[clug] pdflatex regressions?

[Hal Ashburner]

Hey people,
I had to do a sever upgrade from sarge->etch->lenny on a machine that 
uses pdflatex to generate some complicated reports as part of an 
application it's running.
There's a .tex template that gets filled repeatedly for different 
reports and when it works you get very pretty pdf documents.
pdflatex appears to have developped some issues over 5 years with 
respect to errors. Now latex errors have always been hideously unhelpful 
but this is a new level of unhelpful. It's reporting errors every time 
\bfseries is used. running latex on the same .tex produces no such 
error. It also produces a wonderful error that reads:
Dimensions can be in units of em, ex, in pt, pc cm, mm, dd, cc, nd, nc, 
bp or sp; but yours is a new one!
It does this for pt and mm for reasons I don't understand.
None of these errors matter at all except that when it gets to 100 it 
executes the
if (errors > 100) { // yay for magic numbers in the middle of source
 stuff_you_I_hate_you();
}

The only solution I can see that doesn't require me to learn tex and 
re-write all of latex, which would be a task worthy of Sisyphus,  is to  
get the source from debian, adjust the magic number to a bazillion and 
install it locally.

Any better ideas?
Anyone seen all kinds of random breakages on upgrade using tabularx and 
colortbl latex packages?

Cheers,
Hal