Clean up "messages mess" where programs PRINT errors (or
just don't compile) but the problem is lost when output is
rerouted.  See XqtS for a logging mechanism.

Fix auxillary routines that DIE when they should just
return errors (and check for those).  [Notably the
"sampling[fs]".]

Put back SmeasS ability to execute the file it compiles
and the -O option to avoid this.

Try to make the trace options work for state including
calculation of profiles w/o execution (see profile.plan).

Incorporate the "infeas" script into SYN or Calc.
Give a new option that omits use of "infeas"
for tutorial purposes.

Finish the "incremental" processing, priority for components
with state (see incremental.plan).

Add some part of linear processing to state, at least enough
to be able to use a linear (ident) for loops.
It may be enough to implement an "I" value for states,
to be converted to an identity value when needed.

Remove many debug warnings involving the piecewise-linear
approximation; replace with a log and "something funny" message.
Maybe use an arbitrary precision package instead?

Standardize option processing and passing options on.

Add range(s) arg to Xcompt, or to Xcomp for both Xcompt and Xcompl.

Rename scripts so user-supplied names can't conflict (add a "-"?).

For Xcutel and XcuteS, use looparound to write new Perl program
and execute.  (Partially done in Xcutel.loop)

Get rid of .ccf "theory" format; use .ccfc.

Redesign .ccfc files to hold indices for input/state rather
than values.

For splots, rotate the z-axis label to read up.

Implement a variation on sequence sampling:  use a random input
sequence, but as each state is found, choose a random input
in each input subdomain at that state.  (Repeating states?)
