Question: Negative italic correction

[Lars Hellström]

This ends with a question, but I probably should start with describing what
I've done since last.

Last Sutarday I completed rewriting fontinst's (V)PL writing routines so
that they use \make_factor instead of \afm_convert, and since then I've
been working on testing the output to see that I didn't make some strange
change to the metrics. It turned out that there were four types of
differences:

1. I had forgot that scaling introduced by \glyph affect also the sizes of
\glyphrule's, \moveup's, and \movert's; this is a bug which I haven't
gotten around to fixing yet, but it's the next thing on my list of things
to do.

2. There were lots of small differences (up to about 0.00015 of the design
size) in the metrics---these were expected, since the old conversion
routine truncated the intermediate and final results, whereas the new one
does not.

3. There were a few differences in character height, depth, and italic
correction that were pretty large (up to 10% of the design size). I believe
these are due the rounding PLtoTF and VPtoVF does to ensure that there are
no more than 15 nonzero heights and depths, and 63 (I think it is) ditto
italic corrections in a TFM file. Funny that the heights are so well spread
out that there is almost always one at the boundary between two "rounded
heights".

And so the thing that actually led me to writing this

4. In one case (the italic correction of underscore in slanted fonts), a
zero italic correction had changed to a negative italic correction. (The
negative IC is present in the MTX; if you positively slant a character with
negative height and zero IC, the IC will become negative.) This turned out
to be my fault: I had changed an "if italic correction > 0" to "if italic
correction not = 0" when fontinst was going to decide whether it was going
to write a CHARIC property or not.

Anyway, what I wanted to ask was: Is there any use for negative italic
corrections, or should they be ignored? In v1.910 and before, negative
italic corrections have been ignored when one is writing a ligful (V)PL,
but not when one is writing a ligless PL, which is probably a bit
inconsistent.

Lars Hellström