Age | Commit message (Collapse) | Author |
|
FreeType now depends upon libpng and zlib. While I'm adding those
dependencies, I might as well also switch libft2 to a shared
library...
Bug: 9603326
Change-Id: Ibe45c186f0c933fbecebbeffc2d990b4e99d913f
|
|
Bug: 8580410
Change-Id: I3f6245111a6bcc662d4c66b3aa15695fa2ef58f8
|
|
|
|
This patch works around a recent change to bionic that was intented to
throw an assert on null pointer, but actually threw on 0 unicode code
point, because it uses a type pun to cast the unicode code point to a
pointer type.
Change-Id: I6ebf2583201fcf8a39490877b74eceb695b8fcd0
|
|
This patch works around a recent change to bionic that was intented to
throw an assert on null pointer, but actually threw on 0 unicode code
point, because it uses a type pun to cast the unicode code point to a
pointer type.
Change-Id: I6ebf2583201fcf8a39490877b74eceb695b8fcd0
|
|
This is a fix for bug 7363916, where the inconsistent breaking of the
text into script runs was leading to a different measured width on paint
than on measure for a TextView, leading to the content being ellipsized.
This patch classifies the space as Latin, rather than the script of the
surrounding text, which means that it's always drawn from the Latin
font, and thus is measured consistently.
Change-Id: I07abfa2918698c93f55333f1e964cc24a18d3def
|
|
This commit cherrypicks the following commit into the harfbuzz.old
repository, and also picks up a simple fix to the Tibetan shaper.
http://cgit.freedesktop.org/harfbuzz.old/commit/?id=3ab7b37bdebf0f8773493a1fee910b151c4de30f
It should fix some memory leaks observed with Arabic-family scripts.
Change-Id: I31ba3f619506e843b10360b733d36097aa22653a
|
|
In fonts (such as Roboto) that have only a DFLT language tag in their
OpenType layout tables, Harfbuzz was skipping those features because it
was only matching against the explicit language tag. This patch makes
Harfbuzz check the DFLT tag too, in case there's not an explicit match.
Behdad looked at this patch and said it looks fine.
Note that committing this without having positioning fully in place will
cause regressions, as kerning makes the advances as generated by
Harfbuzz inconsistent with the advances actually used in drawText.
Change-Id: I6ea2eb2a580d00410c05c748e2db577072c25534
|
|
Change-Id: I38b145aed01968fe9895b3d629dc0ab48ccbff97
|
|
Duplicated from COPYING.
Bug: 6299628
Change-Id: If37162b8c311200a1bdf6eff38be580e2b3511d8
Signed-off-by: Wei Zhong <wzhong@google.com>
(cherry picked from commit a2ca78ca7d69124d9b3f37aea0dd2920274bf0d5)
|
|
|
|
- Don't force using packed structs:
http://cgit.freedesktop.org/harfbuzz.old/commit/?id=624feeed252024dcc076903e26f6b625f2c93c7f
- Fix RTL GPOS issue
http://cgit.freedesktop.org/harfbuzz.old/commit/?id=4717a185539d678b2040820e05bfe065d5d7c16d
Change-Id: I2461d9916855e09135517d3ca6138865c2d655a9
|
|
Change-Id: I680fb93599192adbf8c9895e98e3b5a98fe26247
|
|
Bug 5775639
In indic_shape_syllable(), control characters are skipped from the output.
However, the offsets and advances data of shaped item are not shifted
when skipping the shaped item data.
Change-Id: If24706e8bffbad533fd97a2ccb030513c4e4694a
|
|
Bug5757588
Cherry-pick the fixes for locale specific bugs and a crash bug in harfbuzz-gpos.c.
Current Android Harfbuzz(http://cgit.freedesktop.org/harfbuzz.old/) is based on
SHA 5a381d21b924da0c0843940d223cd6fff767efdc(Sep. 09 2010).
I cherry pick the following CLs:
93d8573ddcf97648983a1c1b5cdda8a3ec6d4432: For general complex scripts. Android has the fixes by Fabrice.
ff0612c2e7df1b86fc702c72e3015a6a5ae39b4c: For Tibetan only. Included in this CL. Android does not support Tibetan, so it does not hurt to include the fixes.
81f2ecafa19b602f950df5a9e6e1b99c4b5ea55e: For Indic only. Android has the fixes by Jungshik/claire.
81c8ef785b079980ad5b46be4fe7c7bf156dbf65: For general. Included in this CL to prevent crash in gpos data processing.
213bd693a55c04ef25f5e8de5f6feefa9473a30a: For Hebrew only. Android has the fixes by Claire since HC release.
3af70a9936269e7af0e7847497292e28065dea28: For Thai only. Included in this CL.
Change-Id: I4a3a8f8b3db504d59f0d042933c96645f0c23b58
|
|
|
|
Fabrice's earlier change to hb_utf16_script_run_prev also applies to
hb_utf16_script_run_next, essentially replacing a == b == c with
a == c && b == c.
Change-Id: I996bdac2837ceb061792d6fc20ce9e2d3f2c9480
|
|
Bug5733237 : Bengali init feature should only apply to beginning of a word.
A bug caused by the truncation loss in isLtter() and isMark() of harfbuzz-indic.cpp.
Change-Id: I23fedf28ff580747a33d4db19402d064cb290cd7
|
|
|
|
|
|
|
|
- add missing unicode point ranges
- put some range into HB_Script_Arabic instead of HB_Script_Inherited
This may break Syriac script as some of the unicode point are shared with Arabic
Change-Id: I4e846c10f530097ad7eed9730cad53d4e1adec78
|
|
- fix the condition for setting the script
Change-Id: Id8d7da34bad485d2e0a1e34cc7ebb831a73a6be6
|
|
Change-Id: I2fe433eaf0057835528b4e01447dc12547993d2a
Signed-off-by: Iliyan Malchev <malchev@google.com>
|
|
Problem: Non-spacing Hebrew discritics is rendered as a spacing
character(base char) instead of positioned to last precediding
base character.
This is a new bug after sync to latest Harfbuzz.old library.
Reverted to the older version of harfbuzz-hebrew.c
Change-Id: I7e9cc5296fdfb0654ab1eeccfce960a3c2ffa5f0
|
|
The Harfbuzz version includes:
1. More bug fixings.
2. Memory reduction - 50 % less.
3. Shapping for Greek.
Note: The following files include the bug fixings from Chrome:
a. contrib/harfbuzz-unicode.c
b. src/harfbuzz-shaper.cpp
Change-Id: I9f5fc16b01c85fb54bbd66c6f9f5526e68e7d01a
|
|
Unicode property table.
Change-Id: I53c6eabd88f1138ddd5ee963d58e87c6ee5f6784
|
|
Change-Id: I9170e4e86620636d0a5d0ad377602439a4c4740d
|
|
Harfbuzz is an open source library for complex scripts such as
Arabic, Thai and Hindi.
This code was taken from b0d396aa88b3cdf8cea896bfeeba197656e1cdb1
(git://anongit.freedesktop.org/harfbuzz) with chrome's patch.
Chrome team will talk with upstream about the patch.
Change-Id: I06f503bbd537a866a2425b9998075e7dfce4f40e
|
|
|