Re: [Devel] bug with MSung-Light-Acro

[David Turner]

Hello,

  It seems after a first analysis that glyph 2028 has a buggy charstring,
  it uses "cntrmask" and "hintmask" operator with invalid bit counts.

  The previous version of the Postscript hinter returned an error when trying
  to process such glyphs. I have modified it to simply ignore faulty operators
  (though it reports an error through FT_ERROR((...))) and it seems to work
  well without distorting the glyph..

  I'll have a deeper look at this problem later. It's now late in Paris
  and I'll probably won't have much time this week-end to work on this
  topic..

  So, merry christmas to you all, and see you on Wednesday, for those
  still on the list ;-)

Cheers,

- David



> > The "double" glyph could come from having glyph 2028 not being properly
> > rendered (for a yet-to-be-determined reason), and the code using a dangling
> > pointer to a previously rendered glyph (i.e. 2027). I wouldn't be surprised
> > if the garbage that appears was due to heap allocated data accessed through
> > the pointer, overwriting freed memory..
> >
> > It's just an hypothesis but it's the best I can come with for now. Does
> > someone knows where to find this font, __please__ (and do not send it to
> > me via e-mail)..
> 
>   ftp://ftp.adobe.com/pub/adobe/acrobatreader/unix/4.x/chtfont.tar.gz
> 
> > I've added error checks to "ftview", could someone test the new code to
> > see if the doublet and garbage still appear ??
> 
>   Thanks.
> 
>   This time miss CNS1_2028, behind the 2027 is 2029. Need ignore hinting
>   (press h)to display correctly(Hit h then 2028 come out).
> 
> --
> Warm Regards,
> Edward G.J. Lee
> 
> _______________________________________________
> Devel mailing list
> address@hidden
> http://www.freetype.org/mailman/listinfo/devel