[Devel] Hinting metrics

[David Chester]

For the record, here's a message I sent directly to Werner while the lists 
were down:

Hello,

I've attached a patch which incorporates both approaches previously discussed.  
This seems to be a good solution.  A comparison graphic can be found at:

http://www.cs.mcgill.ca/~dchest/kern2.png

You will notice that the problem of the fused 'c' and 'o' is fixed, and 
spacing is otherwise slightly more consistent.

When I asked about implementation before, I was really wondering whether this 
should be somehow done through FT_Get_Kerning.  That seems like it would be 
difficult, because the both glyphs in the kern pair would have to go all the 
way through the autohinting process before the correct kerning vector could 
be calculated.  Would the cache alleviate the need for redundant autohinting 
in this situation, or is it much more complicated than that?

On the other hand, there is the note in the documentation of FT_Get_Kerning, 
that says that some alternate kerning functions may be "implemented through 
format-specific interfaces."  So maybe the current implementation is alright.

Also, if kern_left_err and kern_right_err are added to FT_Glyph_Metrics, each 
driver needs to initialize them.  Might other drivers/formats be able to use 
them in the same way as the autohinter?

I also fixed the ft2demos patch so it applies from where it should.

I'll be interested to hear your response.

David Chester

ft2demos-kern2.diff
Description: Text Data

freetype-kern2.diff
Description: Text Data