extern const FT_Module_Class
(or similar for C++). However, the actual types of the variables
being declared are often different, e.g., FT_Driver_ClassRec or
FT_Renderer_Class. (Some are, indeed, FT_Module_Class.)
This works with most C compilers (since those structs begin with an
FT_Module_Class struct), but technically it's undefined behavior.
To quote the ISO/IEC 9899:TC2 final committee draft, section 6.2.7
paragraph 2:
All declarations that refer to the same object or function shall
have compatible type; otherwise, the behavior is undefined.
(And they are not compatible types.)
Most C compilers don't reject (or even detect!) code which has this
issue, but the GCC LTO development branch compiler does. (It
outputs the types of the objects while generating .o files, along
with a bunch of other information, then compares them when doing the
final link-time code generation pass.)
Patch from Savannah bug #25133.
* src/base/ftinit.c (FT_USE_MODULE): Include variable type.
* builds/amiga/include/freetype/config/ftmodule.h,
include/freetype/config/ftmodule.h, */module.mk: Updated to declare
pass correct types to FT_USE_MODULE.
sufficient to modify a single file, `modules.cfg', to control the
inclusion of modules and base extension files.
This change also fixes the creation of ftmodule.h; it now depends on
`modules.cfg' and thus is rebuilt only if necessary.
Finally, a version of `ftoption.h' in OBJ_DIR is preferred over the
default location.
* modules.cfg: New file.
* builds/freetype.mk: Don't include `modules.mk'.
Include all `rules.mk' files as specified in `modules.cfg'.
(FTOPTION_FLAG, FTOPTION_H): New variables.
(FT_CFLAGS): Add macro definition for FT_CONFIG_MODULES_H.
Add FTOPTION_FLAG.
($(FT_INIT_OBJ)): Don't use FT_MODULE_LIST.
(CONFIG_H): Add FTMODULE_H and FTOPTION_H.
(INCLUDES): Add DEVEL_DIR.
(INCLUDE_FLAGS, FTSYS_SRC, FTSYS_OBJ, FTDEBUG_SRC, FTDEBUG_OBJ,
OBJ_M, OBJ_S): Use `:=', not `='.
(remove_ftmodule_h): New phony target to delete `ftmodule.h'.
(distclean): Add remove_ftmodule_h.
* builds/modules.mk: (MODULE_LIST): Removed.
(make_module_list, clean_module_list): Replace targets
with...
(FTMODULE_H_INIT, FTMODULE_H_CREATE, FTMODULE_H_DONE): New
variables. Reason for the change is that it is not possible to have
a phony prerequisite which is run only if the target file must be
rebuilt (phony prerequisites act like subroutines and are *always*
executed). We only want to rebuild `ftmodule.h' if `module.cfg' is
changed.
Update all callers.
($FTMODULE_H)): Rule to create `ftmodule.h', depending on
`modules.cfg'.
* builds/toplevel.mk: Rewrite and simplify module handling.
(MODULES_CFG, FTMODULE_H): New variables.
Include MODULES_CFG.
(MODULES): New variable to include all `module.mk' and `rules.mk'
files. We no longer use make's `wildcard' function for this.
* Makefile (USE_MODULES): Remove. Update all users.
(OBJ_DIR): Define it here.
* src/*/module.mk: Change
make_module_list: foo
foo: ...
to
FTMODULE_H_COMMANDS += FOO
define FOO
...
endef
in all files. `FTMODULE_H_COMMANDS' is used in `FTMODULE_H_CREATE'.
* src/base/rules.mk (BASE_EXT_SRC): Use BASE_EXTENSIONS.
* builds/unix/detect.mk (setup): Always execute `configure' script.
(have_mk): Rename to...
(have_Makefile): This.
Don't use `strip' function.
* builds/unix/unix.mk: Include `install.mk' only if BUILD_PROJECT is
defined.
(have_mk): Don't use `strip' function.
Test for unix-def.mk in OBJ_DIR, not BUILD_DIR (and invert the test
accordingly).
* builds/unix/install.mk (install, uninstall): Handle `ftmodule.h'.
* builds/os2/os2-dev.mk, builds/unix/unix-dev.mk,
builds/win32/w32-bccd.mk, builds/win32/w32-dev.mk: Don't define
BUILD_DIR but DEVEL_DIR for development header files.
* builds/ansi/ansi-def.mk (TOP_DIR, OBJ_DIR),
builds/beos/beos-def.mk (TOP_DIR, OBJ_DIR), builds/unix/unix-def.in
(TOP_DIR, OBJ_DIR): Removed. Defined elsewhere.
* builds/dos/dos-def.mk (OBJ_DIR), builds/os2/os2-def.mk (OBJ_DIR),
builds/win32/win32-def.mk (OBJ_DIR): Removed. Defined elsewhere.
* builds/unix/unixddef.mk: Don't define BUILD_DIR but DEVEL_DIR for
development header files.
Don't define PLATFORM.
* configure: Copy `modules.cfg' to builddir if builddir != srcdir.
Update snippet taken from autoconf's m4sh.m4 to current CVS version.
Be more verbose.
* include/freetype/config/ftmodule.h: Add comments -- this file is
no longer used if FreeType is built with GNU make.
* docs/CHANGES, docs/CUSTOMIZE, docs/INSTALL, docs/INSTALL.ANY,
docs/INSTALL.GNU, docs/INSTALL.UNX: Document new build mechanism.
Other minor updates.
* modules.txt: Removed. Contents included in `modules.cfg'.
* include/freetype/internal/ftmemory.h (FT_QAlloc_Debug,
FT_Free_Debug) [FT_STRICT_ALIASING]: Fix typos.
* src/base/ftdbgmem.c (FT_Alloc_Debug, FT_Realloc_Debug,
FT_QAlloc_Debug, FT_QRealloc_Debug, FT_Free_Debug)
[FT_STRICT_ALIASING]: Implement.
A lot of internal modifications have been performed lately on the
source in order to provide the following enhancements:
- more generic module support:
The FT_Module type is now defined to represent a handle to a given
module. The file <freetype/ftmodule.h> contains the FT_Module_Class
definition, as well as the module-loading public API
The FT_Driver type is still defined, and still represents a pointer
to a font driver. Note that FT_Add_Driver is replaced by FT_Add_Module,
FT_Get_Driver by FT_Get_Module, etc..
- support for generic glyph image types:
The FT_Renderer type is a pointer to a module used to perform various
operations on glyph image.
Each renderer is capable of handling images in a single format
(e.g. ft_glyph_format_outline). Its functions are used to:
- transform an glyph image
- render a glyph image into a bitmap
- return the control box (dimensions) of a given glyph image
The scan converters "ftraster.c" and "ftgrays.c" have been moved
to the new directory "src/renderer", and are used to provide two
default renderer modules.
One corresponds to the "standard" scan-converter, the other to the
"smooth" one.
The current renderer can be set through the new function
FT_Set_Renderer.
The old raster-related function FT_Set_Raster, FT_Get_Raster and
FT_Set_Raster_Mode have now disappeared, in favor of the new:
FT_Get_Renderer
FT_Set_Renderer
see the file <freetype/ftrender.h> for more details..
These changes were necessary to properly support different scalable
formats in the future, like bi-color glyphs, etc..
- glyph loader object:
A new internal object, called a 'glyph loader' has been introduced
in the base layer. It is used by all scalable format font drivers
to load glyphs and composites.
This object has been created to reduce the code size of each driver,
as each one of them basically re-implemented its functionality.
See <freetype/internal/ftobjs.h> and the FT_GlyphLoader type for
more information..
- FT_GlyphSlot had new fields:
In order to support extended features (see below), the FT_GlyphSlot
structure has a few new fields:
linearHoriAdvance: this field gives the linearly scaled (i.e.
scaled but unhinted) advance width for the glyph,
expressed as a 16.16 fixed pixel value. This
is useful to perform WYSIWYG text.
linearVertAdvance: this field gives the linearly scaled advance
height for the glyph (relevant in vertical glyph
layouts only). This is useful to perform
WYSIWYG text.
Note that the two above field replace the removed "metrics2" field
in the glyph slot.
advance: this field is a vector that gives the transformed
advance for the glyph. By default, it corresponds
to the advance width, unless FT_LOAD_VERTICAL_LAYOUT
was specified when calling FT_Load_Glyph or FT_Load_Char
bitmap_left: this field gives the distance in integer pixels from
the current pen position to the left-most pixel of
a glyph image WHEN IT IS A BITMAP. It is only valid
when the "format" field is set to
"ft_glyph_format_bitmap", for example, after calling
the new function FT_Render_Glyph.
bitmap_top: this field gives the distance in integer pixels from
the current pen position (located on the baseline) to
the top-most pixel of the glyph image WHEN IT IS A
BITMAP. Positive values correspond to upwards Y.
loader: this is a new private field for the glyph slot. Client
applications should not touch it..
- support for transforms and direct rendering in FT_Load_Glyph:
Most of the functionality found in <freetype/ftglyph.h> has been
moved to the core library. Hence, the following:
- a transform can be specified for a face through FT_Set_Transform.
this transform is applied by FT_Load_Glyph to scalable glyph images
(i.e. NOT TO BITMAPS) before the function returns, unless the
bit flag FT_LOAD_IGNORE_TRANSFORM was set in the load flags..
- once a glyph image has been loaded, it can be directly converted to
a bitmap by using the new FT_Render_Glyph function. Note that this
function takes the glyph image from the glyph slot, and converts
it to a bitmap whose properties are returned in "face.glyph.bitmap",
"face.glyph.bitmap_left" and "face.glyph.bitmap_top". The original
native image might be lost after the conversion.
- when using the new bit flag FT_LOAD_RENDER, the FT_Load_Glyph
and FT_Load_Char functions will call FT_Render_Glyph automatically
when needed.
fixed:
. Removing unused variables.
. detect.mk files now must provide $(CONFIG_FILE) and not $(CONFIG_RULE).
. ansi.mk will now be really used as a fallback if the detect mechanism
fails.
. ANSIFLAGS will now be really used (fixing a typo).
. `make clean' now works (again two typos).
. Detection of gcc on Unix has been fixed (using the `-v' option instead of
`--version').
. `make devel' now works (on Unix).
. Fixing *again* a bug in demos/graph/x11/rules.mk to allow multiple use
of `-L' compiler options.
. $(BASE_H) now contains a few more header files.
As usual, a lot of formatting (not finished yet).
reflect the new modules/drivers list management performed through the
file `freetype2/config/modules.mk'
Changed the driver header files to reflect the new modules/drivers
list management. We get rid, at last, of the infamous pre-processor
tricks used to build the list at compile time.
`src/base/ftinit.c' is also modified to reflect the changes..
Werner Lemberg