Closed
Bug 517283
Opened 15 years ago
Closed 15 years ago
fix Core Text reftest failures
Categories
(Core :: Graphics, defect)
Tracking
()
RESOLVED
FIXED
People
(Reporter: jfkthame, Unassigned)
Details
Attachments
(1 file)
3.95 KB,
patch
|
roc
:
review+
|
Details | Diff | Splinter Review |
Preparing for Core Text to become the default text back-end on OS X, I've been re-running reftests, and identified a couple of issues.
(1) The Core Text path could drop an initial glyph when rendering text with bidi-override: rtl, because of an incorrect test for the edges of the range being rendered. This caused a bidi reftest failure. Fixed in gfxCoreTextFonts.cpp by the attached patch.
(2) There is a bug in Core Text when rendering Arabic with left-to-right override and an OpenType (rather than AAT) font. (Reported to Apple as rdar://7231250.) This caused a couple of bidi reftests to fail. Fixed by specifying a standard AAT font in the reftest, to ensure this will be used rather than an OpenType one. (This will not affect other platforms where Geeza Pro is not present; they will simply fall back on the default serif font.)
With these fixes, I get the exact same reftest results with either ATSUI or Core Text. (I'm still seeing some failures, but as these are consistent regardless of the back-end used, and are obviously not happening on Tinderbox, I'm assuming they are due to local config issues such as available fonts.)
Attachment #401256 -
Flags: review?(roc)
Attachment #401256 -
Flags: review?(roc) → review+
We really should have zero reftest failures. Reftests should not depend on available fonts. I get some local failures too and I've been lazy about diagnosing and fixing them, but we should do better! Please figure out your failures and file bugs or fix them if you can :-).
Reporter | ||
Comment 2•15 years ago
|
||
Status: NEW → RESOLVED
Closed: 15 years ago
Resolution: --- → FIXED
You need to log in
before you can comment on or make changes to this bug.
Description
•