* src/pfr/pfrobjs.c (pfr_face_get_kerning): Use `FT_MSB' instead of
`ft_highpow2'.
* src/base/ftutil.c, include/internal/ftobjs.h (ft_highpow2): Remove
it.
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.
In case of an error in the `prep' table, no longer try to execute it
again and again. This makes FreeType handle endless loops in buggy
fonts much faster.
* src/truetype/ttobjs.h (TT_SizeRec): The fields `bytecode_ready'
and `cvt_ready' are now negative if not initialized yet, otherwise
they indicate the error code of the last run.
* src/truetype/ttobjs.c (tt_size_run_fpgm, tt_size_run_prep,
tt_size_done_bytecode, tt_size_init_bytecode,
tt_size_ready_bytecode, tt_size_init, tt_size_done, tt_size_reset):
Updated.
* src/truetype/ttgload.c (tt_loader_init): Updated.
* src/truetype/ttinterp.c (TT_RunIns): Force reexecution of `fpgm'
and `prep' only if we are in the `glyf' table.
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.
This spot has been missed while introducing support for various
drop-out modes years ago (including no drop-out mode, which this
commit fixes).
Problem reported by Patrick Thomas <pthomas505@gmail.com>.
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.
Work around a bug in pkg-config version 0.28 and earlier: If a
variable value gets surrounded by doublequotes (in particular values
for the `prefix' variable), the prefix override mechanism fails.
* builds/unix/freetype2.in: Don't use doublequotes.
* builds/unix/unix-def.in (freetype.pc): Escape spaces in directory
names with backslashes.
Based on a patch from Marek Kašík <mkasik@redhat.com>.
* builds/unix/configure.raw (LIBS_CONFIG): Remove.
* builds/unix/freetype-config.in (libs): Hard-code value.
* builds/unix/unix-def.in: 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.
=========================
Tag sources with `VER-2-5-3'.
* docs/VERSION.DLL: Update documentation and bump version number to
2.5.3.
* README, Jamfile (RefDoc), builds/windows/vc2005/freetype.vcproj,
builds/windows/vc2005/index.html,
builds/windows/vc2008/freetype.vcproj,
builds/windows/vc2008/index.html,
builds/windows/vc2010/freetype.vcxproj,
builds/windows/vc2010/index.html,
builds/windows/visualc/freetype.dsp,
builds/windows/visualc/freetype.vcproj,
builds/windows/visualc/index.html,
builds/windows/visualce/freetype.dsp,
builds/windows/visualce/freetype.vcproj,
builds/windows/visualce/index.html,
builds/wince/vc2005-ce/freetype.vcproj,
builds/wince/vc2005-ce/index.html,
builds/wince/vc2008-ce/freetype.vcproj,
builds/wince/vc2008-ce/index.html: s/2.5.2/2.5.3/, s/252/253/.
* include/freetype/freetype.h (FREETYPE_PATCH): Set to 3.
* builds/unix/configure.raw (version_info): Set to 17:2:11.
* CMakeLists.txt (VERSION_PATCH): Set to 3.
* docs/CHANGES: Updated.
* builds/freetype.mk (INCLUDE_FLAGS) [DEVEL_DIR]: Don't use
pkg-config for bzip2 since not all GNU/Linux distributions have
`bzip2.pc' (and the header file `bzlib.h' is located in /usr/include
normally).
o Introduce `auto' value for `--with-XXX' library options; this is
now the default.
o First use `pkg-config' for library detection, then fall back to
other tests.
* builds/unix/configure.raw (--with-zlib, --with-bzip2, --with-png,
--with-harfbuzz): Rewrite.
Use new `xxx_reqpriv', `xxx_libpriv', and `xxx_libstaticconf'
variables to collect data for `freetype2.pc' and `freetype-config'.
(FT2_EXTRA_LIBS): Renamed to ...
(ft2_extra_libs): This since it gets no longer substituted.
(REQUIRES_PRIVATE, LIBS_PRIVATE, LIBS_CONFIG, LIBSSTATIC_CONFIG):
New output variables, replacing `XXX_PKG' and `LIBXXX'.
Add notice at the end of `configure' showing the library
configuration.
* builds/unix/freetype-config.in (--static): New command line
option.
(libs): Updated.
(staticlibs): New variable, to be used if `--static' is given.
* builds/doc/freetype-config.1: Document `--static'.
* builds/unix/freetype2.in, builds/unix/unix-def.in: Updated.
Problem reported by Hin-Tak Leung <htl10@users.sourceforge.net>.
* builds/unix/configure.raw: Don't use gcc's `-pedantic' flag for
versions < 4.6. This is especially needed for Max OS X since this
OS runs a gcc variant (or emulation) based on version 4.2.1.
* src/cff/cf2ft.c (cf2_initLocalRegionBuffer,
cf2_initGlobalRegionBuffer): It is possible for a charstring to call
a subroutine if no subroutines exist. This is an error but should
not trigger an assert. Split the assert to account for this.
* src/cff/cf2hints.c (cf2_hintmap_build): Return when `hintMask' is
invalid. In this case, it is not safe to use the length of
`hStemHintArray'; the exception has already been recorded in
`hintMask'.
bdflib puts data from the input stream into a buffer in chunks of
1024 bytes. The data itself gets then parsed line by line, simply
increasing the current pointer into the buffer; if the search for
the final newline character exceeds the buffer size, more data gets
read.
However, in case the current line's end is very near to the buffer
end, and the keyword to compare with is longer than the current
line's length, an out-of-bounds read might happen since `memcmp'
doesn't stop properly at the string end.
* src/bdf/bdflib.c: s/ft_memcmp/ft_strncmp/ to make comparisons
stop at string ends.
The build target is now `windows' instead of `win32'.
Problem reported by Nickolas George <darknova.clan@gmail.com>.
* builds/modules.mk: Don't use `win32' and `win16' (!) but
`windows'.
* builds/windows/detect.mk, builds/windows/win32-def.mk:
s/win32/windows/.
* src/autofit/afcjk.c (af_cjk_metrics_init_widths),
src/autofit/aflatin.c (af_latin_metrics_init_widths): Fix handling
of alternative standard characters.
This also fixes a compilation warning in non-debug mode.
* src/cff/cf2ft.c (cf2_checkTransform): Convert assertion into
parameter check.
(cf2_decoder_parse_charstrings): Call `cf2_checkTransform' only if
we are scaling the outline.
(cf2_getPpemY): Remove problematic assertion.
This is useful for OpenType features like `c2sc' (caps to small
caps) that don't have lowercase letters by definition, or other
features that mainly operate on numerals.
* src/autofit/afscript.h: Add more standard characters.
* src/autofit/aftypes.h: Update use of `SCRIPT' macro.
(AF_ScriptClassRec): Add members to hold two more standard
characters.
(AF_DEFINE_SCRIPT_CLASS): Updated.
* src/autofit/afglobal.c, src/autofit/afglobal.h,
* src/autofit/afpic.c, src/autofit/afranges.h, src/autofit/hbshim.c:
Update use of `SCRIPT' macro.
* src/autofit/afcjk.c (af_cjk_metrics_init_widths),
src/autofit/aflatin.c (af_latin_metrics_init_widths): Scan two more
standard characters.
* src/sfnt/ttsbit.c (tt_sbit_decoder_load_bitmap) <glyph_format==5>:
Don't check metrics, which this format doesn't have.
This is another correction to the commit from 2013-11-21.
* src/type1/t1load.c (t1_parse_font_matrix): Properly handle result
of `T1_ToFixedArray'.
* src/cid/cidload.c (cid_parse_font_matrix): Synchronize with
`t1_parse_font_matrix'.
* src/type42/t42parse.c (t42_parse_font_matrix): Synchronize with
`t1_parse_font_matrix'.
(t42_parse_encoding): Synchronize with `t1_parse_encoding'.
* src/psaux/psobjs.c (ps_parser_load_field) <T1_FIELD_TYPE_BBOX>,
<T1_FIELD_TYPE_MMOX>: Properly handle result of `ps_tofixedarray'.
Split the single, incorrect Indic entry into separate scripts so
that the covered ranges are the same: Bengali, Devanagari, Gujarati,
Gurmukhi, Kannada, Limbu, Malayalam, Oriya, Sinhala, Sundanese,
Syloti Nagri, Tamil, Telugu, and Tibetan. At the same time, remove
entries for Meetai Mayak and Sharada – the Unicode ranges were
incorrect (and nobody has complained about that), fonts are scarce
for those scripts, and the Indic auto-hinter support is rudimentary
anyways.
* src/autofit/afscript.h: Updated, using AF_CONFIG_OPTION_INDIC and
AF_CONFIG_OPTION_CJK.
* src/autofit/afstyles.h (STYLE_DEFAULT_INDIC): New auxiliary macro.
Use it, together with AF_CONFIG_OPTION_INDIC and
AF_CONFIG_OPTION_CJK, to update.
* src/autofit/afranges.c [AF_CONFIG_OPTION_INDIC]: Updated.
[!AF_CONFIG_OPTION_INDIC, !AF_CONFIG_OPTION_CJK]: Removed.
Sort entries by tags.
With this commit, the implementation of coverage handling is
completed.
* src/autofit/hbshim.c (af_get_coverage): Exit early if nothing to
do.
Reject coverages which don't contain appropriate glyphs for blue
zones.
* src/autofit/afglobal.c (af_face_globals_compute_style_coverage):
First handle non-default coverages, then the default coverage of the
default script, and finally the other default coverages.
* src/autofit/afcover.h: Remove coverages for alternative fractions,
denominators, numerators, and fractions.
* src/autofit/afstyles.h (META_STYLE_LATIN): Updated.
* src/autofit/afstyles.h (STYLE_LATIN, META_STYLE_LATIN): New
auxiliary macros; use them to define styles for Cyrillic, Greek, and
Latin.
* src/autofit/afcover.h: Remove coverage for oldstyle figures.
Since those digits are used in combination with ordinary letters, it
makes no sense to handle them separately.
* src/autofit/afglobal.c (af_face_globals_get_metrics): Don't limit
`options' parameter to 4 bits.
* src/autofit/hbshim.c (af_get_coverage)
[FT_CONFIG_OPTION_USE_HARFBUZZ]: Scan GPOS coverage of features also
so that we can skip glyphs that have both GSUB and GPOS data.
* src/autofit/hbshim.c (af_get_char_index): Return error value.
Add argument for y offset (to be used in a yet-to-come patch).
* src/autofit/hbshim.h, src/autofit/afcjk.c,
src/autofit/aflatin.c: Updated.
Combining them, as originally envisioned, would lead to much more
complicated code, as investigations have shown meanwhile. The major
drawback is that we run out of available style slots much earlier.
However, this is only a theoretical issue since we don't support a
large number of scripts currently.
* src/autofit/afcover.h: Replace `COVERAGE_{1,2,3}' macros with
a single-element `COVERAGE' macro, sort the elements by the feature
tags, and add entry for `ruby'.
* src/autofit/aftypes.h: Updated.
* src/autofit/hbshim.c: Updated.
Yet-to-come changes will provide HarfBuzz functionality for the new
function.
* src/autofit/hbshim.c (af_get_char_index): New function.
* src/autofit/hbshim.h: Updated.
* src/autofit/afcjk.c (af_cjk_metrics_init_widths,
af_cjk_metrics_init_blues, af_cjk_metrics_check_digits): Updated.
* src/autofit/aflatin.c (af_latin_metrics_init_widths,
af_latin_metrics_init_blues, af_latin_metrics_check_digits):
Updated.
We now use `hb_font' instead of `hb_face' since yet-to-come changes
need this.
* src/autofit/afglobal.h: Include `hbshim.h'.
(AF_FaceGlobalsRec) [FT_CONFIG_OPTION_USE_HARFBUZZ]: New member
`hb_font'.
* src/autofit/afglobal.c (af_face_globals_new)
[FT_CONFIG_OPTION_USE_HARFBUZZ]: Create `hb_font'.
(af_face_globals_free) [FT_CONFIG_OPTION_USE_HARFBUZZ]: Destroy
`hb_font'.
* src/autofit/hbshim.h: Include HarfBuzz headers.
* src/autofit/hbshim.c: Include `hbshim.h' instead of HarfBuzz
headers.
(af_get_coverage): Updated.
The fragmented resources in Suitcase and .dfont should be reordered
when post resource for Type1 is being restored, but reordering of
sfnt resources induces the different face order. Now the ordering
is restricted to post resource only, to prevent the different order
issue (e.g. the face index in the fontconfig cache generated with
Carbon framework is incompatible with that by FreeType2 without
Carbon framework.) Found by Khaled Hosny and Hin-Tak Leung.
http://lists.gnu.org/archive/html/freetype-devel/2013-02/msg00035.htmlhttp://lists.gnu.org/archive/html/freetype-devel/2013-12/msg00027.html
* src/base/ftrfork.c (FT_Raccess_Get_DataOffsets): Add a switch
`sort_by_res_id' to control the fragmented resource ordering.
* include/internal/ftrfork.h: Daclare new switch.
* src/base/ftobjs.c (IsMacResource): Enable the soring for post
resource, and disable the sorting for sfnt resource.
Coverages are the interface to the HarfBuzz library to access
OpenType features for handling glyphs not addressable by the cmap.
Right now, compilation of HarfBuzz is only added to the development
build. A solution for standard build mode will be delayed until
HarfBuzz gets split into two libraries to avoid mutual dependencies
between FreeType and HarfBuzz.
Note that this is only a first step in handling coverages, basically
providing the framework only. Code for handling selected OpenType
features (this is, actually using the data in `afcover.h') will
follow.
* devel/ftoption.h, include/config/ftoption.h
(FT_CONFIG_OPTION_USE_HARFBUZZ): New macro.
* src/autofit/hbshim.c, src/autofit/hbshim.h, src/autofit/afcover.h:
New files.
* src/autofit/afscript.h: Add HarfBuzz script name tags.
* src/autofit/afstyles.h: Add default coverage enumeration values.
* src/autofit/aftypes.h: Update use of `SCRIPT' and `STYLE' macros.
(AF_Coverage): New enumeration (generated by `afcover.h').
(AF_StyleClassRec): New member `coverage'.
(AF_DEFINE_STYLE_CLASS): Updated.
* include/internal/fttrace.h: Add `afharfbuzz' for tracing coverage
data.
* src/autofit/afglobal.h: Update use of `SCRIPT' and `STYLE' macros.
(AF_SCRIPT_FALLBACK): Renamed to ...
(AF_STYLE_FALLBACK): ... this.
* src/autofit/afglobal.c: Include `hbshim.c'.
Update use of `SCRIPT' and `STYLE' macros.
(af_face_globals_compute_style_coverage)
[FT_CONFIG_OPTION_USE_HARFBUZZ]: Call `af_get_coverage'.
Update.
* src/autofit/afmodule.h (AF_ModuleRec):
s/fallback_script/fallback_style/.
* src/autofit/afmodule.c (af_property_set): Adapt handling of
`fallback-script' property to set a fallback style.
(af_property_get, af_autofitter_init): Updated.
* src/autofit/afpic.c: Update use of `SCRIPT' and `STYLE' macros.
* src/autofit/afranges.h: Update use of `SCRIPT' macro.
* src/autofit/autofit.c [FT_CONFIG_OPTION_USE_HARFBUZZ]: Include
`hbshim.c'.
* src/autofit/rules.mk (AUTOF_DRV_SRC): Add `hbshim.c'.
(AUTOF_DRV_H): Add `afcover.h'.
* builds/freetype.mk (INCLUDE_FLAGS) [DEVEL_DIR]: Use pkg-config for
all libraries needed by FreeType.
* src/autofit/afpic.h (AFModulePIC): Fix array sizes to fit the
enumeration values automatically generated by including `afscript.h'
and friends.
* src/autofit/afpic.c (autofit_module_class_pic_init): Updated.
* include/internal/ftrfork.h (CONST_FT_RFORK_RULE_ARRAY_BEGIN): Fix
generated function name.
* src/base/basepic.c (FT_Init_Table_raccess_guess_table): Rename
to ...
(FT_Init_Table_ft_raccess_guess_table): ... this so that the
function name correctly corresponds to what the macro framework
expects.
* src/psnames/rules.mk (PSNAMES_DRV_SRC_S): Use correct file name so
that PIC functions are compiled also.
This is the first commit of a series to create a new top-level
structure (a `style') for handling scripts, writing_systems, and
soon-to-be-added coverages.
This patch unifies the subpixel and non-subpixel cases.
* src/truetype/ttinterp.h (TT_ExecContextRec): Remove
`grayscale_hinting'; all code should refer to `grayscale' instead.
Remove unused `native_hinting' member.
Rename `subpixel_hinting' member to `subpixel.
* src/truetype/ttgload.c (TT_LOADER_SET_PP): Updated.
(tt_loader_init): Updated.
* src/truetype/ttinterp.c (Ins_GETINFO): Simplify.
Updated.
Problem reported by David Kastrup <dak@gnu.org>.
Surprisingly, a description how to do that was completely missing in
the API reference.
* include/freetype.h, include/ftchapters.h: New documentation
section `header_inclusion'.
Problem reported by Nigel Tao <nigeltao@golang.org>.
* src/truetype/ttgload.c (TT_Hint_Glyph): Remove code that shifts
the glyph (component) by a fractional value computed from the LSB
phantom point. This is wrong, since the horizontal phantom points
get rounded horizontally later on.
Greg Hitchcock provided very interesting insights into the
complicated history of the horizontal positions of the TSB and BSB
phantom points.
* src/truetype/ttgload.c (TT_LOADER_SET_PP)
[TT_CONFIG_OPTION_SUBPIXEL_HINTING]: Use `subpixel_hinting' and
`grayscale_hinting' flags as conditionals for the x position of TSB
and BSB.
This corrects the commit from 2013-11-21. Problem reported by
Andrey Panov <panov@canopus.iacp.dvo.ru>.
* src/sfnt/ttsbit.c (tt_sbit_decoder_load_bitmap): Fix logic to
detect excessive bytes for bit-aligned bitmaps.
This is a further improvement to the changes from 2013-11-06.
* src/truetype/ttgload.c (TT_Hint_Glyph): Horizontal phantom points
are rounded horizontally, vertical ones are rounded vertically.
(TT_LOADER_SET_PP): The horizontal position of vertical phantom
points in pre-ClearType mode is zero, as shown in the OpenType
specification.
Problem reported by Akira Kakuto <kakuto@fuk.kindai.ac.jp>.
* src/truetype/ttgload.c (TT_Load_Simple_Glyph): Protect call to
`Update_Max' with both a TT_USE_BYTECODE_INTERPRETER guard and a
`IS_HINTED' clause.
Also remove redundant check using `maxSizeOfInstructions' – in
simple glyphs, the bytecode data comes before the outline data, and
a validity test for this is already present.
Problem reported by Hin-Tak Leung <htl10@users.sourceforge.net>.
* src/sfnt/sfobjs.c (sfnt_load_face): Return correct `bsize->width'
value if the font lacks an `OS/2' table.
We are now able to successfully load `AppleMyoungJo.ttf'.
Problem reported by Hin-Tak Leung <htl10@users.sourceforge.net>.
* src/sfnt/ttsbit.c (tt_sbit_decoder_load_bitmap): Don't trust glyph
format.
Problem reported by Hin-Tak Leung <htl10@users.sourceforge.net>; see
http://lists.nongnu.org/archive/html/freetype-devel/2013-08/msg00005.html
for details.
* src/base/ftobjs.c (FT_Load_Glyph): Check size of `fpgm' and `prep'
tables also for setting `autohint'.
* src/truetype/ttgload.c (TT_Load_Simple_Glyph): Use code from
`TT_Process_Composite_Glyph' for handling unreliable values of
`maxSizeOfInstructions'.
We now follow the `official' announcement from Microsoft (on the
OpenType mailing list, which unfortunately hasn't a public archive).
* include/freetype/tttables.h (TT_OS2):
s/usLowerPointSize/usLowerOpticalPointSize/,
s/usUpperPointSize/usUpperOpticalPointSize/.
* src/sfnt/ttload.c (tt_face_load_os2): Update, and set correct
default values.
This large patch changes the header file directory layout from
`include/freetype/...' to `include/...', effectively removing one
level. Since the file `ft2build.h' is also located in `include'
(and it stays there even after installation), all FreeType header
files are now in a single directory.
Applications that use (a) `freetype-config' or FreeType's
`pkg-config' file to get the include directory for the compiler, and
(b) the documented way for header inclusion like
#include <ft2build.h>
#include FT_FREETYPE_H
...
don't need any change to the source code.
* include/freetype/*: Move up to...
* include/*: ... this directory.
* builds/amiga/include/freetype/*: Move up to...
* builds/amiga/include/*: ... this directory.
*/*: Essentially do `s@/freetype/@/@' where appropriate.
* CMakeList.txt: Simplify.
* builds/unix/freetype-config.in, builds/unix/freetype2.in: For
`--cflags', return a single directory.
* builds/unix/install.mk (install): No longer try to remove `cache'
and `internal' subdirectories; instead, remove the `freetype'
subdirectory.
Problem reported by Nigel Tao <nigeltao@golang.org>.
This is a follow-up commit to the previous one.
* src/truetype/ttgload.c (load_truetype_glyph): Call
`tt_get_metrics' after loading the glyph header.
This commit also improves the start values of vertical phantom
points. Kudos to Greg Hitchcock for help.
* src/truetype/ttgload.c (TT_Get_VMetrics): Add parameter to pass
`yMax' value. Replace code with fixed Microsoft definition.
(tt_get_metrics): Updated.
(TT_LOADER_SET_PP): Add explanation how to initialize phantom
points, taken from both the OpenType specification and private
communication with Greg (which will eventually be added to the
standard).
Fix horizontal position of `pp3' and `pp4'.
* src/truetype/ttgload.h: Updated.
* src/truetype/ttdriver.c (tt_get_advances): Updated.
* docs/CHANGES: Updated.
Simply apply the patch from the bug report.
* builds/unix/ftconfig.in, builds/vms/ftconfig.h,
include/freetype/config/ftconfig.h: The used #pragma directives only
work with gcc versions 4.6 and higher.
Two benefits: The allocated FDEF (and IDEF) array gets slightly
smaller, and the `ttdebug' demo program has access to function
numbers without additional costs.
Fortunately, no changes to FontForge are necessary – this is the
only external TrueType debugger I know of, but others may exist and
should check the code accordingly.
* src/truetype/ttinterp.h (TT_CallRec): Replace `Cur_Restart' and
`Cur_End' with a pointer to the corresponding `TT_DefRecord'
structure.
* src/truetype/ttinterp.c (DO_JROT, DO_JMPR, DO_JROF, Ins_ENDF,
Ins_CALL, Ins_LOOPCALL, Ins_UNKNOWN, TT_RunIns <Invalid_Opcode>):
Updated.
See
http://typedrawers.com/discussion/470/new-microsoft-size-specific-design-selection-mechanism
for the announcement.
* include/freetype/tttables.h (TT_OS2): Add fields
`usLowerPointSize' and `usUpperPointSize'. Since FreeType returns
this structure only as a pointer through `FT_Get_Sfnt_Table', there
shouldn't be any ABI problems.
* src/sfnt/ttload.c (tt_face_load_os2): Implement it.
* docs/CHANGES: Updated.
When using `ADD' with an immediate operand, the instruction is
actually `ADD Rd, Rn, #<imm12>', that is, the maximum of the
immediate operand cannot exceed 4095. It will fail to compile with
LLVM.
However, in GCC, due to some legacy compatibility considerations,
`ADD.W' will be automatically emitted when the immediate operand is
larger than 4095.
* builds/unix/ftconfig.in, include/freetype/config/ftconfig.h
(FT_MulFix_arm) [__GNUC__]: Support clang compiler.
* src/truetype/ttinterp.c (TT_MulFix14_arm) [__GNUC__]: Ditto.
* src/autofit/aflatin.c (af_latin_metrics_init_blues): Report blue
zone types.
(af_latin_metrics_scale_dim): Report scaling changes due to x height
alignment.
Report scaled stroke width and blue zone values.
This line was lost in commit 89ca1fd6 (from 2013-06-25). The effect
is to use a previous darkening amount when producing an unhinted,
unscaled outline. This can cause autohint samples in ftgrid and
ftview to be based on darkened CFF outlines instead of unhinted,
undarkened ones.
The bug was caused by switching to the initial hintmap (the one in
effect when `moveto' executes) just before drawing the final element
in the charstring. This ensured that the path was closed (in both
Character Space and Device Space). But if the final element was a
curve and if the final hintmap was different enough from the initial
one, then the curve was visibly distorted.
The first part of the fix is to draw the final curve using the final
hintmap as specified by the charstring. This corrects the
distortion but does not ensure closing in Device Space. It may
require the rasterizer to automatically generate an extra closing
line. Depending on the hintmap differences, this line could be from
zero to a couple pixels in length.
The second part of the fix covers the case where the charstring
subpath is closed with an explicit line. We now modify that line's
end point to avoid the distortion.
Some glyphs in the bug report font (TexGyreHeros-Regular) that show
the change are:
25ppem S (98)
24ppem eight (52)
25.5ppem p (85)
Curves at the *end* of a subpath are no longer distorted. However,
some of these glyphs have bad hint substitutions in the middle of a
subpath, and these are not affected.
The patch has been tested with a set of 106 fonts that shipped with
Adobe Creative Suite 4, together with 756 Open Source CFF fonts from
Google Fonts. There are 1.5 million glyphs, of which some 20k are
changed with the fix. A sampling of a few hundred of these changes
have been examined more closely, and the changes look good (or at
least acceptable).
* src/cff/cf2hints.h (CF2_GlyphPathRec): New element `pathIsClosing'
to indicate that we synthesize a closepath line.
* src/cff/cf2hints.c (cf2_glyphpath_init): Updated.
(cf2_glyphpath_pushPrevElem): If closing, use first hint map (for
`lineto' operator) and adjust hint zone.
For synthesized closing lines, use end point in first hint zone.
(cf2_glyphpath_lineTo): Take care of synthesized closing lines. In
particular, shift the detection of zero-length lines from character
space to device space.
(cf2_glyphpath_closeOpenPath): Remove assertion.
Updated.
Suggested by Akira Tagoh, see
http://lists.gnu.org/archive/html/freetype/2013-09/msg00030.html
* src/bdf/bdfdrivr.c (BDF_Face_Init): Return Invalid_Argument error
when the font could be opened but non-zero face_index is given.
* src/pcf/pcfdrivr.c (PCF_Face_Init): Ditto.
* src/type42/t42objs.c (T42_Face_Init): Remove unrequired FT_UNUSED
macro for face_index because it is validated in later.
This change introduces a new blue zone property
`AF_BLUE_PROPERTY_LATIN_LONG' to make the auto-hinter ignore short
top segments.
* src/autofit/afblue.dat: Fix Hebrew blue strings.
Use AF_BLUE_PROPERTY_LATIN_LONG for AF_BLUE_STRING_HEBREW_TOP.
* src/autofit/afblue.hin (AF_BLUE_PROPERTY_LATIN_LONG): New macro.
* src/autofit/afblue.c, src/autofit/afblue.h: Updated.
* src/autofit/aflatin.c (af_latin_metrics_init_blues): Handle
`AF_LATIN_IS_LONG_BLUE'.
* src/autofit/aflatin.h (AF_LATIN_IS_LONG_BLUE): New macro.
We simply synthesize a SFNT from the WOFF, create a memory stream
for the new data, and load the SFNT as usual.
Does NOT add any API to access WOFF metadata or private blocks.
* include/freetype/internal/tttypes.h (WOFF_HeaderRec,
WOFF_TableRec): New structures.
* include/freetype/tttags.h (TTAG_wOFF): New macro.
* src/base/ftobjs.c (FT_Open_Face): Set `stream' after calling
`open_face'.
* src/sfnt/sfobjs.c [FT_CONFIG_OPTION_SYSTEM_ZLIB]: Include
`FT_GZIP_H'.
(WRITE_BYTE, WRITE_USHORT, WRITE_ULONG): New temporary macros for
writing to a stream.
(sfnt_stream_close, compare_offsets, woff_open_font): New functions.
(sfnt_open_font): Handle `TTAG_wOFF'.
(sfnt_init_face): Set `stream' after calling `sfnt_open_font'.
* src/truetype/ttobjs.c (tt_face_init): Set `stream' after calling
`sfnt->init_face'.
* src/base/ftobjs.c (open_face): Use a pointer to FT_Stream as an
argument so that a changed stream survives.
Update callers.
This is modeled after zlib's `uncompress' function. We need this
for WOFF support.
* include/freetype/ftgzip.h, src/gzip/ftgzip.c (FT_Gzip_Uncompress):
New function.
* src/gzip/rules.mk: Rewrite to better reflect dependencies.
* src/autofit/afblue.cin, src/autofit/afblue.c: Don't include
`afblue.h' but `aftypes.h'.
* src/autofit/afcjk.c: Don't include `aftypes.h' but `afglobal.h'.
If the font's `clazz->init_face' function wants to swap to new
stream, handling of whether original stream was external could
result to either memory leak or double free. Mark externality into
face flags before calling `init_face' such that the clazz can handle
external streams properly.
* src/base/ftobjs.c (FT_Open_Face): Move code to set
FT_FACE_FLAG_EXTERNAL_STREAM to...
(open_face): This function.
Previously, the loading of a glyph was traced at level 4, if at all.
With this change, all font loading routines emit a tracing message
at level 1, making it easier to select tracing output (for example
using F2_DEBUG="any:1 afhints:7 aflatin:7").
* src/bdf/bdfdrivr.c (BDF_Glyph_Load): Add tracing message.
* src/cff/cffdrivr.c (cff_glyph_load): Ditto.
* src/cff/cffgload.c (cff_decoder_prepare): Improve tracing
messages.
* src/cid/cidgload.c (cid_load_glyph): Use level 1 for tracing
message.
* src/pcf/pcfdrivr.c (PCF_Glyph_Load): Ditto.
* src/pfr/pfrobjs.c (pfr_slot_load): Add tracing message.
* src/truetype/ttgload.c (TT_Load_Glyph): Ditto.
* src/type1/t1gload.c (T1_Load_Glyph): Ditto.
* src/type42/t42objs.c (T42_GlyphSlot_Load): Ditto.
* src/winfonts/winfnt.c (FNT_Load_Glyph): Ditto.
* src/autofit/afglobal.c (af_face_globals_get_metrics): Use
`AF_SCRIPT_DFLT', not value 0.
Simplify code.
* src/autofit/afscript.h: Sort by script name.
* src/autofit/afcjk.c (AF_CJK_MAX_TEST_CHARACTERS): Removed.
(af_cjk_hani_blue_chars): Removed.
(AF_CJK_BLUE_TYPE_*): Removed.
(af_cjk_metrics_init_blues): Replace AF_CJK_MAX_TEST_CHARACTERS with
AF_BLUE_STRING_MAX_LEN.
Change loops to use offsets (in file `afblue.h') into the new arrays
`af_blue_stringsets' and `af_blue_strings' (in file `afblue.c').
Instead of three dimensions (as used in the old blue string array)
we now use properties to do the same, saving one loop nesting level.
* src/autofit/afcjk.h: Remove old enumeration values superseded by
the new data in `afblue.h'.
(AF_CJK_IS_TOP_BLUE, AF_CJK_IS_HORIZ_BLUE, AF_CJK_IS_FILLED_BLUE,
AF_CJK_IS_RIGHT_BLUE): New macros, to be used in
`af_cjk_metrics_init_blues'.
(AF_CJK_BLUE_IS_RIGHT): Remove this now redundant enum value.
(AF_CJK_BLUE_IS_TOP): Renamed to...
(AF_CJK_BLUE_TOP): This.
(AF_CJK_MAX_BLUES): Remove.
(AF_CJKAxisRec): Updated.
* src/autofit/aflatin.c (AF_LATIN_MAX_TEST_CHARACTERS): Removed.
(af_latin_blue_chars): Removed.
(af_latin_metrics_init_blues): Replace AF_LATIN_MAX_TEST_CHARACTERS
with AF_BLUE_STRING_MAX_LEN.
Change loops to use offsets (in file `afblue.h') into the new arrays
`af_blue_stringsets' and `af_blue_strings' (in file `afblue.c').
Use `AF_LATIN_IS_SMALL_TOP_BLUE' macro.
* src/autofit/aflatin.h: Remove old enumeration values superseded by
the new data in `afblue.h'.
(AF_LATIN_IS_TOP_BLUE): Updated definition.
(AF_LATIN_IS_SMALL_TOP_BLUE): New macro.
(AF_LATIN_MAX_BLUES): Remove.
(AF_LatinAxisRec): Updated.
The idea is to have a central file which gets processed by a Perl
script to create proper `.c' and `.h' files using templates. There
are two other reasons to do that:
. The data file should be easily readable. We use UTF-8 encoding
which then gets converted to single bytes.
. Since the number of supported scripts will increase soon, the
current usage of blue string arrays is a waste of space. Using
the Perl script it is possible to imitate jagged arrays,
defining enumeration constants as offsets into the arrays.
This commit only adds files without changing any functionality.
* src/autofit/afblue.dat: New data file.
* src/tools/afblue.pl: New Perl script for processing `afblue.dat'.
* src/autofit/afblue.cin, src/autofit/afblue.hin: New template files
for...
* src/autofit/afblue.c, src/autofit/afblue.c: New source files.
To avoid a dependency on Perl, we add them too.
* src/base/ftbbox.c (BBox_Conic_Check): Remove redundant checks for
extremum at the segment ends, which are already within the bbox.
Slightly modify calculations.
* src/cff/cffload.c (cff_index_get_pointers): Check for `cur_offset
!= 0'; this stronger test is mandated by the CFF specification.
Fix test for INDEX structures which have one or more empty entries
at the end.
* src/truetype/ttinterp.c (TT_MulFix14_long_long,
TT_DotFix14_long_long): `#pragma gcc diagnostic {push,pop}' has been
introduced with gcc version 4.6.
* builds/unix/ftconfig.h: Synchronize with
`include/freetype/config/ftconfig.h'.
* builds/vms/ftconfig.h: Ditto.
Make the differences to the master `ftconfig.h' file as small as
possible for easier maintainance.
Points which are very near to each other are now marked as such.
The `weak' flag is then computed by using the `in' vector of the
first and the `out' vector of the last point of a group of near
points.
For example, this fixes the rendering of glyph `Oslash' in
`Roboto-Thin.ttf'.
* src/autofit/afhints.h (AF_Flags): New value `AF_FLAGS_NEAR'.
* src/autofit/afhints.c (af_glyph_hints_reload): Introduce
the heuristic value `near_limit' to decide whether the current point
is near to the previous one, then set `AF_FLAG_NEAR' accordingly.
Store good `in' vector (of last non-near point) in
`last_good_in_{x,y}' and use it as an argument to
`ft_corner_is_flat' if necessary.
This has the benefit that we don't need to duplicate the data at
different places.
* src/autofit/afscript.h: New file.
* src/autofit/aftypes.h (AF_Script): Include `afscript.h' to define
the enumeration values.
* src/autofit/afglobal.c: Include `afscript.h' to get the script
specific header files.
(af_script_classes): Include `afscript.h' to fill this array.
* src/autofit/afpic.c: Include `afscript.h' to get the script
specific header files.
(autofit_module_class_pic_init): Include `afscript.h' for
initialization.
* src/autofit/afpic.h (AF_SCRIPT_CLASSES_COUNT,
AF_SCRIPT_CLASSES_REC_COUNT): Removed. Use `AF_SCRIPT_MAX' instead.
* src/autofit/rules.mk (AUTOF_DRV_H): Updated.
This has the benefit that we don't need to duplicate the data at
different places.
* src/autofit/afwrtsys.h: New file.
* src/autofit/aftypes.h (AF_WritingSystem): Include `afwrtsys.h' to
define the enumeration values.
* src/autofit/afglobal.c: Include `afwrtsys.h' to get the writing
system specific header files.
Include `afpic.h'.
(af_writing_system_classes): Include `afwrtsys.h' to fill this
array.
* src/autofit/afpic.c: Include `afwrtsys.h' to get the writing
system specific header files.
(autofit_module_class_pic_init): Include `afwrtsys.h' for
initialization.
* src/autofit/afpic.h (AF_WRITING_SYSTEM_CLASSES_COUNT,
AF_WRITING_SYSTEM_CLASSES_REC_COUNT): Removed. Use
`AF_WRITING_SYSTEM_MAX' instead.
The call was (from the top-level of the FreeType tree):
cppcheck --force \
--enable=all \
-I /usr/include \
-I /usr/local/include \
-I /usr/lib/gcc/i586-suse-linux/4.7/include \
-I include \
-I include/freetype \
-I include/freetype/config \
-I include/freetype/internal \
-DFT2_BUILD_LIBRARY \
. &> cppcheck.log
using cppcheck git commit f7e93f99.
Note that cppcheck still can't handle `#include FOO' (with `FOO' a
macro).
*/* Improve variable scopes.
*/* Remove redundant initializations which get overwritten.
* src/gxvalid/*: Comment out redundant code or guard it with
FT_DEBUG_LEVEL_TRACE.
This patch adds a new top level to the auto-hinter's script class
hierarchy. It defines `writing systems' which can contain multiple
scripts.
For example, the `latin' writing system (in file `aflatin.c') is
able to support scripts like Latin, Cyrillic, Armenian, etc., which
can be handled similarly.
Scripts are now named using four-letter OpenType tags.
* src/autofit/aftypes.h (AF_ScriptClassRec): Move relevant members
to...
(AF_WritingSystemClassRec): This new structure. It holds pointers
to functions which can be shared among related scripts.
(AF_WritingSystem): New enumeration.
(AF_Script): Revised values using four-letter tags.
(AF_DEFINE_WRITING_SYSTEM_CLASS): New macro.
(AF_DEFINE_SCRIPT_CLASS): Updated.
* src/autofit/afglobal.c (af_writing_system_classes): New global,
constant array.
(af_script_classes): Updated.
(af_face_globals_free): Updated.
Remove assertion.
(af_face_globals_get_metrics): Updated.
* src/autofit/afglobal.h (AF_SCRIPT_FALLBACK)
[!AF_CONFIG_OPTION_CJK]: Handle this case.
* src/autofit/afloader.c (af_loader_load_g, af_loader_load_glyph):
Updated.
* src/autofit/afpic.c (autofit_module_class_pic_init): Updated;
initialize structures for both writing systems and scripts.
* src/autofit/afpic.h: Updated.
(AF_WRITING_SYSTEM_CLASSES_GET): New macro.
* src/autofit/afcjk.c (af_cjk_writing_system_class): New writing
system.
(af_cjk_uniranges): Renamed to...
(af_hani_uniranges): This.
(af_cjk_script_class): Reduced and renamed to...
(af_hani_script_class): This.
* src/autofit/afcjk.h: Updated.
* src/autofit/afdummy.c (af_dummy_writing_system_class): New writing
system.
(af_dummy_script_class): Reduced and renamed to...
(af_dflt_script_class): This.
* src/autofit/afdummy.h: Updated.
* src/autofit/afindic.c (af_indic_writing_system_class): New writing
system.
(af_indic_uniranges): Renamed to...
(af_deva_uniranges): This.
(af_indic_script_class): Reduced and renamed to...
(af_deva_script_class): This.
* src/autofit/afcjk.h: Updated.
* src/autofit/aflatin.c (af_latin_writing_system_class): New writing
system.
(af_latin_uniranges): Renamed to...
(af_latn_uniranges): This.
(af_latin_script_class): Reduced and renamed to...
(af_latn_script_class): This.
* src/autofit/aflatin.h: Updated.
* src/autofit/aflatin2.c (af_latin2_writing_system_class): New
writing system.
(af_latin2_uniranges): Renamed to...
(af_ltn2_uniranges): This.
Synchronize ranges with `latin'.
(af_latin2_script_class): Reduced and renamed to...
(af_ltn2_script_class): This.
* src/autofit/aflatin2.h: Updated.
Prevent division by zero by a transparent color.
* src/base/ftbitmap.c (ft_gray_for_premultiplied_srgb_bgra):
Return 0 immediately, when alpha channel is zero.
The old code (essentially unchanged since the very beginning)
incorrectly handled this configuration
x -o- x
/ \
/ \
/ \
o o
as flat and this
o o
/ /
x| x|
| |
o---------------o
as round. (`o' and `x' are on and off points, respectively).
This is a major change which should improve the rendering results
enormously for many TrueType fonts, especially in the range approx.
20-40ppem, fixing the appearance of many overshoots.
* src/autofit/aflatin.c (af_latin_metrics_init_blues): Look at the
first and last points of the segment, not the points right before
and after.
* src/sfnt/sfobjs.c (sfnt_load_face): Apple's `sbix' color bitmaps
are rendered scaled and then the `glyf' outline rendered on top. We
don't support that yet, so just ignore the `glyf' outline and
advertise it as a bitmap-only font.
* src/sfnt/ttsbit.c (tt_face_load_strike_metrics)
[TT_SBIT_TABLE_TYPE_SBIX]: Return metrics in 26.6 units.
(tt_face_load_sbix_image): Typo.
Other font engines seem to ignore it also. Problem reported by
Hin-Tak Leung <htl10@users.sourceforge.net>.
* src/sfnt/ttload.c (check_table_dir): Don't abort but warn only if
we have an invalid magic number.
Provide a work-around for an ARM-specific performance bug in GCC.
This speeds up the rasterizer by more than 5%.
Also slightly optimize `set_gray_cell' and `gray_record_cell' (which
also improves performance on other platforms by a tiny bit (<1%).
* src/smooth/ftgrays.c (FT_DIV_MOD): New macro.
Use it where appropriate.
(gray_record_cell, gray_set_cell, gray_move_to,
gray_convert_glyph_inner): Streamline condition handling.
This patch provides slightly optimized versions for ARM, x86, and
x86_64 CPUs if built with GCC.
Also remove some dead code.
* src/truetype/ttinterp.c (TT_MulFix14_arm, TT_MulFix14_long_long,
TT_DotFix14_long_long): New functions.
When building for ARMv7 with thumb2 instructions, the optimized
`FT_MulFix_arm' assembly routine was not being used.
The reason for this is in the `ftconfig.h' header, namely:
- The assembly routine uses the `smull' instruction which is not
available when generating Thumb-1 machine code. It is available
in Thumb-2 mode, though.
- The header was written a long time ago before Thumb-2 became
widely popular (e.g. with Android). So it simply doesn't use the
assembly routine if the `__thumb__' built-in macro is defined.
- When compiling in Thumb-2 mode, the compiler will define both
`__thumb__' and `__thumb2__'.
By checking for `(__thumb2__ || !__thumb__)', we ensure that the
assembly routine is only avoided when generating Thumb-1 code.
Given that this is performance-sensitive function, this improves
`ftbench' as follows on a Galaxy Nexus:
Before (us/op) After (us/op)
- loading Arial.ttf glyphs at 14 ppem [1]
Load 34.285 33.098
- same operation with the light auto-hinter [2]
Load 31.317 29.590
- same operation without hinting [3]
Load 6.143 5.376
- loading Arial.ttf advances at 14 ppem [4]
Load_Advances (normal) 34.216 33.016
Load_Advances (fast) 0.176 0.176
[1] ftbench -t 5 -p -s 14 -b a -f 0008 Arial.ttf
[2] ftbench -t 5 -p -s 14 -b a -r 1 -f 0028 Arial.ttf
[3] ftbench -t 5 -p -s 14 -b a -f 000a Arial.ttf
[4] ftbench -t 5 -p -s 14 -b b -f 0008 Arial.ttf
* builds/unix/ftconfig.in, include/freetype/config/ftconfig.h
(FT_MULFIX_ASSEMBLER): Fix handling for ARMv7.
Werner Lemberg