version 0.003; 2009-04-29

  * bugfix: when loading pure Perl implementation, check for numeric
    semantics that make it impossible to operate correctly (perl 5.6
    with integers longer than floats), and refuse to load if the module
    won't work right

  * bugfix: in sclnum_val_cmp(), work around perl bug #41202 which
    caused some comparisons on values originating as strings to yield
    the wrong answers

  * bugfix: in scalar_num_part(), handle numeric conversions of non-number
    types correctly

  * bugfix: require bugfixed version of Data::Integer (it gave incorrect
    constants on perl 5.6)

  * bugfix: require bugfixed version of Data::Float (for $SIG{__DIE__}
    handling)

  * XS implementation, used if available with fallback to existing pure
    Perl implementation if XS is not available; the XS implementation
    works on Perl 5.6 with wide integers

  * test classification and comparison functions with values that are
    internally expressed in each possible way

  * test classification and value comparison functions with a set of
    test values that adapts to the format widths encountered, rather
    than using separate test scripts for each combination of format widths

  * change very lengthy value comparison test function to not run as
    part of the default test suite

  * document that sclnum_*() functions must be given only numeric
    arguments

  * use simpler "parent" pragma in place of "base"

  * in documentation, use the term "truth value" instead of the less
    precise "boolean"

  * test POD syntax and coverage

  * build with Module::Build instead of ExtUtils::MakeMaker

  * complete dependency list

  * include signature in distribution

  * in documentation, separate "license" section from "copyright" section

version 0.002; 2007-09-02

  * in tests, catch exceptions from sprintf("%.1f", $_), to work around a
    "panic: frexp" problem seen on BSD

  * in documentation, note fundamental problem with Perl 5.6 with wide
    integers

version 0.001; 2007-02-05

  * bugfix: in comparison, handle all cases of integers not representable
    as floats (sclnum_val_cmp() and sclnum_id_cmp() were broken for
    systems where natint_bits > significand_bits+2)

  * bugfix: in scalar_num_part(), sclnum_is_natint(), sclnum_is_float(),
    sclnum_val_cmp(), and sclnum_id_cmp(), avoid side effects on zero
    arguments (sign of a zero can be lost via perl bug #39875 in perl-5.8
    and a similar problem in earlier perls)

  * express large test values in hexadecimal to avoid bugs in decimal
    conversions

  * test handling of zero arguments in all functions, to ensure that
    they don't destroy their signs via perl bug #39875 or the perl-5.6
    equivalent

  * loosen tests on zeroes in t/part.t to cope with perl-5.6's differing
    behaviour for numeric values of strings and dualvar(0, $_)

  * test classification with values originating both as integers and
    as floats

  * revised set of classification and comparison test cases

version 0.000; 2007-01-23

  * initial released version