XSB Src (SVN)-old

Marking (in mark_root) tries to check whether structure-tagged pointer
indeed points to a PSC pointer. That test however would let pass non-reference
values; clearly strange. This situation just arose (in the compiler), so I added that check.
So it may be that this is necessary, but just never arose before, or it could
be that there is another bug in constructing structure pointers that this has
uncovered.
I've generated a warning if this case arises. This may just be kicking the
problem down the road.

Uncommented out the definition of undefined.

Added definition of undefined. (Unclear, to me, why it wasn't here.)

Added to support testing of subsumptive tabling.

Partially fixed the treatment of the -only parameter after the subsumptive
tests were added. Including an -only parameter for a test file that has
a subsumptive version tests both the variant and subsumptive version.

A couple of minor changes to make these files compile under MSVC.
One change in slgdelay.c to add coercion (through using a macro)
may be a problem. Perhaps Terry can check.

In some Flora programs with defeasible reasoning, a
mysterious "null node" warning had been printed out. It turned
out that this was being emitted by find_answers_for_subgoal(),
which is used in cascading abolishes. The problem arose for
deleting calls that had empty answer tries, where a stack was
checked, a null node immediately found, and the routine exited.
In this case, it was harmless, but I added a check to get rid of
the warning in this case, and made the warning itself more
descriptive.

Fixed typo in tables.c

In tst_aux.h fixed so that tables created by call subsumption allow
attributed variable constraints in their answers (including in delay
elements). Calls in call subsumption still cannot include attributed
variables -- I need to understand more about Bao's algorithm before
I'll be ready to add them there. But having them in answers takes us
at least part of the way.

Changes were mostly for documentation and code refactoring.

However, also fixed a latent bug in simplification where
was_simplifiable was not properly checking negative delay elements
corresponding to subsumed subgoal frames. This led to the possibility
of adding un-needed delay elements that might not have been simplified
away. Surprisingly enough, however, none of the tests in the test
suite checked for this.

Adding a new test for copying attributed variables into delay elements
/ lists. I hadn't thought this worked, but apparently it does !

Also, I'm adding a very simple test for incremental evaluation -- more
cases need to be added here, but at least this is a start.