[XeTeX] Problems with spacing in math formulae

Khaled Hosny khaledhosny at eglug.org
Wed Oct 20 11:34:42 CEST 2010

On Wed, Oct 20, 2010 at 07:31:13PM +1030, Will Robertson wrote:
> On 2010-10-20 19:09:52 +1030, Taco Hoekwater <taco at elvenkind.com>
> said:
> >On 10/20/2010 10:35 AM, Khaled Hosny wrote:
> >>
> >>Very useful! Knowing that there were a working version I was able to
> >>track this down (with some git bisect magic), the culprit is the
> >>DisplayOperatorMinHeight math parameter, while ago I increased it from
> >>1400 to 2180 and XeTeX seems not to like that. Now I need to remember
> >>why this was done in the first place (never use terse commit messages,
> >>one day you will regret it).
> >
> >That was because otherwise the display style operator size in luatex
> >was too small (essentially outputting a text style operator in all
> >cases).
> Is it possible to set a happy medium here or is there an
> inconsistency (or bug) in XeTeX in how this value is being used? If
> necessary, unicode-math can manually override whichever fontdimen
> setting this is in XeTeX...

Setting it back to 1400 seems OK so far, with both XeTeX and LuaTeX.

Now I vaguely remember why this value were changed, initially I had a
composite integral as well as the display variant so that one can get
integral of any size (theoretically at lest), but due to a bug in
LuaTeX (by then I was not aware it is a bug, now fixed) I was always
getting the composite integral for display size and with such small min
display operator value it was smaller than the actual display integral,
so I'd to increase so it becomes as large as original one (this explains
the odd value, why 2180 and not 2200). Then I decided the composite
integral is making no good, so I disabled it, so resetting it back to
1400 is not a problem.

That being said, I still think it is an engine bug, I can not find I
reasonable explanation of such odd behaviour.


 Khaled Hosny
 Arabic localiser and member of Arabeyes.org team
 Free font developer

More information about the XeTeX mailing list