···<date: 2013-07-26, Friday>···<from: Khaled Hosny>···
> On Fri, Jul 26, 2013 at 12:59:55PM +0200, Philipp Gesang wrote: > > ···<Datum: Friday, 26. July 2013>···<Von: Ulrike Fischer>··· > > > > > Am Fri, 26 Jul 2013 12:19:23 +0200 schrieb Javier Bezos: > > > > > > > Ulrike, > > > > > > > >> The new babel redefines the \TeX command and surrounds it by > > > >> \textlatin and this reveals a bug: > > > > > > > > Do you know if it works with the old (3.8) babel? (I've not tried > > > > it out yet -- just back from holidays). I have not changed this part > > > > (afaik :-)) because I'm still investigating how to fix the whole > > > > \text<script> stuff in a backward compatible way. > > > > > > miktex still has Babel <v3.8m>, so I could test: There is the same > > > problem, the fontdimen is 0 and there is to much space between X and > > > n. > > > > > > Imho the correct solution would be that luatex/luaotfload sets > > > \fontdimen1 for italic fonts to a sensible positiv value. > > > > What’s “a sensible value” if italic correction is done on node > > level? > > Note that \fontdimen1 is *not* italic correction, but the italic angle > of the font i.e. its slant amount. > > > Someone knowledgeable about font parameters and how they translate to > > the original values would have to supply the code, though, > > ConTeXt code in luaotfload already does that, at least I see that > copytotfm() sets parameters.slant for italic fonts, so someone needs to > check why this is not working as expected. > > > especially when it comes to the more delicate parts described in > > appendix G. > > You don’t need to worry about math font dimentions, the OpenType math > machinary does not need them (they might even confuse LuaTeX to thing it > is dealing with an old font) and old math fonts should have them in the > loaded TFM files already. This line: http://repo.or.cz/w/context.git/blob/refs/heads/origin:/tex/context/base/font-otf.lua#l1954 rounds the value of parameters.slant to -0. If I scale it to the point size: parameters.slant=- math.round(math.tan(italicangle*math.pi/180) * 2^16) I get the same output visually with babel as with polyglossia in Herbert’s original example. I’ll put Hans in the CC since he know’s best. Best, Philipp
pgpThujASvgj8.pgp
Description: PGP signature
