bug in fontinst 1.801: duplicated kernings (was: ae fonts Q)

[Lars Hellström]

Bernd Raichle wrote:
  [snip]
>Can you please keep it in such a way, that at most one kerning
>instruction and at most one ligature instruction for each pair of
>characters will be possible?  I.e., it will be still possible to have
>one kern _and_ one ligature instruction for the same glyph pair?

That's no problem, I just move the beginning of a group and snip some code
(as for ligatures, I would believe it is rather a responsibility of the ETX
file writer to see to that a ligature isn't repeated).

>Reason: This ``feature'' will be needed in the future.  Currently I'm
>implementing a replacement of TeX's internal ligature and kerning
>routines for the next e-TeX version.

This complicates things. One would still want the exclusive behaviour for
fonts that should be used with normal TeX.

>  The new ligature/kerning builder
>is able to first create ligatures (ignoring all kerning instructions
>in the font's lig/kern table) and then insert appropriate kerning
>between the resulting char resp. ligature nodes (ignoring all ligature
>instructions).  As one positive side effect of the new routines this
>will allow you to insert kerning between two ligatures which is
>currently not possible with TeX.

Let me see if I understand this correctly. You add the capability to kern
between two ligatures, whereas with standard TeX you could only add a kern
between the first ligature glyph and the glyph which is the first character
of the second ligature. This does not call for both a ligature and a kern.

>  Another feature of the new
>implementation will be the possibility to cleanly prevent the creation
>of a ligature even if the ligature is declared in the font's lig/kern
>table but in this case it will sometimes be necessary to add kerning
>between the two glyphs.  Thus the ability to have a kerning _and_ a
>ligature instruction for a glyph pair should be kept.

I see. If the ligature is overrided by the user, then the font might also
contain a kern, which is then inserted by TeX instead of making a ligature.

Hmm... Due to the fact that it primarily is normal TeX that should be
supported, I believe it actually would be simpler to implement a version
that would impose the one ligature limit as well (i.e., as you described
above). Switching between normal TeX settings and e-TeX settings can be
simplified to redefining one control sequence.

I will think about it.

/Lars Hellström