[XeTeX] Weaving xetex.web

Jonathan Kew jonathan_kew at sil.org
Sat Mar 17 10:13:23 CET 2007

On 17 Mar 2007, at 7:22 am, Micah Cowan wrote:

> Akira Kakuto wrote:
>>> I'm interested in studying XeTeX's source to understand it  
>>> better, and
>>> to take full advantage of some of its features. The best way to  
>>> do this
>>> would seem to me to be to obtain a weave from xetex.web.

Possibly, though you'll find that many parts are sadly lacking in  
internal documentation. :(

>>> However, when I
>>> attempt to do this, I obtain only the Index and Table of  
>>> Contents, with
>>> no actual content/nodes.
>> I don't know web details, but it seems that, e.g.,
>> {@"10000 to separate char and command code}
>> gives an error "! Illegal use of @ in comment."
>> {|@"10000| to separate char and command code}
>> seems to be ok. Jonathan ?
> Yes, I was encountering those; however, they're just warnings and  
> don't
> actually halt processing. I fixed them primarily by doubling @@s  
> within
> comments, where that hadn't been done, and processing went through  
> with
> only a warning about an unused node (something about |k|...),

Yes, I'm afraid a lot of the xetex changes are not carefully  
commented in Weave-compatible Knuth style; they lack the proper  
markup on names, etc., as well as doubling of @ signs. (I've never  
intended to weave this web, I just work with the actual source files.)

> but the
> resulting DVI/PDF still had only Index and ToC.
> Interestingly, the .tex result itself seems to contain all the
> descriptions and whatnot, so I don't know if something is silently
> halting the process or whatnot. Perhaps enabling \tracingall and doing
> some major rooting around could illuminate things, but I thought  
> someone
> might already know what the problem might be.

I don't know, but if I have a little time I might give it a try.

On 17 Mar 2007, at 12:37 am, Micah Cowan wrote:

> Is there a separate dev list/forum, or is there no real public  
> interface
> to current XeTeX development?

There's no separate dev list at the moment, just this all-purpose  
list. If the level of technical discussion gets high, though, then a  
separate list would make sense, so that users don't get swamped with  
dev stuff.

When it comes to detailed discussion of implementation, patches,  
etc., I've usually just had direct email from the people involved,  
but of course that means it's not open for wider review and input  
(until patches make their way into Subversion, which I try to do  
promptly in most cases). I'd be happy for there to be a developer's  
forum, public bug tracking DB, etc.; I just don't have the time to  
set up and manage these things. I know they could be hosted on  
sourceforge or some other such place (as we don't [yet] have  
appropriate infrastructure running on our own server), but it still  
takes someone's work to make it happen.

(On the specific topic of Unicode math extensions -- the XeTeX  
features involved, many of which are coming in the next release, as  
well as (La)TeX macro support -- there is a discussion group  
<unimath at googlegroups.com>.)

> I might be interested in helping in the
> effort to have XeTeX potentially generate PDFs directly, without dvi
> intermediates.

That's certainly an interesting possibility, though it would be a  
pretty major undertaking. I've wondered about it from time to time,  
but not attempted to go there yet. For now, I'd rather focus on  
remaining issues in xdvipdfmx.

Note that there is actually a performance *advantage* from the  
current architecture on multi-CPU machines, which are becoming  
increasingly common. XeTeX + xdvipdfmx is faster overall than pdfTeX  
on my dual-Xeon desktop, when processing a large Plain TeX document  
(The TeXbook is my usual test for this) -- and also produces more  
compact PDF. YMMV, of course.

More information about the XeTeX mailing list