XSB Src (SVN)-old

Fixed an invalid read in remove_incomplete_tries

Just adding a little documentation.

Avoiding spurious noise in the testsuite.

Added lock extension, for locking when compiling XSB files.

Small documentation changes.

1. Added creation of a lockfile when compiling to avoid multiple
processes trying to compile a .P file at the same time. This is not
guaranteed to avoid all kinds of collisions, just simple ones I've encountered
when starting multiple processes at the same time, when they use the same
files that need re-compiling.
2. Added code to catch compiler errors and close the object file. On windows
this lack had sometimes kept corrupted .xwam files from being deleted, and
resulted in confusing results, fixable only by explicitly deleting the .xwam
file. This may only be a windows issue.
3. In looking at compiler errors, I noticed that in the move from written
messages to thrown errors, we lost some possibly useful information in the
resulting error messages, so I added it back in. (And did a couple of
other small improvements in the area of compiler errors.)

Changes due to new max_arity and also took out restrict.tex from
manual1 as this chapter was obsolete.

Fixed yet another bug in incremental table view maintenance,
though this one was obscure. The function
get_list_of_answers_from_answer_list was core dumping. This
function takes the answer list which was produced by
traverse_variant_trie, and copies their corresponding answers to
a list on the heap, constructing the answers as it goes. The
problem arose in cases where no answers needed to be copied, in
which case uninitialized garbage was returned to find_the
visitors, rather than a pointer to a list. The fix was simple --
properly initialize the variable whose value is returned, and
check the return when the function was called.

I didn't think that such a case could happen, so I looked into
why it did, and it turns out it depended on a feature of XSB's
trie hash nodes. A call to a hash node may have the
corresponding position free or bound. If it is free, then when
the hash_handle code executes, it finds the next value to return
by traversing the hash table. It can be the case where there is
no next value. In certain cases, this leads directly to
list_of_answers_for_answer_list returning no answers and to the
problem we had.

Along the way, I also created a debugging macro,
print_hash_handle that helped me verify why things were
happening.
y

Fix one place where parameterized module names were not handled correctly.
(And cleanup patch in cp_opt.P to make testing of internal variables more
precise.)

File had syntax error and wouldn't read correctly. Don't know exactly what was
meant for it to compile (or maybe not?), but tired to do minimal change to
make it syntactially correct.