/layout/reftests/text/emoji-03.html
>12:08:59     INFO -  REFTEST TEST-LOAD | http://10.26.132.20:30481/tests/layout/reftests/text/emoji-03-notref.html | 163 / 1549 (10%)
>12:08:59     INFO -  REFTEST TEST-UNEXPECTED-PASS | http://10.26.132.20:30481/tests/layout/reftests/text/emoji-03.html | image comparison (!=), max difference: 90, number of differing pixels: 1
>12:08:59     INFO -  REFTEST   IMAGE 1 (TEST): ...
>12:08:59     INFO -  REFTEST   IMAGE 2 (REFERENCE): ...

Comment 3

[Mats Palmgren (inactive)]

Yeah, I agree this is bad RAM or interference of some kind.  It's not uncommon
for these errors to occur in Android reftests (every other week or so).