been touched, while keeping everything backwards-compatible..
* include/freetype/t1tables.h: re-naming structure types. This done
basically:
typedef T1_Struct_
{
} T1_Struct;
becomes:
typedef PS_StructRec_
{
} PS_StructRec, *PS_Struct;
typedef PS_StructRec T1_Struct; /* backwards-compatibility */
hence, we increase the coherency of the source code by effectuively
using the 'Rec' prefix for structure types..
names table loader. Invalid individual name entries are now handled
correctly. This allows the loading of very buggy fonts like
"foxjump.ttf" without allocating tons of memory and causing crashes..
the debug sub-system initialization. trace levels can now be specified
within the "FT2_DEBUG" environment variable. See the comments within
"ftdebug.c" for more details
* include/freetype/internal/fttrace.h: new file to define the trace
levels used for debugging. it is used both to define enums and
toggle names for FT2_DEBUG
* src/base/ftobjs.c, src/base/ftstream.c: FT_Assert renamed to
FT_ASSERT
* include/freetype/internal/ftextend.h, src/base/ftextend.c,
src/base/Jamfile, src/base/rules.mk: removing "ftextend" from the
library, since it is now completely obsolete..
w.r.t. the selected charmap.
* include/freetype/freetype.h: Add prototype.
* include/freetype/internal/ftdriver.h: Add `FTDriver_getNextChar'
typedef.
(FT_Driver_Class): Use it.
* include/freetype/internal/psnames.h: Add `PS_Next_Unicode_Func'
typedef.
(PSNames_Interface): Use it.
* include/freetype/internal/tttypes.h: Add `TT_CharNext_Func'
typedef.
(TT_CMapTable): Use it.
* src/base/ftobjs.c (FT_Get_Next_Char): New function, implementing
high-level API.
* src/cff/cffdrivr.c (cff_get_next_char): New function.
(cff_driver_class): Add it.
* src/cid/cidriver.c (Cid_Get_Next_Char): New function.
(t1cid_driver_class): Add it.
* src/pcf/pcfdriver.c (PCF_Get_Next_Char): New function.
(pcf_driver_class): Add it.
* src/psnames/psmodule.c (PS_Next_Unicode): New function.
(psnames_interface): Add it.
* src/sfnt/ttcmap.c (code_to_next0, code_to_next2, code_to_next4,
code_to_next6, code_to_next_8_12, code_to_next_10): New auxiliary
functions.
(TT_CharMap_Load): Use them.
* src/truetype/ttdriver.c (Get_Next_Char): New function.
(tt_driver_class): Add it.
* src/type1/t1driver.c (Get_Next_Char): New function.
(t1_driver_class): Add it.
* src/winfnt/winfnt.c (FNT_Get_Next_Char): New function.
(winfnt_driver_class): Add it.
* src/pcf/pcfread.c (pcf_load_font): For now, report Unicode for
Unicode and Latin 1 encodings.
It didn't work correctly when several arguments were passed
to it. For example, when typing:
./configure --disable-shared --prefix=/usr
the "--prefix=/usr" was incorrectly sent to the "make" program..
the face object (face->toc.tables, face->root.family_name,
face->root.available_size, face->charset_encoding,
face->charset_registry are now freed). Thanks to Niels Moseley.
unpleasant overwrites in the case of invalid glyph data (found in
the output of buggy font converters)
the computation of auto-hinted stem widths has been modified to
avoid certain color fringes in LCD-decimation rendering (a.k.a.
"ClearType")
library when dealing with certain weird fonts (like "Stalingrad",
in "sadn.pfb". This font has no full font name entry.. )
* src/base/ftoutln.c, include/freetype/ftoutln.h: added the
FT_Outline_Check API to check the consistency of outline data
* src/base/ftobjs.c (FT_Load_Glyph): added a call to the new
FT_Outline_Check to ensure that loaded glyphs are valid. This
allows certain fonts like "tt1095m_.ttf" to be loaded even though
it appears they contain really funky glyphs..
there still is a bug there though.. !!
* src/pshinter/pshrec.c: now ignores invalid "hintmask" and "cntrmask"
operators (instead of returning an error). Glyph 2028 of the CFF font
"MSung-Light-Acro" couldn't be rendered otherwise (it seems its
charstring is buggy, though this requires more analysis)..
src/pshinter/ahalgo2.c, src/pshinter/pshglob.h: fixed a bug where
the X and Y axis where inversed in the postscript hinter. this
caused problem when displaying on non-square surfaces..
constant with a fixed-float equivalent. For some reason, some compilers
aren't capable of directly computing a floating pointer constant casted
to FT_Fixed, and will link a math library instead !!
* src/cff/cffload.h, src/cff/cffload.c, src/cff/cffgload.c: updated
to mode the definition of encoding tables within "cffload.c" instead
of making them part of a shared header (causing problems in "multi"
builds)
for the Postscript hinter
* docs/BUGS: closed the AUTOHINT-NO-SBITS bug.
* src/pshinter/pshrec.c (t2_hint_stems), src/cff/cffobjs.h,
src/cff/cffobjs.c, src/cff/cffload.c, src/cff/cffload.h,
src/cff/cffgload.c, src/cff/cffgload.h, src/cff/cffdriver.c,
include/freetype/internal/cfftypes.h: added Postscript hinter support
to the CFF driver
* src/base/ftobjs.c (FT_Done_Library): fixed a stupid bug that crashed
the library on exit
on hinted glyphs..
* src/cid/cidgload.c, src/cid/cidobjs.c, src/cid/cidobjs.h,
src/cid/cidriver.c, include/freetype/internal/t1types.h: added
Postscript hinter support to the CID font driver !!
(FT_Load_Glyph): "fixed" the bug that prevented embedded bitmaps from
begin loaded when the auto-hinter is used.. This actually is a hack
but will be enough until the internal re-design scheduled for
FreeType 2.1
some of the exported functions should only be used by applications
that need to implement custom cache types
* src/truetype/ttgload.c: fixed a nasty bug that prevent composites
from loading correctly. Believe it or not, this was due to an invalid
macro definition !!
passing "-L/usr/lib" to gcc
* docs/FTL.TXT: simple fix (change "LICENSE.TXT" to "FTL.TXT")
* builds/unix/freetype2.m4: added autoconf macro, we need to install
it in $(prefix)/share/aclocal/freetype2.m4 but I didn't modified
builds/unix/install.mk yet..
GET_LongLE and GET_ULongLE which where incorrect (creating problems
in the pcf driver)..
* INSTALL: updated the instructions to build shared libraries with
Jam.. they were simply erroneous..
purposes..
* src/smooth/ftsmooth.c (ft_smooth_render): fixed a nasty hidden bug where
outline shifting wasn't correctly undone after bitmap rasterization. this
created problems with certain glyphs (like '"' of certain fonts..) and
the cache system..
* src/pshinter/pshalgo2.c (psh2_hint_table_init),
src/pshinter/pshalgo1.c (psh1_hint_table_init): removed compiler
warnings
* include/freetype/cache/*, src/cache/*: yet another massive rewrite of
the caching sub-system, in order to both increase performance and allow
simpler cache sub-classing. As an example, the code for the image and
sbit caches is now much simpler
I still need to update the documentation in www/freetype2/docs/cache.html
to reflect the new design though..
codes and LCIDs as found in MSDN (Passport SDK). Also added
comments about the meaning of bit 57 of OS/2 (TT_UCR_SURROGATES)
which with OpenType v.1.3 now means "there is a character beyond
FFFF in this font". Thanks to Detlef Wuerkner <TetiSoft@apg.lahn.de>
for noticing this.
routine that created nasty alignment artefacts.
* src/pshinter/pshrec.c, tests/gview.c: debugging updates..
* src/smooth/ftgrays.c: de-activated experimental gamme support,
apparently, "optimal" gamma tables depend on the monitor type,
resolution and general karma, so it's better to compute them outside
of the rasterizer itself..
produces smoother glyphs at small sizes for very little cost
* src/autohint/ahglyph.c, src/autohint/ahhint.c: various fixes to
the auto-hinter. They merely improve the output of sans-serif fonts.
Note that there are still problems with serifed fonts and composites
(accented characters)
* tests/gview.c: updated the debugging glyph viewer to show the
hints generated by the "autohint" module
improvements to the memory debugger to report more information in
case of errors. Also, some allocations that occured through
REALLOC couldn't be previously catched correctly..
the debugging memory manager to report the location (source file name
+ line number) where leaked memory blocks are allocated in the source
file.. funny, isn't it ??
* src/cff/cffload.c (CFF_Done_Font): Free subfonts array only if
we are working with a CID keyed CFF font. Otherwise, a variable
that was never allocated memory might freed. This is a correction
to the previous patch for freeing subfonts.
David Turner