summaryrefslogtreecommitdiff
path: root/xdelta3
AgeCommit message (Collapse)Author
2008-04-15Fixes for issue 70. The test inputs tickled a case where thejosh.macdonald
non-blocking API would not make progress, searching for match on the page boundary and repeatedly asking the application to get another source block. This is only a non-blocking issue, glad it's fixed. I'll add encode_decode_test to my test suite before another release.
2008-04-09Add mechanism for processing merge arguments in order, memoryjosh.macdonald
management for xd3_whole_target structures being passed between arguments-processing and main merge_output function.
2008-04-02merge command now successfully rewrites a single input. success!josh.macdonald
2008-04-02Merge command passes its first smoke test, reproduces a delta byjosh.macdonald
instruction-level rewriting of the input to the output. Hits an assertion due to target-copies that cross window boundaries--next step is to fix the output window size to the last input window size. This fixes an important bug in xd3_choose_instruction, which assumed that copy instructions were at least four bytes, but this is no longer the case for merged outputs (at least for the case where window size changes, which will not be allowed in future).
2008-03-31Implement more of "xdelta3 merge" command. Compile fixes.josh.macdonald
Work on xdelta3-regtest.py.
2008-03-14Fix for issue 67, test now fails as root with an error message.josh.macdonald
2008-02-11Can't explain how the usize_t->uint32_t changes here solve the problemjosh.macdonald
described by issue 61, but it's likely the same as issue 66 and I'm going to apply them blindly with TODOs to try and understand it later.
2007-12-28Fix for DJGPP (1/4)?josh.macdonald
2007-12-28Fixes issue 58. Incorrect looping condition in xd3_smatch() causedjosh.macdonald
checking for matches with incorrect checksums, detected by debug builds. This also improves speed slightly and potentially degrades compression where the sprevsz array had spurious matches beyond its actual coverage.
2007-12-27Adding more support for "merge" command.josh.macdonald
2007-12-17Call main_input() recursively for main_merge_arguments.josh.macdonald
2007-12-17Fix.josh.macdonald
2007-12-17Patches for Makefile.mingw and xdelta3.h comments, work on "merge"josh.macdonald
command. Fix for new return value in stream_overflow.
2007-12-15Patches for mingw.josh.macdonald
2007-12-15Compile on OS X, better implementation of UNALIGNED_OK=0 small checksum.josh.macdonald
2007-12-14Add skeleton for "merge" command.josh.macdonald
2007-12-14Fix another snprintf issue. (Wish I had "patch" on my cygwin box... Wishjosh.macdonald
I wasn't using a Cygwin box for dev purposes.)
2007-12-14Fix a few more uint uses.josh.macdonald
2007-12-14Replace uint with usize_t, more portable.josh.macdonald
2007-12-14Fixes for Win32 and vsnprintf.josh.macdonald
2007-12-1480-col fixes.josh.macdonald
2007-12-14Implement UNALIGNED_OK for xd3_small_cksum_update, hoping to fix josh.macdonald
test failure on Xscale/ARM platform.
2007-12-13A few cleanups -- xd3_scksum should have been using uint32_t, but Ijosh.macdonald
doubt that is the real problem. I suspect the problem on ARM/Xscale has to do with unaligned access issues.
2007-12-07Release 3.0tjosh.macdonald
2007-12-07Patches from A. Mennucc for Debian. Fixing makfile for Cygwinjosh.macdonald
conditional compilation, which is irregular.
2007-12-06Adjust the low-end range for command_line_arguments unittest. I wish thisjosh.macdonald
didn't depend on the system rand() function, but it's portable at least.
2007-12-05Fix bug in xd3_set_source, the optimization for shiftby/maskby werejosh.macdonald
in some cases not setting those variables, which are used by xd3_blksize_div
2007-12-02Fix recode, recode test.josh.macdonald
2007-12-02Fix an off-by-one bug lurking in the array initialization josh.macdonald
in djw_compute_mtf_1_2
2007-12-02Working on the test failure in dbeug mode, partial recode test written.josh.macdonald
2007-11-27Diagnose expected non-failures test failure introfuced in 3.0s relatingjosh.macdonald
to DJW and compiler optimization level. Mysterious.
2007-11-26The rabin-karp checksum looks better in testing but doesn't reallyjosh.macdonald
seem to improve things in practice. Removed HASH_PRIME.
2007-11-15Get uniqueness and fullness metrics working, add compaction parameter forjosh.macdonald
hash-table size bias.
2007-11-13Learned a lot about Rabin-Karp performance.josh.macdonald
2007-11-13(no commit message)josh.macdonald
2007-11-11Compile with g++ 3.4.4 and add C++ checksum_test.ccjosh.macdonald
2007-11-10Adds output regarding inefficient copy instructions to "printdelta".josh.macdonald
Adds crude inefficiency check to xd3_smatch(), preventing 4-byte matches if (here-addr) >= 2^14 and 5-byte matches if (here-addr) >= 2^21. Nice improvement.
2007-11-09Release 3.0sjosh.macdonald
2007-11-07A few optimizations: -1 (fastest) is faster than before. -fno-builtinjosh.macdonald
makes a difference. New implementation of xd3_forward_match() is written and tested, but not enabled, needs a little more benchmarking.
2007-11-07Adds compare_test for benchmarking xd3_forward_match, the optimizedjosh.macdonald
version is not yet enabled in xdelta3.c
2007-10-30Update FASTEST config based on 50-round test.josh.macdonald
2007-10-29Running experiments for new -1 and -2josh.macdonald
2007-10-29Restore old secondary djw defaults.josh.macdonald
2007-10-29Add a new COMPLEVEL_1, previous -1 becomes -2.josh.macdonald
2007-10-29Add a new fastest, fastest->faster (-2)josh.macdonald
2007-10-29Inline some functions.josh.macdonald
2007-10-29Use maskby/shiftby when source->size <= source->blksize, which helpsjosh.macdonald
xd3_process_memory() uses.
2007-10-28New xd3_blksize_div() function used everywhere to avoid __umoddi3josh.macdonald
2007-10-28Eliminate calls to __umoddi3 (compiler generated for 64-bit % 32-bit).josh.macdonald
More 80col reformatting.
2007-10-2880col changes.josh.macdonald
generated by cgit v1.2.3 (git 2.39.1) at 2024-11-14 04:12:02 +0000