SGML docs can't be built as .PDF ...

I've confirmed recent reports of trouble building PDF output from our
CVS HEAD SGML. Using Fedora 8 openjade (plus slow-lists patch),
with the TeX parameter size increases recommended in our docs, it dies
here:

LaTeX Warning: Reference `PGCRYPTO-WITH-WITHOUT-OPENSSL' on page xlvii undefine
d on input line 55630.

LaTeX Warning: Reference `101017' on page xlvii undefined on input line 55655.

! TeX capacity exceeded, sorry [grouping levels=255].
\rlap #1->\hb(at)xt@ \z@
{#1\hss }
l.55663 {start}}
\Link%
! ==> Fatal error occurred, the output PDF file is not finished!

There does not seem to be any parameter controlling "grouping levels",
but in any case it seems that this is not so much TeX's problem as a bug
in the TeX code emitted by openjade. The failure is occurring in the
list-of-tables (presumably at the 255'th table, though I didn't count),
and at the end of the list of tables I find

{\def\Label%
{PL-LANGUAGE-TABLE}}\endNode{}\endSeq{}\endLink{}\endPar{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\end!
Seq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\e!
ndSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}
\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSeq{}\endSpS{}\Node%
{\def\Element%

All those \endSeq{}'s correspond to unclosed \Seq macros in the code for
each list-of-tables entry. So this looks to me like a stylesheet bug,
but I don't know the DSSSL stuff well enough to know where to look.
Anyone want to tackle it?

(A workaround might be to suppress the list-of-tables in PDF output,
but I dunno how to do that either.)

regards, tom lane