------------
Version 0.43
------------
* fixed: undef values returned by unpack were not writable
  and caused problems with almost all versions of perl
* fixed: sizeof() could return garbage when used on array
  types
* fixed bogus warning generated by pack when explicitly
  passing 'undef'
* improved warning messages generated by pack
* added error when using offsetof on bitfields
* added new initializer method
* the offset argument to the member method is now optional;
  without the offset argument, a list of all possible
  members is returned
* code cleanups
------------
Version 0.42
------------
* added new 'ieeefp' feature
* new floating point conversion code to support both IEEE
  and non-IEEE sytems
  o on machines using the IEEE format to store floating
    point numbers, you can pack/unpack arbitrary byteorder
    floating point values for all supported native sizes
  o on machines not using the IEEE format, packing and
    unpacking of floating point values is restricted to
    the types natively available
* floating point conversion checks and warnings have been
  improved
* added floating point format detection to ccconfig
* some code cleanups
------------
Version 0.41
------------
* added support for gcc's asm syntax
* some cleanups
------------
Version 0.40
------------
* same as 0.13, just to make CPAN(?:PLUS|-indexer)? happy
------------
Version 0.13
------------
* changed behaviour of member method in list context
  
  o doesn't return the member's type any longer
  o returns all matching members instead, if multiple
    members match the given offset
  o this change is NOT backward compatible
  o to get the old behaviour, turn
      ($member, $typeof) = $c->member( $type, $offset );
    into
      $member = $c->member( $type, $offset );
      $typeof = $c->typeof( $type.$member );
* fixed lots of memory leaks resulting from using sv_dec()
  where SvREFCNT_dec() should have been used instead
* fixed memory leak when object construction fails
* fixed sourcify for enums in special cases
* fixed segfault in debugging mode
* removed lots of limitations and fixed lots of bugs when
  dealing with member expressions
* changed module/ccconfig versioning; now
    $Convert::Binary::C::VERSION
    $Convert::Binary::C::Cached::VERSION
  as well as the version of ccconfig are all set to the
  version of the distribution to avoid confusion
* added new typeof method
* added dwimmy handling of offset suffixes
* added basic type support for many methods
* allowed basic types to be used without parsing any code
* allowed enum/struct/union prefixes for enum, compound,
  struct and union methods
* improved performance when using a threaded perl binary
  by about 20%; performance is now only 5% worse than
  with a non-threaded binary
* added new "understanding types" section to documentation
* fixed some inconsistencies in documentation
* added some new warnings
* added lots of new tests, some based on coverage analysis
  using gcov
* fixed some tests that could fail due to random binary
  input when (un)packing floating point values
* used ppport.h for compatibility
* added check for unsupported perl binary configuration
  with usemymalloc=y _and_ threads
* lots of code cleanups
------------
Version 0.12
------------
* fixed some bugs in ccconfig
* fixed dealing with cache file corruption
* minor fixes in documentation
* use Devel::Tokenizer::C for tokenizer generation
------------
Version 0.11
------------
* fixed problem with recursive data structures in sourcify
* minor cleanups
------------
Version 0.10
------------
* added --version for ccconfig
* cleaned up and fixed format strings
* updated README
------------
Version 0.09
------------
* fixed problems with va_list on Linux-PPC that caused some
  tests to segfault
* updated preprocessor output detection in ccconfig for
  "smart" preprocessors
* minor fixes and cleanups
* minor documentation updates
------------
Version 0.08
------------
* added ccconfig --nodelete option
* added context info for structs/unions/enums
* removed 'c99' feature
* documentation updates and fixes
------------
Version 0.07
------------
* fixed packing/unpacking for #pragma pack
* complete rewrite of member() algorithm
* added 'KeywordMap' option to add keywords directly to the parser
* added new behaviour for EnumSize => -1
* extended ranges of some integer options
* integrated new version of ucpp
* added 'ccconfig' script
* example scripts are now generated
* added and improved tests
* documentation updates
* lots of other tiny tweaks
* deprecated 'c99' feature, will be removed in 0.08
------------
Version 0.06
------------
* removed options 'HasVOID' and 'HasC99Keywords'
* added 'DisabledKeywords' option, which replaces both 'HasVOID' and
  'HasC99Keywords', and allows to disable a lot of other keywords, too:
    HasVOID => 0         -->  DisabledKeywords => ['void']
    HasC99Keywords => 0  -->  DisabledKeywords => [qw(inline restrict)]
* updated parser to ignore type definitions within code blocks
* fixed an access violation when cleaning up the preprocessor
* removed warning when using sourcify method
* added some more tests
* added examples directory and some example scripts
* added more examples to documentation
* documentation fixes/updates
------------
Version 0.05
------------
* fixed a bug that suppressed the error message
  when the cache file could not be created, and
  added a test for it
------------
Version 0.04
------------
* support for 64-bit data types on all platforms
* additional data types
  o long long
  o long double
* new methods
  o clean -- cleaning an object
  o clone -- cloning an object
  o depencencies -- retrieve an object's dependencies
  o sourcify -- emit an object's parsed data as C source
* changed method behaviour
  o configure
    - The 'HashSize' option has been removed. I didn't like
      it anyway, and since the hash tables are now growing
      automatically, it's completely useless.
  o def
    - Instead of undef, 0 and 1, the def method will now return
      undef, '' (the empty string) or the kind of the type you
      passed as an argument, e.g. 'typedef' or 'union'.
* added method behaviour
  o configure
    - Calls to configure (and other configuration methods) can
      be chained together when they're 'set' method calls.
  o parse, parse_file
    - All method calls that have no return value will return a
      reference to their object, so method chaining is possible.
  o pack, unpack, sizeof, offsetof, member
    - These methods now can all operate on compound members, i.e.
      you could say $obj->sizeof( 'type.foo.x' ).
  o member
    - The member method now can optionally return the type of the
      requested member.
* new Convert::Binary::C::Cached module
  o cached Convert::Binary::C objects
  o identical interface
  o can speed up parsing by up to 10 - 20 times
* increased speed for packing and unpacking arrays
* increased portability
* added new and improved existing tests
* some bug fixes and a lot of code cleanups
* upgraded to bison 1.75
------------
Version 0.03
------------
* fixed default sizes for floats and double in object constructor
* fixed some 64-bit architecture issues
* fixed bug in bison parser
* fixed problems with some tests
* added support for dec_osf platform
------------
Version 0.02
------------
* decreased the complex test file, which speeds up the complex test
  and decreases the size of the distribution file
* decreased the number of threads in the threads test
* added copyright notice to the README file
* minor corrections/cleanups
------------
Version 0.01
------------
* initial release