0.0086
GJL 
 - Fixed some remaining int-as-pointer bugs that prevented working on
   64 bit linux.
 - Some tests failed on some machines. Removed one test and relaxed precision
   on two other tests. The two tests that needed the relaxed condition use
   floats, rather than doubles: tapprox is replaced by tapprox_cruder for
   these. In levmar.t, the second to last test is now commented out. It
   fails on some debian machines, but not others. Also removed routine
   thread1 in file thread.t for case of floats. This should be fixed at
   some point

 - Minor edits of documentation.
0.0085
GJL  04-Sep-2006
 - Removed non-fatal warnings from liblevmar 'starting point not feasible'
 - Removed all uses of int's as pointers, as this is not portable to,
   for instance, AMD64. Also did explicit casts to remove 'int to pointer without
   cast' type warnings. (I don't see whats happening here, but its casting perl IVs,
   which are supposed to be able to hold pointers.

0.0084
GJL  24-Jul-2006
 - Documentation fixes.
 - Added (a working) 'call' method to Levmar::Func. It currently will not
   work with perl subs. Added Func/t/call.t .

0.0083
GJL  20-Jul-2006
 - Fixed bug that prevented x and t from having different dimensions.
   This allows multi-dimensional data where t represents the same data
   for each of the co-ordinates axes. Added an example to the docs and
   the test directory using this feature.

0.0082
GJL  18-Jul-2006
 - Added debian directory for building deb package.

0.0081
GJL  17-Jul-2006
 - More cleaning of documentation.
 - Removed -lg2c from Makefile.PL. It is apparantly not needed (with
   recent versions of gcc at least.)

0.008
GJL 16-Jul-2006
 - Changes since .007. Added FIXB option. (and a couple of tests for it to levmar.t)
 - Tried to improve documentation.

0.002
GJL 5-Jul-2006
 - levmar.pd: made levmar() threading aware.
 - t/liblevmar.t: changed to agree with the new inteface required by threading.