588684 - layout/reftests/svg/objectBoundingBox-and-fePointLight-01.svg fails on linux with 130078

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Timothy Nikkel (:tnikkel)]

layout/reftests/svg/objectBoundingBox-and-fePointLight-01.svg fails on linux on tinderbox with 130078 applied. I don't think making this pass should block landing of 130078 for the following reasons. I can not tell apart the test and reference using my eyes alone, even after reftest-analyzer points out the differences to me. Using reftest-analyzer to circle the differences, the failure mode with 130078 is the same as the random orange bug 584624. I can not reproduce the failure on my Ubuntu box. Before this test was just randomly failing it was perma-orange (bug 558977), and then was fixed (aka made randomly orange) by a cairo update.

We'll keep this bug open to track marking this test as passing on linux.

Comment 1

[Timothy Nikkel (:tnikkel)]

Attachment: mark the test as failing on linux

Comment 2

[Timothy Nikkel (:tnikkel)]

http://hg.mozilla.org/mozilla-central/rev/fa55d98c6552

Comment 3

[Robert Longson [:longsonr]]

Does it pass now that bug 556487 landed? That marked the test as random but the patch in the bug suggested that it now passed.

Comment 4

[Timothy Nikkel (:tnikkel)]

We could try looking at some logs to see if there were any failures, or just marking it as passing and seeing if we still get intermittent failures.

Comment 5

[Daniel Holbert [:dholbert]]

Looks like it's still failing sometimes, at least.  I just glanced at one log and found this:

REFTEST TEST-START | file:///home/cltbld/talos-slave/mozilla-central_fedora-debug_test-reftest/build/reftest/tests/layout/reftests/svg/objectBoundingBox-and-fePointLight-01.svg
++DOMWINDOW == 153 (0x122570c8) [serial = 12985] [outer = 0x86e0278]
REFTEST TEST-START | file:///home/cltbld/talos-slave/mozilla-central_fedora-debug_test-reftest/build/reftest/tests/layout/reftests/svg/objectBoundingBox-and-fePointLight-01-ref.svg
++DOMWINDOW == 154 (0x9fffa18) [serial = 12986] [outer = 0x86e0278]
REFTEST TEST-KNOWN-FAIL(EXPECTED RANDOM) | file:///home/cltbld/talos-slave/mozilla-central_fedora-debug_test-reftest/build/reftest/tests/layout/reftests/svg/objectBoundingBox-and-fePointLight-01.svg |

Source: full-log version of 
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1285008125.1285010252.5515.gz
Rev3 Fedora 12 mozilla-central debug test reftest on 2010/09/20 11:42:05
If the test had passed, that would be KNOWN-PASS(EXPECTED RANDOM) in that last line.

Comment 6

[Robert Longson [:longsonr]]

Attachment: Bug 588684 - Mark objectBoundingBox-and-fePointLight reftests passing on linux

We can mark this fuzzy if necessary but it does seem OK now

Comment 7

[Robert Longson [:longsonr]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=789b600241d3130d76c56744d6f6b76c3d341a0e

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/1c5975c43b7c
Mark objectBoundingBox-and-fePointLight reftests passing on linux r=dholbert

Comment 9

[Andreea Pavel [:apavel]]

https://hg.mozilla.org/mozilla-central/rev/1c5975c43b7c

Comment 10

[Timothy Nikkel (:tnikkel)]

It's not every day a regression from bug 130078 gets resolved!