|
2.00 Wed Apr 13 16:33:35 2005
This is a complete re-write of Compress::Bzip2. I left the in-memory
compress routines much as is, compress and decompress.
(re)created by h2xs 1.23 with options
-b 5.6.0 -n Compress::Bzip2 bzlib.h
I wanted Compress::Zlib compatibility cause I have s/w that uses
Compress::Zlib and I wanted an easy migration path, since bzlib gives
about 20-30% better compression on ascii files than gzip.
This is a 2.x release. Since there was a change in authorship, I
figured I should bump the major version number. I could have easily
given this a slightly different name. Given the minimal functionality
of version 1, that didn't seem like a useful thing to do. Heck,
version 1 doesn't even work with bzlib 1.x, you have to patch in the
BZ2_ prefixes, so my guess is that Compress::Bzip2 isn't being used
in a lot of places.
2.01 Sat Apr 23 2005
- cleaned up for 64bit, HPUX PA-RISC 2.0W and itanium
- fleshed out the Compress::Zlib compatibility, added a
couple of tests for it.
- added bzip2, bunzip2 and bzcat, which act like the bzip
commands.
- added some compatibility for windows, can't test it out really
- added the bzlib library underneath
- changed the test suite so that it does not rely on bzip2
being installed.
- added a test to the Makefile.PL to determine whether or
not the internal Bzlib is necessary.
2.02 Sat Apr 23 2005
- 2.01 rejected by PAUSE because the META.yml was invalid
- learned how to do "make dist"
2.03 Sun Apr 24 2005
- didn't have config.in in the MANIFEST, the tarball was unbuildable.
- rewrote the INSTALL, README to give help on the config.in options
for locating your bzip2 installation.
2.04 Mon Apr 25 2005
- tested different styles of builds. confirmed some errors in the
Makefile.PL test for bzip2 that were pointed out by Sisyphus.
- Sisyphus reported some changes necessary for windows builds. I
injected those changes into the code. one of them, adding
aTHX_ to the call to PerlIO_binmode seems like a perl version
thing.
- added tests for internal_bzlib, copied from bzip2 1.0.2.
2.05 Wed Apr 27 2005
- cleaned up the pod so it's more presentable online.
- win32 builds - the eye of the needle. Sisyphus has been doing most of
the work.
- cater to win32 nmake - :: ok but $^ not
- cater to win32 dmake - :: bad but $^ ok
- cater to win32 - use File::Spec catfile and rel2abs all over the place
2.06 Fri Apr 29 2005
- build for bzip2.exe fine tuning for win32 nmake
2.07 Sun May 1 2005
- bzlib-src/Makefile.PL had a stupid mistake in it that shows up for
win32 nmake. I can't test that stuff out. Plus, got some new
info from Sisyphus, so tossed my scribble and put in something
known to work.
- implemented Compress::Bzip2 1.03 compatibility layer to encompass
the post-1.00 development thread taken by Marco Carnut and David
Robins. Made an effort to include the three .t test files asis.
- Why? Cause I found that PerlIO::via::Bzip2, by Arjen Laarhoven,
is built on the streaming interface of 1.03. From looking over
it, I gather it redirects a 1.03 compression/decompression stream
to/from a file. It doesn't do the extra bit of fiddling with the
trailers to make the resultant file usable by the bzip2 commands.
Additionally, the 1.03 decompress stream doesn't handle the pause
when an embedded EOF marker is found, which is an issue for files
larger than the buffer size (900k by default).
So, I figured it would be easy to provide a compatibility layer
for packages that use 1.03. By using my stuff, the headers and
trailers would be automatically taken care of, as would the EOF
marker. Something like PerlIO::via::Bzip2 would then
interoperate with bzip2 commands. Sure, I think bzread and bzwrite
are better choices for file IO, but this is perl, right? There has
to be at least 10 ways to do the same thing.
2.08 Wed May 11 2005
- Sisyphus got the win32 nmake problem licked. merging his stuff in.
- perldoc changes for bzip2 1.03 compatibility.
- got ownership of the module on cpan
2.09 Tue Aug 9 2005
- change email addy
- patch for test 041-error.t, thanks to Dave Evans
- prototype error for bzwrite, limit variable was not optional, thanks
to Christian Drauch for pointing this out.
|
