Closed Bug 593330 Opened 9 years ago Closed 9 years ago

reftest layout/reftests/transform/abspos-1* fails on win7

Categories

(Core :: Layout, defect)

x86
Windows 7
defect
Not set

Tracking

()

RESOLVED FIXED
Tracking Status
blocking2.0 --- beta7+

People

(Reporter: jfkthame, Unassigned)

References

Details

(Keywords: intermittent-failure)

This seems to be a consistent failure from bug 579276 or one of the other layers-related patches that landed at the same time (592985, 590367, 593224, 590735, 589632).

http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1283508339.1283509646.6601.gz
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1283502122.1283503543.10296.gz

REFTEST TEST-UNEXPECTED-FAIL | file:///c:/talos-slave/mozilla-central_win7_test-reftest/build/reftest/tests/layout/reftests/transform/abspos-1a.html |
REFTEST TEST-UNEXPECTED-FAIL | file:///c:/talos-slave/mozilla-central_win7_test-reftest/build/reftest/tests/layout/reftests/transform/abspos-1b.html |
REFTEST TEST-UNEXPECTED-FAIL | file:///c:/talos-slave/mozilla-central_win7_test-reftest/build/reftest/tests/layout/reftests/transform/abspos-1c.html |
REFTEST TEST-UNEXPECTED-FAIL | file:///c:/talos-slave/mozilla-central_win7_test-reftest/build/reftest/tests/layout/reftests/transform/abspos-1d.html |
Pushed http://hg.mozilla.org/mozilla-central/rev/b92757b90f6f to disable the failing tests for now (marked as random on windows).
Thanks. I pushed 525563b639f1 to disable the tests just for D2D.
There were two problems in the tests. One of the problems is that with D2D/DirectWrite, there are antialiased pixels at the left edge of the "A" which, with changeset c888cdffd617, get clipped in the test but not in the reference. Adding some padding to the test fixes that. However, there's another problem which is that the rasterization of the "N" is somehow affected by the clipping region even though the N is not near the edge of the clipping region. I don't know how to get around that, which is why I disabled the tests.
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
(In reply to comment #3)
> However, there's another
> problem which is that the rasterization of the "N" is somehow affected by the
> clipping region even though the N is not near the edge of the clipping region.

I'm not sure exactly why this happens, but it's similar to some of the other issues we've seen where glyph rasterization under D2D is affected by some apparently-unrelated aspect of the context in which the glyphs are being painted. (See examples and discussion in bug 578118, for example.)

> I don't know how to get around that, which is why I disabled the tests.

Yeah. (Sigh.) It's likely that experimenting with different fonts, sizes, or text characters could give you a testcase that passes for now - but it also seems likely to be very fragile, subject to breakage at any time as a result of changes to the underlying D2D rasterization. So not really a good solution.
Blocks: 438871
Whiteboard: [orange]
You need to log in before you can comment on or make changes to this bug.