Re: [Devel] Preparing the 2.1.5 release

[Werner LEMBERG]

Dear friends,


digging in old emails I've found the stuff attached below.  Are there
volunteers who want to look at some of the problems?


    Werner


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

The importance is from top to down.

. From the ChangeLog (compressed to a single entry):

    2003-06-04  Werner Lemberg  <address@hidden>

      * include/freetype/config/ftoption.h: Define
        TT_CONFIG_OPTION_BYTECODE_INTERPRETER and
        TT_CONFIG_OPTION_COMPILE_UNPATENTED_HINTING to make it the new
        default.

  What I really want for the default is the following: autohinting for
  everything except for glyphs which are rendered with
  FT_PARAM_TAG_UNPATENTED_HINTING, and patented hinting should not be
  available.  It should be rather straightforward to implement this.
  Otherwise, the bytecode interpreter should be disabled again by
  default.

. The native PS hinter sucks (based on ftview results).  There seems
  to be a fundamental bug somewhere: I have the feeling that hinting
  is done before scaling.  The bigger a glyph is in size, the larger
  are the effects of the hinter.  It should be exactly the opposite.
  Take any well-hinted font (e.g. the Type 1 version pf TeX's cmr10)
  and you'll see what I mean.

  Also the URW variant of Zapf Dingbats, uzdr.pfb, is rendered
  extremely badly if Type 1 hinting is performed.  On the other hand,
  autohinting produces quite fine results.

. The code for computing the glyph advance width of the autohinter has
  a minor bug IMHO.  Take for example pala.ttf, scale it at 239pt, and
  look at the glyphs `IJK'.  Switching on and off the autohinter makes
  the advance width jump very noticeable.  Now repeat the same while
  scaling at 240pt: No jumps at all.

. Perhaps related to the previous item is the patch from David
  Chester, sent to the list on Jul 24th, which I think is a real
  improvement to apply -- please check and commit if you approve.  I
  don't think it can remove the big advance width jumps since it
  handles the case where the correction is less than a single pixel.

. Another minor issue: The encoding issue of FNT fonts isn't solved
  yet.  Since it isn't generally possible to exactly determine the
  real encoding of a FNT font I suggest that we introduce new encoding
  tags to directly map the FNT character set byte without further
  interpretation.  A reliable synthetic Unicode mapping can only be
  done for a subset of the 7bit ASCII...

. Please also have a look at the following mails (no order of
  importance):

    03/06/14: [Devel] drop-out control
    03/06/07: [Devel] FT_LruList_Remove_Selection doesn't update
                      list->num_nodes
    03/06/11: [Devel] Split_Cubic and gray_split_cubic
    03/06/05: Re: [Devel] another face->postscript.isFixedPitch :-(
    03/05/03: [Devel] ad-hoc constants in ahglyph.c
    02/09/06: [Devel] Bug in ttinterp.c ???