Bug 66376

Summary:

REGRESSION (r93140?): fast/text/international/text-combine-image-test.html failing on Windows

Product:

WebKit

Reporter:

Adam Roben (:aroben) <aroben>

Component:

Tools / Tests

Assignee:

Nobody <webkit-unassigned>

Status:

NEW ---

Severity:

Normal

CC:

joepeck, mitz

Priority:

Keywords:

LayoutTestFailure, PlatformOnly, Regression

Version:

528+ (Nightly build)

Hardware:

OS:

Windows 7

URL:

http://build.webkit.org/results/Windows%207%20Release%20(Tests)/r93147%20(15530)/results.html

Attachments:

Description	Flags
before image	none
after image	none

Description Adam Roben (:aroben) 2011-08-17 06:51:03 PDT

fast/text/international/text-combine-image-test.html started failing on Windows between r93140 and r93147 (inclusive). I suspect r93140 is to blame.

http://trac.webkit.org/log/trunk?rev=93147&stop_rev=93140

http://build.webkit.org/results/Windows%207%20Release%20(Tests)/r93139%20(15529)/results.html passed
http://build.webkit.org/results/Windows%207%20Release%20(Tests)/r93147%20(15530)/results.html failed

Comment 1 Adam Roben (:aroben) 2011-08-17 06:51:58 PDT

Created attachment 104169 [details]
before image

Comment 2 Adam Roben (:aroben) 2011-08-17 06:52:11 PDT

Created attachment 104170 [details]
after image

Comment 3 Adam Roben (:aroben) 2011-08-17 06:53:25 PDT

You can see from the attached images that some of the vertical characters now have a different font. The new font doesn't seem to match the horizontal characters.

Comment 4 Adam Roben (:aroben) 2011-08-17 08:22:47 PDT

Added expected failure results in http://trac.webkit.org/changeset/93209

Comment 5 Joseph Pecoraro 2011-08-17 10:29:30 PDT

Thanks for pointing this out. My change fixed a leak, where a font that had
previously been used as a remote CSS Font Face's temporary transparent font,
(while the font face is loading), would be considered active forever.

I just looked over the Windows related changes I made in r93140 and r93147
and they look correct. So, since this could be a cache issue that persisted
between tests. To try to verify that I would check out if this test has the same
expected results now if it was run individually before r93140.

Unfortunately I don't have any other ideas why these changes would have
negatively affected this test. This test does not use font face, it uses a font
named "HiraMinProN-W3", which I assume is a system font because I see
no SVG/CSS3 font-face declarations or included declarations.

Comment 6 mitz 2011-08-17 12:04:31 PDT

(In reply to comment #5)
> This test does not use font face, it uses a font
> named "HiraMinProN-W3", which I assume is a system font because I see
> no SVG/CSS3 font-face declarations or included declarations.

The Windows system running the test doesn’t have a font by that name installed, so it proceeds with system fallback.