[Devel] Re: CFF problems

[Tom Kacvinsky]

I spoke too soon.

1.  I found the problem with ftview and CFF fonts (basically, the variable
num_faces was not set when we load a pure CFF font).  I'll commit the fix
presently.

2. I found some errors in the function parse_t2_real, which is in t2parse.c
I won't commit this; I want other eyes to see it first.  The "fixes" I made
work, but I had to do some things I didn't understand.  For instance, the byte
stream seems to start at "start + 1" instead of "start".  So I incremented the
byte stream pointer to get around this.  Why?  I'll send the patch later after
I do some more grubbing around.

3.  The heuristic for figuring out the EM square is fubar (WWII sense).  Sorry 
David!
There is just now way to figure out the EM square based on the FontMatrix 
(because
of scaling, rotation, translating, shifiting, etc...).  I think it best to 
leave the
EM square at 100 as a good default.  Where is the EM square used anyway, David? 
 It
doesn't get set for Open Type fonts (at least not in any of cff/*.c).

Tom

On Tue, 21 Nov 2000, Tom Kacvinsky wrote:

> OK,
> 
> This font came from a Distiller produced PDF file, didn't it?
> 
> Next, the FontMatrix in the font is [.001 0 0 -.001 0 0], so this font is
> supposed to be flipped over the horizontal axis.  As I am not a user of 
> FreeType
> (only a debugger), perhaps David can point you to the correct function for
> transforming a font (it is best not to mess with the CFf guts -- they are
> working properly in this case).
> 
> The default grid for this ont is not 32768 units by 32768 units.  The numbers
> you see are getting shifted right 16, so the number is much smaller.  However,
> it makes the assumption that dict->font_matrix.yy is postive, which for your
> font, it is not.  I will committ a patch for this.
> 
> Finally, left side bearings can be negative.  This usually means the left edge
> of the glyph is to the left of the glyph origin.  I rarely see this, but I 
> know
> it is done.
> 
> I'll look into why ftview doesn't work with CFF fonts...
> 
> Tom
> 
> On Tue, 21 Nov 2000, Blinn, Eric W wrote:
> 
> > 
> > I'm seeing some problems with CFF fonts. I have some fixes
> > or workarounds that work reasonably with my application, 
> > but I've also been digging down into the FreeType code to 
> > understand what is going on and wanted to go ahead and 
> > bounce the problems off the FreeType and/or CFF gurus 
> > on the list.
> > 
> > I've attached a small CFF font file to demonstrate the 
> > problems. Load it into ftview with a ppem of 72 (note that 
> > the ftview from ft2demos-current fails to open a CFF file 
> > but the one from ft2demos-beta8 works). Once the font is 
> > brought up in ftview, the problems are as follows:
> > 
> > 1) The glyphs are tiny. In T2_Init_Face() the units_per_EM
> > is being set extremely high, at 32768 grid units. Standard 
> > Type 1 Postscript fonts should be fixed at 1000 grid units, 
> > right? Setting units_per_EM to 1000 seems to get the CFF 
> > glyphs rendered at the proper size.
> > 
> > 2) The glyphs are flipped (reflected over a horizontal 
> > midline). There is some disabled code in T2_Load_Glyph() 
> > where FT_Outline_Transform() would otherwise be called. 
> > Enabling that code and feeding it the appropriate transform 
> > matrix values flips the glyphs into the orientation I would 
> > expect.
> > 
> > 3) The pen position is not always getting advanced correctly. 
> > Glyphs often sit on top of one another. This appears to be 
> > caused by incorrect glyph metrics, particularly a negative 
> > left side bearing (horiBearingX). The left side bearing 
> > should always be positive for a horizontal layout, right? I 
> > thought maybe this was happening because the metrics were 
> > being calculated using a control box. You might expect this
> > behavior if some control points lie far enough outside the 
> > glyph's bounding box. However, I've tried using the bounding 
> > box to calculate the metrics instead, but saw no improvement.
> > 
> > Thanks,
> > Eric Blinn
> > 
> >  <<example.cff>> 
> > 
> 
> 
> 
> _______________________________________________
> Devel mailing list
> address@hidden
> http://www.freetype.org/mailman/listinfo/devel
>