commit d345669c9b65345553bec948c39d2d32b803770b
Author: Brian Harring <ferringb@gmail.com>

    mark 1.0.1

commit b55b9282b2fe3632cf876093f45a3d16a559b4ec
Author: Brian Harring <ferringb@gmail.com>

    use the slightly cleaner googlecode URL

commit 50a6bdfdcc4a3de57a8d93fa08a36a816c956a80
Author: Brian Harring <ferringb@gmail.com>

    drop the nocxx patch, and update log generation to git

commit 33a695eec090da92e755127d46cb5503c4797ab8
Author: Brian Harring <ferringb@gmail.com>

    update news w/ xz/lzma support, and version of gcc -Wall is clean on

commit 86e9b1fecceb7d4ecffe9d9286a58836623e6faf
Author: Rafael G. Martins <rafael@rafaelmartins.eng.br>

    added XZ auto-detection support.

commit c6565e078c89e2f7428cc99135810b006ae13f95
Author: Rafael G. Martins <rafael@rafaelmartins.eng.br>

    added XZ functionality to libcfile.

commit 2f23f1115ab784876d9b794a7ae6ff35f80b9eca
Author: Rafael G. Martins <rafael@rafaelmartins.eng.br>

    added liblzma build system checks, using pkgconfig

commit 4cb354f1616e0d097f6b59635accbd4bec68dde9
Author: Rafael G. Martins <rafael@rafaelmartins.eng.br>

    fixed typo when enabling dcbuffer debug

commit d526b75308a42c6259d78ad61f85c531f99e2868
Author: Brian Harring <ferringb@gmail.com>

    update to automake 1.11

commit 03fa8290635a0c3cf8f77d18bb70d8ab24dcb4fd
Author: Brian Harring <ferringb@gmail.com>

    no longer needed

commit 723f3b32ba776ef2afca8b197b649d8c142debf5
Author: Brian Harring <ferringb@gmail.com>

    fix whitespace

commit 4bb2045d41b87cdae73a9e2858b13bd21afabcda
Author: Brian Harring <brian.harring@intel.com>

    update copyright/license specifics

commit 95c18477dea9c9dd2b362ff40e5a7762195e8d98
Author: Brian Harring <ferringb@gmail.com>

    being that I wrote this in full, I thus deem it to be BSD 3 clause.

commit 66d9049a804c5d7f1cd41813a865046c1a14df6e
Author: Brian Harring <ferringb@gmail.com>

    update the url away from sf.net to code.google

commit 8110be79d61b4e9a8b09e029672d0729c3c667cb
Author: Brian Harring <ferringb@gmail.com>

    add .gitignore

commit 5445782aecc5eca7fe37e508fcd030fe9c9c5504
Author: Brian Harring <ferringb@gmail.com>

    placate automake; touch ChangeLog

commit 6c3bf03829ce592a306c7930ae172efa37181c29
Author: Brian Harring <ferringb@gmail.com>

    remove pointless ChangeLog touching

commit f6d0c71278f7536bd3811fbf2caec06fd09ea349
Author: Brian Harring <ferringb@gmail.com>

    add initial io backing layout to the struct; unused

commit e5210d9329336f5660c3b6000e10217033578fad
Author: Brian Harring <ferringb@gmail.com>

    add note about warning cleanup

commit 234ec297c437f2d7bc591cce760a65e166a548e3
Author: Brian Harring <ferringb@gmail.com>

    invoke autoheader after autoconf, else config.h/config.h.in are stale

commit 0bc32afddaf76e9e9644d9094ac2c13b6e1dbd36
Author: Brian Harring <ferringb@gmail.com>

    config.h is autogen'd... thus don't hold onto it

commit c3ed8a418ed5d842e23b257069b252e590a6b843
Author: Brian Harring <ferringb@gmail.com>

    remove ChangeLog... autogenerated

commit dfdbd358f959de96a9eac917104a8a9a772f5692
Author: Brian Harring <ferringb@gmail.com>

    fixup random spaces that slipped in

commit b3dab4f21707bc28cab4d2459c0a1bdfc180fe9a
Author: Brian Harring <ferringb@gmail.com>

    add install-sh to ignored list

commit 672bcb74bca79c239c3e1149c060ed1bbd6d5f80
Author: Brian Harring <ferringb@gmail.com>

    bug fix... check the identify_format result rather than the array address. this closes up convert_delta ignoring a failed format identification

commit 1d6ec2c86fc22936f212c59f4ba6dae86ed79836
Author: Brian Harring <ferringb@gmail.com>

    force to long unsigned int since the cl range may be particularly large

commit 5a584ff6d9f359f97e8e034191b59f9afd48c36c
Author: Brian Harring <ferringb@gmail.com>

    add missing string.h import for memset definition

commit ae428e74094ad4baa77ff40316637108b14df762
Author: Brian Harring <ferringb@gmail.com>

    guess it was %d, not %z...

commit 6d6a00f981d3e2fc3e67e9fa2235597cac66cf86
Author: Brian Harring <ferringb@gmail.com>

    remove pointless assign

commit 933818edf6fba1ebd3b4d0f43a3978ba35c4555a
Author: Brian Harring <ferringb@gmail.com>

    fixup size_t printf complaints

commit a4fba5d4bf033f2f0d07aeb9247013d187cefc58
Author: Brian Harring <ferringb@gmail.com>

    more cleanup of autogen'd files

commit 4b21bb72b59512d254bbfd984ef770e3095f8336
Author: Brian Harring <ferringb@gmail.com>

    autogen'd

commit c2c60e030b47a8e5477b2705cfdf17db1b3503b5
Author: Brian Harring <ferringb@gmail.com>

    shift to using memset during initialization for speed reasons

commit b7905c3c83a54d95aeabe88b49beb3b308414ea8
Author: Brian Harring <ferringb@gmail.com>

    minor hack... convert get_chksum to a macro

commit df3334b2ce382de3e9c24d87292bd3e777bfb4fd
Author: Brian Harring <ferringb@gmail.com>

    only calculate chksum when needed

commit 8a49c244854ed4e3e614e165b56922c9dfe9b5db
Author: Brian Harring <ferringb@gmail.com>

    add initial set of ignores

commit f1aae4d1114f53ace3bae1d5934a886917d47cda
Author: Brian Harring <ferringb@gmail.com>

    remove generated file

commit 6ac73f5746017138cbea9cf2693e67dad268123b
Author: Brian Harring <ferringb@gmail.com>

    update to automake 1.10...

commit dd7bc52a86f5f5a0fee5b1778a1ce473fad374f8
Author: Brian Harring <ferringb@gmail.com>

    fix up signage of types

commit b805363e13447cdae002b1f18cdf6eca29b391e4
Author: Brian Harring <ferringb@gentoo.org>

    remove generated files from vcs

commit f183d1ff9a908b55395477846c3730f0aacbfe62
Author: Brian Harring <ferringb@gentoo.org>

    add explict -lm linkage

commit 1802d5d0d928ee2fb328eab8be6b332f3e788da3
Author: Brian Harring <ferringb@gentoo.org>

    remove macro hack, and just use an intermediate func for dcb_add_add

commit dec59e83f232a9afb8fab03a7776259c14ff1fd8
Author: Brian Harring <ferringb@gentoo.org>

    version 1.0

commit 7fcaa4b4fd4083a544e742d1fd4969f182d77442
Author: Brian Harring <ferringb@gentoo.org>

    update to NEWS

commit 647c34882d2980b11a9e4a03fcea9608fdb52722
Author: Brian Harring <ferringb@gentoo.org>

    tweak...

commit 8883386b5fb211aa0a7bc2cad1c7ee7e0d09da60
Author: Brian Harring <ferringb@gentoo.org>

    update it...

commit 92e8b18f0365d9d84002cf89a448b6f1a083448c
Author: Brian Harring <ferringb@gentoo.org>

    gentoo.org -> gmail.com, oopyright updates I missed in earlier commits...

commit 67a00ba7cc76d7a5d79c82e98fc10ca55ea4b763
Author: Brian Harring <ferringb@gentoo.org>

    cleanup todo

commit bd5cab154bd41c865b2336921b340247dc84430d
Author: Brian Harring <ferringb@gentoo.org>

    updates

commit 9f2fc402addf71899170d54d587a5551f5f5c5b8
Author: Brian Harring <ferringb@gentoo.org>

    invalid memory access, if it's end of bucket insertion (and resize is needed), don't try a chksum check

commit 0ed62b6892b98fae75aff1b322c691ed3babee5d
Author: Brian Harring <ferringb@gentoo.org>

    remove a daft partial comment

commit 560993a4f67f68b293d0ea3dbb642c0dc120eea5
Author: Brian Harring <ferringb@gentoo.org>

    missing stdlib import for malloc

commit d0bb90f3cd862ef4912d8c57bce1c65968d7019f
Author: Brian Harring <ferringb@gentoo.org>

    patch used during make dist to gut the cxx check from configure

commit 0c602ff81d4822a237c66a6926d7932191c73018
Author: Brian Harring <ferringb@gentoo.org>

    make hr_size (hash ref size) abided by, and use the defaults for diffball binary

commit 8406fd0e394c6cdd3d30cde91669aeaacfe7748d
Author: Brian Harring <ferringb@gentoo.org>

    cvs2cl -> bzr log

commit 8dda930c46be7524c4c9f518dc0557d1f5e94e7a
Author: Brian Harring <ferringb@gentoo.org>

    heavy cleanup; leveled quite a few more asserts to ensure that the buckets are properly ordered and contain no duplicates; improved search funcs to do a switchover dependant upon # of elements to search also

commit 9d97b29d26deb7c885150729d7a611a85c6caa2c
Author: Brian Harring <ferringb@gentoo.org>

    drop unused chunks of the adler32 struct (parity mainly), nuke dead code, and shift to general types where applicable for compiler to be able to optimize

commit 5105fcfa8e8c7d8ae34f06c5ac9e2ee2802e62c9
Author: Brian Harring <ferringb@gentoo.org>

    few long -> int changes, broke find chksum insert position function out of base_insert, and fixed a long standing (but undetected) bug for insertion resulting in an invalidly ordered chksum array. levelled few asserts to make sure this doesn't occur again...

commit 3f8fdacb058858f6a5375e73ce1d0eca996aed8e
Author: Brian Harring <ferringb@gentoo.org>

    remove uneeded config.h

commit 7a0f7e37d7f07c8f37b21da03db0bca9d2d2bdf3
Author: Brian Harring <ferringb@gentoo.org>

    inline internal_loop_block, and find_chksum

commit 7e18377212ade9126ef3d637bd2413e938cd41e3
Author: Brian Harring <ferringb@gentoo.org>

    inline update

commit 8d116b187c9d86569d597ddbe815fb2f005941f7
Author: Brian Harring <ferringb@gentoo.org>

    nuked intermediate build files

commit 9fc459d90e8a6220f9de31bdfa8d7e60d2df795f
Author: Brian Harring <ferringb@gentoo.org>

    rebuild...

commit 41690627881530031f57bd6cedc0cb7e64e838c8
Author: Brian Harring <ferringb@gmail.com>

    cvs->bzr conversion

commit 8c9f56b83d87f7dd4a400b2646b48dd1a7c78c44
Author: bharring <bharring>

    bumping to 0.7.2

commit 601357530e5e34a42e98cfd66aca460906aa7237
Author: bharring <bharring>

    alright... bit of cleanup/fixups.
    rh_rbucket_resize no longer allocs, then stupidly reallocs the same size
    different hash_size's are now possible (code handles it).
    Currently controlled via headers...

commit a2fc9ad5f7f380373f478107630c6337d9de804f
Author: bharring <bharring>

    hand off to the backend, let it decide on it's own for appropriate hash size

commit d8efd51028c99aad7856abaa88fd46f5cdad876b
Author: bharring <bharring>

    addition of configure.orig removal note

commit f6aca78a6dd54a5b7b69d5334e86cebbdd993417
Author: bharring <bharring>

    email address updates, and addition of 0.7.1 notes

commit 61a43a5fdd9595f1a771df8b0db01ab66b59b2ab
Author: bharring <bharring>

    updated manpages with appropriate version

commit 406cd624138b2f2a6f33344d90b118bcb3ea9920
Author: bharring <bharring>

    include release tag's into ChangeLog, and a quick change to my local testing script

commit 8ea3262503d4766c6e55d8ce97803de3619a4e25
Author: bharring <bharring>

    updates for 0.7.1, plus ynking of configure.orig from make dist

commit 43805b6a83742331b42360884173cc4fcecba4b5
Author: bharring <bharring>

    warning cleanups

commit 7e2f596ce0be1d015df07f689aa10e9535d32ce4
Author: bharring <bharring>

    whitespace cleanup, and io fix.
    flag_lseek_needed was supposed to reset the last lseeker to a different id, but via a typo was setting last lseeker to that cfh_id.  exactly opposite of what's desired.

commit c7b32015bc60d43290f85bceb49b8a4c3f9c7606
Author: bharring <bharring>

    make -Wall default cflag

commit f61a02e649a656309346cda38bd08fa8339454a8
Author: bharring <bharring>

    meh, typo slipped in, corrected it

commit 95078d1722ec130ea61cb872bbc6591c185addc2
Author: bharring <bharring>

    exit call that slipped by, removed it

commit 1cd624f0b1da5dc608ce6ae812161125809c0f15
Author: bharring <bharring>

    updated email addie

commit c4e342e33a7a93659540176ef86d2bfbf98d3111
Author: bharring <bharring>

    addition of warnings info

commit 0e94f8bc2ef26de4a4c46ee111784f601952f9ea
Author: bharring <bharring>

    updated comment about configure and nocxx punts.

commit b979ca88ce273cac16a693558536154b86cba661
Author: bharring <bharring>

    updates/corrections date timeline (since I'm a slacker), and addition of a tentative 0.7 news section.

commit f6a9a83da7b786795093e689da0a144032522b25
Author: bharring <bharring>

    auto ChangeLog generation for make dist (dist-hook rules)

commit f69143243be750e3768017fdfe4434bd88ea0bdc
Author: bharring <bharring>

    meh. fixed the last commit, so that it would only try free'ing when a CommandBuffer was in the ring buff.

commit febfda00916b0cfd51e647317d86daa665bf65f1
Author: bharring <bharring>

    cleanup of memory leaks due to lifting the code straight out of patcher; only leaks possible were during errors, but they're covered now.

commit 742b3342905ba24a46b704ce84db8745479e88ee
Author: bharring <bharring>

    conversion over to the simple_difference api.

commit 62ac8b091c0d574caf69dc9b2a9008ac8e7d10bd
Author: bharring <bharring>

    free the CommandBuffer on the way out, lest you leak.

commit bc63a7bc78ee46c7daeec3bba2e3b54f4a0dce07
Author: bharring <bharring>

    convert patcher over to api.c:simple_reconstruct

commit 2f2e44e8c0a61ea45fca548553b2b7d67604effb
Author: bharring <bharring>

    whitespace cleanup, and addition of simple_reconstruct

commit 55c1b1c97b303ad03e139d5a87d2007f49914adb
Author: bharring <bharring>

    cleanup; if cfile_WONLY, buff must be null.

commit bfacd601cfc865a7130d83deb72305c3afb5993e
Author: bharring <bharring>

    corrected bug from previous change, if the window is empty, null the cfile window so the loop ends.

commit 0f50168182701810517eca104bf4afeb575c8909
Author: bharring <bharring>

    lil trick from the portage early autotooling, disabling the cxx check that libtool screws up and auto-enforces (even though this source is c, through and through).

commit de1597641f900173ccfeb580175f8f945a9857fa
Author: bharring <bharring>

    personal utility for make dist'ing && updating a local ebuild for merging for testing on my box.
    rather gentoo specific I'm afraid

commit 31330425423e44b0bb097c7221161ac1302adbd5
Author: bharring <bharring>

    better cfile_mem_alias handling, plus a buttload of asserts.
    cleaned up the remaining v0printfs, and converted remaining abort to a UNSUPPORTED_OPT return

commit 7bee657a8ee7ea71d2adb45b5241989bdf41363e
Author: bharring <bharring>

    added copen_mem prototype, correct IS_LAST_LSEEKER so that if it's a mem alias, it's always last lseeker.

commit 6e5936e2809097e7a76e1499c1f0d67257c0ba68
Author: bharring <bharring>

    whitespace cleanup (my original sed'ing seemed to have some troubles with my old, insane, code formatting approach).
    Aside from that, converted for onehalfpass vcfh usage over to next_page, instead of a cseek + expose_page.

commit 4f6f575f8fb3ee0ac79fe7cbb4acd029e681ab9b
Author: bharring <bharring>

    whitespace cleanup, and tagged a few comments about questionables

commit f466517b4019f95ef40952d612b52eefd0d20cbe
Author: bharring <bharring>

    initial write memory target, plus an explicit memory read
    copen_mem is what to look into

commit ab0b45f97f79ec816fcfc46fafeca524a4c3fee2
Author: bharring <bharring>

    Shifting around the error defines, so they're usable from the appropriate code, rather then pulling all in (as was previously the case)

commit 6a39b127a5200f148034173287fab5d1ada4ace1
Author: bharring <bharring>

    no need for the errors include.

commit 779e20e8fa9459c5a34d878be518fb06c750394e
Author: bharring <bharring>

    error checking of init_adler32 return

commit f8a6c4605f99deb315cb0d29d9976564129c9159
Author: bharring <bharring>

    removal of hash_stats

commit 60dc27ccd102147162c9958449d3a7ed2f379312
Author: bharring <bharring>

    better error handling in get_nearest_prime

commit 3c1c5b1afc9edabfdb79bc5af3b8182c99831990
Author: bharring <bharring>

    addition of a basic api header, plus differ algo slapped in.

commit 2ec4a8b4b00a8bb1bf3684734872a1ff0e15517a
Author: bharring <bharring>

    for the moment, distribute the config alongside, since it's used in defs.h

commit cbf16510693801a3047269f88242f1c3f364c9b4
Author: bharring <bharring>

    distribute the manpages (didn't I commit this once already!?)

commit 195211dacccdd7a2a04f0d2a0a7bfc75e626d0f7
Author: bharring <bharring>

    addition of central global_verbosity

commit 9cccad55ca3e1a69f3cadf46a7c4e10930affa25
Author: bharring <bharring>

    clenup of global_verbosity handling.

commit aeb694f349083b9f60daa5a39d23400f855e58cc
Author: bharring <bharring>

    no longer relevant, can test via python bindings I'm working.

commit bea4b443a0cd8725a32f58cf48f83962162f586a
Author: bharring <bharring>

    bundle the man pages, bugger.

commit 6fcc8db693b48f4e234f21dc9ad8914cb597a216
Author: bharring <bharring>

    *cough* generate the man/Makefile when running configure...

commit 1dd2ccaeda8928b996ac6215bf60ba50e62f1a83
Author: bharring <bharring>

    man pages added back in.

commit 03f101032a5d88400158fdf2733a20bb38cc4376
Author: bharring <bharring>

    change of the options code, s/Example-/Example:/

commit 3b95cbecefe68b29afa8e47f89c6429792ac699e
Author: bharring <bharring>

    bug in looping code for handling N patches

commit 90b9fb36b243b35ed5b0c6e7b83e9db5c7b55512
Author: bharring <bharring>

    if a search func returns -1 on unfound, don't assign the results of it to an unsigned var. (I know I've fixed this specific bug before :/ )

commit bab4e4cc9799e7e683908bb66baff08c319582fc
Author: bharring <bharring>

    hey hey. it works once again.
    wasn't deref'ing a ptr, as it should've been.

commit 8bfa581fe647f1ea1c7c26a422201c2cefb12aa4
Author: bharring <bharring>

    compilation corrections for gcc-3.4/4.0 (whenever they break it).  condition lvalues assignment, replaced with inline funcs
    converted ENSURE_LSEEK_POSITION macro to ensure_lseek_position function, since the macro was forcing raw_ensure_position calls (not desirable).

commit 8776021b3fef316e2c0fed075268637332edd7fb
Author: bharring <bharring>

    DCB_insert removal, since it was combined with DCB_finalize
    DCB_insert -> DCB_finalize

commit bc2d1f54c18e5b7429acad11b73b5d5fc5cfe569
Author: bharring <bharring>

    fixed DCB_finalize's macro, since it had the nasty habit of lieing.

commit 0e68e215a5da5ed6222d5b267e7098b082294ad9
Author: bharring <bharring>

    largefile var in libdiffball (compilation time), and fixup so user cflags are abided by (again).
    stupid AC_PROG_CC

commit 2ff6bce553ccfa8632408705e7ab67f2d336e14a
Author: bharring <bharring>

    further cleanup, whitespace for string-misc.*, _GNU_SOURCE , not __USE_GNU for strnlen errors, and fix of hash.h so it includes
    the prototype for base_rh_bucket_hash_init

commit 4512b70cded8a0cf5766399fe6236a51926f29ed
Author: bharring <bharring>

    cleanup, conversion away from off_u64/off_u32/signed long/unsigned long to size_t and ssize_t

commit bd9e99ebd1ae853d025cbfead06ad5b0802e01cf
Author: bharring <bharring>

    removed unused compressor var; don't yet allow compressing for writing...

commit 5a05b9f895173745336537ecb236d2192df632d1
Author: bharring <bharring>

    missed one

commit 411c502eb2f77dcb3f40bb55036478eeaaf4d0ea
Author: bharring <bharring>

    major change up of autotooling

commit 5f32c21b5ae2abff0a18fdb7b44692bca9180b0d
Author: bharring <bharring>

    moving em.

commit 0a3d93aa1a129ff883aca5f2ebe4d235a20c367c
Author: bharring <bharring>

    AM_CONFIG_HEADER -> AC_CONFIG_HEADER

commit 59dd5d1e58dd546e9b6409afb70fe65f4607979a
Author: bharring <bharring>

    tweaks to make it a bit more verbose about how/where it died, if it dies

commit 4b1ef24ffa429430191254c0e41438236dbd96c1
Author: bharring <bharring>

    code cleanup (ongoing).
    mostly whitespace correction, and yanking dead code.

commit 8f5111324e7ca50785dc85d83c4177589a6b2ba0
Author: bharring <bharring>

    whitespace corrections

commit 76a50f03c6402459e47cd5622abfeed27a6de3a7
Author: bharring <bharring>

    tweaks so it works properly still

commit db16868858d4e129e73da7582896f593d1f0188b
Author: bharring <bharring>

    alright... cfile changes, mainly.
    changed copen so it opens the fd on it's own, added copen_dup_fd that has the fd handed in.

commit c32242226beeabb625b1d077661d2695fee596c5
Author: bharring <bharring>

    autogenerated, no need to store it.

commit 40e0c17a65804024c72ba830fa8a941ce30b1ebd
Author: bharring <bharring>

    no need to bundle this in cvs

commit c2178854b1a816c304a50316b8412f492f7525d5
Author: bharring <bharring>

    conversion to sane tab usage instead of previous fucked up style

commit 2aa4703178215c12bc9deaf6dde47a38759092d8
Author: bharring <bharring>

    final man page updates.  0.6.5 is ready.

commit 2e92314d291190287277dc8e267c98d957496c9d
Author: bharring <bharring>

    warning cleanup.

commit 01b167372b744059dcf192bf441c710b7b08adb5
Author: bharring <bharring>

    man page updates.

commit 117b854854ef54f53f6b45ec89db5cb2bbb4f18d
Author: bharring <bharring>

    Cleanup.  branding it 0.6.5, have DCommand_collapsed fully working (so convert_delta works across the board now), fixed a couple of
    bufferless logic bugs in patcher, updated news.

commit 48b19657e4ba1d5f9087ca4610098e927f5041bf
Author: bharring <bharring>

    whitespace cleanup.

commit 09e2a9b73fa10a93272336d3189bfcec565d1033
Author: bharring <bharring>

    basic code for collapsed dcommand.  Cleanup in patcher also, and changes to init func for no_buff.

commit e4d612747cf57c73b848d3e3a75459ba50c9c896
Author: bharring <bharring>

    ability to collapse multiple patches into one, sans switching.

commit 2da638cc9e5cd4032b53881fd5d8402ea2bb4fdb
Author: bharring <bharring>

    could've sworn I commited these printf fixes for recent changes... meh.

commit 097d0e6d95c33f8e634919a9a4718fad495290cf
Author: bharring <bharring>

    pointed at berlios

commit 8e52171a8ef79b0c2d918299ad7b4daddd0c6aa4
Author: bharring <bharring>

    update to authors email

commit 63d2b04d54b47963cef89411b39c415c09a6e34c
Author: bharring <bharring>

    Collection of updates for largefile crap, thanks to wizeman for amd64 specific info (and pointing out the previous index/quanta bug
    via hitting the segfault :)

commit b4b052955b7e1332eee026df25a40926fcf0d188
Author: bharring <bharring>

    off by one alloc error for dcbsearch

commit b3934afed8d2f85efab05d83f1ac4ffff437f1df
Author: bharring <bharring>

    remaining silencing of handle registration.

commit d4e78b7539f42023cd76158d89ea0af6fad12c62
Author: bharring <bharring>

    silenced registration of handles slightly.

commit 7c92d624cd80097c2c3ed4ad5d75e584ad549293
Author: bharring <bharring>

    install libdiffball.so if profiling isn't on (had it inverted).

commit 0f5fe02c3a10d641f40a894d6728da310b21ceb8
Author: bharring <bharring>

    used for autogenerating autotools crap.

commit 3d5b6b849d960591cb24062deb26ecc808f17771
Author: bharring <bharring>

    Fixes to Makefile.am, added buffersize command opt for patcher for read_seq_write_rand.

commit f90168e637d027abca7a38a2f7ab307dbdafe92c
Author: bharring <bharring>

    copyright header update, shifting DCB_SRC_ID type (instead of unsigned char) into use, fixed a long standing (and stupid) bug in
    multipass for first run against a file.

commit ea8d57a0dd7bdf9b949abc25d606e1be48a9e3a8
Author: bharring <bharring>

    Ugh... shoot me if I ever let if/else if/else if|switch type support accrue again.
    Broke it all down into individual functions, abusing function ptrs (yes c++ might be wise here, but I also care to do
    without the c++ req at this point due to reasons tied to gentoo's stage1's).

commit fee0e9316a9ad798ecfc9fbb7f64a2aaca54099f
Author: bharring <bharring>

    broke RefHash down so new hash implementations are pluggable via func ptrs.
    collapsed insert_block and find_matches down to common code, and updated code as needed to use the new code.

commit 8eac5fbfd911dfae4995fbfddd8f97982608ea6f
Author: bharring <bharring>

    Collection of tweaks, fixed a missing include in errors.c, bumped vers # to 0.6.2 (will be tagged and bagged in cvs when released), usage
    output/missing files specified as args error output cleaned up, and oh yeah, all diff-algs/formats code is now schlopped into a lib, libdiffball

commit b71e37d91443d4a1f05ab0fa6ae5a8add4aa1142
Author: bharring <bharring>

    fixed the usage string

commit 773194d222f1fb92a6d7db5128ca65c8d263fcbd
Author: bharring <bharring>

    Removed libpopt dependency for arg parsing, fixed a segfault for handling multiple patches.

commit 8736ca9387bd7644f003f7d16d9b3a31228ec619
Author: bharring <bharring>

    patcher.c contains a fix for reorder/bufferless, and cfile.h is just a tweak of a comment.

commit dbd867b77f1885380c5a7fd21c9e00624591bd51
Author: bharring <bharring>

    Cleaning up top level return value checks, updated copyright year, added usage/help options.  Cleanup somewhat.

commit 93ee6b6f9bbb3f2c3a0ade32d233b762c2bc84bc
Author: bharring <bharring>

    Quieted an obnoxious warning, that wasn't relevant.

commit 50d51d8b268d022d7db8866586a41ba686e92f2e
Author: bharring <bharring>

    Removed the old OneHalfPassCorrectingAlg implementation.

commit b894e75da76d63f16b869bba6530c16ea66d90dd
Author: bharring <bharring>

    K.  bufferless reconstruction is now added, including necessary checks/handling in patcher.c.
    Added _REGISTER_VOLATILE_(ADD|COPY)_SRC to be used as a hint that during reconstruction of a command buffer, the code adding commands
    may potentially be seeking in the cfh (in other words, don't screw with it's position- use cfile_dup_cfh (new function), duplicate the
    handle, and work w/in the dup letting the reconstruct code do it's thing w/ it's original cfh).
    Duplication of the cfh only occurs when the dcb is a bufferless type, rest of the time it just functions as if DCB_REGISTER_(ADD|COPY)_SRC
    was called.  Like I said, it's used as a hint.
    Aside from that, old commented code cleaning.

commit 5218703db61ed726e321df6f15f1640ae0193524
Author: bharring <bharring>

    the need for read_seq_write_rand is now determined in patcher.c, dependant on files involved being compressed, formats, etc.
    quieted bsdiff.c's patch processing function, added --help option to options.h, added commentary to ChangeLog regarding
    intent on using cvs2cl.pl (the file *should* be regenerated via cvs2cl prior to make dist's)

commit 4a2b6dabf986cdfe380a89a60f99201b836b4c2b
Author: bharring <bharring>

    Banishing them to the attic.  tar-cmp/dump-tar-headers are useful on occasion, update-cvs I haven't used in ages.

commit d42a3f68b3fa758a4245093b0dc4a9fa46e46ecf
Author: bharring <bharring>

    updates to the basic info regarding the plans for diffball (roadmap, todo, new features, etc)

commit 720d31fe212ca4f69d7edd23d5a157bde06ee4aa
Author: bharring <bharring>

    err, forgot to include these...
    Rather important, last few commits fail w/out these files.

commit 886633c456de21f3cb595e037b22193bb995c0dc
Author: bharring <bharring>

    Bumped to 0.6.  Intended release version, after things have been completely stabled.
    Need to move the looping/collapse code into patcher (out of apply-patch.c, also rename apply-patch.*).
    sorting is a bit quicker, DCB_collapse now dumps commands into their own command_list dependant on their src_id, avoiding that comparison,
    also giving finer grained control over releasing of memory.

commit 3c91d40eff9aef1b17a794bdd10db0e612e2e345
Author: bharring <bharring>

    forgot to remove data-struct.h include and header entry in makefile.am.

commit 020e98f914e6ca62cc56c6b6f0da82d763620567
Author: bharring <bharring>

    Another long dead file.

commit f4a26e241afc4ab7745eee5d222f4286765beb9b
Author: bharring <bharring>

    Hmm. This file has been dead a *long* time.

commit 5042a29850e21154913f1ef741b8b7ac33ea0629
Author: bharring <bharring>

    Ohk.  DCB_rec_copy_DCB_src, add_overlay, and add_copy have been modified so that overlay srcs are correctly
    processed.  so far, ./patcher ccache-2.2.tar cc.bsd ccr.bsd cc.bsd ccr.bsd == ccache-2.3.tar.
    So I've got that going for me (enough caddieshack references).
    Cleanup here and there, levelled *shitload* more asserts, looks like merge version support is pretty much finished.

commit 78e968326689b1aa6d51c5a4bb9c70c6e9498b58
Author: bharring <bharring>

    Closed a minor 484 byte segment not being free'd (poptFreeContext).

commit e9c385ce96cfd5aaf16efdf554d805e423b844d7
Author: bharring <bharring>

    Lots o changes.
    DCB_register_dcb_src now does something (specifically register a dcb as a merge key,
    on the fly looking up and inserting the appropriate chains).
    So I've got that going for me.
    Also, ought to  works across all ov_chains (recursive).
    Problem is, bsdiff's read function isn't recursive in pulling commands. need
    to do something about that.

commit b4a2667fd49f655f53aebd80c90cd513ed34dcca
Author: bharring <bharring>

    Long standing warning being fixed; basically the function definition for
    default_dcb_src_cfh_read_func was unsigned len, rather then unsigned long len.

commit d61360489053ae59e0c3c0ec93008a499973eaed
Author: bharring <bharring>

    merge versions is temporarily broke, rearranged dcb->DCB.full.(lb|buffer)* into struct command_list, typedef'd
    the same as ov_chain's struct.  Best to keep things similar, easier to manage.
    moving version merging into DCB_add_(overlay|copy), using a registered DCB_DCBUFFER.

commit 5bc5c1bd156fd101c546b674b946c723c5c79403
Author: bharring <bharring>

    Cleaning up dcb->srcs[x].ov

commit e4fe356dc30a0d5fa9e3cd84cd5ba3ab53d61b0f
Author: bharring <bharring>

    Re-worked how overlays are represented in the buffer.
    now all copies are in ov_chain for an overlay; the entry in full is just offset=ov_chain offset, len=# of commands in ov_chain to use.

commit a9ae08eb05ed646a3cdf61bba76c3984acd3d69d
Author: bharring <bharring>

    Fix to cseek for the 'avoid lseeking by next_page if applicable' such that behaves correctly w/ CFILE_WRITEABLE.
    Basically, disable next_page cseek for writeable.  better method I'm sure, but this works.

commit 9bc974804f0e353453b0891b6383e0567a53cf99
Author: bharring <bharring>

    Happy documentation/commentary updates.
    Yay.

commit 067a0abd3fb533af4244d4a21256f37a1e5db519
Author: bharring <bharring>

    Remaining commits, Makefile.am == search-dcb.* additions, convert_delta.c == adjustments for the encoding function changes (passing an int
    src_id rather then the actual cfile).

commit 9383c0f7431ac74ad755f0f424c2eacc52b50ccc
Author: bharring <bharring>

    Missed this commit- clean up for DCB_collapse, basically don't alloc a crapload of memory for an array that isn't used (dc_array).
    This is due to DCB_collapse passing back a malloced section.

commit 4ee8133e2d6d75d74029012a7c2feac473e92b50
Author: bharring <bharring>

    modifications to use multiple patches for reconstruction (both command line control, and actual code for handling it).
    search-dcb.* will likely be merged into the dcbuffer mess (which later should be split up :), but that'll come after applying n patches is
    fully supported- as always, support for overlay formats is lagging behind non-overlay :)

commit bc3f1193fb502be44d64ea7ffec7fc963c4fffa3
Author: bharring <bharring>

    split tar.c into a looping function call for retrieving tar entries, and an actual tar_entry function (for parsing the entry).
    Also removed the requisite to read through the whole file- that's goofy.
    diffball has had a couple of tweaks, main one being addition of a defacto DCB_insert call (should check buffer type first thinking about it),
    and making the target tar_entries array into (tar_entry *), rather then (tar_entry **).  no point in wasting space allocating that extra
    layer of pointers, since there isn't any sorting done to targets.

commit ee19eee5ccde34345f34a10814d52f1c36552133
Author: bharring <bharring>

    Changes to the encoding calls (now a dcbuffer src_id int is passed in, with actually registering the src being upto the callee).
    This seems backwards, but allows for multiple-patch-in-single-run support, and gives control to the callee.
    Also, cleanup in get_next_command, and minor adjustments to dcbuffer structs.

commit 777d4cbd767530840ba17516d6a671286ef20657
Author: bharring <bharring>

    autohell update.

commit 8d92e62f72a526cf069d87955d823bd83b552b56
Author: bharring <bharring>

    rewrote OneHalfPassCorrecting to use the expose_page framework (one less buffer)- also fixed a bug where onehalf would call update_adler32
    specifying a block that extended past the actual allocated end.
    Seems a bit quicker.

commit 9c1e842ec0bcea1bff81a2ab57146740f5e4d98b
Author: bharring <bharring>

    3 tweaks- one to to cseek to reduce lseek calls (if the desired pos is in the next page, refill),
    second tweak is to crefill- for bzip2/gzip it was returning after *any* data had been written, rather then upon a full refill.  This
    screwed with the first tweak, and seemed suboptimal behaviour, so now it refills data buff fully (looping as needed)
    third tweak is to prev_page- it would on occasion attempt to seek into negative numbers, due to assuming the page was always aligned on
    size boundaries.  if current_offset - size <= 0, it cseek_set's to 0, otherwise, current_offset -size cseek_cur.

commit c393775f9e91049320623e4cbdc467a3c031288b
Author: bharring <bharring>

    Should be a fully working version of reorder_for_seq_read now, including
    overlays.  cfile now has RW support, which needs clean up/optimization.

commit 8659719b66733ebd014a9936946a91d380ea2d72
Author: bharring <bharring>

    Was calling cclose, but not close for fh's.  Fixed now.

commit aa1d967424ed4daf2368e1cbe8880ff95a85ec68
Author: bharring <bharring>

    bsdiff reconstruction back up and working now, although it doesn't work w/ reordering command access for seq src reads.
    need to write bsdiff_read_func.
    From there, adjust register_overlay_src so that it takes a function ptr to a src_overlay_func, so that for patching/reordering the read
    can be ran, but the mask is applied seperately.
    That and reorder_for_seq_access needs to be adjusted to *recursively* pull all overlay_chain srcs.
    Busy busy. either way, got the general lay of the land back to where finishing reorder_for_seq and applying multiple patches in one run
    is *much* easier.

commit cad324b3a63a99c321f6a967dc262942a916d188
Author: bharring <bharring>

    An attempt at being witty.

commit 2f70ee706a051f218df67f063a0f5c2ce592fd32
Author: bharring <bharring>

    cfile tests need redoing. later.

commit e6f187716d01f29b274425ae317fd79ec23eda63
Author: bharring <bharring>

    *Cough*.
    Had a check wrong, so w/ cfw's it never refilled the buff upon expose_page/next_page calls.
    Pays to check prior to committing...

commit e829ea1a8bdbf0551660abf4c2046db4c3d4b2db
Author: bharring <bharring>

    Continuing work to bring overlay's back into the fold.
    Minor adjustment to the cfile_window mechanism- need to go through and get it ready for read/write access to a file, currently it
    thinks one or the other.
    Continued internal adjustments to dcbuffer's structure, nothing worth noting.

commit 4937090646b7e12cb46ffda2f7d1915902882134
Author: bharring <bharring>

    Updated readme/todo, and split the goals into roadmap.
    Grouped all registered commandbuffer srcs under a single struct, cleaned up that area.
    Fixing overlay support in read_seq_write_rand is next, possible due to cleanups.

commit 44e83ae10f560d9987bd5304671e72fcd315327c
Author: bharring <bharring>

    Not incredibly useful extra_offsets addition to dcbuffer. using it for overlays, but first cleansing dcb.full a bit.

commit 5e298432659ac45c38c35b254621f5ab895f8547
Author: bharring <bharring>

    Err, added a few missing function prototypes, and commited what was
    missed in the earlier register_dcb_src_dcb related commit...

commit b50bcac1b6b4b809e75eb894fe9ae83a15f832f5
Author: bharring <bharring>

    Basic register functions for register_DCB_DCB

commit 4619909b387d4dd31fd8830aa527e92c2436cc43
Author: bharring <bharring>

    Continued tweaks to attempt and allow for different srcs to be used in dcb->src.

commit 89ceee3aa6f7f00ea6fecfe97e5c488d1398e420
Author: bharring <bharring>

    Tweaks moving dcb->src_cfh to dcb->src.cfh.

commit a51338eb33758e3fc36eef6043f8f93077940cfd
Author: bharring <bharring>

    Few tweaks, cleanup mostly.

commit 7b4667b084abbfa7903eb58291b63d169cdbbc2d
Author: bharring <bharring>

    Minor tweak, returning 0L rather 0.

commit 112d14324bf579b1fdefd2ef931010a4245ef912
Author: bharring <bharring>

    Configure checks weren't puking when libs weren't found.  That's been
    rectified.

commit d6c20aeb9fd85609542ae3e134370f28c1295199
Author: bharring <bharring>

    moved the default_dcb functions to dcb-cfh-funcs, with the makefile.am modifications commited earlier.

commit 5ce09696a98b817b1782027c1af57f72e658fab0
Author: bharring <bharring>

    bsdiff.c support tweaked, using the new system.
    dcb_src_copy appears to be working perfectly for bsdiff, read_seq_blah support I've yet to test.
    Side note, read_seq support appears to be working nicely.

commit bad095d7b461a652e9d957b70ae0051ab50422d7
Author: bharring <bharring>

    Working version of reordering alg for sequential accessing of src_cfh's.
    Set as the default at the moment due to testing.

commit 506fffbc1a109f6a4dc183c03517ccb645b65afc
Author: bharring <bharring>

    Fairly massive change to dcbuffer's cfh storage mechanism.  Removed DCB.full.cb*, opting for DCB.full.src_id, eg opening up to 256
    different src cfh's.
    Currently, the add/copy abstraction still exists.  This can (and likely will) be removed, and have the encoding'ing algorithm pester
    dcbuffer, which makes a on the fly determination if it's an add src or not.
    Think that's a better solution, since it open's up the full 256, since now, if a src is both add and copy, dual entries are needed.
    Either way, needs some tweaking. apply-patch.c contains the basic start of a reording attempt to sequential read from src's, random
    writing to the reconstructed file.

commit 469c64eb656b5b717afd7113a8fefec352bf1057
Author: bharring <bharring>

    Added a seekable flag to access_flags, used for determining if a write handle is seekable (conceivably read handle also).

commit 61a20df1d64c76e33f865e4792bf61ed3ac79e7e
Author: bharring <bharring>

    Converted to using a DCLoc_match insted of DCLoc for DCommand.
    This is primarily to adjust things so that reconstruct_file can write the output in a non-sequential manner.

commit 3a0fc9e68d8822326afeb5ced2ebfa28fb17351d
Author: bharring <bharring>

    Seekable write support is working now, although not optimal.
    Need to convert from using pos to indicate how much data is to be written, end should be used for that.
    Once that's done, the code for seeking w/in the buffer can be used for writing, improving that corner case.

commit 6310346f7a8b1d516f047920e177c8c90aabf0cf
Author: bharring <bharring>

    Updated view-patch perl script I've been using.

commit b23e6e26242497c60d532f2be2a4864879847e2d
Author: bharring <bharring>

    Minor modifications to enable CSEEKING on a WONLY handle.

commit 3ccc7ab9d6b0b8b8b0788fd3dfdb609b1106ea9e
Author: bharring <bharring>

    Cleanup of v0->v2printf, 2 mild fixes in cfile to catch when an error is
    returned, and modified patcher to allow using compressed reference
    files.

commit 87c4309f9847d4ad8e5a796bfaf7a98804fc0b26
Author: bharring <bharring>

    Hokay, gzip/zlib should be working perfectly for reads, exempting corner
    cases (not sure on them).  Fixed a couple of *very* dumb mistakes in
    internal_gzopen, stuff that is blatantly stupid (mind you it is a
    follow the general method check_header laid down in gzio.c in zlib, yet
    somehow I screwed it up rather good).
    Gah.

commit 6e14371cb6fe597cf3397dcdd206482aaa85a556
Author: bharring <bharring>

    General commentary about diffball, added dep information, updated formats, mild rant about fdtu (hey, I wrote the support, I'm
    entitled), pretty much some basic documenation/general info.

commit a93575e7e7a67e84c75daf03e639cb572d490137
Author: bharring <bharring>

    Fixed a few warnings, printf(%lu, (unsigned short)var) as an example.

commit 99d1780f24dae0a5610bfa1e18272c316cf9fc65
Author: bharring <bharring>

    Tweaks, bsdiff appears to be working fine now.

commit 0c1e354d0be7bb3877bba9f6451acb84a22038f3
Author: bharring <bharring>

    Minor adjustments, basically eliminating possibility of getting stuck in infinite loop when crefill returns niadda.
    gzip reading needs work, unable to be used w/ patches while bzip2 can be.

commit bd46c0b5d2bb20185488bd7092c5b2387a681270
Author: bharring <bharring>

    Added info regarding bdelta's breakage.

commit cb155b03fec90292a382439732bc3df4c5679421
Author: bharring <bharring>

    Two fixes- Fixed -to-stdout option for all bins (set it 1 for STDOUT,
    don't think that is correct/best way).
    Fixed bdelta, last command if size2 != processed size is actually an
    add, tiz a bug.

commit 2337d96fbf8c4ade2701892b88a06da991382488
Author: bharring <bharring>

    Modified bdelta reconstruction to use the new add_src construct for
    handling when it leaves off trailing nulls.

commit b757da21a6ed964eb32cf8ed63b913be16f2c7a9
Author: bharring <bharring>

    Iffy on if these functions should exist, but for now they do.

commit c5c3e8491fbd1ca1e843c55934749a4cd4487f45
Author: bharring <bharring>

    Gah. Forgot to include these.  Basically wrapper functions for xdelta,
    just creating a new cfile that is limited to the xdelta patch (ignoring
    all of the deltup header crap).  supports deltup v3 and v4

commit b2b8daf28a6b5996cde738a224d7c1b457980ded
Author: bharring <bharring>

    Removal of fprintf, using v0printf instead, work on errorhandling.

commit 7414c49750d41f87c12d19f5d933a431405a215a
Author: bharring <bharring>

    Modification to enable flagging if a cfile should be freed or not via the DCBUFFER_REGISTER_* macros.
    Modified CommandBuffer to support it, need to clean that bugger up.

commit ec1fe34e27b75e8c0efd8988dde4d2a74e983ecf
Author: bharring <bharring>

    Basic error handling groundwork.  Needs finishing/checking.

commit 605d8e80d1d9f72bcc813b1951e6caf79a4ebcc0
Author: bharring <bharring>

    Error handling/printfs.

commit be26498164ff2579d9a2a8e6a9c1250f1dfb9540
Author: bharring <bharring>

    Error handling work.

commit 26bc41321606398820123bdc6cfe561af7276c05
Author: bharring <bharring>

    Needs better error handling.  Fixed an issue w/ bzip2 specifying a data
    offset for a compressed file (eg offset=0 is actually 0 +
    data_fh_offset, rather then 0)

commit 9f334227f332d4e88348a8d323099facc44f6b1b
Author: bharring <bharring>

    Updates for the --enable-asserts/what not configure option

commit b12ace5f209dcfde6b77640fbff61eb2ff629df6
Author: bharring <bharring>

    General modifications for error handling, not necessarily perfect (they
    aren't).

commit 56ff872ff9fced3c0658f42292fb03639c391b55
Author: bharring <bharring>

    Commented out the main len1 code, and just did a simple add_add until
    convert_delta supports converting bsdiff patches (eg src option via
    commandline).

commit bd6e52096d15289985950d81803fe17d845ae46f
Author: bharring <bharring>

    Fixed the previous attempt at a fix.  Now works correctly.

commit 8d8a17b7584ea126f760d6a383dcf06c7e5d188f
Author: bharring <bharring>

    added --enable-profiling for AM_CFLAGS modification to compile w/
    -ftest-coverate... blah blah, gcov crap.

commit 6288480cd4af98cdbffe2f706ebdff405b3509a7
Author: bharring <bharring>

    update for typedef struct tar_entry tar_entry

commit d3cf520760c594b300c49cd090a8617fa3c51f95
Author: bharring <bharring>

    Fixes for make dist, adding bsdiff.h into noinst_HEADERS, and the
    ctest/hash_stats source to be distributed along w/ the dist tarball
    reguardless of configure options.

commit fa61adba7fe35a706848576e9dd25c85becde3b5
Author: bharring <bharring>

    Ahem.  Removed all *_ABORT() functions, using return *_ERROR instead.
    so.  a lot of functions need general cleanup, return's being called from
    void functions for instance.
    Fixed a longstanding bug in hash.c, basically not checking to see if
    there is enough data in a buffer window to call update on it all, or
    piecemeal.

commit 9dc20cb64ce13e7ed875e62ea6bb74d273b46c74
Author: bharring <bharring>

    bsdiff/fdtu added support.

commit e5fd089a2e698929512432906555625020370be2
Author: bharring <bharring>

    removal of a duplicate include

commit 5ea2dde398832288b15307dec6a7908969e31e98
Author: bharring <bharring>

    Update for Makefile.am/configure.in changes.

commit 1632a2ee27869360b4a8c9a146b11d3f94785d64
Author: bharring <bharring>

    clean up of unused variables.

commit d2c295137e671ef7158e565f52c2cf7a5f2f44cd
Author: bharring <bharring>

    Fix for disabling -O2 during debugging. tweaks here and there also.

commit 8c7ec38002ebd1a382863170f2d85b8a516086bf
Author: bharring <bharring>

    enable a bevy of debug options via configure.  eg ./configure
    --enable-debug-cfile...

commit 341eabbab333f46e25932e04ac45135783834e1a
Author: bharring <bharring>

    Again, updates for new cfile setup, and changed encoding options for formats, along w/ new dcbuffer changes (multi-add).
    Borderline 0.4, next updates are just fixes (add fdtu support also), then tagging as 0.4.

commit d6898c076227d065c01fef94d7d0edd9026a9231
Author: bharring <bharring>

    Update for new multi-add_srcs.

commit f087ebb632ac8a4959f4608d9825e1a18a5c4ab4
Author: bharring <bharring>

    Added entry for bsdiff, stating 'unwilling to convert' due to req of having src file on hand (overlay format).
    Add option down the line to specify src file, so that conversion of overlay formats is allowed.

commit 32fb735578c1b9e80d3850866992d1e8550f55f2
Author: bharring <bharring>

    Added bsdiff support.

commit 2a7c33c5df64d4aabec103024c0e7d87936cc6b6
Author: bharring <bharring>

    Adjustments to use new cfile/dcbuffer multiple possible add_srcs.  function ptrs are fun...

commit cdb3412d3aea0d01c273c42a5e5d53863a2e7425
Author: bharring <bharring>

    initial setup for multiple add_src's implemented.  ain't great, but it's a start.

commit ddf6430a142b6b45dcbaae68979dee9171dc8fe3
Author: bharring <bharring>

    Updates for changes to cfile.  bsdiff is now working also.

commit 7ff11f8dfea25424dc7057485e593c185a1db719
Author: bharring <bharring>

    Converted to using zlib deflate interface rather then gzdopen, complete tracking of multiple cfh's into one filehandle, fixed up raw
    read errors (reading to much, rather then to the cfh limit), bzip2 completely working.  need check for when cfh limits are too small
    for a compressor, eg, in crefill possible to go into permenant loop if cfh length set wrong (I think).

commit 8fa266ad30b67867e990e0a554e185b033567a1e
Author: bharring <bharring>

    Initial framework for supporting multiple add/copy srcs (along w/
    definable function for copying blocks out of said cfh).  Testing it, and
    cleaning it up to fit in existing work.

commit f163c82fac356dd199230461864fb7d1a96d054a
Author: bharring <bharring>

    cleanupt

commit aeacd8924aabbc5b7f285c214cb7f4881b893ef6
Author: bharring <bharring>

    tweak

commit befe8d48513a29b05ffd10b5f2b32548453efab5
Author: bharring <bharring>

    autodetection of compressor

commit 429ddfe793586108d26201809e741225b0f7f769
Author: bharring <bharring>

    Testing out autodetection of compression type.

commit ba586229a47e8bb2eb4921e727985e6cdad9664f
Author: bharring <bharring>

    Backhanding the crystal ball...

commit c1456af6d5cd396dd60431341488fe86c27ef967
Author: bharring <bharring>

    Added AUTODETECT_COMPRESSOR as a valid compressor, along w/ a quick
    check for known formats.  Works fine, although need to watch the whole
    decompressing data that gets decompressed again.  need to abstract
    cfile...

commit 18d5eeb0cf5e0da8349c1445209bfc1a343c5574
Author: bharring <bharring>

    Working bzip2 support.

commit 81a4f2b0f8c844fb6eff85d9ee9931ac6386e901
Author: bharring <bharring>

    Cleanup, modification to bsdiff.c, fix in formats.{c,h} w/ addition for
    bsdiff spec.
    Side note, bsdiff's format uses add overlays.  Cool.

commit 68f362d9ac5f895c64a9a5ddff3eab9b10769b9b
Author: bharring <bharring>

    Rather dumb mistake in formats, fixed now.

commit 06cf190f4d4261d09d662ee9bc34da9c370525a4
Author: bharring <bharring>

    Additional entry for bsdiff format.

commit 0b939b37f815c9183cb31b023d0fde3b3a905e6f
Author: bharring <bharring>

    Gutted bsdiff format, rewriting it.

commit 92eb6c42ca26f54cd939632cf59d37eda6910e9f
Author: bharring <bharring>

    sane renaming of typedefs for off_(u|s)(32|64).  Fixed, no breakage.

commit 09821eca7fe92b44fe93d3c4f4de63d325c1c19e
Author: bharring <bharring>

    Initial work towards adding bzip2 as a supported compressor_type to
    cfile.  Should be ready to go, just need to test it, then test my
    existing bsdiff code (then commit).

commit 080db2fe47f0d4f12d09fbf50ae49a5de9ab48de
Author: bharring <bharring>

    Fixed a dumb assert mistake.

commit 1e9377f5f6501626656c1556b6f67c86127af6ce
Author: bharring <bharring>

    Clean up, fixed the free'ing tar struct issue.
    Basically needed general cleanup for long link's in tar.c .
    diff-algs.c: minor adjustments to the v?printf's, along w/ calling
    print_refhash to make things a bit saner.

commit e57a6d36a20b5cc59d7feb365ff8c12c487b8139
Author: bharring <bharring>

    The 0.4_pre4 release.
    so *far*, have eliminated or at least reduced the nasty i/o deadlock to
    either differ or diffball for a tarball...
    That and performance now is pretty much top notch, loosing normally to
    bsdiff (it is suffix based after all), but neh, no complaints, I can go
    after bsdiff later.
    Either way, changes to the general strategy- diffball uses a 24 seed
    scan on matched files, then passes the dcb off to multipass to process
    the skipped entries/gaps.  Performs well, using a reverse bucket hash
    (RBUCKET_HASH), which performs quite well for eliminating i/o blocking
    for lseeks.
    Need to optimize BUCKET_HASH/RBUCKET_HASH for the bucket insertion, and
    specifically the R*_HASH types since right now it runs through the
    versionned gaps, the *full* reference usually, then again through the
    versionned gaps.
    Performs well, but run-time could quite likely be improved w/out
    compromising compression.
    Also, fixed/re-added some code to bdelta.c's encoding function to
    produce the correct size2 (version file size)... so conversion is a go
    again (and the warning was removed from convert_delta.c)
    Few fixes to dcbuffer.c, bout it.  mostly performance/strategy
    adjustments.

commit 72b3d6d34496438896698978381b821f6c56d013
Author: bharring <bharring>

    heh, another fix.

commit 2ab43c9786948a2cb5e9feeba7bbd27f3c98d24e
Author: bharring <bharring>

    fixed stupid, *stupid* fuck ups in RH_BUCKET_HASH... also added
    RH_RBUCKET_HASH... I'd split the commits normally, but halfway through
    RBUCKET I noticed (and fixed) some dumb screwups...

commit dd4a91768cbfdcdbf09de8cf40dbfadc34fb76b7
Author: bharring <bharring>

    Shut the fucking thing up a bit...

commit e715cf0b4a0b0c5a6ccf92c0bb6a74e24147e52c
Author: bharring <bharring>

    RH_BUCKET_HASH completed, and should be pretty much bug free-
    optimizations need to be done of course, but trying out RH_RBUCKET_HASH
    first...

commit 25b6aee1dd4376a72eea1592775e14d2bf7b2601
Author: bharring <bharring>

    Added RH_CMOD_HASH in my never ending quest to make a mess of hash.*...

commit fc120c0f7c0fbba56a0b35e90a4c8eee6573973f
Author: bharring <bharring>

    update

commit 022a5a2191af0d57416fa44cd0f279e9b07c0a5f
Author: bharring <bharring>

    Adjustment to get diffball using the new hash api.

commit 66095aad0d99f327c5edead697add73d71e9e7f1
Author: bharring <bharring>

    Adjustements, trying out the bdelta hashing alg.  Looks
    decent, although I think it's probably going to result in clusters.
    Other then that, fixes to diff-algs.c and hash.c, using the newly split
    hash functions.

commit bf37c4537f01260be9e0ee098a41da2a9443553a
Author: bharring <bharring>

    Update info

commit 88f00bf1ee1a09660a2f2836480322cd67f904ec
Author: bharring <bharring>

    Created a new function a while back, DCB_get_next_gap .  So far looks
    like I've shaken the bugs out, so commiting it.
    Basically, it's the gap logic moved out of multipass into dcbuffer.

commit 6c9c9ffd87b2c4e6aa11005a308a279da19e2523
Author: bharring <bharring>

    did away with the bastardly rbuff_end in insert_block, now using
    cfile_window via expose_page.
    No doubt, error checks needed...

commit 76d7b7291d8b65744cda1577038f80ac6a1fc014
Author: bharring <bharring>

    Added RMOD_HASH type, which is chksum_ent hashed.
    Should prove useful for dealing w/ large datasets (the bastard OO for
    instance), by storing the checksum, thus allowing to check and see if
    the actual chksum is correct.
    Not positive it works *exactly* as I want yet... split hash_it and
    get_offset into a 3rd function, lookup_offset which does both hash_it
    and get_offset's job.
    Did this since diff-algs is after just the offset, while hash.c may be
    after more (RMOD_HASH for instance is).
    Testing OO against RMOD_HASH now, shoudl be good shit.

commit 8e4acb763a4c083b8a150126760101497e36836d
Author: bharring <bharring>

    Adjusted param's for functions to off_u64... dcbuffer.c reflects this
    already, missed importing this alongside.

commit abd51dd3ae78a4176bbba89df43c896ed1bd4e88
Author: bharring <bharring>

    Changed appropriate types to off_u64, upped multipass to 512 beginning
    seed len.

commit ad2eb982a97f07c78b2e0860b94f4c05955ee377
Author: bharring <bharring>

    Fix to patcher.c, for bdelta format (that format is getting annoying).

commit 8954be4d0ebe4804e5d18f6cac7d8847f67dc67c
Author: bharring <bharring>

    bdelta conversion warning messages...

commit 6a8f4f12215ffa7238a3966631f6fbed94fc14c3
Author: bharring <bharring>

    Broke init_refhash into 3 functions.
    init, which stores sample rate, prime's the hr, etc.
    insert, processing and inserting a block of a cfile into the hr
    finalize, does nothing for mod_hash, for sort_hash, does the appropriate
    sorting.

commit 271bec2499723f372749b96bafacbfee7ce6f0b5
Author: bharring <bharring>

    Fixed a bug w/ matches and llm where if the last command was an add,
    either an assert got flagged or it got stupid.

commit 0ffc67fdb5208de4ed3027a01fa147c3d5688e70
Author: bharring <bharring>

    Adjusted sanity check/assert, since convert_delta has issues convert'ing
    to bdelta format (no knowledge of size1/size2).
    Loosened the assert, and added a v0 level complaint about it...

commit a508a762fc498427f4064b4ebb1ac9598a9ac6ea
Author: bharring <bharring>

    FUCKING diffball's free crapped out my run against oo.
    bastard, I'm fixing that soon... not even sure why the bitch segfaults,
    but she does...
    turned down multipass from starting at 512 to starting at 128.  possibly
    result's in a less optimal, but betting the correcting nature of
    onehalfpass will handle that fine.
    Unless a source has had *very* little changes, finding a 512|256 is a
    bit hard, going by my tests w/ oo.
    Make it dynamic I guess, prob best method.

commit 1a94fe812e0bb19c224595fb9394a6b77b38294a
Author: bharring <bharring>

    Hokay, rather major tweaks here and there, although the abstraction of
    everything has made it *much* easier.
    cfile.c: few assertions, fixes, etc. nothing major.
    dcbuffer.c: added 2 new types of commandbuffer's, matches, and llm.
      llm is matches, just linked list... needed for multipass.
      assertions galore also.
    diff-algs.c: added multipass, which currently calls just onehalfpass
      multiple times w/ ever decreasing seed_len.  works well, although it
      gets hammered by usual chksum getting neutered by the modulus op.
    added hash.{c,h}.  this is splitting the actual hash array out of
      diff-algs, since now it has two types- original, long modulus method.
      second is a chksum_ent (chksum, offset) stuffed in an array, and
      qsorted.  works decently to get around the neutering by modulus op.
    adler32.c: currently using basically the exact settings of xdelta.
      don't really see any benefit, my original primes based and xdelta both
      get neutured by the modulus hash type.
    hash_stats.{c,h}: simple lil bin that just init's a refhash, and dumps
      the print_refhash_stats.  good for data collecting.
    defs.h: added a few new types, off_u32/64.  intended for largefile
      support, although it isn't finished yet.
    dcbuffer.c, differ.c: tweaks to use the new additions.  did fix that
      stupid segfault caused by freeing the src/trg tar_ent array though.
    delta.c: didn't belong, old dcbuffer.c; *really* old dcbuffer.c.

commit 2311ef8ee4fb05c1f31634ace0a14660421cac28
Author: bharring <bharring>

    tweaks, cleanup, bout it.

commit 001c778d9b0c9e7d35f78c9833fe1e4c3fd58a07
Author: bharring <bharring>

    diffball.c is starting to look more and more like a rat's nest compared
    to the rest of the code...
    Fixed a bug where tar for whatever reason, doesn't store the trailing
    slash on a directory name...

commit ae29f09814734ab59098b290b1e9d81d0d877eea
Author: bharring <bharring>

    Shut cfile's debug info up, even if -vvvv.  Only possible to enable it
    via the configure.in option.
    Original intention...

commit 9777d6f27c276f87d56234226d7b10bfedd2cf62
Author: bharring <bharring>

    Fix to writeSBytesBE ; basically doing some val >> 32 results in a
    positive number, which wasn't what I expecting...

commit 79702f9c1bfabe1ff54308e80fdd29903fa7d409
Author: bharring <bharring>

    auto(conf|make) suck.

commit 788fe6b1542285d6fda0ec26b07a0d4b403614fe
Author: bharring <bharring>

    Hmm, got removed somehow. shit.

commit 86ac0048328c5ad54e2c603d8bccf3d7ebe42dd4
Author: bharring <bharring>

    Briefly, all access methods of accessing/reading from CommandBuffer now
    use functions rather then direct accesing.
    Did this so I could a more efficient structure of just copies for the
    command buffer, also for the upcoming multi-pass.
    Multi-pass struct in dcbuffer.h is not final, or even likely a
    prototype.
    All formats were modified to use the new functions, so any changes to
    the CommandBuffer/dcbuffer won't affect them now.
    diff-algs was also modified to use the new functions- it calls the add
    DC_ADD function regardless if the CommandBuffer actually stores add's or
    not, figure it's easier just to let the call be made and ignore it since
    nothing is done.
    There should be *no* direct DCBuffer(Incr|Decr) call's anymore.  Need a
    count function, but it's minor.

commit 7fc5206690e3f08b9d725af84e1d1dde602d99e4
Author: bharring <bharring>

    Friends don't let friends use BBEdit.
    Still cleaning up formatting fuck up's from it (tab's vs 4 spaces), and
    I only used it for like 2-3 weeks back at the start of this project...
    Either way, fixing formatting, commiting it so have a *Good* formatted
    version of this file, about to start conversions to new dcbuffer.{c,h}
    api.

commit aa548550d64f4422a7b56ab497dc9e014acac9b0
Author: bharring <bharring>

    quick tweaks, about it.
    Moving towards multi-pass.

commit e9b98d7d87e3fdb6ed6cefee26d70822cfabe978
Author: bharring <bharring>

    branding the 0.40_pre3 tag shortly...

commit 1cd6557cc5330ac79f82bcf597331b1945ebfd79
Author: bharring <bharring>

    Major changes to dcbuffer.
    Add DCBF_cur_(len|off) for pulling the current offset/len.
    Modified CommandBuffer so that there are 2 types of stored buffers, one
    full (original) and one for multi-passing.

commit bcc63ad50f8b815e66d18c2921671973be803ded
Author: bharring <bharring>

    Getting ready to add multi-pass.

commit fc0335d1f00eeb91693b7b4f4725a3168c5d90d8
Author: bharring <bharring>

    Fixed a bug in cfile.{c,h}, which was why diffball/differ were
    segfaulting *HARDCORE* on large files if the user was root.
    Side note, stopped using root as my default account, cause of those
    segfaults.
    Long story short, access_flag CFILE_BUFFER_ALL and CFILE_RONLY shared a
    bit- so it lead to every file being buffered.
    Which is really bad when the two files are half a gig each...
    Either way, fixed it.
    Enabled parity in adler32 again, prob just going to throw it out.  might
    be useful for non-prime, but neh, who knows.

commit 798aeb87a2afa14791cd2f3bf0e5320fcf255b66
Author: bharring <bharring>

    Import of my files, since the laptop appears dead (taking my most recent
    repository with it).
    Shit.

commit 76d478cdea9c4e03d976147e7892651695ded03d
Author: bharring <bharring>

    minor modification, added v0printf (basically a fprintf(stderr, expr)
    alias).

commit c9c24e378c7810e349a4e5c51c5ddae0579f8f2e
Author: bharring <bharring>

    Hokay.  Tweaked DCBufferAddCmd to automatically realloc the
    CommandBuffer if more space is needed.
    Took a crap load of pain, but got it.

commit 3d12755322b8d2f16923de650a3fa2270853570f
Author: bharring <bharring>

    fix to apply-patch.h
    Rest is adding a new debug flag, DEBUG_HASH , which enables checking of
    duplicate hash entries for good vs bad (bad being where checksum's don't
    match).
    Enabled via the usual configure.in method.

commit 5ed5cc0c76d93bf58f1ce0c0c6fb99566f952401
Author: bharring <bharring>

    modifications here and there, starting up the cfile compressor setup.

commit 220023d807e106bc69ad2237029485054b949ba6
Author: bharring <bharring>

    starting cleanup of tar.c

commit a317715c8d4d544a88a4f76393bdd9192d26fd45
Author: bharring <bharring>

    tweak disabling ctest, useful, but not needed for make dist...

commit 4609c4ba00f7018ac046c2b7cf0ab6e4b5fe0f2e
Author: bharring <bharring>

    Fixes.

commit b6208f26a9d5ffa00800b42c914a4751b4209bc0
Author: bharring <bharring>

    Tweak to backhand automake into including the man pages when doing a
    'make dist'.
    That and get the generated makefile to install said man pages...

commit c436245f910943fff40435c64fa3696c962902a5
Author: bharring <bharring>

    tweaks to the crystal ball vision.

commit cd2dccc811e567ffe0b33908f6b94359a82d62d9
Author: bharring <bharring>

    Hoo Ha.  Got man pages wrote, although they are a wee bit sparse.
    Updated Makefile.am to make automake aware of the new files, and install
    em.

commit 554eca90035c2cbdedf600c89968486d26b09158
Author: bharring <bharring>

    tweaks to correct/shut up -Wall.
    Not so much shut it up, as fix the problem (if any) or quiet it if no
    reason to complain.

commit f62c6504b1d72a02ef73db788c201fa9350ea05f
Author: bharring <bharring>

    Tweaks.

commit f680eb1493dc5963d6f803244836321f78e9e2e4
Author: bharring <bharring>

    hokay.  Command line Options are now basically completed for each
    binary.
    Cleaned up the check_for_format and identify_format functions.
    Things seem to be going fairly well, cleaned up the output for the
    binaries/code.
    Adjustment to Makefile.am... um yeah, a lot of work here and there.

commit bd86833b5fa988c48129ced17a6690a5e67d92b2
Author: bharring <bharring>

    Tweaks here and there, getting close to switching over to run time
    selection of format to use.
    Necessary functions are in place, just need to right the code.

commit 5ec54c96f4d994cdd2cec4e0bf0d3955e4fba4f5
Author: bharring <bharring>

    Fixed a dumb mistake, dawned on me while attempting to sleep- basically
    the macro definitions (FORMAT_OPTIONS for instance) should have they're
    'variables' wrapped in ()'s, since the dev might pass a statement like
    buff +3 , which should be interpreted as (buff + 3) where ever used.
    Dumb mistake, fixed though.

commit a0b8b54c4e4412b72f6e8e1c7f26b476b677723c
Author: bharring <bharring>

    hokay.  More option crap, added a check_*_magic function to each format,
    used via identify_format (in formats.c) for identifying the patch
    format.
    Also added check_for_format, which is a command line option checker-
    yeah, may belong in options.c, but this (at the moment, 4:00am) makes
    more sense to me.
    Either way, getting there.
    Need to finish this options crap (including adding argInfo and Descr to
    options.h), then refactor cfile to support true windowing (for xdelta
    format).
    Oy.

commit 742385aa47c68ad29880cafe71c624edf76e4956
Author: bharring <bharring>

    tweaks as I've gotten a little more used to auto(conf|make).

commit b3945fad1c1c150b454328457e562c720fd62458
Author: bharring <bharring>

    Hokay.  Bit o work here.
    formats.h is a header that includes all format headers, centralizing it.
    formats.c (not added yet, nor wrote) will be for identifying formats via
    calls to format magic sig checkers.
    options.c is a standard usage message for puking the usage out to the
    user upon option error's.  options.h include the prototype, and several
    macro define's that are popt specific options.
    STD_OPTIONS is the standard options to each binary, DIFF_OPTIONS is
    differencing specific (eg differ/diffball), etc.
    Cleaned up the code in general, more work to do.  nearing initial
    release..

commit 195b21b79c43f239267b60a06279fc50130bd379
Author: bharring <bharring>

    cleaned up the makefile something fierce, modifications here and there,
    differ now uses popt (as shall the rest shortly) for option parsing.
    going to add options.c shortly, once certain things are worked out
    (automatic variable updating for instance).

commit 2b525bcba2fda53b0d37ff68f48241dc3021855f
Author: bharring <bharring>

    tweaks

commit 96c892e3045a626491db75f5aad4f8272a310657
Author: bharring <bharring>

    I have this sneaking suspicion I'm going to be messing with
    verbosity/debugging printfs for a long time...

commit b9eb79ee4cc46b146fb684f3626e5bdedfaacdc2
Author: bharring <bharring>

    tweak for asserts and such.

commit a8b569273baa252c85a228192817f0fc6ec280c7
Author: bharring <bharring>

    started work on removing annoying aborts, and having that function
    return appropriatly.

commit ea41b9841dd6869dec6b2b7f31f3cadd1902e0b5
Author: bharring <bharring>

    Hokay, widespread changes.  added defs.h, which basically contains a
    crapload of printf specifications, each controlled by preprocessor
    definitions.  Modified configure.in to control the definitions.
    Either way, clean up here and there.  Getting ready initial
    release version.

commit f3eb51a6569a3294c54b0e57ce02d1c339f51e3d
Author: bharring <bharring>

    starting rather wide spread modifications, basically cleaning up the
    code so that debug d efinitions enable printfs, and enable assertion's.

commit e5cb4f0ae2305422ef436ebfae233801cdeba178
Author: bharring <bharring>

    Tweaks here and there.  Minor adjustment to udiff.c, basically puking
    when it encounters the lovely 'Binary file x and y differ'

commit d57d2dd782887bd9937d513f30fdfa0e71af6a3b
Author: bharring <bharring>

    Added a CRAPPY kludge (think duct tape) that ought to solve the lack of
    newline at end of file issue.
    It's geared such that they're must be two such errors, which likely
    ain't a correct assumption.
    Either way, it is solving the problem for binary files.  Text on the
    other hand...
    Well, I'm sure the cvs log's will have more about this down the line...

commit 7087efb12cc4757a0793d1160d20d47eb2dd63e5
Author: bharring <bharring>

    getting closer.  Removed an assert, that frankly is pointless at the
    moment.
    It was designed to ensure cflush sends out all data... which is fine and
    well for uncompressed (where all data can be written out).
    Either way, removed it, left a suitable note there.
    Got unified gnudiff close, producing off by one (just to piss me off
    I'm betting).
    The problem is an unusual one, basically diff doesn't like working on
    binary files.  Forcing it to work on binary files, there are issues with
    lack of newline at the end of the file.
    Soo... should be fixed, mainly since the lack of a newline at the end of
    a valid text file ought to bring this problem out.
    Hopefully fixed in the next commit.

commit 245e5a420ff50eb3982cf485d9e90f202add8d68
Author: bharring <bharring>

    Partially working multiple segment handling.
    For segment I mean specification of modification areas.  It works on
    some files, not on others...
    So.. the whole ordering of ' ', '-', '+' is probably bringing out a
    logic error in the code.

commit c45598b822b08ff9fea97a99a8205298e96e5ae7
Author: bharring <bharring>

    work in progress, udiff (unified gnu diffs) doesn't work perfectly yet.
    Heh.
    Does work for one segment patches (eg, one area of modification).
    Multi-segment is soon to follow, just need to identify where I'm off by
    a line.
    From there, need to create solution to handle recursive patches.
    That's where it'll get fun.

commit 6ce442f5a3ee7f8938073e58d5a12f151fc8f5b1
Author: bharring <bharring>

    Temporary commit to show greg about cvs.

commit 679d98016f7b10f6f1f0ae7d3c3843af8223bb0f
Author: bharring <bharring>

    adjustments here and there.

commit 99f42d708bd3974808bdc486553ec75943ddadd4
Author: bharring <bharring>

    test file tweaks, beginning work on expose_page (exposing the buffering
    system's data buffer)

commit ce555bec95bd7368c0e192af010f912e3e04a0ee
Author: bharring <bharring>

    Quick fix to ctell when using the CSEEK_END flag.
    Updated test to reflect.

commit 007ee1833127deb9ab6c93319c8e81d6fd58133f
Author: bharring <bharring>

    Love the new buffering system... quickly added (eg needs some
    optimization) md5 support in about 15 minutes.
    Nice...

commit 878b35a719e136b0f7580913ac40a3ebcbbff9ef
Author: bharring <bharring>

    minor tweak here and there.

commit 36a7d79e81367bfd8a338a5122a31edfaf5d8411
Author: bharring <bharring>

    Hokay.  Buffer system got yanked and rewrote.
    cread/cwrite/cseek all function the same *externally* internally, it's
    a slightly different matter.  Basically I simplified it all (hopefully),
    and made it a much cleaner design.
    It'll be quite a bit easier to add md5 and transparent
    compression/decompression.
    Also.  Added ctest- it's basically a crapload of assert's that abuse the
    buffer system attempting to break it.
    Still needs more tests added, but has been quite valuable in rewriting
    the buffer system to the original external api, and being able to do it
    *right* this time.
    cheers.

commit 2e10cd3220dceec9630f498687924e09ac33ac6e
Author: bharring <bharring>

    Fixed a rather lovely memory leak in cfile, related to cclose and state
    flag BUFFER_ALL...
    Basically resulted in all matched tar entries, not having their cfile
    buffer's deallocated- and since BUFFER_ALL was in effect, say you
    matched 80% of the entries on two 500mb files, you'd end up w/ 800mb
    leaked (roughly)... Math may be a bit off there, but you get the jist.

commit 9e92c3e91b28dc515ffcc41652fe7c36aa5a5976
Author: bharring <bharring>

    Been using valgrind to identify memory leaks.
    Looks like there are areas where I'm not freeing alloced memory... nifty
    tool btw.

commit 5db761e1208276af2000f2ca8e9278d1307ab4a3
Author: bharring <bharring>

    Tweaks here and there, adjustments to planned release.

commit 9ae1f0e05427a4e79843b34eb10e45bb32fa4aca
Author: bharring <bharring>

    Adjusted AUTHORS formatting (no actual content change).
    Added what I've figured for a development/release schedule to TODO.

commit 616407b49ec3b37cbaaba5c5441a417ca4db4850
Author: bharring <bharring>

    Hokay.  Code cleanup.  Again.
    Could've sworn I commited these changes earlier, although I think it may
    be completion of that earlier endeavor...

commit 6f021261bce803698dc5572481229f663a5df2c2
Author: bharring <bharring>

    code clean up.

commit 1edcecbec41f3c6f7028ae675fbf978d547cf51c
Author: bharring <bharring>

    Well.
    This format was the code I wrote out originally that was to be close to
    bdelta's.
    Got bdelta support built in now, although it is a bit dumb.
    Sorry fellas....

commit ff80050db9f4a6937361c60e3840268c85c7acfb
Author: bharring <bharring>

    trivial stuff.
    bdelta format has basically been verified that it works now, for large
    files too.
    Still have the issue of reading bdelta format, and dealing w/ delta's
    that don't reconstruct the full file, and leave null at the end of it.

commit c9bb8a989e6513ba2faf141b2f83e19e1a93e30a
Author: bharring <bharring>

    hokay.  Got write support pretty much finished for bdelta.  Verified it
    works via testing w/ bdelta.
    Niete.  file.{c,h} has some nice additional functions, basically a copy
    block function, standarized function for copying to/from cfiles...
    basically just a simple function that handles the buffering and such,
    worth converting other formats to use.

commit 363a5ec0b8af50f1277a635e3927d3494b6a47b1
Author: bharring <bharring>

    stupid god damned le functions...

commit 129e704c0a5cc708bdb3cdbe1716fb827d6e32b0
Author: bharring <bharring>

    minor commit, annoying bug in writeUBytesLE that's been making bdelta
    writing a pain...

commit f9046035dca902d8db3ccdde65b0ada3cd18bf0e
Author: bharring <bharring>

    Minor tweak

commit 2aeb87a746e52242d94aa715e9d942bca8ade200
Author: bharring <bharring>

    Somewhat minor change, although widespread- basically went through and
    adjusted all defined struct's so that they were typedef's, so I don't
    have to have ugly-ass (as an example) struct cfile ... in each source.
    Yes, trivial, but a but load of changes by hand.
    Modified cfile fairly heavily due to complication, function arguement
    definition (struct cfile *cfile) doesn't really work, since now w/
    typedef it's (cfile *cfile)... had to make it (cfile *cfh), and modify
    all use of that variable by hand (write a script to do this next time).
    So... uncommenting old code in cfile may cause problems... should have
    got em all, but who knows.
    Changed adler32_seed to ADLER32_SEED_CTX, and prime_ctx to PRIME_CTX .
    Other then that, modified dcbuffer's struct definition to allow for the
    coming attempt of spliting functions into multiple parts...
    Oy.
    ~bdh

commit e0117855de0c5ef7b34b1fcbf5463e8c6685b835
Author: bharring <bharring>

    hokay.  Modified bit-functions for big/lil endian considerations-
    basically you must specify the endian-ness.  Yay.
    So yeah, modified all formats for it.  add bdelta support, although it's
    got an annoying quality that fscks up my setup a bit- basically, the
    bdelta set of commands may specify additions/copy's for a total of only
    (say) 75 bytes out of a 100 byte target if the remainder is null padded
    (?).
    So that's annoying, cause I'll have to either modify dcbuff, or
    everytime a bdelta patch is in use, take that into consideration...

commit 09178a507a1caa5997c736af4613a3473e022dee
Author: bharring <bharring>

    Now have *working/beta* xdelta-1.1.3 read support.  I don't know what a
    couple of bytes in the format are used for, so bit wary there.  That and
    not doing md5 verification, since I need to add md5 support to the buffering
    system... oy.
    Either way, she comes out md5 correct  against mozilla-1.2.1
    mozilla-1.4.1 (200+mb files), so looks like things are correct.
    In serious need of writing options into the commands.  That and dcbuffer
    needs completion, and need to start breaking large functions (eg
    differencing or encoding) up so that they can be run piece meal, eg
    encode 100k of this, reconstruct part of the file w/ those commands,
    flush commands, load next 100k of commands, wash, rinse, repeat.

commit 78d2a5ec34e381e74d0b02f0ff676e4533e78440
Author: bharring <bharring>

    Hokay.  Now can output in bdiff format.  Still need to add md5 support.
    So... modification of cfile, and allow it to pass data through evp's md5
    function.
    Not looking forward to that..

commit 7499e713ec2ac1b7fb8af4431b2a270906aff157
Author: bharring <bharring>

    Added a new format that can be read, bdiff's format (bdiff.{c,h})
    Reconstruction seems to go fine, pretty straightforward format.  Has
    md5 checksum's in it which need to be checked (eg reconstruct ain't
    done), but looks to be pretty make done aside from that.

commit d8a6142816f3131a564f61bdc9b1c36ff7649c0f
Author: bharring <bharring>

    Lots of adjustments here and there, clean up and such.
    Been attempting to decipher the xdelta-1.13 format.  Which sucks.
    Hard.

commit 7285148cf2f261af9d9fb39b694ccfd79278e792
Author: bharring <bharring>

    Hokay, aside from tweaking diffball/patcher for choosing the output
    type, raw format has had it's reconstruct function finished so it
    supports reconstruction of the format now.
    Makefile.am had a few tweaks, about it.
    Really, *REALLY* need to adjust the Makefile.am and clean it up.

commit 00b904a4aa651d3cbf4a7a3e18a20c36a9d90061
Author: bharring <bharring>

    Tweaks here and there working towards making outputting an encoding
    leaving the command buffer unmodified/same as prior (since DCBufferIncr
    is used).
    Also modified convert_delta to output switching and raw if given a 3rd
    filename.

commit 3ebda00e062be60ae86dfec365c91add5466e0fa
Author: bharring <bharring>

    ...and copyright notice on header's.

commit 17efbfcd51eef4a595380a36348df52421aebcdb
Author: bharring <bharring>

    Tacked on copyright notice at top of c files.

commit 751fd14e8efc6d8443b5f584e6c77372c7294a61
Author: bharring <bharring>

    Hokay.  Quickly scratched out the basic framework for a binary for
    converting between delta formats (convert_delta), also added the
    beginning of a new delta output- raw.  Basically bdelta's format,
    although will need tweaking, plus needs to have it's reconstruct
    function gutted and rewritten.
    Tweaks here and there...

commit 339e4d4180b20286aa64452d6efd815cdfa85cec
Author: bharring <bharring>

    Um... yeah.
    Lot of adjustments here and there, fixes, improvements, etc.
    Check TODO for where it stands right now...

commit ac1c81a755f8421bf34bc5e925c16a45913ff3f1
Author: bharring <bharring>

    check adler32.c.
    Suddenly diffball is outputing slightly smaller patches, I'm thinking
    either due to an error in the parity work I'm commiting or a bug I
    corrected unknowingly.
    Either way, still md5 correct.
    Also, modified switching to skip blank len commands.

commit a2a4430b10c9ee2f27ea312e63e58ab6b9171be3
Author: bharring <bharring>

    improved adler32's speed, added profiling flags to Makefile.am for
    current speed tests.

commit 99e763889ff164be4fe939b54d80c13c5422857e
Author: bharring <bharring>

    minor screw up adding un-matched blocks at end of a
    onehalfpasscorrecting run is now fixed, producing what patches that now
    results in md5 correct tarball's *should be* 100% of the time.
    Other then that, cfile needs heavy optimization.

commit 5073646e073461dc870424996a6b2969ba0b693e
Author: bharring <bharring>

    tweak to makefile.am to add diffball's needed files...

commit 6b544e3d591179c08ca8d35080ca48351d5959fe
Author: bharring <bharring>

    added a quick hack to handle unmatched files in diffball.c
    Impressive so far, although I think switching encoding will be where it shines better...
    Just need to implement a correcting alg for the dcbuffer for switching, since it *will not* allow for add's in a row.  Probably will add this into the switching encoding as a function, from there, clean up after I implement an optimizing function for gdiff and switching.
    Hoozah.

commit d41dba2e8e46ee57f1c5f0520285ac633802ee09
Author: bharring <bharring>

    kind of a proxy/placeholder for string.h, this *should* from now on be
    the included rather then string.h
    Main reason being that this holds missing string functions (such as
    strnlen).

commit 553a50b200d8be334b4d55c8905d31526e0a89e6
Author: bharring <bharring>

    Hokay.  Pulled an all-nighter (I was intending to anyways due to
    circadian rythym fun), and built upon the work I'd done in diff-algs to
    make onehalfpasscorrecting position independent (via cfile).
    Long story short, cleaned up diffball.c (and implemented string-misc.c's
    use of config.h), tacked on differencing support, and have it now able
    to do per-file/tar-entry differencing.
    Nice.  So far offset_start results in a somewhat equal/slightly worse
    output, dc_pos on the other hand results in an output that *so far*
    looks to beat the shit out of differ.c (mine), xdelta (macdonalds), and
    bdelta (jjw).
    Nice.  Need's further work though, haven't yet implemented the ability
    to handle *unmatched* files.  That's not hard though...

commit 3a246487bd23e57af5609916b47d29b0dee4c357
Author: bharring <bharring>

    oh yeah.  Finally got cfile's cseeking and all whipped fairly well into shape.
    BOOYAH!

commit 80057857171368967fc08b4b0d2c0f14a69eaf6f
Author: bharring <bharring>

    work in progress.  do *not* use this revision.

commit c8aba059a67dd7dad615566c9d753887d0f8c17d
Author: bharring <bharring>

    hokay.  trying something, so bear with me here.
    basically working to enable diff-algs.c (onehalfpasscorrecting) to work correctly w/ my rather hackney cfile.{c,h}
    frankly cfile scares me a bit, and that ain't good.  right now it looks like the ducks are lined up and everything works, although running multiple onehalfpasscorrecting against a file, then outputing the patch can cause some issues with switching encoding.
    Mainly since it expects to *never* see 2 adds in a row.  Soo.  Ought to write something there.
    Course I dislike that format right now cause it doesn't quite meet the standards I had deemed for it, so neh.

commit 59434cf7f78e6ff6eb1f47f42bebf8f2267f2b1c
Author: bharring <bharring>

    nips tucks tweaks abound.  Suffice it to say, work here and there.
    Annoying, switching did not behave as well as I'd like... smaller output, compresses equal or slightly worse.
    Crap.

commit 1cd951264834fc2bb7706e7a699458ac2e463583
Author: bharring <bharring>

    fixed bug in it that cause a bit o problems...

commit 9ff7583b7806869643b3778372918888d60ac3c1
Author: bharring <bharring>

    fucking accidentally wiped working directory, eg bugs might abound.
    seperated diff-algs.c's onehalfcorrectingpass into an init hr and the actual pass.
    needs work.

commit 66bf3076c5db45350266d74d93e2fa894b4f2219
Author: bharring <bharring>

    adding snapshot... heavy modifications, not finished yet, but getting rather close.
    primes.{c,h}= quick lil ctx setup for finding primes under ulong.  Nice...
    switching.{c,h}= rather similar to gdiff, but compresses better and is smaller typically.  Nice...
    rest of the stuff is modifications= gdiff I'm experimenting w/ rearranging the data (add data) to improve compression, unlikely to continue.
    It'll be reverted to gdiff v4, w/ the extra fun attempts split off as v5 along w/ some definite improvements (expanding the allowed range for a copy command and add len- look in switching if curious).
    Moving to enabling verbosity, and adding options.  further, enable prime usage for diff-algs (for hr_size), etc.  Basically some tweaks are in order...
    Also need to split diff-algs up a bit, initializing, running, etc.  Likely branch that, since unsure that splitting a tarball up into entries will result in better compression although I'm fairely sure it will.
    Need to work on hashing alg a bit, testing so far indicates for large tarball files using the whole dataset using 32 works typically great, although 24 sometimes edges out.  Worth using in tarball implementation.
    ~bd

commit c77e6563df283b3e8bb860e141ef26cb17678fe7
Author: bharring <bharring>

    buffer/assert bug.  fixed.

commit c5f28922242d874c35aa79f6a0db26d319b3a8c7
Author: bharring <bharring>

    rewrote authors, todo got updated, fixed a bug in cfile w/ memalias cfiles.

commit d91fce088098e86ff9ab940848cdb9b4fb92b592
Author: bharring <bharring>

    HOKAY.
    this ought to be the final correct version for autoconf and automake.
    it'll require usual aclocal/autoconf/automake/autoheader bolsh, but it ought to work.

commit 998bb24626c5b980b987525eb808efd4303a6d52
Author: bharring <bharring>

    fucking autoconf....

commit e37a3d0225eece6c292e77010302ee202c954b6e
Author: bharring <bharring>

    continued tweaking of the modifications to the adler32 hashing algorithm using a prime table.
    the table likely needs better optimization, since it was created rather haphazardly- also fixed a problem w/ parity (eg the checksum was always even).

commit 9d6a85245a75e29143bcc73862454665261b5b5a
Author: bharring <bharring>

    stupid frigging files...

commit 6dd673e53ac5dd9fe13251ed4f71472eb2093453
Author: bharring <bharring>

    misc, credit to those where I got the info/algs for the implementation is in authors, copying is the gpl 2 license..................

commit 3be865cce4b67bf3eb962ff067410c5588e9f869
Author: bharring <bharring>

    continued attempts w/ autoconf and automake... should be there now.

commit ea4200eca6e592e51877067f968c35b8fb156a7d
Author: bharring <bharring>

    heh.
    missed one.

commit f6101adbd88a4be611d02b33c72a38373bca2d37
Author: bharring <bharring>

    files (hopefully all that are needed for autoconf and automake move.

commit eeb0c9af1e15d745ead02afcbc1b1daec80d36fd
Author: bharring <bharring>

    *** empty log message ***

commit 44d36e29dbad2d65fdea5d007d84cc0a18fe7588
Author: bharring <bharring>

    prepping to use autoconf and automake, about there.

commit ebd82cdf8aa39dccbf237e4208b8bf95df3c0ea4
Author: bharring <bharring>

    Ok.   cfile mem aliasing works now, although I don't doubt there are conditions where it may fail. last commit had a rather *dumb* little mistake involving leaving an exit(0) in place.
    heh.  either way, cfile.c needs *serious* clean up.

commit ef0631ca063a32545bc759c481ba59cd5528ec56
Author: bharring <bharring>

    abstracted the rolling/flushing of adler32, added multiplier for some experiments.
    not sure what I modified in gdiff.c to be honest, likely just friggin printf debugging...
    Ah yes, works still btw.

commit 63282197701f00be6c435b364a686b6430981d15
Author: bharring <bharring>

    added gpl licensing header to *.{c,h}

commit ebaccc607aff452bdc6620ac5b790658b84cb62e
Author: bharring <bharring>

    code needs cleanup, but cut normal memory usage down by a third (removed reliance on having files loaded into memory).
    Performance issues exist- haven't optimized the buffers at all, so that needs to be handled...
    further, enable the ability to work off a file *in* memory- basically allow for that which I just removed. heh heh.
    ugh.

commit 52abea36bd120c2ab0bb62d4c4d788845494379f
Author: bharring <bharring>

    Hokay.  The long stupid road of dealing w/ buffers in cfile is now (99.999% likely) fixed.  All knowns faults/problems are fixed.
    next is clean up of diff-algs, further tar specific diff-alg needs to be added.

commit 6e0522224d9058c9a043336bc1f5c6df572da142
Author: bharring <bharring>

    Add DEBUG_CWRITE definition for debugging cwrite...

commit 979d0c7b59e0ff8c2e6c1e4e310507ff693941a5
Author: bharring <bharring>

    commented out code removal continues I'm afraid...

commit 75344f4f92a8a2ea22e9d0453e8a0abbaaa69ce8
Author: bharring <bharring>

    cleaning up commented out code, need to add verbosity options to the multitude of printfs...

commit a7536a091b4aa66158ce838f2d18dbc079169fc0
Author: bharring <bharring>

    Well, forgot a file.
    shite.

commit f06a44927245223b3e1886df7ecf2536099f147a
Author: bharring <bharring>

    Hokay.  GDIFF reconstruction is now seperated from the actual writing, it basically recreates the delta commandbuffer, and a standard function interpreates the dcbuffer's commands and does the file reconstruction.
    This will allow for doing inplace reconstruction much easier, although streams/stdin files ought to make it fun...

commit 15af6a38d7cd956880ac10a9591ef8c24c326c13
Author: bharring <bharring>

    Heh. Forgot to add those puppies...

commit abaa7617d7d9ab0165b28f4deb37288e9097033a
Author: bharring <bharring>

    delta.{c,h}=>diff-algs.{c,h}, apply-patch.c temporary name for reconstruction code.  Continuing on that endeavour

commit e566a60b1542b8f482bf00aa7a1456abc0fca1aa
Author: bharring <bharring>

    *** empty log message ***

commit 6db32e49406634754c36afce8f11d713fd5b0226
Author: bharring <bharring>

    fixed warnings.

commit 0793c13dff4285e7bcf7a8c09288fefdc68fe21d
Author: bharring <bharring>

    fixed yet another annoying bug w/ 247 248 in the encoding.  either way, tiz fixed, moved the dcbuff output to cfile.  next need to move the decoder to create a DCBuff, and isolate that actual file manipulation into a common area.
    lovely.  the dumb dog just puked...

commit e24d0a448ff38013df590441075b875403406d4d
Author: bharring <bharring>

    Lots and lots of changes.  View-patch is basically a simple perl version of od, since OSX's od/hexdump are completely and utterly brain dead (seems to be a common affliction of their ported utilities...).
    Either way, modified gdiff to use a new struct cfile, which is basically a layer of functions over read/write/lseek... reason for it's uses is it simplifies gdiff (cuts down on # of buffers), and allow for me to enable compression/decompression w/out having to write it into each encoding scheme.  Needs some ironing out I'm sure, but so far it's behaving well.
    note, only reconstruction uses it at the moment- I need to move outputing over also.
    Finally, moving the various encodings over to a different setup- reconstruction function is giving the delta, and it should in turn recreate the deltacommandbuffer- output, giving a commandbuffer, outputs it.  pretty straightforward.
    I want to do this so I can move the actual copy/add portions into a seperate, common file- makes controlling/debugging easier for reconstruction.
    That and it makes adding a new encoding much easier- just fit it into the cfile/DCBuffer setup, and the rest will be handled (including compression). nifty.
    NOTE- for next month/month-and-a-half I'll be developing on OSX.  I intend to obviously keep this code so it works w/ true linux machines (since this is mach), but likely there will be screw-ups.  At least it'll make porting this beast easier...
    ~harring

commit 17484a221711078c6a06ed84fe5834efb829b857
Author: bharring <bharring>

    moving uncompleted/non-compiling files into cvs due to fire, transferring cvs to backup-repository and will be working from there while
    displaced.

commit c4a69590df14a002e56d0a81fe3d8edcaac6d7ad
Author: bharring <bharring>

    Ahem. Huge amount of moving/tweaking/fixing. Patcher now is basically a simple options handler. It calls the appropriate encoding
    decoder/reconstructer, which uses a buffer structure provided by pdbuff.{c,h} that will allow me to basically handle stdin for
    a delta stream without having to be aware of if it's stdin or from a file. Accomplished via refreshPDBuffer, and initPDBuffer.
    Haven't implemented stdin support, but it shouldn't be hard- I've abstracted it enough to allow for it.
    Either way, plugging new encodings/reconstructors in shouldn't be too hard- what will suck is plugging GNUDiff style encodings in,
    specifically unified.

commit 73f2d8f9bf888080385b0b64fd6bbc4062b24e99
Author: bharring <bharring>

    added check for unknown offset type (and defined ver5/6).  Side note not listed in previous revision, Makefile has been completely
    rewrote 3 times now- 1st was crappy, 2nd/3rd were basically 'gcc -MM', w/ third being better target definition for the binaries.

commit 5822f9c2faa8266cbc6af4c3a14c16213fa77652
Author: bharring <bharring>

    fairly major work- got a good start on having delta.c able to handle multiple encodings, added gdiff.{c,h} which is the gdiff encoding
    still need to modify patcher.c to handle multiple encodings (based off of the files magic), need to modify differ.c so it works off of
    file rather then loading the entire file into memory, and need to do checkpointing (still).
    next thing to be commited will be the patcher modifications, and (likely) any fixes to gdiff.c related to positioning.

commit 0a874acbfff66bbfd87c3c015abfe2ff706ddc2b
Author: bharring <bharring>

    code-clean up, abstracting/moving hardcoded values/functions/transformations into bit-functions.c...

commit b8f57bc52366ddec29670a8c2c771e1dfc0b6971
Author: bharring <bharring>

    note, last revision didn't had a flaw in the encoding since I half changed it and didn't finish it prior to commiting..

commit 69136fb3209e508462189929cb2f76dbd330bf0c
Author: bharring <bharring>

    *** empty log message ***

commit 8022aa6a387e1966b3da8609fad3db27bcc0d5b9
Author: bharring <bharring>

    Fixed update-cvs, since it wasn't picking up the second line of files to commit.

commit 562ff48b093a6469474fa42eac4f62907a657061
Author: bharring <bharring>

    Rearranged/recreated Makefile, fixed up complaints from -Wall, split off bit-functions.{c,h}

commit 47114ec18e240ce151d9931c3db3b29ada7e937a
Author: bharring <bharring>

    going to add relative addressing to last dc offset, and then try jjw's encoding.

commit 1d36afc1df0453b30b4731a9f02e86a605f96528
Author: bharring <bharring>

    rearranging, basically tacking stats into the diffing for A) information B) adapative encoding down the line

commit 26e80bcdf140c7a0bbe7420e5471743644da336e
Author: bharring <bharring>

    fixed a buffer bug, basically added a check to the while loop for the patcher buffer in case it's used up all of the command data in
    the buffer, but hasn't hit a refresh yet.  basically no data buffered, which could (and did) result in it exiting early on
    evolution-1.2.4/1.3.1 tests.

commit 46bbbf2b8196bcc0c6376dfe947ae58e5c62cf2a
Author: bharring <bharring>

    for statistical info.
    try diffing/patching evolution.  It fails the md5summing, and I want to know why.

commit 040deef1aa0576e87d013c7ffbd6bfd62af0e1e8
Author: bharring <bharring>

    hokay, tweaks, about it.

commit 241bfb6c3e74fc4d03dd1dcbb13b33a0a22c1a5a
Author: bharring <bharring>

    Finally getting somewhere here people.
    differ produces (from what I can see) a correct (eg no fuck-ups) encoding.
    patcher applies (from what I an see) correctly (eg no fuck-ups) the patch file.
    Hoozah.  Tested it against debianutils-{3,7} uncompressed tarballs (roughly 100k), it behaves.
    Now for some serious clean up/refactoring.

commit f1ac6e9e44050d16c319890eae27cb086f14a542
Author: bharring <bharring>

    still at it.

commit b9ab3cb48bacd28e44b618d0b2cf58f26eb0c98c
Author: bharring <bharring>

    ironing out the lovely encoding/patching.  mostly fixing offset errors now.

commit 672fbb195e2102e2145a154f02357a054c818cd3
Author: bharring <bharring>

    not needed anymore- I already have a *mostly* correct tarball handling code.  Extensions need to be added though.

commit 473d8f93e0f01629ccf8122faf30addf95183b3e
Author: bharring <bharring>

    *** empty log message ***

commit dc5def9463fcf3dd631364e7de01497d8e7eb755
Author: bharring <bharring>

    patcher is to be the main source for using a delta to patch the reference.  Fixing fuck-ups in the encoding, adjusted it to be
    reference based from the current position in the version file- offset no matter what continues to grow, using reference (and losing a
    bit to sign) this runs at worst equivalent to doing absolute.
    That said and done, this needs heavy tweaking to remove hardcoded limits (using long's for file size for instance).

commit 355938ba15488781046c052e6b3dd1386b1ced70
Author: bharring <bharring>

    Hokay.  So far, exempting possible len errors, outputing what should be a correct encoding.
    Watch the command bits in lookback- it will need to be modified to handle commands being pushed out of the buffer.

commit 4639e126b1466e0afc3ae1799cde882b11644759
Author: bharring <bharring>

    encoding output is *mostly* right, although final copies (at the tail for the file) don't appear to be coming through correct

commit 19f7da91d872923daca0ee6753f9f9686579a928
Author: bharring <bharring>

    adding encoding output.  Looks to be a bit fucky...

commit 4669acf2f2c97741769ddf486382e3da645592a6
Author: bharring <bharring>

    adding lookback buffer, likely major clean-up/refactoring to come after. getting a bit messy.  also added update-cvs, simple bash
    script for commiting ONLY the files listed within.  since my working dir is a bit messy, this saves on cvs headaches.

commit 83a8a2afac5eb32d5fee8030bd03c3815f7bf989
Author: bharring <bharring>

    got hashing going well (65% load for under 200,000 elements usually).
    I've got forward and backwards matching going also, need to implement correction, and encoding.
    Obviously a delta buffer will be needed, so that needs to be implemented (llist? malloc block, w/ two pointers?  structs needed?)

commit 0981490eee62b81d76fc3e8663d72741fcb08b05
Author: bharring <bharring>

    Wrote rolling adler32 into delta.c alongside a chcksum array generator.
    Unfortunately, probably will need to code a rolling adler32 function, cause for each chksum computation you can save 2*(seed_length -1)
    additions as long as the offset for the footprint is +1.  Can do it further (eg now bound by starting pts being adjacent), just haven't
    wrote out the algs.
    Prob something like L=seed_length, for orig chksum starting at 2 len=L, and computing starting at 4 len=L.
    s1.total@4 = s1.total@2 - s1.2 - s1.3 + s1.(3+L) + s1.(4+L);
    s2.total@4 = s2.total@4 - (L*s1.2) - (L*s1.3) + (2*s1.(3+L)) + s1.(4+L);

commit c24ec15683d3465f7636ccbc6d5ea19c3efbf0ef
Author: bharring <bharring>

    delta.* should, assuming I ever get that far, become a true lib interface.
    For now, it holds the majority of the delta functions.

commit 0c8813a0ffe811747727fda17d3d209a2ce73523
Author: bharring <bharring>

    *** empty log message ***

commit 0640827e1368a94af478ca5a87ea441128f30ecd
Author: bharring <bharring>

    *** empty log message ***

commit 5a6d1e0e536c4b515a1bf2617410ce868d8925f6
Author: bharring <bharring>

    added check to tarball file for an entry that stops checking for most common field entry when the most common (Thus Far) has a count >
    (the array being checked+1)/2.  The plus one is for odd files, it insures that it only kills when an entry has > half.

commit be957927e71af5a1bdea5f9dec4381f388fe4adf
Author: bharring <bharring>

    added len fields to common strings in structs to improve speed, fixed a bug in the common-field alg- basically, a fair portion of the
    strings past in (mtime for instance) may not be null trailled, soo... it never matched. Always went quadratic.  not good.  Fixed it,
    now it first checks the stored/passed strnlen, if they match, strcmp's only the stored/passed len (which should be equal).  Cut running
    time down on openoffice's tarballs from something like upwards of 10 minutes, to the limitations of disk i/o.  Like a minute or so.
    Memory usage went up from 24 mb for total (that includes arrays), to in the range of 25-27mb, all dependent on how many different
    versions there are in the fields.

commit 56150bd0abb3960092d2616346e8391af460c702
Author: bharring <bharring>

    Common tarball entry processing working.
    However.  It's slow as Shit.  With a capital S here people, it's that bad.
    Needs to be improved... try it on openoffice tarballs, it is fricking slow.
    Possibly, push this processing to the dumping of files, or have it as a check done when processing
    each target tarball entry?
    Also, could use hashes for it, but I'm not really up for making it *that* complex.

commit bf5e33adcbe94059e1d08daa38d5251fb4b64fd0
Author: bharring <bharring>

    Slowed it down quite a bit.  It's now reading every tarball entry, and finding the common ones.
    Could speed it up quite a bit by doing strnlen check once, and storing it.

commit 01f641b65c61e9fb4f9eb8bce80b618b17ee0a8d
Author: bharring <bharring>

    updated/modified llist structs, for identifying most common entry of certain fields.

commit f7d9111a38540182903c1874f958c7ad79b207d1
Author: bharring <bharring>

    *** empty log message ***

commit e9f47afaef4cf767987a0c069ce092dc4dd63a2b
Author: bharring <bharring>

    to hold doubly linked list structs, single linked.. etc.

commit 3b44ec8bc48b099bf1f2bd0d955ac29a7e9eb945
Author: bharring <bharring>

    Tweaks here and there- added (source|target)_matches, adding libmagic support next.

commit 3a71936f00beb0982acfaee860b8682ec308e128
Author: bharring <bharring>

    Removed reliance on binary md5sum- now using openssl's evp functionality (openssl/evp.h)
    updated Makefile to reflect the new library need.

commit e5ca118ec57817a84df7ab856747fc3a045447e8
Author: bharring <bharring>

    added offset adjustment for gnu extension longlink.  needs verification

commit 38256e4268d40b27c0265b29b819f3d89e721667
Author: bharring <bharring>

    Added Makefile, moved tar constants/functions into tar.c and tar.h.
    diffball.c minor modifications, gnu-tar.h is /usr/include/tar.h

commit cb3a0c0d1b61e973e23a9aa2c6c21b7726919073
Author: bharring <bharring>

    fixed issue w/ mallocing one byte to small- it was leading to diffball going quadratic in memory usage.
    For 60,000 files, it now takes around 8mb to keep it in memory.  Nice...

commit 430b0804fc0695528e7687dd31263a8aab1d76a0
Author: bharring <bharring>

    stupid fucker is taking a shitload of memory.  what the hell.

commit 0e88747730e9b06ee36342fa5aaefd9fecd14409
Author: bharring <bharring>

    Tweaked for better memory usage (save 29,000+% for best case, and likely over 400% for worst case);

commit 867b06618ee6a9c49c2faa27051ac1a23a8ba894
Author: bharring <bharring>

    *** empty log message ***

commit d1249e7d2327ab3025bd02604f70812d4587e2a8
Author: bharring <bharring>

    Issue w/ ././@LongLink still exists in oo-1.0.3_src
    Beginning attempt at matching tar_entries, and entry_header diffing.

commit 98b5ca4896fad42f34b56f12735f266a78a945bb
Author: bharring <bharring>

    *** empty log message ***

commit 6308b02ca9a8dbe2a43a4165aecb54b53c28f78d
Author: bharring <bharring>

    *** empty log message ***

commit 59d0f5c1d5af6872ae72e520a77836b2125700df
Author: bharring <bharring>

    *** empty log message ***

commit 421f07809d607aeb1a1d4e80e66b769a2ac70dde
Author: bharring <bharring>

    *** empty log message ***

commit feadb1ee2e2d8fc58c070a2abb3dbb2d069469ee
Author: bharring <bharring>

    *** empty log message ***

commit 7751353ede103f8b540b59c1f3aa17b70fecc189
Author: root <root>

    Initial revision
