While this doesn't show up with FreeType, exactly the same code
leaks with ttfautohint's modified auto-hinter code (which gets used
in a slightly different way).
It certainly doesn't harm since it is similar to already existing
checks in the code for embedded arrays.
* src/autofit/afhints.c (af_glyph_hints_reload): Set `max_contours'
and `max_points' for all cases.
Thanks to Ben Mitchell <ben@rosettatype.com> for guidance with blue
zone characters!
* src/autofit/afblue.dat: Add blue zone data for Thai.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
* src/autofit/afscript.h: Add Thai standard characters.
* src/autofit/afranges.c: Add Thai data.
* src/autofit/afstyles.h: Add Thai data.
Avoids multiple mallocs per typical glyphs.
With this and recent changes to avoid mallocs, the thread-safe
stack-based loader is now as fast as the previous model that had one
cached singleton.
* src/autofit/afhints.h (AF_SEGMENTS_EMBEDDED, AF_EDGES_EMBEDDED):
New macros.
(AF_AxisHintsRec): Add two arrays for segments and edges.
* src/autofit/afhints.c (af_axis_hints_new_segment): Only allocate
data if number of segments exceeds given threshold value.
(af_axis_hints_new_edge): Only allocate data if number of edges
exceeds given threshold value.
(af_glyph_hints_done): Updated.
This avoids at least two malloc calls for typical glyphs.
* src/autofit/afhints.h (AF_POINTS_EMBEDDED, AF_CONTOURS_EMBEDDED):
New macros.
(AF_GlyphHintsRec): Add two arrays for contours and points.
* src/autofit/afhints.c (af_glyph_hints_init, af_glyph_hints_done):
Updated.
(af_glyph_hints_reload): Only allocate data if number of contours or
points exceeds given threshold values.
This avoids one malloc per load.
* src/autofit/afloader.h (AF_LoaderRec): Change type of `hints' to
`AF_GlyphHints'.
Update prototype.
* src/autofit/afloader.c (af_loader_init): Use `AF_GlyphHints'
parameter instead of `FT_Memory'.
(af_loader_done): Directly reset `load_hints'.
(af_loader_load_g): Updated.
* src/autofit/afmodule.c (af_autofitter_load_glyph): Use local
`hints' object.
No need to create a new glyph loader; we can reuse the one from
`slot->internal->loader'. It's hard to tell why it was written that
way originally, but new code looks sound and correct to me, and
avoids lots of allocations.
* src/autofit/afloader.c (af_loader_init): Change return type to
`void'.
Don't call `FT_GlyphLoader_New'.
(af_loader_reset): Don't call `FT_GlyphLoader_Rewind'.
(af_loader_load_g): Update code to use `internal->loader', which
doesn't need copying of data.
* src/autofit/afloader.h (AF_LoaderRec): Remove `gloader' member.
Update prototype.
* src/autofit/afmodule.c (af_autofitter_load_glyph): Updated.
We never have to deal with composite glyphs in the autohinter, as
those will be loaded into FORMAT_OUTLINE by the recursed
`FT_Load_Glyph' function.
In the rare cases that FT_LOAD_NO_RECURSE is set, it will imply
FT_LOAD_NO_SCALE as per `FT_Load_Glyph', which then implies
FT_LOAD_NO_HINTING:
/* resolve load flags dependencies */
if ( load_flags & FT_LOAD_NO_RECURSE )
load_flags |= FT_LOAD_NO_SCALE |
FT_LOAD_IGNORE_TRANSFORM;
if ( load_flags & FT_LOAD_NO_SCALE )
{
load_flags |= FT_LOAD_NO_HINTING |
FT_LOAD_NO_BITMAP;
load_flags &= ~FT_LOAD_RENDER;
}
and as such the auto-hinter is never called. Thus, the recursion in
`af_loader_load_g' never actually happens. So remove the depth
counter as well.
* src/autofit/afloader.c (af_loader_load_g): Remove `depth'
parameter.
<FT_GLYPH_FORMAT_COMPOSITE>: Remove associated code.
(af_loader_load_glyph): Updated.
Stop sharing a global `AF_Loader'. Allocate one on the stack during
glyph load.
Right now this results in about 25% slowdown, to be fixed in a
following commit.
With this patch loading glyphs from different faces from different
threads doesn't immediately crash in the autohinting loader code.
Bugs:
https://bugzilla.redhat.com/show_bug.cgi?id=1164941
* src/autofit/afloader.c (af_loader_init): Pass
`AF_Loader' and `FT_Memory' instead of `AF_Module' as arguments.
(af_loader_reset, af_loader_load_glyph): Also pass `loader' as
argument.
(af_loader_done): Use `AF_Loader' instead of `AF_Module' as
argument.
* src/autofit/afmodule.c (af_autofitter_init): Don't call
`af_loader_init'.
(af_autofitter_done): Don't call `af_loader_done'.
(af_autofitter_load_glyph): Use a local `AF_Loader' object.
* src/autofit/afloader.h: Include `afmodule.h'.
Update prototypes.
Move typedef for `AF_Module' to...
* src/autofit/afmodule.h: ... this place.
No longer include `afloader.h'.
Assure that FreeType's internal include directories are found before
`CPPFLAGS' (which might be set by the user in the environment), and
`CPPFLAGS' before `CFLAGS'.
* builds/freetype.mk (FT_CFLAGS): Don't add `INCLUDE_FLAGS'.
(FT_COMPILE): Make this a special variable for compiling only the
files handled in `freetype.mk'.
(.c.$O): Removed, unused.
* src/*/rules.mk (*_COMPILE): Fix order of include directories.
Using `%ld' for pointer differences causes warnings on 32bit
platforms. The correct type would be (the relatively new) `%td',
however, this is missing on some important platforms.
This patch improves the change from 2014-11-28.
* src/autofit/afhints.c (AF_INDEX_NUM): Use `int' typecast. Our
pointer differences are always sufficiently small.
(af_glyph_hints_dump_points, af_glyph_hints_dump_segments,
af_glyph_hints_dump_edge): Revert to `%d' and use `AF_INDEX_NUM'.
The functions in this patch *do* return non-trivial errors that must
be taken care of.
* src/autofit/afloader.c (af_loader_load_g), src/base/ftobjs.c
(FT_Render_Glyph_Internal), src/base/ftoutln.c (FT_Outline_Render),
src/cff/cffgload.c (cff_decoder_parse_charstrings) <cff_op_endchar>,
src/psaux/psobjs.c (ps_parser_load_field_table), src/psaux/t1decode
(t1_decoder_parse_charstrings) <op_endchar>, src/truetype/ttgload.c
(load_truetype_glyph <subglyph loop>, tt_loader_init,
TT_Load_Glyph), src/truetype/ttgxvar.c (TT_Set_MM_Blend),
src/truetype/ttobjs.c (tt_size_run_fpgm, tt_size_run_prep): Do it.
This essentially moves the Telugu script from the `Indic' hinter to
the `Latin' hinter.
Note that this is a first shot and quite certainly needs
refinements.
* src/autofit/afblue.dat: Add blue zone data for Telugu.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
* src/autofit/afscript.h: Add Telugu standard characters and move
data out of AF_CONFIG_OPTION_INDIC block.
* src/autofit/afranges.c: Move Telugu data out of
AF_CONFIG_OPTION_INDIC block.
* src/autofit/afstyles.h: Update Telugu data; in particular, use
AF_WRITING_SYSTEM_LATIN.
Problem reported by Ralf S. Engelschall <rse@engelschall.com>.
* src/autofit/afhints.c (af_glyph_hints_reload): Fix threshold for
finding first non-near point.
Properly initialize non-near point deltas.
This essentially moves the Devanagari script from the `Indic' hinter
to the `Latin' hinter. Thanks to Girish Dalvi
<girish.dalvi@gmail.com> for guidance with blue zone characters!
* src/autofit/afblue.dat: Add blue zone data for Devanagari.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
* src/autofit/afscript.h: Add Devanagari standard characters and
move data out of AF_CONFIG_OPTION_INDIC block.
* src/autofit/afranges.c: Move Devanagari data out of
AF_CONFIG_OPTION_INDIC block.
Move U+20B9, (new) Rupee sign, from Latin to Devanagari.
* src/autofit/afstyles.h: Update Devanagari data; in particular, use
AF_WRITING_SYSTEM_LATIN.
* src/autofit/afhints.h (AF_Edge_Flags): New value
`AF_EDGE_NEUTRAL'.
* src/autofit/aflatin.c (af_latin_hints_compute_blue_edges): Trace
neutral blue zones with AF_EDGE_NEUTRAL.
(af_latin_hint_edges): Skip neutral blue zones if necessary.
Such blue zones match either the top or the bottom of a contour. We
need them for scripts where accent-like elements directly touch the
base character (for example, some vowel signs in Devanagari, cf.
U+0913 or U+0914).
* src/autofit/afblue.hin (AF_BLUE_PROPERTY_LATIN_NEUTRAL): New
property.
* src/autofit/afblue.h: Regenerated.
* src/autofit/aflatin.h (AF_LATIN_IS_NEUTRAL_BLUE): New macro.
(AF_LATIN_BLUE_NEUTRAL): New enumeration value.
* src/autofit/aflatin.c (af_latin_metrics_init_blues,
af_latin_hints_compute_blue_edges): Handle neutral blue zones.
The adaptation of the cjk auto-hinter module to blue stringsets in
2013-08-25 had three severe bugs. Mea culpa.
1. Contrary to the latin auto-hinter, characters for reference and
overshoot values of a blue zone are specified separately. Due to
the screwed-up change it didn't work at all.
2. A boolean comparison was erroneously replaced with a cast,
causing invalid results with the `^' operator later on. The
visual artifact caused by this problem is the topic of the bug
report.
3. Two flag values were inverted, causing incorrect assignment of
reference and overshoot values.
* src/autofit/afblue.dat: Fix CJK bluestrings, introducing a new
syntax to have both reference and overshoot characters in a single
string. This is error #1.
Add extensive comments.
* src/autofit/afblue.hin (AF_BLUE_PROPERTY_CJK_FILL): Removed, no
longer used.
(AF_BLUE_PROPERTY_CJK_TOP, AF_BLUE_PROPERTY_CJK_HORIZ): Fix values.
This is error #3.
* src/autofit/afblue.c, src/autofit/afblue.h: Regenerated.
* src/autofit/afcjk.c (af_cjk_metrics_init_blues): Correct error #1.
Use character `|' to separate characters for reference and overshoot
values.
Improve tracing messages, synchronizing them with the latin
auto-hinter.
(af_cjk_hints_compute_blue_edges): Fix value of `is_top_right_blue'.
This is error #2.
(af_cjk_align_linked_edge): Add tracing message.
* src/autofit/afcjk.h (AF_CJK_IS_FILLED_BLUE): Removed, no longer
used.
* src/autofit/hbshim.c (af_get_coverage): Merge input glyph coverage
of GSUB lookups into output coverage. Otherwise, ligatures are not
handled properly.
Don't check blue zone characters for default coverage.
In particular, local extrema without horizontal or vertical segments
are better recognized:
+ A + D
\ /
\ /
\ /
\ /
\ + C
\ /
B +/
If the distances AB and CD are large, point B wasn't previously
detected as an extremum since the `ft_corner_is_flat' function
`swallowed' BC regardless of its direction, tagging point B as weak.
The next iteration started at B and made `ft_corner_is_flat' swallow
point C, tagging it as weak also, et voilà.
To improve that, another pass gets now performed before calling
`ft_corner_is_flat' to improve the `topology' of an outline: A
sequence of non-horizontal or non-vertical vectors that point into
the same quadrant are handled as a single, large vector.
Additionally, distances of near points are now accumulated, which
makes the auto-hinter handle them as if they were prepended to the
next non-near vector.
This generally improves the auto-hinter's rendering results.
* src/autofit/afhints.c (af_glyph_hints_reload): Implement it.
* src/autofit/afhints.h (AF_FLAGS): Remove no longer used flag
`AF_FLAG_NEAR'.
Problem reported by Karsten Lücke <karsten.luecke@kltf.de>.
The new algorithm takes care of the width of stems: If the distance
between two segments is larger than the largest stem width, the
demerits quickly increase for larger distances. This improves
hinting of slanted fonts (especially if the inner parts of serifs
have non-horizontal `shoulders'), avoiding false stem links.
* src/autofit/aflatin.c (af_latin_hints_link_segments): Use largest
stem width (if available) to compute better demerits for distances
between stems.
(af_latin_hints_detect_features): Pass stem width array and array
size.
(af_latin_metrics_init_widths): Updated to use original algorithm.
(af_latin_hints_apply): Updated to use new algorithm.
* src/autofit/aflatin.h: Updated.
* src/autofit/afcjk.c: Updated.
This works around a problem with HarfBuzz (<= 0.9.26), which doesn't
validate glyph indices returned by
`hb_ot_layout_lookup_collect_glyphs'.
* src/autofit/hbshim.c (af_get_coverage): Guard `idx'.
* docs/CHANGES: Updated.
Werner Lemberg