Update from version 1.17 to 1.22.

Pkgsrc change:
o Canonicalize HOMEPAGE pointer.

Upstream changes:

2007-10-23  Jarkko Hietaniemi  <jhi@iki.fi>

* Add cartesian_product() and power_set(), both as full
  constructors and as iterators.

* Add empty_clone().

* Makefile.PL not requiring Test::More since we are not using it.

* Released as 1.22.

2007-10-04  Jarkko Hietaniemi  <jhi@iki.fi>

* Made to work with the upcoming Perl 5.005_05
  (yes, you read that right), most importantly
  Scalar::Util made optional (we fall back to
  pure Perl emulations for blessed() and refaddr()
  if necessary).  Everything else already worked.

* Released as 1.21.

Sat Aug 6 12:39:43 2005  Jarkko Hietaniemi  <jhi@iki.fi>

* [cpan #13816] Set::Scalar blesses unblessed refs

  A genuine bug, the suggsted fix used, but then again Set::Scalar
  was never designed or tested be used with references as the set
  members.  I would not recommend doing that unless much more
  testing has been conducted.  A test added to misc.t for that,
  and a warning about using references as the set members added
  to the pod.

* [cpan #13856] funny behavior in
          Set::Scalar::Base::intersection and Set::Scalar::Base::union

  A genuine bug, but the suggested fix as-is would break
  the special cases of intersecting with the null set and
  unioning with the universal set.
  A slightly enhanced version of the fix used instead,
  tests added to intersection.t and union.t.

* [cpan #13857] Set::Scalar::Base::_binary_underload bug

  Not really a bug.  The _binary_underload() method is only
  ever going to be called by Set::Scalar itself, or by classes
  derived from Set::Scalar, so the assumption that all references
  are object and that they are capable of calling the new() method
  (i.e. being instance of Set::Scalar) is completely valid.
  Trying to use the suggested fix also badly breaks the
  laws.t when the universal and null sets are present.

* Released as 1.20.

Sun Mar 28 17:16:26 2004  Jarkko Hietaniemi  <jhi@iki.fi>

* [cpan #5829] When is_disjoint() was called in list
  context, and the result was disjoint (not disjoint
  universes), the return value was a list with one
  undef element, from Alexei.

* Released as 1.19.

Sat Oct 4 17:56:24 2003  Jarkko Hietaniemi  <jhi@iki.fi>

* Removed a cut-and-paste bug from symmetric_difference();
  from frederik.

* Released as 1.18.

(he)