[luatex] Unicode math issue with LuaLaTeX

[Ulrik Vieth]

On 06/13/2010 05:28 PM, Khaled Hosny wrote:
> On Sun, Jun 13, 2010 at 02:52:15PM +0200, Taco Hoekwater wrote:
>> Ulrik Vieth wrote:
>>>
>>> Somehow the parsing of these values seems to be very wrong.
>>> If I understand correctly, that's a matter of luaotfload. Right?!
>>
>> Something is wrong in luaotfload, yes.
>
> For the record, I once imported the whole ConTeXt math lua modules into
> luaotfload, some issues were fixed, but I got a bunch of new issues that
> I reverted that since I wasn't sure what was actually needed.
>
> Now if some one can tell me what luaotfload is supposed to do, e.g. what
> TFM tables are missing, I can work on that.

 From my understanding, there may be actually two problems:

1. If an OT math font is loaded in \fam 4 (as in unicode-math.sty),
its OT math parameters are not used, so LuaTeX ends up applying
the TFM-based parameters from CM fonts in \fam2+3, which implies
using the height of the radical glyph to determine the rule thickness.

2. If an OT math font is loaded in \fam 2+3, some of its OT math 
parameters are not loaded correctly and some are getting strange values.

For example, in XITS Math you have the following values:

RadicalExtraAscender=50,
RadicalRuleThickness=66,
RadicalVerticalGap=50,

I would expect those values to be converted as follows
during the loading of \textfont3:

{assigning \Umathradicalkern\textstyle=32768} % 0.5 pt
{assigning \Umathradicalrule\textstyle=43254} % 0.66 pt
{assigning \Umathradicalvgap\textstyle=32768} % 0.5 pt

However, I actually end up with the following values:

{assigning \Umathradicalkern\textstyle=433848} % 6.62 pt
{assigning \Umathradicavgap\textstyle=542310} % 8.275 pt
- no value for \Umathradicalrule -> fallback to TFM method

Somehow, deeply buried inside luatofload, soemthing goes wrong,
but I have no idea what really happens or how to trace it.

All I can see are the implicit assignments of \Umath ... parameters
shown in the output of \tracingall.

Regards, Ulrik