Re: How do you use Lout?

[Jeffrey Howard Kingston]

This is a special case of a proposal made by some bright person (sorry,
forget who at the moment) to generate a signature on each run and
declare convergence when two successive signatures are equal, except
that this proposal will also declare convergence on the very first
run if there are no error messages.

It would probably work well in practice.  However, it is a little known
fact that even when there are no error messages at all, the document
might still not have converged.  The error messages point to missing
values, which are not the same thing as unfinalized values.  I have
occasionally been caught out with this myself, with index entries
having page numbers off by one owing to very late small changes in
the positions of page breaks, caused by the insertion of running
header lines.

The only sure way to test for convergence is to see whether the entire
cross reference database is identical after the run to what it was
before the run.  Then, clearly, nothing can ever change (OK, except the
value of @Date).

Jeff Kingston


On Fri, 19 May 2000, W. L. Estes wrote:

> > The most common complaint I've heard from people who use Lout as
> > part of a larger system is the difficulty in knowing how many
> > runs it will take to resolve cross references.  Just running it
> > six times seems overkill, and very slow, but there is no way to
> > predict how many runs it will take, and no indication given that
> > it has converged, even when it has.  Of course, there are good
> 
> lout outputs a message for each unresolved cross reference when it
> finds one. if one counts the number of such messages, at some point,
> either the count is zero or does not change from the immediately
> preceding value of the count. In each case, would it be correct to
> assume that all cross references have been resolved?
> 
> --Will
>