MathJax halts on long page in IE 8

[Brion Vibber]

Hi all --

One of our Wikipedia testers put a nice big test page on my MediaWiki test site, and found that in IE 8 MathJax typesetting only gets up to a certain point and then stops:

http://leuksman.com/mw/index.php/Help,_Formula#.5Coiint_and_.5Coiiint_using_MathJax

In our bug tracker:
https://bugzilla.wikimedia.org/show_bug.cgi?id=35219

All equations appear to render through to completion in Firefox for me; I don't have my IE VMs with me for a couple days so can't test it myself just now.

Apparently under IE 8 it's consistently stopping at the same place, and there's no visible error messages.

Thanks for any ideas or help!

-- brion vibber (brion @ pobox.com / brion @ wikimedia.org)

[Davide P. Cervone]

OK, I am able to reproduce the problem in IE9 when in IE8 standards mode.  From a quick look it appears that MathJax has actually typeset everything, but has failed to reveal the last block of equations that it processed (it does it in chunks to improve performance).  I'll check into it.  It is fine in IE9 standards mode, but fails for IE8 standards or below.

I'll get back to you.

Davide

[Davide P. Cervone]

I figured out what the issue is, here.  It turns out that there are a few "[Math Processing Errors]" earlier in the page, and that is throwing off MathJax's equation count, and so it is failing to reveal the last n equations (where n is the number of processing errors earlier on).  The reason there are processing errors is that IE8 and below throw a run-time error when you try to set an elements style to a named color that it doesn't recognize.  The way the default \color macro works is just to set the element's CSS color to the given name, and that is failing for some of the colors in the table of accepted colors.  I believe that if you include the color extension (which implements the LaTeX-compatible \color command), the colors are handled through a lookup table, and so that should get rid of the processing errors in this case.

I'll look into having it handle the equation counts properly even when there are math processing errors, but it looks to me like the issue should not really be a show-stopper, as it is based on the presence of errors (and not TeX errors, but MathJax errors), not on the length of the page.

Davide

On Mar 14, 2012, at 4:26 PM, Brion Vibber wrote: