The dot product between freeVector and projVector or cosine of
the angle between these FT_F2Dot14 unit vectors used to be scaled up
by 4 and routinely occupied 32 bits in an FT_Long field F_dot_P.
This patch scales the value down by 2^14 instead, which simplifies
its use throughout the bytecode interpreter.
This does not lead to the loss of precision because the lower bits
are unreliable anyway. Consider two unit vectors (1,0) and (.6,.8)
for which the true value of F_dot_P is .6 * 0x40000000 = 0x26666666.
These vectors are stored as (0x4000,0) and (0x2666,0x3333) after
rounding and F_dot_P is assigned 0x26660000. The lower bits were
already lost while rounding the unit vector components.
Besides code simplification, this change can lead to better
performance when FT_MulDiv with the scaled-down F_dot_P is less
likely to use the costly 64-bit path. We are not changing the type
of F_dot_P to FT_F2Dot14 at this point.
* src/truetype/ttinterp.c (Compute_Funcs): Scale F_dot_P down by 14
bits and modify its use accordingly.
(Direct_Move, Direct_Move_Orig, Compute_Point_Displacement): Modify
the use of F_dot_P field.
* src/truetype/ttobjs.c (tt_size_run_fpgm): Change arbitrary
assignment of F_dot_P to its theoretical maximum in case we decide
to scale back its type later.
* src/truetype/ttinterp.c (TT_DivFix14): New macro.
(Normalize): Use it here.
(Current_Ratio): Use TT_MulFix14 instead of FT_MulDiv.
(Ins_SHPIX): Cancel out two TT_MulFix14 calls.
* src/truetype/ttinterp.c (Ins_ISECT): Use angle between vectors to
avoid grazing intersections. The previous threshold was too coarse,
incorrectly rejecting short but valid vectors.
* include/freetype/internal/ftcalc.h (FT_MulDiv_No_Round): Don't
enclose with `TT_USE_BYTECODE_INTERPRETER'; we now need the function
elsewhere also.
* src/autofit/afcjk.h: Use AF_CONFIG_OPTION_CJK.
* src/truetype/ttgload.c (tt_loader_init): Fix compiler warning.
* src/truetype/ttinterp.c (Ins_MSIRP): Fix compiler warning.
* src/truetype/ttinterp.h: Use
TT_CONFIG_OPTION_BYTECODE_INTERPRETER.
This is the large, famous `Infinality' patch to support ClearType
bytecode which has been available from
http://www.infinality.net/blog/ for some time, and which has been
refined over the last years. While still experimental, it is now
mature enough to be included directly into FreeType.
Most of the code is based on the ClearType whitepaper written by
Greg Hitchcock
http://www.microsoft.com/typography/cleartype/truetypecleartype.aspx
which gives a detailed overview of the necessary changes to the
Microsoft rasterizer so that older fonts are supported. However, a
lot of details are still missing, and this patches provides a
framework to easily handle rendering issues down to the glyph level
of certain fonts.
Note that ClearType support is not completely implemented! In
particular, full support for the options `compatible_widths',
`symmetrical_smoothing, and `bgr' (via the GETINFO bytecode
instruction) is missing.
* src/truetype/ttsubpix.c: New file, providing code to handle
`tweaks', this is, rules for certain glyphs in certain fonts
(including wildcards) which need a special treatment.
* src/truetype/ttsubpix.h: New file, holding the tweaking rules.
* include/freetype/config/ftoption.h, src/devel/ftoption.h
(TT_CONFIG_OPTION_SUBPIXEL_HINTING): New macro.
* include/freetype/internal/ftobjs.h (FT_PIX_FLOOR_GRID,
FT_PIX_ROUND_GRID, FT_PIX_CEIL_GRID): New macros.
* src/truetype/truetype.c [TT_USE_BYTECODE_INTERPRETER]: Include
`ttsubpix.c'.
* src/truetype/ttgload.c: Include `ttsubpix.h'.
[All changes below are guarded by TT_CONFIG_OPTION_SUBPIXEL_HINTING.]
(tt_get_metrics): Set tweak flags.
(TT_Hint_Glyph): Call `FT_Outline_EmboldenXY' if necessary.
(TT_Process_Simple_Glyph): Compensate emboldening if necessary.
(compute_glyph_metrics): Handle `compatible widths' option.
(tt_loader_init): Handle ClearType GETINFO information bits.
* src/truetype/rules.mk (TT_DRC_SRC): Updated.
* src/truetype/ttinterp.c: Include `ttsubpix.h'.
[Where necessary, changes below are guarded by
TT_CONFIG_OPTION_SUBPIXEL_HINTING.]
(Direct_Move, Direct_Move_X): Extended.
(Round_None, Round_To_Grid, Round_To_Half_Grid, Round_Down_To_Grid,
Round_Up_To_Grid, Round_To_Double_Grid, Round_Super, Round_Super_45,
SetSuperRound): Add parameter to handle the number of grid lines per
pixel.
(SET_SuperRound, ROUND_None, CUR_Func_round): Updated.
(DO_SROUND, DOS45ROUND, DO_ODD, DO_EVEN): Updated.
(DO_ROUND, DO_NROUND): Updated.
(DO_RS): Take care of `Typeman' bytecode patterns.
(Ins_FDEF): Add some debugging code. Commented out.
(Ins_ENDF): Restore state.
(Ins_CALL, Ins_LOOPCALL): Handle inline delta functions.
(Ins_MD): Handle `Vacuform' rounds.
(Move_Zp2_Point, Ins_SHPIX, Ins_MSIRP, Ins_MDAP, Ins_MIAP,
Ins_MDRP, Ins_MIRP): Handle tweaks.
(Ins_ALIGNRP): Add tweak guard.
(Ins_IUP, Ins_DELTAP): Handle tweaks.
(Ins_GETINFO): Handle new ClearType bits.
(TT_RunIns): Handle tweaks.
* src/truetype/ttinterp.h: Updated.
(SPH_TweakRule, SPH_ScaleRule): New structures for tweaks.
(TT_ExecContextRec): Add members for subpixel hinting support.
* src/truetype/ttobjs.h (TT_DefRecord): Add `inline_delta' member.
* src/truetype/ttinterp.c (DO_SSW): SSW *does* use font units. For
verification, it took some time to find a font which actually uses
this instruction.
* src/truetype/ttinterp.c (Ins_MIRP): Typo, present since ages. The
code is now in sync with the other operators (e.g. MSIRP) which
modify twilight points.
Jump instructions are now bound to the current function. The MS
Windows rasterizer behaves the same, as confirmed by Greg Hitchcock.
* src/truetype/ttinterp.h (TT_CallRec): Add `Cur_End' element.
* src/truetype/ttobjs.h (TT_DefRecord): Add `end' element.
* src/truetype/ttinterp.c (DO_JROT, DO_JMPR, DO_JROF): Check upper
bound of jump address.
(Ins_FDEF, Ins_CALL, Ins_LOOPCALL, Ins_UNKNOWN, TT_RunIns): Updated.
* src/truetype/interp.c: Set even more TT_Err_Invalid_Reference
error codes only if pedantic hinting is active. At the same time,
try to provide sane values which hopefully allow useful
continuation. Exception to this is CALL and LOOPCALL – due to
possible stack corruption it is necessary to bail out.
* src/truetype/ttinterp.c (TT_RunIns, Ins_FLIPPT, Ins_DELTAP,
Ins_DELTAC): Exit with error only if `pedantic_hinting' is set.
Otherwise, try to do something sane.
Thanks to Greg Hitchcock who explained the issue.
* src/truetype/ttinterp.c (Ins_MIRP): Replace a `>=' operator with
`>' since the description in the specification is incorrect.
This fixes, for example, glyph `two' in font `Helvetica Neue LT Com
65 medium' at 15ppem.
On LLP64 platforms (e.g. Win64), long (32-bit) cannot cover
the memory address (64-bit). Also the casts from the pointer
type to long int should be removed to preserve the address
correctly.
* src/raster/ftraster.c (New_Profile): Replace "%lx" by "%p".
(End_Profile) Ditto.
* src/truetype/ttinterp.c (Init_Context): Ditto.
* src/truetype/ttinterp.c (free_buffer_in_size): Don't duplicate
FT_GlyphZoneRec size->twilight to be freed. If duplicated,
FT_FREE() erases the duplicated pointers only and leave original
pointers. They can cause the double-free crash when the burst
errors occur in TrueType interpreter and free_buffer_in_size()
is invoked repeatedly. See Savannah bug #31040 for detail.
* src/truetype/ttinterp.c (TT_RunIns): Decrease the trace level
showing the error when the interpreter returns with an error,
from FT_TRACE7() to FT_TRACE1().
* src/truetype/ttinterp.c (free_buffer_in_size): New function to
free the buffer allocated during the interpretation of this glyph.
(TT_RunIns): Unset FT_Face->size->{cvt_ready,bytecode_ready} if
an error occurs in the bytecode interpretation. The interpretation
of invalid bytecode may break the function definitions and referring
them in later interpretation is danger. By unsetting these flags,
`fpgm' and `prep' tables are executed again in next interpretation.
Fix Savannah bug #30798, reported by Robert Swiecki.
* src/truetype/ttinterp.c (BOUNDSL): New macro.
Change `BOUNDS' to `BOUNDSL' where appropriate.
* src/truetype/ttinterp.h (TT_ExecContextRec): Fix type of
`cvtSize'.
