[MiKTeX] Origin of nulls?

[Walter E B. <wb...@fn...>]

I use the latest MiKTeX version under Cygwin to produce genealogy 
reports, often many hundreds of pages long.  The .tex source files are 
auto-generated by software that interfaces with my genealogy database.

Most of the time, these reports are beautifully produced.  However, on 
rare occasions some of these reports fail.  I've come to understand the 
reason for the failure, but would appreciate some insights as to the cause.

The diagnostic is "TeX capacity exceeded, sorry [.2=200000]" -- but this 
is deceptive.  What I have discovered, in every case that leads to this 
diagnostic, is a problem in each of four .ind files (I have four indeces 
in each of my genealogy reports).  Specifically, each .ind file ends 
with a sequence of numerous null characters (my editor shows each such 
character as ^@).  No two .ind files end with the same number of nulls; 
the shortest has about three dozen, while the longest has many hundreds, 
perhaps over 1000 (sorry, I didn't count).

I do not understand the origin of these null characters.  Here is the 
fragment of my GNUmakefile with the code that establishes the .ind files:

#-----------------
override define make-indexes
for F in bday date indi plac; do \
   sed -e '/|see/s/{[0-9]*}$$/{see}/' $$F.idx | makeindex -c -s $$F.mst 
-i -t $$F.ilg -o $$F.tmp; \
   ( test -f $$F.ind  &&  cmp -s $$F.tmp $$F.ind )  ||  cp $$F.tmp 
$$F.ind; \
done
endef
#------------------

As I say, this problem is intermittent in that it does not manifest for 
most reports I produce.  Once it arises for a given report, however, it 
is 100% reproducible -- at least until the database is updated with new 
information to be reflected in the next version of the report.

I would be pleased to receive suggestions toward understanding (and 
fixing!) the true underlying cause of this vexing problem.

Best,

-- WEB