This avoids additional calls to `tt_face_lookup_table' for the
`glyf' table, which can be expensive.
* include/freetype/internal/tttypes.h (TT_LoaderRec): Move
`glyf_offset' field to ...
(TT_FaceRec): ... this structure.
* src/truetype/ttgload.c (load_truetype_glyph): Updated.
(tt_loader_init): Move initialization of `glyf_offset' to ...
* src/truetype/ttpload.c (tt_face_load_loca): .. this function.
This speeds up FreeType's handling of malformed fonts.
* src/truetype/ttinterp.c (TT_RunIns): Set up limits for the number
of twilight points, the total number of negative jumps, and the
total number of loops in LOOPCALL opcodes. The values are based on
the number of points and entries in the CVT table.
(Ins_JMPR): Test negative jump counter.
(Ins_LOOPCALL): Test loopcall counter.
* src/truetype/ttinterp.h (TT_ExecContext): Updated.
* docs/CHANGES: Updated.
Without this patch, a loca sequence like `0 100000 0 100000 ...',
where value 100000 is larger than the `glyf' table size, makes
FreeType handle the whole `glyf' table as a single glyph again and
again, which is certainly invalid (and can be very slow, too).
* src/truetype/ttpload.c (tt_face_get_location): Implement.
Improve tracing messages.
We don't need some divisions if a line segments stays within a single
row or a single column of pixels.
* src/smooth/ftgrays.c (gray_render_line) [FT_LONG64]: Make divisions
conditional.
The algorithm calls `gray_set_cell' at the start of each new contour
or when the contours cross the cell boundaries. Double-checking for
that is wasteful.
* src/smooth/ftgrays.c (gray_set_cell): Remove check for a new cell.
(gray_convert_glyph): Remove initialization introduced by 44b172e88.
Problems reported as
https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=40
We now map the strike index right before accessing the physical
data, not earlier.
* src/sfnt/sfobjs.c (sfnt_load_face): Set `face->sbit_strike_map'
after creating the map so that...
* src/sfnt/ttsbit.c (tt_face_load_strike_metrics): ... this function
can be used before and after setting up `sbit_strike_map'.
(tt_face_set_sbit_strike): Revert change.
(tt_sbit_decoder_init, tt_face_load_sbix_image): Map strike index.
* src/truetype/ttdriver.c (tt_size_select): Revert change.
FreeType tries to sanitize strike header data; we now reject
completely broken ones.
* include/freetype/internal/tttypes.h (TT_FaceRec): New
`sbit_strike_map' array pointer.
* src/base/ftobjs.c (FT_Match_Size): Reject matches where either
width or height would be zero.
Add tracing message in case of error.
* src/sfnt/sfobjs.c (sfnt_load_face): Populate `sbit_strike_map',
only using (more or less) valid strike header data for
FT_Face's `available_sizes' array.
(sfnt_done_face): Updated.
* src/sfnt/ttsbit.c (tt_face_set_sbit_strike): Use
`sbit_strike_map'.
(tt_face_load_strike_metrics): Improve tracing.
* src/truetype/ttdriver.c (tt_size_select): Use `sbit_strike_map'.
* src/smooth/ftgrays.c (gray_move_to): Call `gray_set_cell' directly
instead of...
(gray_start_cell): ... this function, which is removed.
(gray_convert_glyph): Make initial y-coordinate invalid.
* src/tools/ftrandom/ftrandom.c (_XOPEN_SOURCE): New macro, set to
500.
* src/tools/ftrandom/Makefile (CFLAGS): Split off include
directories to ...
(INCLUDES): ... this new variable.
(LDFLAGS): New variable.
(ftrandom.o, ftrandom): Updated.
Thanks to Hrant H Papazian <hpapazian@gmail.com> for help.
* src/autofit/afblue.dat (AF_BLUE_STRING_ARMENIAN_*): Improve
selection of characters.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
It now supports both a normal build (`./configure && make') and a
development build (`make devel').
* src/tools/ftrandom/Makefile (VPATH): Set it so that
`libfreetype.a' gets searched in both `objs' (for the development
build) and `objs/.libs' (for a normal build which uses libtool).
(LIBS): Add missing libraries.
(ftrandom.o): New rule.
(ftrandom): Use automatic variables.
Problems reported by Bob Taylor <Bob.Taylor@monotype.com>.
* src/truetype/ttgxvar.c (TT_Vary_Apply_Glyph_Deltas): Fix rough
sanity test for glyph variation array header size.
Always set stream position before reading packed x and y deltas.
Fix thinko w.r.t. `localpoints' array.
* src/tools/ftrandom/ftrandom.c (GOOD_FONTS_DIR): Provide better
default.
(error_fraction): Make it of type `double' to work as advertized –
this was completely broken.
Update all related code.
(error_count, fcnt): Make it unsigned to fix compiler warnings.
Update all related code.
(fontlist): Change `len' member to `long' to fix compiler warnings.
(FT_MoveTo, FT_LineTo, FT_ConicTo, FT_CubicTo, abort_test): Tag
unused variables.
(TestFace, FindFonts, copyfont, do_test): Fix compiler warnings.
(ExecuteTest): Ditto.
Call `FT_Done_FreeType'.
(getErrorCnt): Replace `ceil' with an ordinary cast to `unsigned
int'.
(usage): Improve output.
(main): Fix compiler warnings.
* src/tools/ftrandom/README: Updated.
* src/base/ftobjs.c (FT_Open_Face): Check whether negation was
actually successful. For example, this can fail for value
-32768 if the type is `signed short'. If there are problems,
disable the stroke.
It turns out that there is significant cost associated with `FT_Span'
creation and calls to `gray_render_span' because it happerns so
frequently. This removes these steps from our internal use but leaves
it alone for `FT_RASTER_FLAG_DIRECT" to preserve API. The speed gain
is about 5%.
* src/smooth/ftgrays.c (gray_render_span): Removed. The code is
migrated to...
(gray_hline): ... here.
Zero coverage is unlikely (1 out of 256) to warrant checking. This
gives 0.5% speed improvement in dendering simple glyphs.
* src/smooth/ftgrays.c (gray_hline, gray_render_span): Remove checks.
This gives 2% speed improvement in rendering simple glyphs.
* src/smooth/ftgrays.c (TPixmap): Reduced pixmap descriptor with a
pointer to its bottom-left and pitch to be used in...
(gray_TWorker): ... here.
(gray_render_span): Move pixmap flow check from here...
(gray_raster_render): .. to here.
This commit avoids `EBDT' and friends being looked up again and
again while loading a single embedded bitmap.
* include/freetype/internal/tttypes.h (TT_FaceRec)
[TT_CONFIG_OPTION_EMBEDDED_BITMAPS]: New fields `ebdt_start' and
`ebdt_size'.
* src/sfnt/ttsbit.c (tt_sbit_decoder_init): Move table lookup to ...
(tt_face_load_sbit): ... this function; also store the table size
and offset.
This removes unnecessary complexity of span merging and buffering.
Instead, the spans are rendered as they come, speeding up the
rendering by about 5% percents as a result.
* src/smooth/ftgrays.c [FT_MAX_GRAY_SPANS]: Macro removed.
(gray_TWorker): Remove span buffer and related fields.
(gray_sweep, gray_hline): Updated.
* include/freetype/ftimage.h: Remove documentation note about
`FT_MAX_GRAY_SPANS', which was never in `ftoption.h' and is now gone.
`FT_Get_Advance' returns 0 for bitmap fonts. It first gets the
advance value from the font table and then scales it by the
`font->size->metrics->x_scale' field. But `FT_Select_Size' doesn't
set that value for bitmap fonts and the advance gets scaled to zero.
Taken from
https://github.com/behdad/harfbuzz/issues/252
* src/sfnt/ttsbit.c (tt_face_load_strike_metrics)
<TT_SBIT_TABLE_TYPE_EBLC>: Set scale values.
Since 2016-05-16 we detect infinite recursion directly.
* src/sfnt/ttload.c (tt_face_load_maxp): Don't adjust
`maxComponentDepth'.
* src/truetype/ttgload.c (load_truetype_glyph): Don't abort if
`maxComponentDepth' is not valid. Instead, simply adjust its value
and emit a tracing message.
To match the logic in FDEF. The value of the end is only used for
bound-checking in `Ins_JMPR', so it may not have been obvious that
it was not recorded. Tested (as part of Font Validator 2.0) all the
fonts on Fedora and did not see any change.
* src/truetype/ttinterp.c (Ins_IDEF): Updated.
GETDATA[], opcode 0x92
bytecode instruction. It always returns 17, and we have absolutely
no idea what it is good for...
* src/truetype/ttinterp.c (Pop_Push_Count, opcode_name): Updated.
(Ins_GETDATA): New function.
(TT_RunIns): Add it.
This commit implements undocumented (but confirmed) stuff from
Apple's old bytecode engine.
GETVARIATION[], opcode 0x91
This opcode pushes normalized variation coordinates for all axes
onto the stack (in 2.14 format). Coordinate of first axis gets
pushed first.
GETINFO[], selector bit 3
If GX variation support is enabled, bit 10 of the result is set
to 1.
* src/truetype/ttinterp.c: Include FT_MULTIPLE_MASTERS_H.
(opcode_name) [TT_CONFIG_OPTION_GX_VAR_SUPPORT]: Updated.
(Ins_GETINFO) [TT_CONFIG_OPTION_GX_VAR_SUPPORT]: Handle selector
bit 3, checking support for variation glyph hinting.
(Ins_GETVARIATION) [TT_CONFIG_OPTION_GX_VAR_SUPPORT]: New function
to implement opcode 0x91.
(TT_RunIns) [TT_CONFIG_OPTION_GX_VAR_SUPPORT]: Handle opcode 0x91.
This finally fixes the rendering of the cyclist and the lizard in
the `Zycon' font.
* src/truetype/ttgxvar.c (ft_var_readpackedpoints): `first' point
index is always cumulative.
(tt_handle_deltas): Rename to...
(tt_interpolate_deltas): ... This.
Add new parameter for output point array.
Update caller.
(TT_Vary_Apply_Glyph_Deltas): Add `points_out' array; it now holds
the intermediate results of `tt_interpolate_deltas' that are to be
added to `outline->points'.
* src/autofit/afhints.c (af_glyph_hints_dump_segments): Trace
`delta' also.
Don't show first point of segment as a replacement for `pos'; this
is (a) misleading, since the difference to `pos' can be almost
arbitrarily large in corner cases, and (b) it is better to have all
segment data in font units instead of a single value given in output
space coordinates.
Improve layout.
(af_glyph_hints_dump_edges): Show px->units and units->px conversion
values for convenience.
Improve layout.
* src/autofit/afhints.h (AF_SegmentRec): New member `delta'.
* src/autofit/aflatin.c (af_latin_hints_compute_segments): Compute
`delta'.
(af_latin_hints_compute_edges): Reject segments with a delta larger
than 0.5px.
The coordinates of the base font should be used for interpolation
purposes, NOT the current points (i.e., the result of accumulation
of previous deltas).
* src/truetype/ttgxvar.c (TT_Vary_Apply_Glyph_Deltas): Initialize
`points_org' before looping over all tuples.
This commit covers the most important one.
* src/autofit/afmodule.c (af_property_set): Handle `warping',
`darkening-parameters', and `no-stem-darkening'.
* src/cff/cffdrivr.c (cff_property_set): Handle
`darkening-parameters', `hinting-engine', and `no-stem-darkening'.
* src/truetype/ttdriver.c (tt_property_set): Handle
`interpreter-version'.
Recognizing properties follows in another commit.
* devel/ftoption.h, include/freetype/config/ftoption.h
(FT_CONFIG_OPTION_ENVIRONMENT_PROPERTIES): New macro.
* include/freetype/config/ftstdlib.h (ft_getenv): New macro.
* src/base/ftinit.c (ft_set_default_properties): New function to
parse `FREETYPE_PROPERTIES' and calling `ft_property_string_set'.
(FT_Init_FreeType): Updated.
This is a preparation for handling an `FREETYPE_PROPERTIES'
environment variable to control (some) driver properties.
No change in functionality.
* src/base/ftobjs.c (ft_property_do): Add `value_is_string'
parameter.
(ft_property_string_set): New function.
(FT_Property_Set, FT_Property_Get): Updated.
* include/freetype/internal/ftobjs.h: Updated.
* include/freetype/internal/services/svprop.h
(FT_Properties_SetFunc): Add `value_is_string' parameter.
* src/autofit/afmodule.c (af_property_set), src/cff/cffdrivr.c
(cff_property_set), src/truetype/ttdriver.c (tt_property_set):
Updated, emitting an error currently if `value_is_string' is set.
Rasterization sub-banding is utilized at large sizes while using
rather small fixed memory pool. Indeed it is possible to make an
educated guess how much memory is necessary at a given size for a
given glyph. It turns out that, for large majority of European glyphs,
you should store about 8 times more boundary pixels than their height.
Or, vice versa, if your memory pool can hold 800 pixels the band
height should be 100 and you should sub-band anything larger than
that. Should you still run out of memory, FreeType bisects the band
but you have wasted some time. This is what has been implemented in
FreeType since the beginning.
It was overlooked, however, that the top band could grow to twice the
default band size leading to unnecessary memory overflows there. This
commit fixes that. Now the bands are distributed more evenly and
cannot exceed the default size.
Now the magic number 8 is really suitable for rather simple European
scripts. For complex Chinese logograms the magic number should be 13
but that is subject for another day.
* src/smooth/ftgrays.c (gray_convert_glyph): Revise sub-banding
protocol.
* src/smooth/ftgrays.c (TArea): Restore original definition as `int'.
(gray_render_line) [FT_LONG64]: Updated.
(gray_convert_glyph): 32-bit band bisection stack should be 32 bands.
(gray_convert_glyph_inner): Trace successes and failures.
Doing so allows us to link them to edges – some fonts like
`NotoSansGurmukhi-Regular' have such isolated points sitting exactly
on other outlines.
* src/autofit/aflatin.c (af_latin_hints_compute_segments): Don't
ignore one-point contours but handle them specially as one-point
segments.
(af_latin_hints_compute_edges): Append one-point segments to edges
if possible.
This patch restores original `TCoord' definition as `int' so that the
rendering pool is used more efficiently on LP64 platforms (unix).
* src/smooth/ftgrays.c (gray_TWorker, TCell, gray_TBand): Switch some
fields to `TCoord'.
(gray_find_cell, gray_render_scanline, gray_render_line, gray_hline,
gray_sweep, gray_convert_glyph): Updated.
* src/smooth/ftgrays.c (gray_convert_glyph): Do not use volatile
qualifier.
(gray_raster_render): Move span initializations from here.
(gray_sweep): ... to here and remove unused `target' argument.
* src/smooth/ftgrays.c (gray_init_cells): Remove function.
(gray_TWorker): Remove fields that become local variables.
(gray_raster_render): Move rendering buffer declaration from here.
(gray_convert_glyph): ... to here and update accordingly.
Removing the checks from `gray_hline' shaves 1% off rendering speed.
* src/smooth/ftgrays.c [STANDALONE_]: Duplicate `FT_MIN' and `FT_MAX'.
(gray_TWorker): No need to store `clip_box'.
(gray_hline): Remove unnecessary boundary checks.
(gray_convert_glyph): Move boundary checks from here.
(gray_raster_render): ... to here and consolidate.
* src/smooth/ftgrays.c [STANDALONE_]: Duplicate `FT_Outline_Get_CBox'.
(gray_compute_cbox): Remove this function.
(gray_convert_glyph): Update to use `FT_Outline_Get_CBox'.
This compiler doesn't recognize the end-of-comment sequence `*/' if
it immediately follows non-ASCII characters.
* src/autofit/afscript.h: Ensure whitespace before `*/'.
* src/truetype/ttinterp.c (Ins_SHPIX): Allow SHPIX to move points in
the twilight zone. Otherwise, treat SHPIX the same as DELTAP.
Unbreaks various fonts such as older versions of Rokkitt and DTL
Argo T Light that would glitch severly after calling ALIGNRP after a
blocked SHPIX.
Suggested by Alexei. `UL' is only needed for 16bit compilers, but
it seems noone is using this anymore (and we no longer test whether
FreeType compiles in such an environment). Otherwise, it is easy to
add the postfix to the `AF_UNICODE_RANGE' macro.
The convergence of Bézier flatteners is fast with the deviation
from straight line being assymptotically cut 4-fold on each bisection.
This justifies smaller bisection stack size.
* src/smooth/ftgrays.c (gray_TWorker): Remove common `bez_stack'.
(gray_render_conic): Create and use conic `bez_stack'. Move back the
band analysis from...
(gray_conic_to): ... here.
(gray_render_cubic): Create and use cubic `bez_stack'. Move back the
band analysis from...
(gray_cubic_to): ... here.
(gray_move_to): Updated.
* src/autofit/afblue.dat: Add blue zone data for Armenian.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
* src/autofit/afscript.h: Add Armenian standard characters.
* src/autofit/afranges.c: Add Armenian data.
* src/autofit/afstyles.h: Add Armenian data.
This essentially moves the Gurmukhi script from the `Indic' hinter to
the `Latin' hinter.
* src/autofit/afblue.dat: Add blue zone data for Gurmukhi.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
* src/autofit/afscript.h: Add Gurmukhi standard characters and move
data out of AF_CONFIG_OPTION_INDIC block.
* src/autofit/afranges.c: Move Gurmukhi data out of
AF_CONFIG_OPTION_INDIC block.
* src/autofit/afstyles.h: Update Gurmukhi data; in particular, use
AF_WRITING_SYSTEM_LATIN.
This patch prepares data structures and the like.
See added comments in `ttinterp.h' for more information on this and
the following commits in the series.
* devel/ftoption.h, include/freetype/config/ftoption.h
(TT_CONFIG_OPTION_SUBPIXEL_HINTING): Assign values to differentiate
between subpixel versions.
(TT_SUPPORT_SUBPIXEL_HINTING_INFINALITY,
TT_SUPPORT_SUBPIXEL_HINTING_MINIMAL): New macros.
* include/freetype/ftttdrv.h (TT_INTERPRETER_VERSION_40): New macro.
* include/freetype/internal/tttypes.h (TT_FaceRec): Updated.
* src/truetype/ttinterp.h (TT_ExecContextRec): Define new fields
`subpixel_hinting_lean', `vertical_lcd_lean',
`backwards_compatibility', `iupx_called', iupy_called', and
`grayscale_cleartype' for new hinting mode.
* src/truetype/ttdriver.c (tt_property_set): Handle v38 and v40
interpreters conditionally.
* src/truetype/ttgload.c (TT_Hint_Glyph): Save phantom points unless
in v38 backwards compatibility mode.
Updated.
(compute_glyph_metrics): Add v38 backwards compatibility mode
constraint for adjusting advance widths.
Updated.
(tt_loader_init): Handle new flags `subpixel_hinting_lean',
`grayscale_cleartype', and `vertical_lcd_lean'.
Updated.
(tt_get_metrics, TT_Process_Simple_Glyph, TT_LOADER_SET_PP):
Updated.
* src/truetype/ttobjs.c (tt_driver_init): Conditionally set
default interpreter version number.
* src/truetype/ttsubpix.c, src/truetype/ttsubpix.h: Updated.
* include/freetype/config/ftstdlib.h (FT_LONG_MIN): New macro.
* src/cff/cffparse.c (cff_parse_font_matrix): Use largest scaling
value of all matrix coefficients to scale matrix.
* src/cff/cffobjs.c (cff_face_init): Use `matrix->yx' member for
matrix normalization if `matrix->yy' is zero.
The previous fix for #46372 misunderstood a composite glyph referring
same component twice as a recursive reference. See the discussion
http://lists.gnu.org/archive/html/freetype/2016-05/msg00000.html
Thanks to Khaled Hosny for finding this issue.
* src/truetype/ttgload.c (ft_list_get_node_at): A function to get
the i-th node from FT_List. (load_truetype_glyph): In the traversal
scan of the reference tree in the composite glyph, we clear the
nodes filled by previous sibling chain.
We never dereference `face_id', and some implementations might use a
running number instead of a pointer. Additionally, disallowing
value zero was undocumented.
* src/cache/ftccmap.c (FTC_CMapCache_Lookup), src/cache/ftcmanag.c
(FTC_Manager_LookupFace, FTC_Manager_RemoveFaceID): Remove test for
`face_id'.
A single decrement counter of segments to draw, instead of an array,
contains all the information necessary to decide when to split and
when to draw a conic segment. The number of splits before each draw is
equal to the number of trailing zeros in the counter.
* src/smooth/ftgrays.c (gray_TWorker): Remove `lev_stack'.
(gray_render_conic): Updated to use decrement counter of segments.
Otherwise some properties could be set to arbitrary values, which is
harmless, but querying could give wrong positive results.
* src/cff/cffdrivr.c (cff_property_set) [hinting-engine],
* src/truetype/ttdriver.c (tt_property_set) [interpreter-version]:
Only allow defined values.
This essentially moves the Gujarati script from the `Indic' hinter to
the `Latin' hinter.
* src/autofit/afblue.dat: Add blue zone data for Gujarati.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
* src/autofit/afscript.h: Add Gujarati standard characters and move
data out of AF_CONFIG_OPTION_INDIC block.
* src/autofit/afranges.c: Move Gujarati data out of
AF_CONFIG_OPTION_INDIC block.
* src/autofit/afstyles.h: Update Gujarati data; in particular, use
AF_WRITING_SYSTEM_LATIN.
Werner Lemberg