Age | Commit message (Collapse) | Author | Files | Lines |
|
https://bugs.freedesktop.org/show_bug.cgi?id=31992
https://bugs.freedesktop.org/show_bug.cgi?id=31993
https://bugs.freedesktop.org/show_bug.cgi?id=31994
https://bugs.freedesktop.org/show_bug.cgi?id=31995
|
|
This was clearly broken in testing. Who knows... Fixes for me.
Test with a Nastaleeq font, or with Arabic Typesetting.
Backporting from Chromium.
|
|
|
|
|
|
|
|
This was causing SIGBUS on some ARM systems.
|
|
Reported in Oct 2011 to the mailing list by John Tapsell and confirmed
by Khaled Hosny.
https://bugreports.qt.nokia.com//browse/QTBUG-4475
http://lists.freedesktop.org/archives/harfbuzz/2011-January/001062.html
|
|
|
|
|
|
As reported by Kenichi Ishibashi.
|
|
|
|
|
|
The enum value doesn't make sense, HB_Other_NotAssigned
should be used for unassigned characters.
|
|
The old code would work for a maximum of one diacritic and
insert dotted circles in valid syllables.
|
|
Reviewed-by: Lars Knoll
|
|
The logClusters local variable is not needed here since we are
shaping syllable by syllable. Remove it to avoid compiler warnings
about set but not used variables.
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
|
|
This fixes a crash in thaiWordBreaks() detected a WebKit layout test.
Specifically, editing/selection/thai-word-at-document-end.html
Regressed with cce760d41f115fecd5b9b6b20b62883b10a9c204
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
Disable data structure packaging with RVCT, it appears to miscompile.
Signed-off-by: Anreas Kling <andreas.kling@nokia.com>
|
|
Regressed with a1714c84f245beb182a2221a1832b041788af200 and
ab9a897b688e991a8405cf938dea9d6a2f1ac072, which removed the handling for value
records that specify an device advance tables but do not actually have an
offset for it.
Previously HB_OPEN_Get_Device handled this through the HB_Device being
allocated as structure and zero initialized. Now this case is represented
by a null pointer HB_Device, which we must handle.
Reviewed-by: Andreas Kling <andreas.kling@nokia.com>
|
|
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
Cleanup was a missing for the case where loading an X placement
device table failed.
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
usage of _HB_OPEN_Free_Device() in CaretValue cleanup
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
Fixes regression from 5ab1e41c + a1714c84
"Get" functions should read from tables, not create them.
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
Saves sizeof(pointer) * 1 when the font doesn't have these tables.
Cuts resident memory consumption by 40kB when loading qt.nokia.com
in QtWebKit.
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
Re-ordered some gpos structures and removed the packing from HB_ShaperItem
as it's not worth it.
Reviewed-by: Andreas Kling
|
|
|
|
Many fonts don't have these tables, so we can save sizeof(pointer) * 3.
This cuts resident memory consumption by 300kB when loading qt.nokia.com
in QtWebKit.
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
Disable structure packaging with RVCT for the moment
|
|
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
|
|
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
Signed-off-by: Lars Knoll <lars.knoll@nokia.com>
|
|
This cuts persistent memory consumption by over 1MB when simply browsing
some popular websites with QtWebKit.
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
Signed-off-by: Lars Knoll <lars.knoll@nokia.com>
|
|
These are often unnecessary and allocating them dynamically will save
5 bytes per instance.
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
Signed-off-by: Lars Knoll <lars.knoll@nokia.com>
|
|
Support for MM fonts was discontinued in OpenType 1.3 (April 2001) and
we can save 8 bytes per ValueRecord by ignoring them.
This does not introduce any behavior change, since Qt never installed a
handler for MM fonts.
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
Signed-off-by: Lars Knoll <lars.knoll@nokia.com>
|
|
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
Signed-off-by: Lars Knoll <lars.knoll@nokia.com>
|
|
* Load libthai.so.0 since libthai.so is not there on all systems
* Remove dependency on codecs. Unicode->TIS620 is so simple we can
simply hardcode it in harbuzz-thai.c
* Speed up detection of word boundaries
* Falback when libthai is not found is now to not break instead of
breaking after every character (in line with recommendations from
unicode.org linebreaking algorithm)
Reviewed-by: Simon Hausmann
|
|
This saves significant memory, as the HB_CharAttributes
and HB_GlyphAttributes structures are not 4 byte aligned
anymore, by aligned to 1 resp. 2 bytes.
The change is not compliant with ISO C, but accepted by
almost all compilers.
|
|
|
|
See http://bugreports.qt.nokia.com/browse/QTBUG-1887.
We were not finding the base character correctly in the case
where the syllable contained a ZWJ.
In addition, the indic OT specs require us to also apply the 'loca',
'cjct' and 'calt' features. They seem to be mostly unused by todays
fonts, but we should better apply them anyways.
|
|
availableGlyphs has to be initialized before we call
stringToCMap.
Added a test case that triggered the problem.
|
|
|
|
Add a greek shaper that maps decomposed greek characters
to their precomposed versions in the greek extended range.
Reviewed-By: Simon Hausmann
|
|
The patch causes crashes using hebrew, and I can't see what
the change is supposed to fix.
See also http://bugreports.qt.nokia.com/browse/QTBUG-6436
|
|
Qt-merge-request: http://qt.gitorious.org/qt/qt/merge_requests/2174
Reviewed-by: Eskil
Signed-off-by: Simon Hausmann <simon.hausmann@nokia.com>
|
|
|
|
This fixes a bug in the rendering of Malayalam
|
|
Long outstanding bug report for Qt. See
http://bugreports.qt.nokia.com/browse/QTBUG-1042
|
|
Extend the arabic shaping module to also support N'Ko.
|
|
As reported on the mailing list on June 10, 2009
by Praveen A.
|
|
See http://bugreports.qt.nokia.com/browse/QTBUG-1802
|