Roll HarfBuzz to 1.2.6

third_party/harfbuzz-ng/NEWS

+Overview of changes leading to 1.2.6
+Friday, April 8, 2016
+====================================
+
+- Blacklist GDEF table of another set of Times New Roman (Bold) Italic.
+- DirectWrite backend improvements.  Note: DirectWrite backend is
+  exclusively for our internal testing and should NOT be used in any
+  production system whatsoever.
+
+
+Overview of changes leading to 1.2.5
+Monday, April 4, 2016
+====================================
+
+- Fix GDEF mark-filtering-set, which was broken in 1.2.3.
+
+
+Overview of changes leading to 1.2.4
+Thursday, March 17, 2016
+====================================
+
+- Synthesize GDEF glyph class for any glyph that does not have one in GDEF.
+  I really hope we don't discover broken fonts that shape badly with this
+  change.
+- Misc build and other minor fixes.
+- API changes:
+  - Added HB_NDEBUG.  It's fine for production systems to define this to
+    disable high-overhead debugging checks.  However, I also reduced the
+    overhead of those checks, so it's a non-issue right now.  You can
+    forget it.  Just not defining anything at all is fine.
+
+
+Overview of changes leading to 1.2.3
+Thursday, February 25, 2016
+====================================
+
+- Blacklist GDEF table of certain versions of Times New Roman (Bold) Italic,
+  due to bug in glyph class of ASCII double-quote character.  This should
+  address "regression" introduced in 1.2.0 when we switched mark zeroing
+  in most shapers from BY_UNICODE_LATE to BY_GDEF_LATE.
+  This fourth release in a week should finally stablize things...
+
+- hb-ot-font's get_glyph() implementation saw some optimizations.  Though,
+  might be really hard to measure in real-world situations.
+
+- Also, two rather small API changes:
+
+We now disable some time-consuming internal bookkeeping if built with NDEBUG
+defined.  This is a first time that we use NDEBUG to disable debug code.  If
+there exist production systems that do NOT want to enable NDEBUG, please let
+me know and I'll add HB_NDEBUG.
+
+Added get_nominal_glyph() and get_variation_glyph() instead of get_glyph()
+
+New API:
+- hb_font_get_nominal_glyph_func_t
+- hb_font_get_variation_glyph_func_t
+- hb_font_funcs_set_nominal_glyph_func()
+- hb_font_funcs_set_variation_glyph_func()
+- hb_font_get_nominal_glyph()
+- hb_font_get_variation_glyph()
+
+Deprecated API:
+- hb_font_get_glyph_func_t
+- hb_font_funcs_set_glyph_func()
+
+Clients that implement their own font-funcs are encouraged to replace
+their get_glyph() implementation with a get_nominal_glyph() and
+get_variation_glyph() pair.  The variation version can assume that
+variation_selector argument is not zero.  Old (deprecated) functions
+will continue working indefinitely using internal gymnastics; it is
+just more efficient to use the new functions.
+
+
+Overview of changes leading to 1.2.2
+Wednesday, February 24, 2016
+====================================
+
+- Fix regression with mark positioning with fonts that have
+  non-zero mark advances.  This was introduced in 1.2.0 while
+  trying to make mark and cursive attachments to work together.
+  I have partially reverted that, so this version is much more
+  like what we had before.  All clients who updated to 1.2.0
+  should update to this version.
+
+
 Overview of changes leading to 1.2.1
 Friday, February 23, 2016
 ====================================
 
 - CoreText: Fix bug with wrong scale if font scale was changed later.
   https://github.com/libass/libass/issues/212
 - CoreText: Drastically speed up font initialization.
 - CoreText: Fix tiny leak.
 - Group ZWJ/ZWNJ with previous syllable under cluster-level=0.
   https://github.com/behdad/harfbuzz/issues/217
@@ skipping 1325 matching lines @@
     HB_VERSION_MAJOR
     HB_VERSION_MINOR
     HB_VERSION_MICRO
     HB_VERSION_STRING
     HB_VERSION_CHECK()
     hb_version()
     hb_version_string()
     hb_version_check()