This fixes a 13-year old bug. The original overflow check should have
been updated when rounding was introduced into this function
(c2cd00443b).
* src/base/ftcalc.c (FT_DivFix) [!FT_LONG64]: Updated.
* include/freetype.h (FT_DivFix): Updated documentation.
* include/freetype.h (FT_MulFix): Unconditionally defined.
* src/base/ftcalc.c [FT_MULFIX_ASSEMBLER]: Move code from here...
* include/internal/ftcalc.h [FT_MULFIX_ASSEMBLER]: ... to here,
which conditionally replaces the function with an inline version
through the macro.
FT_MulFix assembler used to reside in ftcalc.c before f47d263f1b.
* include/config/ftconfig.h, builds/unix/ftconfig.in,
builds/vms/ftconfig.h [FT_MULFIX_ASSEMBLER]: Move code from here...
* src/base/ftcalc.c [FT_MULFIX_ASSEMBLER]: ... to here.
* src/base/ftcalc.c (FT_DivFix): Use unsigned values for
computations which use the left shift operator and convert to signed
as the last step.
* src/base/fttrigon.c (ft_trig_prenorm, FT_Vector_Rotate,
FT_Vector_Length, FT_Vector_Polarize): Ditto.
* src/cff/cffgload.c (cff_decoder_parse_charstrings): Simplify.
* src/cff/cffload.c (cff_subfont_load): Fix constant.
* src/cff/cffparse.c (cff_parse_integer, cff_parse_real, do_fixed,
cff_parse_fixed_dynamic): Use unsigned values for computations which
use the left shift operator and convert to signed as the last step.
* src/cid/cidload.c (cid_get_offset): Ditto.
* src/psaux/psconv.c (PS_Conv_ToFixed): Ditto.
* src/psaux/t1decode.c (t1_decoder_parse_charstrings): Ditto.
* src/truetype/ttinterp.c (TT_MulFix14, TT_DotFix14): Ditto.
This is essentially a mechanical conversion, adding inclusion of
`FT_INTERNAL_DEBUG_H' where necessary, and providing the macros for
stand-alone compiling modes of the rasterizer modules.
To convert the remaining occurrences of FT_Err_XXX and friends it is
necessary to rewrite the code. Note, however, that it doesn't harm
if some cases are not handled since FT_THROW is a no-op.
For some divisions, we use casts to 32bit entities. Always guard
against division by zero with these casts also.
* src/base/ftcalc.c (ft_div64by32): Remove redundant cast.
(FT_MulDiv, FT_MulDiv_No_Round): Add 32bit cast.
(FT_DivFix): Add 32bit cast (this omission triggered the bug).
include/freetype/config/ftconfig.h, builds/unix/ftconfig.in: Minor
beautifying.
* include/freetype/ftadvanc.h, include/freetype/ftgasp.h,
include/freetype/ftlcdfil.h: Protect against FreeType 1.
Some other minor fixes.
* devel/ftoption.h: Synchronize with
include/freetype/config/ftoption.h.
Formatting, documentation improvements.
builds/unix/ftconfig.in, include/freetype/freetype.h,
src/base/ftcalc.c:
Make FT_MulFix an inlined function. Also provide an assembler
implementation for ARM architectures. this is done to speedup
FreeType a little (on x86 3% when loading+hinting, 10% when
rendering, ARM savings are more important though).
Disable this by undefining FT_CONFIG_OPTION_INLINE_MULFIX in
ftconfig.h
* include/freetype/internal/ftcalc.h, src/base/ftcalc.c
(FT_Matrix_Multiply_Scaled, FT_Vector_Transform_Scaled): New
functions.
* src/cff/cffobjs.h (CFF_Internal): New struct. It is used to
provide global hinting data for both the top-font and all subfonts
(with proper scaling).
* src/cff/cffobjs.c (cff_make_private_dict): New function, using
code from `cff_size_init'.
(cff_size_init, cff_size_done, cff_size_select, cff_size_request):
Use CFF_Internal and handle subfonts.
(cff_face_init): Handle top-dict and subfont matrices correctly;
apply some heuristic in case of unlikely matrix concatenation
results. This has been discussed with people from Adobe (thanks
goes mainly to David Lemon) who confirm that the CFF specs are fuzzy
and not correct.
* src/cff/cffgload.h (cff_decoder_prepare): Add `size' argument.
* src/cff/cffgload.c (cff_builder_init): Updated.
(cff_decoder_prepare): Handle hints globals for subfonts.
Update all callers.
(cff_slot_load): Handling scaling of subfonts properly.
* src/cff/cffparse.c (cff_parse_fixed_dynamic): New function.
(cff_parse_font_matrix): Use it.
* src/cff/cfftypes.h (CFF_FontDictRec): Make `units_per_em'
FT_ULong.
* docs/CHANGES: Document it.
test for a valid glyph index must be deferred to the font drivers.
This patch fixes Savannah bug #18301.
* src/base/ftobjs.c (FT_Load_Glyph): Don't check `glyph_index'.
* src/bdf/bdfdrivr.c (BDF_Glyph_Load), src/cff/cffgload.c
(cff_slot_load), src/cid/cidgload.c (cid_slot_load_glyph),
src/pcf/pcfdrivr.c (PCF_Glyph_Load), src/pfr/pfrobjs.c
(pfr_slot_load), src/truetype/ttdriver.c (Load_Glyph),
src/type1/t1gload.c (T1_Load_Glyph), src/winfonts/winfnt.c
(FNT_Load_Glyph): Check validity of `glyph_index'.
src/autofit/afcjk.c, src/autofit/afhints.c, src/autofit/aflatin.c,
src/autofit/aftypes.h, src/base/ftcalc.c, src/base/ftoutln.c,
src/gzip/ftgzip.c, src/psaux/psconv.c, src/truetype/ttgload.c,
src/type1/t1gload.c:
this is a major patch used to drastically improve the performance
of loading glyphs. This both speeds up loading the glypn vector
themselves and the auto-fitter.
note that we've started using inline assembler with GCC to
implement FT_MulFix, given that this function is so damn
important for the engine's performance.
the resulting speed-up is about 25%.
- fixing apinames.c, adding support for Watcom and Borland compilers
- adding generation of exported symbols list to the build system, including the Unix one !!
sorry Werner, I have no time to document this in ChangeLog at the moment
64 bit version).
* include/freetype/internal/ftcalc.h: Updated.
* src/truetype/ttinterp.c (TT_MULDIV_NO_ROUND): New macro.
(TT_INT64): Removed.
(DO_DIV): Use TT_MULDIV_NO_ROUND.
* src/pfr/pfrdrivr.c (pfr_get_metrics): Directly use
metrics->x_scale and metrics->y_scale.
use the new trigonometric functions provided in "fttrigon.h". This
gets rid of some old 64-bit computation routines, as well as many
warnings when compiling the library with the "long long" 64-bit
integer type.
created incorrect scale factors !!
* src/autohint/ahglobal.c, src/autohint/ahglyph.c, src/autohint/ahhint.c:
fixed the incorrect blue zone computations, and improved the composite
support. Note that these changes result in improved rendering, while
sometimes introducing their own artefacts. That's probably the last
big change to the autohinter before the introduction of its complete
replacement..
* docs/docmaker.py: updated the DocMaker script in order to add
command line options (--output,--prefix,--title), fix the erroneous
line numbers reported during errors and warnings, and other formatting
issues..
* src/base/ftcalc.c: various tiny fixes related to rounding in 64-bits
routines and pseudo"optimisations" :-)
latest versions from gnu.org.
* builds/compiler/gcc-dev.mk: Add `-Wno-long-long' flag.
* include/freetype/internal/ftcalc.h: Define FT_SqrtFixed()
uncoditionally.
* src/base/ftbbox.c: Include FT_INTERNAL_CALC_H.
Fix compiler warnings.
* src/base/ftcalc.c: Fix (potential) compiler warnings.
* src/base/ftcalc.c (FT_SqrtFixed): Corrected/optimized the 32-bit
fixed-point square root computation. It is now used even with
64-bits integers, as it is _much_ faster than calling FT_Sqrt64 :-)
* src/base/ftbbox.c: Removed invalid "#include FT_BEZIER_H" line.
* src/base/ftbbox.c (BBox_Cubic_Check): Rewrote function to use
direct computations with 16.16 values instead of sub-divisions. It
is now slower, but proves a point :-)
* src/raster/ftraster.c, src/smooth/ftgrays.c, src/base/ftbbox.c:
Fixed the bezier stack depths.
* src/base/ftcalc.c (FT_MulFix): Minor rounding fix.
* builds/beos: Added BeOS-specific files to the old build system
(no changes were necessary to support BeOS in the Jamfile though).
* ftconfig.h, ftoption.h: Updated "ftconfig.h" to detect 64-bit int
types on platforms where Autoconf is not available). Also removed
FTCALC_USE_LONG_LONG and replaced it with
FT_CONFIG_OPTION_FORCE_INT64.
* builds/win32/freetype.dsp: Updated the Visual C++ project file.
Doesn't create a DLL yet.
* cffgload.c: Removed a compilation warning.
fixed-point square root. it is now used even with 64-bits
ints, as it's simply _much_ faster than calling FT_Sqrt64 :-)
* src/base/ftbbox.c : removed invalid "#include FT_BEZIER_H" line
direct computations with 16.16 values instead of sub-divisions.
It is now slower, but proves a point :-)
* src/raster/ftraster.c, src/smooth/ftgrays.c, src/base/ftbbox.c :
fixed the bezier stack depths..
* src/base/ftcalc.c (FT_MulFix): minor rounding fix
* builds/beos : added BeOS-specific files to the old build system
(no changes were necessary to support BeOS in the Jamfile though)
16.16 square root of a 16.16 number (this could come handy
in a later version of the auto-hinter)
- small fixes to the smooth renderer. It used to use way too
much line segments when drawing beziers !!
Moving some internal structures and constants from freetype.h to ftobjs.h.
Finally removing FT_LOAD_ANTI_ALIAS.
Cleaning up all error codes. Only the used ones have survived :-)
Removed unused FT_MAX_GLYPH_FORMATS constant.
T2 error codes are now in the range 0x500-0x5FF (instead of `TrueDoc').
Some minor improvements of error return values.
Finally fixing error code values in ftraster and ftgrays to be compliant
with all other FT error codes.
C++ programs easier:
Renamed FT_WordXX to FT_UIntXX.
Changed a lot of void* to FT_Byte* if related to i/o streams -- FreeType
always accesses streams byte-wise, so this makes sense IMHO.
Added a lot of #ifdef __cplusplus to header files (and removed a few).
Other minor syntax fixes (mostly casts).
Replaced the variable `private' with `private_dict' -- `private' is
reserved in C++.
- using FT_UNUSED instead of UNUSED
- using FT_LONG64 and FT_INT64 instead of LONG64 & INT64
- using FT_SIZEOF_INT & FT_SIZEOF_LONG instead of...
- removed the #ifdefs that used SIZEOF_INT, instead we now
use FT_Int32 and FT_UInt32 when needed to support
32-bits quantity correctly on 64-bits systems..
Alexei Podtelezhnikov