freetype/docs/CHANGES

LATEST CHANGES BETWEEN 2.0.5 and 2.0.4

  NOTE THAT 2.0.5 DOES NOT CONTAIN THE POSTSCRIPT HINTER. THIS MODULE WILL
  BE PART OF THE NEXT RELEASE (EITHER 2.0.6 or 2.1)


  - fixed a bug that made certain glyphs, like "Cacute", "cacute" and "lslash"
    unavailable from Unicode charmaps of Postscript fonts. This prevented the
    correct display of Polish text, for example.

  - the kerning table of Type 1 fonts was loaded by FreeType, when its AFM
    file was attached to its face, but the FT_FACE_FLAG_HAS_KERNING bit flags
    was not set correctly, preventing FT_Get_Kerning to return meaningful
    values..
  
  - improved SFNT (TrueType & OpenType) charmap support. Slightly better
    performance, as well as support for the new formats defined by the
    OpenType 1.3 specification (8, 10 and 12)
  
  - fixed a serious typo in "src/base/ftcalc.c" which caused invalid
    computations in certain rare cases, producing ugly artefacts..
  
  - the size of the EM square is computed with a more accurate algorithm
    for Postscript fonts. The old one caused slight errors with embedded
    fonts found in PDF documents..

  - fixed a bug in the cache manager that prevented normal LRU behaviour
    within the cache manager, causing unecessary reloads (for FT_Face and
    FT_Size objects only)
  
  - added a new function named "FT_Get_Name_Index" to retrieve the glyph
    index of a given glyph name, when found in a face.

  - fixed a reallocation bug that generated a dangling pointer (and possibly
    memory leaks) with Postscript fonts (in src/psaux/psobjs.c)
    
  - many 16-bit correctness fixes
  
  - removed many pedantic compiler warnings from the sources

  - added an Amiga build directory in "builds/amiga"


============================================================================

LATEST CHANGES BETWEEN 2.0.4 and 2.0.3

  - Fixed a rather annoying bug that was introduced in 2.0.3.  Namely, the
    font transformation set through FT_Set_Transform was applied twice to
    auto-hinted glyphs, resulting in incorrectly rotated text output.
    
  - Fixed _many_ compiler warnings.  FT2 should now compile cleanly with
    Visual C++'s most pedantic warning level (/W4).  It already compiled
    fine with GCC and a few other compilers.

  - fixed a bug that prevented the linear advance width of composite
    TrueType glyphs to be correctly returned

  - Fixed the Visual C++ project files located in "builds/win32/visualc"
    (previous versions used older names of the library).
  
  - Many 32-bit constants have an "L" appended to their value, in order
    to improve the 16-bitness of the code.  Someone is actually trying to
    use FT2 on an Atari ST machine!

  - Updated the "builds/detect.mk" file in order to automatically build
    FT2 on AIX systems.  AIX uses "/usr/sbin/init" instead of "/sbin/init"
    and wasn't previously detected as a Unix platform by the FreeType build
    system.
                                                       
  - Updated the Unix-specific portions of the build system (new libtool
    version, etc.).

  - The SFNT kerning lodaer now ensures that the table is sorted (since some
    problem fonts do not meet this requirement).

============================================================================

LATEST CHANGES BETWEEN 2.0.3 and 2.0.2

  I. CHANGES TO THE MODULES / FONT DRIVERS
  
    - THE AUTO-HINTER HAS BEEN SLIGHTLY IMPROVED, in order to fix several
      annoying artefacts, mainly:
      
        - Blue zone alignement of horizontal stems wasn't performed
          correctly, resulting in artefacts like the "d" being placed one
          pixel below the "b" in some fonts like Time New Roman.
          
        - Overshoot thresholding wasn't performed correctly, creating
          unpleasant artefacts at large character pixel sizes.
          
        - Composite glyph loading has been simplified.  This gets rid
          of various artefacts where the components of a composite glyphs
          were not correctly spaced.
          
      These are the last changes to the current auto-hinting module.  A new
      hinting sub-system is currently in the work in order to support native
      hints in Type 1 / CFF / OpenType fonts, as well as globally improve
      rendering.

    - The PCF driver has been fixed.  It reported invalid glyph dimensions
      for the fonts available on Solaris.
    
    - The Type 1, CID and CFF drivers have been modified to fix the
      computation of the EM size.

    - The Type 1 driver has been fixed to avoid a dangerous bug that
      crashed the library with non-conforming fonts (i.e. ones that do not
      place the .notdef glyph at position 0).

    - The TrueType driver had a rather subtle bug (dangling pointer when
      loading composite glyphs) that could crash the library in rare
      occasions!


  II. HIGH-LEVEL API CHANGES
                              
    - The error code enumeration values have been changed.  An error value
      is decomposed in a generic error code, and a module number.  see
      <freetype/fterrors.h> for details.

    - A new public header file has been introduced, named FT_TRIGONOMETRY_H
      (include/freetype/fttrig.h), providing trigonometric functions to
      compute sines, cosines, arctangents, etc. with 16.16 fixed precision. 
      The implementation is based on the CORDIC algorithm and is very fast
      while being sufficiently accurate.


  III. INTERNALS
             
    - Added BeOS-specific files in the old build sub-system.  Note that
      no changes were required to compile the library with Jam.

    - The configuration is now capable of automatically detecting 64-bit
      integers on a set of predefined compilers (GCC, Visual C++, Borland
      C++) and will use them by default.  This provides a small performance
      boost.

    - A small memory leak that happened when opening 0-sized files (duh!)
      have been fixed.
          
    - Fixed bezier stack depth bug in the routines provided by the
      FT_BBOX_H header file.  Also fixed similar bugs in the rasterizers.

    - The outline bounding box code has been rewritten to use direct
      computations, instead of bezier sub-division, to compute the exact
      bounding box of glyphs.  This is slightly slower but more accurate.

    - The build system has been improved and fixed, mainly to support "make"
      on Windows 2000 correctly, avoid problems with "make distclean" on non
      Unix systems, etc.
      
    - Hexadecimal constants have been suffixed with "U" to avoid problems
      with certain compilers on 64-bit platforms.

    - A new directory named "src/tools" has been created.  It contains
      Python scripts and simple unit test programs used to develop the
      library.
      
    - The DocMaker tool has been moved from "docs" to "src/tools" and
      has been updated with the following:
      
         - Now accepts the "--title=XXXX" or "-t XXXX" option from the
           command line to set the project's name in the generated API
           reference.
           
         - Now accepts the "--output=DIR" or "-o DIR" option from the
           command line to set the output directory for all generated HTML
           files.
           
         - Now accepts the "--prefix=XXXX" or "-p XXX" option from the
           command line to set the file prefix to use for all generated HTML
           files.

         - Now generates the current time/data on each generated page
           in order to distinguish between versions.

      DocMaker can be used with other projects now, not only FT2 (e.g. MLib,
      FTLayout, etc.).
                                                                
===========================================================================

LATEST CHANGES BETWEEN 2.0.2 and 2.0.1


  I. CHANGES TO THE MODULES / FONT DRIVERS

    - THE TRUETYPE BYTECODE INTERPRETER IS NOW TURNED OFF, in order to
      avoid legal problems  with the Apple patents.  It  seems that we
      mistakenly  turned this option  on in  previous releases  of the
      build.

      Note that if  you want to use the  bytecode interpreter in order
      to get high-quality TrueType  rendering, you will need to toggle
      by        hand        the        definition        of        the
      TT_CONFIG_OPTION_BYTECODE_INTERPRETER   macro    in   the   file
      "include/freetype/config/ftoption.h".

    - The CFF driver has been improved by Tom Kacvinsky and Sander van
      der Wal:

      * Support for "seac" emulation.
      * Support for "dotsection".
      * Support for retrieving glyph names through
        "FT_Get_Glyph_Name".

      The first two items are necessary to correctly a large number of
      Type 1 fonts converted to the CFF formats by Adobe Acrobat.

    - The Type 1 driver was also improved by Tom & others:

      * Better EM size computation.
      * Better support for synthetic (transformed) fonts.
      * The  Type 1  driver returns  the charstrings  corresponding to
        each glyph in the  "glyph->control_data" field after a call to
        "FT_Load_Glyph" (thanks Ha Shao).

    - Various other bugfixes, including the following:

      * Fixed a nasty memory leak in the Type 1 driver.
      * The autohinter  and the pcf  driver used static  writable data
        when they shouldn't.
      * Many casts were added to  make the code more 64-bits safe.  It
        also now compiles on Windows XP 64-bits without warnings.
      * Some incorrect writable statics were removed in the "autohint"
        and "pcf" drivers.  FreeType 2 now compiles on Epoc again.


  II. CHANGES TO THE HIGH-LEVEL API

    - The library header files inclusion scheme has been changed.  The
      old scheme looked like:

        #include <freetype/freetype.h>
        #include <freetype/ftglyph.h>
        #include <freetype/ftcache.h>
        #include <freetype/cache/ftimage.h>

      Now you should use:

        #include <ft2build.h>
        #include FT_FREETYPE_H
        #include FT_GLYPH_H
        #include FT_CACHE_H
        #include FT_CACHE_IMAGE_H

      NOTE THAT  THE OLD  INCLUSION SCHEME WILL  STILL WORK  WITH THIS
      RELEASE.  HOWEVER, WE  DO NOT GUARANTEE THAT THIS  WILL STILL BE
      TRUE IN THE NEXT ONE (A.K.A. FREETYPE 2.1).

      The  file <ft2build.h>  is used  to define  the  header filename
      macros.  The complete and  commented list of macros is available
      in the API reference under the section name "Header File Macros"
      in Chapter I.

      For more information, see section I of the following document:

        http://www.freetype.org/
          freetype2/docs/tutorial/step1.html

      or

        http://freetype.sourceforge.net/
          freetype2/docs/tutorial/step1.html

    - Many, many comments have been added to the public source file in
      order to  automatically generate  the API Reference  through the
      "docmaker.py" Python script.

      The latter has been updated  to support the grouping of sections
      in chapters and better index sort.  See:

        http://www.freetype.org/freetype2/docs/reference/ft2-toc.html


  III. CHANGES TO THE BUILD PROCESS

    - If you  are not  building FreeType 2  with its own  build system
      (but with your own Makefiles or project files), you will need to
      be  aware that  the  build  process has  changed  a little  bit.

      You don't  need to put the  "src" directory in  the include path
      when  compiling  any FT2  component.   Instead,  simply put  the
      component's directory in the current include path.

      So, if you were doing something like:

        cc -c -Iinclude -Isrc src/base/ftbase.c

      change the line to:

        cc -c -Iinclude -Isrc/base src/base/ftbase.c

      If you were doing something like:

        cd src/base
        cc -c -I../../include -I.. ftbase.c

      change it to:

        cd src/base
        cc -c -I../../include ftbase.c


======================================================================

LATEST CHANGES BETWEEN 2.0.1 and 2.0

  2.0.1 introduces a few changes:

    - Fixed many bugs related to  the support of CFF / OpenType fonts.
      These  formats are  now much  better supported  though  there is
      still work planned to  deal with charset tables and PDF-embedded
      CFF files that use the old "seac" command.

    - The  library could not  be compiled  in debug  mode with  a very
      small  number   of  C  compilers   whose  pre-processors  didn't
      implement the "##"  directive correctly (i.e. per se  the ANSI C
      specification!)  An elegant fix was found.

    - Added  support for  the  free Borland  command-line C++  Builder
      compiler.   Use "make  setup bcc32".   Also fixed  a  few source
      lines that generated new warnings with BCC32.

    - Fixed a bug in FT_Outline_Get_BBox when computing the extrema of
      a conic Bezier arc.

    - Updated the INSTALL file to add IDE compilation.

    - Other  minor bug  fixes,  from  invalid Type  1  style flags  to
      correct   support   of  synthetic   (obliqued)   fonts  in   the
      auto-hinter, better support for embedded bitmaps in a SFNT font.

    - Fixed some problems with "freetype-config".

  Finally, the "standard" scheme for including FreeType headers is now
  gradually changing,  but this will  be explained in a  later release
  (probably 2.0.2).

  And very  special thanks to Tom Kacvinsky  and YAMANO-UCHI Hidetoshi
  for their contributions!


======================================================================

CHANGES BETWEEN beta8 and 2.0

  - Changed  the default  installation  path for  public headers  from
    "include/freetype" to "include/freetype2".

    Also added a new "freetype-config" that is automatically generated
    and installed  on Unix and  Cygwin systems.  The script  itself is
    used to retrieve the current  install path, C compilation flags as
    well as linker flags.

  - Fixed several small bugs:

    * Incorrect max advance width for fixed-pitch Type 1 fonts.
    * Incorrect glyph names for certain TrueType fonts.
    * The  glyph advance  was not  copied when  FT_Glyph_To_Bitmap was
      called.
    * The  linearHoriAdvance  and  linerVertAdvance  fields  were  not
      correctly returned for glyphs processed by the auto-hinter.
    * "type1z"  renamed back to  "type1"; the  old "type1"  module has
      been removed.

  - Revamped the  build system  to make it  a lot more  generic.  This
    will  allow us  to  re-use  nearly un-modified  in  lots of  other
    projects (including FreeType Layout).

  - Changed "cid" to use "psaux" too.

  - Added the  cache sub-system.  See <freetype/ftcache.h>  as well as
    the sources  in "src/cache".  Note  that it compiles but  is still
    untested for now.

  - Updated "docs/docmaker.py", a draft  API reference is available at
    http://www.freetype.org/ft2api.html.

  - Changed "type1" to use "psaux".

  - Created a  new module named  "psaux" to hold  the Type 1 &  Type 2
    parsing routines.  It should be  used by "type1", "cid", and "cff"
    in the future.

  - Fixed an important bug in "FT_Glyph_Get_CBox".

  - Fixed  some compiler  warnings  that happened  since the  TrueType
    bytecode decoder was deactivated by default.

  - Fixed two memory leaks:

    * The    memory   manager   (16    bytes)   isn't    released   in
      FT_Done_FreeType!
    * Using custom input streams, the  copy of the original stream was
      never released.

  - Fixed the  auto-hinter by performing automatic  computation of the
    "filling direction" of each glyph.   This is done through a simple
    and  fast approximation, and  seems to  work (problems  spotted by
    Werner though).  The Arphic fonts are a lot nicer though there are
    still a lot of things to do to handle Asian fonts correctly.


======================================================================

BETA-8 (RELEASE CANDIDATE) CHANGES

  - Deactivated the TrueType bytecode interpreter by default.

  - Deactivated the "src/type1" font driver.  Now "src/type1z" is used
    by default.

  - Updates to the build system.  We now compile the library correctly
    under  Unix  system  through  "configure" which  is  automatically
    called on the first "make" invocation.

  - Added the auto-hinting module!  Fixing some bugs here and there.

  - Found some bugs in the  composite loader (seac) of the Type1-based
    font drivers.

  - Renamed the directory "freetype2/config" to "freetype2/builds" and
    updated all relevant files.

  - Found a memory leak in the "type1" driver.

  - Incorporated Tom's patches to  support flex operators correctly in
    OpenType/CFF fonts.  Now all I need is to support pure CFF and CEF
    fonts to be done with this driver :-)

  - Added the  Windows FNT/FON driver in "src/winfonts".   For now, it
    always  "simulates"   a  Unicode  charmap,  so   it  shouldn't  be
    considered completed right now.

    It  is there  to be  more a  proof of  concept than  anything else
    anyway.  The driver is a single  C source file, that compiles to 3
    Kb of code.

    I'm  still working on  the PCF/BDF  drivers, but  I'm too  lazy to
    finish them now.

  - CHANGES TO THE HIGH-LEVEL API

    * FT_Get_Kerning has a new parameter that allows you to select the
      coordinates of the kerning  vector (font units, scaled, scaled +
      grid-fitted).
    * The  outline functions are  now in <freetype/ftoutln.h>  and not
      part of <freetype/freetype.h> anymore.
    * <freetype/ftmodule.h>    now     contains    declarations    for
       FT_New_Library, FT_Done_Library, FT_Add_Default_Modules.
    * The so-called convenience  functions have moved from "ftoutln.c"
      to  "ftglyph.c",  and  are  thus available  with  this  optional
      component    of   the   library.     They   are    declared   in
      <freetype/ftglyph.h> now.
    * Anti-aliased  rendering is now  the default  for FT_Render_Glyph
      (i.e. corresponds to render_mode == 0 == ft_render_mode_normal).
      To generate a monochrome bitmap, use ft_render_mode_mono, or the
      FT_LOAD_MONOCHROME     flag    in    FT_Load_Glyph/FT_Load_Char.
      FT_LOAD_ANTI_ALIAS is still defined, but values to 0.
    * <freetype/freetype.h>  now include <freetype/config/ftconfig.h>,
      solving a few headaches :-)
    * The type FT_GlyphSlotRec has now a "library" field.

  - CHANGES TO THE "ftglyph.h" API

    This API has  been severely modified in order  to make it simpler,
    clearer, and more  efficient.  It certainly now looks  like a real
    "glyph factory"  object, and allows client  applications to manage
    (i.e.  transform,  bbox  and  render) glyph  images  without  ever
    knowing their original format.

  - Added  support  for CID-keyed  fonts  to  the  CFF driver.   Maybe
    support for pure CFF + CEF fonts should come in?

  - Cleaned up  source code in order  to avoid two  functions with the
    same name.  Also  changed the names of the  files in "type1z" from
    "t1XXXX" to "z1XXXX" in order to avoid any conflicts.

    "make multi" now works well :-)

    Also removed the use of "cidafm" for now, even if the source files
    are  still there.  This  functionality will  certainly  go into  a
    specific module.

  - ADDED SUPPORT FOR THE AUTO-HINTER

    It  works :-) I  have a  demo program  which simply  is a  copy of
    "ftview"       that      does       a      `FT_Add_Module(library,
    &autohinter_module_class)' after  library initialization, and Type
    1 & OpenType/CFF fonts are now hinted.

    CID  fonts are  not hinted,  as they  include no  charmap  and the
    auto-hinter doesn't include  "generic" global metrics computations
    yet.

    Now, I need to release this thing to the FreeType 2 source.

  - CHANGES TO THE RENDERER MODULES

    The  monochrome  and smooth  renderers  are  now  in two  distinct
    directories, namely "src/raster1" and "src/smooth".  Note that the
    old "src/renderer" is now gone.

    I ditched  the 5-gray-levels renderers.  Basically,  it involved a
    simple #define toggle in 'src/raster1/ftraster.c'.

    FT_Render_Glyph,  FT_Outline_Render  &  FT_Outline_Get_Bitmap  now
    select the best renderer  available, depending on render mode.  If
    the current renderer for a  given glyph image format isn't capable
    of supporting  the render mode, another  one will be  found in the
    library's list.   This means that client applications  do not need
    to  switch or  set  the  renderers themselves  (as  in the  latest
    change), they'll get what they want automatically.  At last.

    Changed the demo programs accordingly.

  - MAJOR INTERNAL REDESIGN:

    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.

      he  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 has 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
        IF 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 IF 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.

  - Reformatted all  modules source  code in order  to get rid  of the
    basic data types redifinitions (i.e. "TT_Int" instead of "FT_Int",
    "T1_Fixed"  instead  of  "FT_Fixed").  Hence  the  format-specific
    prefixes like  "TT_", "T1_",  "T2_" and "CID_"  are only  used for
    relevant structures.


======================================================================

OLD CHANGES FOR BETA 7

  - bug-fixed the  OpenType/CFF parser. It  now loads and  displays my
    two  fonts nicely,  but I'm  pretty certain  that more  testing is
    needed :-)

  - fixed the crummy Type 1 hinter, it now handles accented characters
    correctly (well, the accent is  not always well placed, but that's
    another problem..)

  - added the CID-keyed Type 1  driver in "src/cid". Works pretty well
    for only 13 Kb of code  ;-) Doesn't read AFM files though, nor the
    really useful CMAP files..

  - fixed two bugs in the smooth renderer (src/base/ftgrays.c). Thanks
    to Boris Letocha for spotting them and providing a fix..

  - fixed potential "divide by zero" bugs in ftcalc.c.. my god..

  - added source  code for  the OpenType/CFF driver  (still incomplete
    though..)

  - modified the  SFNT driver slightly  to perform more  robust header
    checks  in TT_Load_SFNT_Header. This  prevents certain  font files
    (e.g.  some  Type  1  Multiple  Masters)  from  being  incorrectly
    "recognized" as TrueType font files..

  - moved a lot of stuff from  the TrueType driver to the SFNT module,
    this   allows   greater   code   re-use   between   font   drivers
    (e.g. TrueType, OpenType, Compact-TrueType, etc..)

  - added a tiny segment cache to the SFNT Charmap 4 decoder, in order
    to minimally speed it up..

  - added support for Multiple Master fonts in "type1z". There is also
    a  new file  named  <freetype/ftmm.h> which  defines functions  to
    manage them from client applications.

    The new file "src/base/ftmm.c" is also optional to the engine..

  - various  formatting changes (e.g.  EXPORT_DEF ->  FT_EXPORT_DEF) +
    small bug fixes in FT_Load_Glyph, the "type1" driver, etc..

  - a minor fix to the Type 1 driver to let them apply the font matrix
    correctly (used for many oblique fonts..)

  - some fixes for 64-bit systems (mainly changing some FT_TRACE calls
    to use %p instead of %lx).. Thanks to Karl Robillard

  - fixed  some bugs  in  the sbit  loader (src/base/sfnt/ttsbit.c)  +
    added  a new flag,  FT_LOAD_CROP_BITMAP to  query that  bitmaps be
    cropped when  loaded from a file  (maybe I should  move the bitmap
    cropper to the base layer ??).

  - changed the default  number of gray levels of  the smooth renderer
    to 256  (instead of  the previous 128).  Of course, the  human eye
    can't see any difference ;-)

  - removed TT_MAX_SUBGLYPHS,  there is no static limit  on the number
    of subglyphs in a TrueType font now..


======================================================================

OLD CHANGES 16 May 2000

  - tagged "BETA-6"  in the  CVS tree. This  one is a  serious release
    candidate even though it doesn't incorporate the auto-hinter yet..

  - various obsolete files were removed, and copyright header updated

  - finally  updated  the  standard   raster  to  fix  the  monochrome
    rendering bug + re-enable  support for 5-gray levels anti-aliasing
    (suck, suck..)

  - created new header files, and modified sources accordingly:

     <freetype/fttypes.h>
       - simple FreeType types, without the API
     <freetype/internal/ftmemory.h>
       - definition of memory-management macros

  - added   the   "DSIG"   (OpenType   Digital   Signature)   tag   to
    <freetype/tttags.h>

  - light update/cleaning of the build system + changes to the sources
    in  order  to  get  rid  of _all_  compiler  warnings  with  three
    compilers, i.e:

    gcc with "-ansi -pedantic -Wall -W", Visual C++ with "/W3 /WX" and
    LCC

    IMPORTANT NOTE FOR WIN32-LCC USERS:
    |
    |  It seems the C pre-processor  that comes with LCC is broken, it
    |  doesn't  recognize  the  ANSI  standard  directives  #  and  ##
    |  correctly when one of the  argument is a macro. Also, something
    |  like:
    |
    |     #define F(x)  print##x
    |
    |     F(("hello"))
    |
    |  will get incorrectly translated to:
    |
    |     print "hello")
    |
    |  by its pre-processor. For  this reason, you simply cannot build
    |  FreeType 2 in debug mode with this compiler..

  - yet another massive grunt work. I've changed the definition of the
    EXPORT_DEF,  EXPORT_FUNC, BASE_DEF &  BASE_FUNC macros.  These now
    take an argument, which is the function's return value type.

    This  is necessary to  compile FreeType  as a  DLL on  Windows and
    OS/2.  Depending on the compiler used, a compiler-specific keyword
    like  __export or __system  must be  placed before  (VisualC++) or
    after (BorlandC++) the type..

    Of course, this needed a lot of changes throughout the source code
    to make it compile again... All cleaned up now, apparently..

    Note also  that there is a  new EXPORT_VAR macro  defined to allow
    the   _declaration_    of   an   exportable    public   (constant)
    variable.  This  is  the   case  of  the  raster  interfaces  (see
    ftraster.h and ftgrays.h), as well as each module's interface (see
    sfdriver.h, psdriver.h, etc..)

  - new feature: it  is now possible to pass  extra parameters to font
                 drivers  when creating  a new  face object.  For now,
                 this capability is unused.  It could however prove to
                 be useful in a near future..

      the FT_Open_Args structure was  changes, as well as the internal
      driver interface  (the specific "init_face"  module function has
      now a different signature).

  - updated the tutorial (not finished though).

  - updated the top-level BUILD  document

  - fixed  a  potential memory  leak  that  could  occur when  loading
    embedded bitmaps.

  - added     the     declaration     of     FT_New_Memory_Face     in
    <freetype/freetype.h>, as  it was  missing from the  public header
    (the implementation was already in "ftobjs.c").

  - the file <freetype/fterrors.h> has been seriously updated in order
    to allow the automatic generation of error message tables. See the
    comments within it for more information.

  - major directory  hierarchy re-organisation. This was  done for two
    things:

      * first,  to ease  the "manual"  compilation of  the  library by
        requiring at lot less include paths :-)

      * second,  to  allow  external  programs to  effectively  access
        internal  data  fields. For  example,  this  can be  extremely
        useful if  someone wants  to write a  font producer or  a font
        manager on top of FreeType.

    Basically, you  should now use  the 'freetype/' prefix  for header
    inclusion, as in:

        #include <freetype/freetype.h>
        #include <freetype/ftglyph.h>

    Some new include sub-directories are available:

     a. the  "freetype/config" directory,  contains two files  used to
        configure the build of the library. Client applications should
        not need to look at these normally, but they can if they want.

        #include <freetype/config/ftoption.h>
        #include <freetype/config/ftconfig.h>

     b. the "freetype/internal"  directory, contains header files that
        describes library  internals. These are the  header files that
        were  previously  found  in  the "src/base"  and  "src/shared"
        directories.


    As  usual, the build  system and  the demos  have been  updated to
    reflect the change..

    Here's a layout of the new directory hierarchy:

    TOP
      include/
         freetype/
            freetype.h
            ...
            config/
              ftoption.h
              ftconfig.h
              ftmodule.h

            internal/
              ftobjs.h
              ftstream.h
              ftcalc.h
              ...

      src/
         base/
            ...

         sfnt/
         psnames/
         truetype/
         type1/
         type1z/


    Compiling a module is now  much easier, for example, the following
    should work when in the TOP directory on an ANSI build:

       gcc -c -I./include -I./src/base src/base/ftbase.c
       gcc -c -I./include -I./src/sfnt src/sfnt/sfnt.c
       etc..

    (of course, using -Iconfig/<system> if you provide system-specific
     configuration files).

  - updated the structure of FT_Outline_Funcs in order to allow direct
    coordinate scaling within  the outline decomposition routine (this
    is  important for virtual  "on" points  with TrueType  outlines) +
    updates to the rasters to support this..

  - updated  the OS/2  table  loading code  in "src/sfnt/ttload.c"  in
    order to support version 2 of the table (see OpenType 1.2 spec)

  - created  "include/tttables.h"  and  "include/t1tables.h" to  allow
    client applications to access some of  the SFNT and T1 tables of a
    face  with  a  procedural  interface (see  FT_Get_Sfnt_Table())  +
    updates to internal source files to reflect the change..

  - some  cleanups in  the source  code to  get rid  of  warnings when
    compiling with the "-Wall -W -ansi -pedantic" options in gcc.

  - debugged and moved the smooth renderer to "src/base/ftgrays.c" and
    its header to "include/ftgrays.h"

  - updated TT_MAX_SUBGLYPHS  to 96 as some CJK  fonts have composites
    with up to 80 sub-glyphs !! Thanks to Werner


======================================================================

OLD CHANGES - 14-apr-2000

  - fixed  a bug  in  the  TrueType glyph  loader  that prevented  the
    correct loading of some CJK glyphs in mingli.ttf

  - improved the standard Type 1 hinter in "src/type1"

  - fixed two bugs  in the experimental Type 1  driver in "src/type1z"
    to handle the new XFree86 4.0 fonts (and a few other ones..)

  - the smooth  renderer is now  complete and supports  sub-banding to
    render large glyphs at high speed. However, it is still located in
    "demos/src/ftgrays.c" and should move to the library itself in the
    next  beta..   NOTE:  The  smooth  renderer   doesn't  compile  in
    stand-alone mode anymore, but this should be fixed RSN..

  - introduced convenience  functions to  more easily deal  with glyph
    images, see  "include/ftglyph.h" for more details, as  well as the
    new  demo program  named "demos/src/ftstring.c"  that demonstrates
    its use

  - implemented  FT_LOAD_NO_RECURSE in  both the  TrueType and  Type 1
    drivers  (this  is required  by  the  auto-hinter  to improve  its
    results).

  - changed   the  raster   interface,  in   order  to   allow  client
    applications to provide their own span-drawing callbacks. However,
    only the smooth renderer  supports this. See "FT_Raster_Params" in
    the file "include/ftimage.h"

  - fixed  a small bug  in FT_MulFix  that caused  incorrect transform
    computation!

  - Note: The tutorial is out-of-date, grumpf.. :-(


======================================================================

OLD CHANGES - 12-mar-2000

  - changed  the  layout  of  configuration  files  :  now,  all  ANSI
    configuration         files         are         located         in
    "freetype2/config".  System-specific over-rides  can be  placed in
    "freetype2/config/<system>".

  - moved all configuration macros to "config/ftoption.h"

  - improvements in the Type 1 driver with AFM support

  - changed the fields  in the FT_Outline structure :  the old "flags"
    array is re-named "tags", while all ancient flags are encoded into
    a single unsigned int named "flags".

  - introduced     new      flags     in     FT_Outline.flags     (see
    ft_outline_.... enums in "ftimage.h").

  - changed outline functions to "FT_Outline_<action>" syntax

  - added a smooth anti-alias renderer to the demonstration programs

  - added Mac graphics driver (thanks Just)

  - FT_Open_Face  changed  in  order   to  received  a  pointer  to  a
    FT_Open_Args descriptor..

  - various  cleanups,  a  few  more API  functions  implemented  (see
    FT_Attach_File)

  - updated some docs


======================================================================

OLD CHANGES - 22-feb-2000

  - introduced the "psnames" module. It is used to:

      o convert  a Postscript glyph  name into the  equivalent Unicode
        character code (used by the  Type 1 driver(s) to synthetize on
        the fly a Unicode charmap).

      o provide an  interface to retrieve the Postscript  names of the
        Macintosh,  Adobe  Standard &  Adobe  Expert character  codes.
        (the Macintosh  names are  used by the  SFNT-module postscript
        names support routines, while the other two tables are used by
        the Type 1 driver(s)).

  - introduced the "type1z" alternate Type  1 driver. This is a (still
    experimental) driver  for the Type  1 format that  will ultimately
    replace the one  in "src/type1". It uses pattern  matching to load
    data from the  font, instead of a finite  state analyzer. It works
    much better than the "old"  driver with "broken" fonts. It is also
    much smaller (under 15 Kb).

  - the  Type 1  drivers (both  in "src/type1"  and  "src/type1z") are
    nearly  complete.  They  both  provide automatic  Unicode  charmap
    synthesis through  the "psnames" module. No  re-encoding vector is
    needed.  (note  that they  still  leak  memory  due to  some  code
    missing, and I'm getting lazy).

    Trivial AFM support has been added to read kerning information but
    wasn't exactly tested as it should ;-)

  - The TrueType  glyph loader has  been seriously rewritten  (see the
    file  "src/truetype/ttgload.c". It  is now  much, much  simpler as
    well as  easier to read,  maintain and understand  :-) Preliminary
    versions introduced a  memory leak that has been  reported by Jack
    Davis, and is now fixed..

  - introduced  the new  "ft_glyph_format_plotter", used  to represent
    stroked outlines  like Windows "Vector" fonts, and  certain Type 1
    fonts  like "Hershey".  The corresponding  raster will  be written
    soon.

  - FT_New_Memory_Face  is  gone.  Likewise,  FT_Open_Face has  a  new
    interface that uses a structure  to describe the input stream, the
    driver (if required), etc..


TODO

  - Write FT_Get_Glyph_Bitmap and FT_Load_Glyph_Bitmap

  - Add a function like FT_Load_Character( face, char_code, load_flags
    )  that  would really  embbed  a  call  to FT_Get_Char_Index  then
    FT_Load_Glyph to ease developer's work.

  - Update the tutorial!

  - consider adding  support for Multiple  Master fonts in the  Type 1
    drivers.

  - Test the AFM routines of the  Type 1 drivers to check that kerning
    information is returned correctly.

  - write a decent auto-gridding component  !! We need this to release
    FreeType 2.0 gold !


less urgent needs:

  - add a CFF/Type2 driver
  - add a BDF driver
  - add a FNT/PCF/HBF driver
  - add a Speedo driver from the X11 sources


======================================================================

OLDER CHANGES - 27-jan-2000

  - updated the  "sfnt" module  interface to allow  several SFNT-based
    drivers to co-exist peacefully

  - updated  the "T1_Face"  type  to better  separate Postscript  font
    content  from the  rest of  the FT_Face  structure. Might  be used
    later by the CFF/Type2 driver..

  - added an experimental replacement Type 1 driver featuring advanced
    (and speedy) pattern matching to retrieve the data from postscript
    fonts.

  - very minor  changes in the implementation  of FT_Set_Char_Size and
    FT_Set_Pixel_Sizes (they now implement default to ligthen the font
    driver's code).


======================================================================

OLD MESSAGE

This file summarizes the changes that occured since the last "beta" of
FreeType 2.  Because  the list is important, it  has been divided into
separate sections:

Table Of Contents:

    I   High-Level Interface (easier !)
   II   Directory Structure
  III   Glyph Image Formats
   IV   Build System
    V   Portability
   VI   Font Drivers


----------------------------------------------------------------------

High-Level Interface:

  The high-level API has been considerably simplified.  Here is how:

    - resource objects have disappeared.  this means that face objects
      can now be created with  a single function call (see FT_New_Face
      and FT_Open_Face)

    - when calling  either FT_New_Face  & FT_Open_Face, a  size object
      and a glyph slot object  are automatically created for the face,
      and can  be accessed  through "face->glyph" and  "face->size" if
      one really  needs to.   In most cases,  there's no need  to call
      FT_New_Size or FT_New_Glyph.

    - similarly,  FT_Load_Glyph  now  only  takes  a  "face"  argument
      (instead  of a  glyph  slot  and a  size).  Also, it's  "result"
      parameter is  gone, as the glyph  image type is  returned in the
      field "face->glyph.format"

    - the list  of available  charmaps is directly  accessible through
      "face->charmaps",  counting "face->num_charmaps"  elements. Each
      charmap  has an  'encoding'  field which  specifies which  known
      encoding it deals with. Valid values are, for example :

          ft_encoding_unicode      (for ASCII, Latin-1 and Unicode)
          ft_encoding_apple_roman
          ft_encoding_sjis
          ft_encoding_adobe_standard
          ft_encoding_adobe_expert

      other  values may  be added  in the  future. Each  charmap still
      holds  its "platform_id"  and "encoding_id"  values in  case the
      encoding is too exotic for the current library


----------------------------------------------------------------------

Directory Structure:

  Should seem obvious to most of you:

     freetype/
         config/        -- configuration sub-makefiles
            ansi/
            unix/       -- platform-specific configuration files
            win32/
            os2/
            msdos/

         include/       -- public header  files, those to  be included
                           directly by client apps

         src/           -- sources of the library
           base/        -- the base layer
           sfnt/        -- the sfnt "driver"  (see the drivers section
                           below)
           truetype/    -- the truetype driver
           type1/       -- the type1 driver
           shared/      -- some header files shared between drivers

         demos/         -- demos/tools

         docs/          -- documentation (a bit empty for now)


----------------------------------------------------------------------

Glyph Image Formats:

  Drivers are now able to  register new glyph image formats within the
  library.  For  now, the  base layer supports  of course  bitmaps and
  vector  outlines, but  one  could imagine  something different  like
  colored bitmaps, bi-color vectors or wathever else (Metafonts anyone
  ??).

  See  the   file  `include/ftimage.h'.   Note  also  that   the  type
  FT_Raster_Map  is gone,  and  is now  replaced  by FT_Bitmap,  which
  should encompass all known bitmap types.

  Each new  image format  must provide at  least one "raster",  i.e. a
  module capable of  transforming the glyph image into  a bitmap. It's
  also possible  to change the default  raster used for  a given glyph
  image format.

  The default outline  scan-converter now uses 128 levels  of grays by
  default,  which tends  to smooth  many  things. Note  that the  demo
  programs have been updated significantly in order to display these..


----------------------------------------------------------------------

Build system:

  You still need  GNU Make to build the library.  The build system has
  been very seriously re-vamped in order to provide things like :

   - automatic host platform  detection (reverting to 'config/ansi' if
     it is not detected, with pseudo-standard compilation flags)

   - the ability to compile from the Makefiles with very different and
     exotic compilers. Note that  linking the library can be difficult
     for some platforms.

     For example, the file `config/win32/lcclib.bat' is invoked by the
     build system to create the ".lib" file with LCC-Win32 because its
     librarian  has too  many flaws  to be  invoked directly  from the
     Makefile.

  Here's how it works:

  - the first time you type `make',  the build system runs a series of
    sub-makefiles in order to detect your host platform. It then dumps
    what  it found,  and  creates  a file  called  `config.mk' in  the
    current  directory. This  is a  sub-Makefile used  to  define many
    important Make variables used to build the library.

  - the second time, the build system detects the `config.mk' then use
    it  to  build the  library.  All object  files  go  into 'obj'  by
    default,  as well  as the  library file,  but this  can  easily be
    changed.

  Note that  you can run "make  setup" to force  another host platform
  detection  even   if  a  `config.mk'  is  present   in  the  current
  directory.  Another solution  is  simply to  delete  the file,  then
  re-run make.

  Finally, the  default compiler  for all platforms  is gcc  (for now,
  this will hopefully changed in  the future). You can however specify
  a different  compiler by specifying  it after the 'setup'  target as
  in:

      gnumake setup lcc         on Win32 to use the LCC compiler
      gnumake setup visualc     on Win32 to use Visual C++

  See  the file  `config/<system>/detect.mk' for  a list  of supported
  compilers for your platforms.

  It should be relatively easy  to write new detection rules files and
  config.mk..

  Finally, to  build the demo programs,  go to `demos'  and launch GNU
  Make, it will use the `config.mk'  in the top directory to build the
  test programs..


----------------------------------------------------------------------

Portability:

  In  the  previous  beta,  a  single FT_System  object  was  used  to
  encompass  all  low-level  operations like  thread  synchronisation,
  memory management and i/o access. This has been greatly simplified:

    - thread synchronisation  has been dropped, for  the simple reason
      that the library  is already re-entrant, and that  if you really
      need  two  threads accessing  the  same  FT_Library, you  should
      really synchronize access to it yourself with a simple mutex.

    - memory  management is  performed  through a  very simple  object
      called "FT_Memory",  which really is a table  containing a table
      of pointers to  functions like malloc, realloc and  free as well
      as some user data (closure).

    - resources have disappeared (they created more problems than they
      solved), and  i/o management have  been simplified greatly  as a
      result. Streams are defined through FT_Stream objects, which can
      be either memory-based or disk-based.

      Note that  each face  has its own  stream, which is  closed only
      when  the  face object  is  destroyed.  Hence,  a function  like
      TT_Flush_Face in  1.x cannot be directly  supported. However, if
      you really need something like  this, you can easily tailor your
      own streams  to achieve the same  feature at a  lower level (and
      use FT_Open_Face instead of FT_New_Face to create the face).

  See the file  "include/ftsystem.h" for more details, as  well as the
  implementations found in "config/unix" and "config/ansi".


----------------------------------------------------------------------

Font Drivers:

  The  Font Driver  interface has  been modified  in order  to support
  extensions & versioning.


  The  list of  the font  drivers that  are statically  linked  to the
  library at compile time is  managed through a new configuration file
  called `config/<platform>/ftmodule.h'.

  This file is autogenerated when invoking `make modules'. This target
  will parse  all sub-directories of 'src', looking  for a "module.mk"
  rules file, used to describe the driver to the build system.

  Hence, one  should call  `make modules' each  time a font  driver is
  added or removed from the `src' directory.

  Finally, this version provides a "pseudo-driver" in `src/sfnt'. This
  driver doesn't  support font  files directly, but  provides services
  used by  all TrueType-like font  drivers. Hence, its code  is shared
  between  the TrueType  & OpenType  font formats,  and  possibly more
  formats to come if we're lucky..


----------------------------------------------------------------------

Extensions support:

  The extensions support is inspired by the one found in 1.x.

  Now, each font driver has  its own "extension registry", which lists
  which extensions  are available  for the font  faces managed  by the
  driver.

  Extension ids are  now strings, rather than 4-byte  tags, as this is
  usually more readable..

  Each extension has:
    - some data, associated to each face object
    - an interface (table of function pointers)

  An extension  that is format-specific should  simply register itself
  to the correct font driver. Here is some example code:

   // Registering an extensions
   //
   FT_Error  FT_Init_XXXX_Extension( FT_Library  library )
   {
     FT_DriverInterface*  tt_driver;

     driver = FT_Get_Driver( library, "truetype" );
     if (!driver) return FT_Err_Unimplemented_Feature;

     return FT_Register_Extension( driver, &extension_class );
   }


   // Implementing the extensions
   //
   FT_Error  FT_Proceed_Extension_XXX( FT_Face  face )
   {
     FT_XXX_Extension            ext;
     FT_XXX_Extension_Interface  ext_interface;

     ext = FT_Get_Extension( face, "extensionid", &ext_interface );
     if (!ext) return error;

     return ext_interface->do_it(ext);
   }

--- end of CHANGES ---