bugs

Lin-Kernighan heuristic for the TSP and minimum weight perfect matching

PENDING:

1998/8/1.  (Feature request) Milestones, probe depths, and move depths aren't 
in the matching code.  1998/10/2 milestones are now there.

FIXED:

2000/9/16.  In src/jbmr.w and src/match.w, the change_log array 
wasn't being grown properly, causing memory corruption
of the dirty set structure.  The result is that LK segfaulted on some 
very large instances, sometimes only with a large number of iterations.
Thank you Bruce Perens and Electric Fence!

Don't worry, the bug doesn't affect the results in my PhD work.

1998/8/28.  Fails with length_t being an int.  Try dsj1000.
Seems to be handling of cluster dist and writing to e[3].  Different
comparisions being made when writing to e[3] and taking the max?
(Fixed 1998/8/29.  Change < to <=.  jbmr.w problem, not in match.w)

1998/8/28.  Random distance matrix fails in construct a tour.
Endless loop when trying to get second batch of neighbours for a city:
always getting the same ones?
(Fixed sometime 1998/10.  Nonsense code with computing maxima is now fixed.)

1998/8/7.  Discovered and removed a space leak type problem in lk.w
regarding the prng handed to the optimizers.  I'm sure there are other
space leaks.  (See 1998/7/24 entry).

1998/7/24.  There's no way to free a tsp_instance_t.  I haven't got a good
 OO interface for the tsp instance type.