1lrzip ChangeLog 2 3JUNE 2016, version 0.630 Con Kolivas, Peter Hyman, Petr Písař, Joel Fredrikson 4 5* checksum.buf should only be changed after the semaphore wait 6* Update README 7* Add documentation for recursive mode 8* Implement gzip compatible -r recursive option 9* Add initial argument processing for recursive option 10* Tidy 11* Add one more verbose for compat mode 12* Add support for various combinations in compat mode 13* models is array of chars. char's signess is implementation specific. It's 14unsigned on ARMv7. Unsigned char cannot represent negative values. GCC 6 15complains about it: 16* Fix decompression of multiple chunk encrypted archives 17* Tidy gotos 18* Show correct lengths during testing on big endian and compressed archives 19* Update copyright dates 20* Allow less than maxram to be malloced for checksum to fix Failed to malloc 21ckbuf in hash_search2 22* Base temporary output buffer on maximum mallocable, not maxram 23* Enable subdir objects for future automake compatibility 24* Add support for -m option in lrztar 25* Big endian fix for Solaris Sparc courtesy of joelfredrikson. 26* Fixed typographical error, changed accomodate to accommodate in README. 27* A whitespace fix on lrztar. 28* Add sanity check to prevent trying to malloc more ram than a 29system/environment is capable of 30* Cosmetic help change for compat 31* Add rudimentary manpage for lrz 32* Fix lrz symbolic linkage 33* Do not fail if we are unable to write temporary files, giving a warning only 34that it might fail if we don't have enough ram 35* Try /tmp/ if none of the temporary environment directories or the current 36directory are writeable 37* Set STDOUT correctly in compat mode 38* Style police 39* Fix false warning on decompressing from stdin without keep files 40* Fix false warning on compressing from stdin without keep files 41* Don't show extra message in compat mode decompress 42* Show correct appname when called in compat mode 43* Add support for progress, fast and best flags in compat mode 44* Add compatibility mode with gzip when called as lrz 45* Correct adding slash to control->tmpdir. off-by-one error. 46* Update manpage for long options 47 48 49MARCH 2015, version 0.621 Con Kolivas, Peter Hyman 50 51* Fix libzpaq.cpp warnings 52* Fix warnings in LzmaLib.c 53* Making the high buffer only one page size is faster for sliding mmap mode 54* Fix incompatible log callback in liblrzip 55* Use PRId64 instead of lld in fscanf 56* Use int64_t for i64 57* Fix ISO C warning in lrzip.c 58* Fix ISO C warning in lrzip.c 59* Fix ISO C warning in runzip.c 60* Fix ISO C warnings in rzip.c 61* Fix iso c warning in util.c 62* Fix EOL to unix on libzpaq 63* control->tmpdir fixes 64* Null terminate long options to find unhandled options 65* Trivial date 66* Add long option support 67* Brace failure lead to broken dump to stdout 68* fflush messages on print_err 69* Fix inverse logic 70* Honour the proper unix environment variable for temporary directory and then 71try variations 72* Attempt to decompress file in ram only if we are unable to create temporary 73files when decompressing to stdout 74* Avoid checking return values repeatedly on compress that only lead to failures 75anyway 76* Microoptimisation 77* Don't check for failure condition that can no longer occur 78* Don't check twice for failure conditions that are fatal in hot paths 79* Cache the chunk bytes value to avoid setting it on each read_header call 80* fake_mremap is only used when defined to mremap 81* Remove unused cksem functions 82* Fix remaining use of mutexes lock/unlocking in different threads with cksems, 83corecting cksem usage on osx 84* Update copyright dates 85* Make match_len a function completely removing all indirect calls to get_sb, 86significantly speeding up the single_get_sb case 87* Make full_tag a pointer allowing us to avoid a function call for get_sb 88* Call sliding_get_sb directly in sliding_next_tag 89* Make next_tag a pointer to allow ordinary mapping to avoid an extra function 90call 91* fix order of lrzip.conf search 92 93 94MARCH 2015, version 0.620 Con Kolivas, Peter Hyman, Chris Spiegel, Elmir Jagudin 95 96* Increase maxram when we abandon use of temporary input/output buffers 97* Don't delete the tmpinfile when decompressing from stdin before allowing seek 98to end to succeed in checking md5 99* Use temporary file from read_seekto when STDIN will not fit in the ram input 100buffer 101* Remove unused read_i64 function 102* Add message about issue tracker in BUGS 103* Use a common exit path iin lrzip_compress/decompress and fix lr leak on 104successful return 105* Fix parenthesis placement inside of unlikely(). 106* Clear sa_mask and sa_handler before calling sigaction(). 107* Fix for lrzip -i. Decompressed size wrong 108* added '-m' command line option 109* Fix wrong README file being included in Makefile 110* Pass strict sizes to decompress length, rounding up only the amount we're 111allocating to not confuse decompression libraries 112* Convert the thread locking to use cksems 113* Add cksems to util.h 114* Fix 'Failed to malloc ckbuf in hash_search2' with very large files. 115* Round up compression and decompression buffers to page size since malloc will 116allocate them that large anyway. 117* Increase the compressed buffer size given to libzpaq in case of incompressible 118data since it does not check if it's trying to write beyond the end of the 119buffer. 120* Provide a helper function to round a value up to the nearest page size for 121malloc optimisations. 122 123 124SEPTEMBER 2013, version 0.616 George Makrydakis 125 126* Making things more homogeneous in checks, some space removed. 127* Simplifying and refactoring logic, alignment edits, empty semantic rewire. 128* Target directories must always exist and -f should not create them. 129* Deferring output directory after -f check has taken place. 130* Preserve extraction semantics without resorting to tar stripping. 131* Making sure last forward slash is removed from input path. 132* Removing pointless forced overwrite point check. 133 134 135SEPTEMBER 2013, version 0.615 Con Kolivas, George Makrydakis 136 137* Adjusting -O flag semantics, options help update. 138* Making -O flag operative for lrztar, whitespace path fix, lrzuntar fix, other 139minor. 140* Further tighten up ram restrictions with stdin/stdout to prevent running out 141of memory with all the buffers involved. 142* Massive files fail with -U due to trying to allocate the whole lot in ram 143while doing checksums. Do it piecemeal to avoid the problem. Patch and debugging 144courtesy of Adam Tkáč. 145* We have to run through the clear buffer function even for empty buffers or 146corrupt archives with empty match streams. 147* MD5 code uses little endian so remove arbitrary SWAP macro and explicitly use 148htole32. 149* Rewrite the magic if we receive lzma properties and have not yet written them 150yet during stdout operation. 151* Set the control lzma properties only once. 152* Add a control lock mutex for protecting certain control variables. 153* Fix stdin fake mremap creating null bytes on osx, patch courtesy of John 154Boyle. 155* Cache frequently used indirectly referenced variables in the sliding mmap 156code. 157* Micro-optimise sliding_get_sb_range. 158* A fix for a bug where large files containing the same non-zero bytes which 159requires a sliding window, courtesy of Serge Belyshev. 160* Put vchar should take a 64 bit integer and is used from more than one call 161site so uninline it. 162* Microoptimise in hash_search. 163* Inline rzip functions used from only one caller. 164* Check for successful calloc of hash table only after performing it. 165* stdin_eof is just a bool 166* hash_bits can only be up to 64 so use a char type. 167* Check endianness of build with autotools to enable md5 support on more 168platforms. 169 170 171JULY 2012, version 0.614 Con Kolivas 172* Fix lrztar failing due to incorrect fix for -S freeing segfault. 173 174 175JULY 2012, version 0.613 Con Kolivas, Serge Belyshev. 176* Fix other non-win32 compilation for libzpaq. 177* Fix libzpaq compilation on other non-win32 OSs. 178* Fix Types.h missing which breaks typedefs on macosx. 179* Cope with missing ffsll() on various platforms. 180* Fix get_ram compilation on osx. 181* Rename lrzip.h to lrzip_core.h to cope with case insensitive filesystems not 182working with Lrzip.h vs lrzip.h 183* Include test for pthread.h and include header in lrzip_private.h. 184* Use system extensions in configure.ac to allow to build on non-gnu-linux 185platforms. 186* Fix invalid free error when -o and -S are used together. 187* Use no back end compression for blocks smaller than 64 bytes to avoid issues 188to do with minimum buffer size and minimum match sizes of the rzip 189preprocessing. 190* Fix md5 calculaton on blocks larger then 2^32. Bug reproduced and debugged by 191Serge Belyshev. Patch by Paul Eggert. 192 193 194MARCH 2012, version 0.612 Con Kolivas 195* Minor formatting improvement to delete misaligned %. 196* Implement progress counter on zpaq compress/decompress per thread within the 197Reader function. 198* Add some lrzip style defines and optimisations to the lrzip functions in 199libzpaq.h 200* Re-instantiate the lzo compresses test for zpaq compression. 201* Remove redundant code. 202* Delete now-unused zpipe code. 203* Implement virtual write support for bufWrite to speed up zpaq support and 204simplify the virtual read function for bufRead. 205* Add virtual read() support to the bufRead class to speed up encoding/decoding 206via the zpaq library. 207* Use libzpaq backend for decompression. 208* Move zpaq compression to new libzpaq library back end. 209* Merge branch 'master' into zpaq 210* Add minimal libzpaq compilation support. 211* Move to libzpaq directory. 212* Import libzapq files. 213* Fix Lrzip.h not being installed into include directory and make distcheck 214failing. 215 216 217MARCH 2012, version 0.611 Con Kolivas. 218* Move the update counter to earlier in the hash_search so it still updates even 219when there are not matches. 220* Perform all checksumming in a separate thread to speed up the hash search in 221the rzip phase. 222* Optimise the hot function empty_hash. 223* Micro-optimise the hot next_tag function. 224* Micro-optimise the match_len function. 225* Pass one less variable to the get_sb function as an optimisation. 226* Cache the hash_entry value passing one less variable to hash_empty as a 227micro-optimisation. 228* Use ffsl for a faster lesser_bitness function. 229* Prevent lrzuntar from automatically overwriting existing directories without 230the -f option being specified. 231* Calling lrzip via lrzcat would fail as the proper parameters weren't being 232passed to the control structure. Fix. 233 234 235MARCH 2012, version 0.610 Con Kolivas, Michael Blumenkrantz 236* Implement complete set of liblrzip libraries, documentation and example uses 237with support for simple lrzip_compress() and lrzip_decompress() or complete 238fine-grained control over all compression and decompression options. 239* Use as much of the low buffer as possible with a single memcopy before going 240fine grained byte by byte. 241* Preserve the compressed time on decompression where suitable. 242* Store a copy of the control struct to be reused on subsequent files to prevent 243variables being modified in the control struct on the first file that corrupt 244compression/decompression of the 2nd file. 245* Explicitly select C99 to avoid certain warnings. 246* Generic modifications to silence -Wextra warnings. 247* Fix typos. 248* Use an array of parameters in lrztar to allow working with directories with 249spaces in their names. 250 251 252SEPTEMBER 2011, version 0.608 Con Kolivas 253* get_sb only allows accessing one byte at a time, yet we don't need that 254functionality when sliding mmap is not in use. Use different versions of the 255function and the larger memcpys depending on whether sliding mmap is in use or 256not. This affords a substantial speedup in the rzip phase of files not requiring 257sliding mmap. 258* A small optimisation of moving the check for remapping the low buffer also 259speeds up the sliding mmap version slightly. 260* Installation of the included bash completion script was disabled to avoid 261conflict with distribution bash completion packages. 262* Minor cleanup of sha4.h 263* Make mremap use the fake mremap on all platforms not linux. 264 265SEPTEMBER 2011, version 0.607 Con Kolivas 266* Updated lzma library to version 920. 267* Fixed a rare unable-to-decompress corner case. 268* Added lrzip completion script. 269* Updated makefile to more portable posix version. 270* Hopefully fixed compilation on FreeBSD not supporting memopen. 271* Added lots more debugging information about offsets in max verbose mode. 272* Removed error and warning messages that are spammy or harmless. 273 274MAY 2011, version 0.606 Con Kolivas 275* lrzuntar broke as lrzip doesn't automatically use stdout now, so use lrzcat 276in the lrztar script instead. 277 278MAY 2011, version 0.605 Con Kolivas 279* Addition of explicit lrzcat symlink which automatically sets "-d -o -" 280* Revert automatic output to stdout due to addition of explicit lrzcat 281executable, and to be consistent with gzip. 282* Fix counter from being zeroed on every loop in rzip.c which was creating 283excess progress output. 284* Update GNU license. 285* Set aside at least page_size of zeroes when compressing a chunk, thus 286allowing files that are otherwise too small to compress with lrzip to be 287compressed. The zeroes will just be dropped on decompression. 288* Add explicit checks to cope with zero sized files as well. 289* Remove the arbitrary 128 byte minimum size restriction on archives now that 290any size file can be safely archived. 291* Set the eof flag when reading older file format files to not hang on zero 292byte sized archives. 293* Update the percentage counter when getting file info to not overflow with 294tiny files. 295* Set the maximum file mode to 0666 to avoid setting the executable bit when 296permissions are not carried over (such as from STDIO). 297 298APRIL 2011, version 0.604 Con Kolivas 299* Detach threads after creating them on the compression side. Not joining them 300meant that compressing massive files requiring hundreds of threads would 301eventually hit the resource limit of number of threads created even though 302the threads themselves would exit. 303 304APRIL 2011, version 0.603 Con Kolivas, George Makrydakis, Jari Aalto. 305* lseek in stream.c wasn't being compiled to the lseek64 variant on Apple 306due to missing includes, breaking >2GB files. Added includes. 307* Detect when stdout is being redirected and automatically direct output to 308stdout unless a filename is specified. 309* Update lrztar to properly support -S -O and -o, and use new syntax not 310requiring '-o -' for stdout. 311* Update lrzip.conf to support encryption. 312* Do a sanity check to ensure lrzip is not attempting to work on a directory. 313* Typo fixes. 314 315APRIL 2011, version 0.602 Con Kolivas 316* Fixed the symlinks breaking package generation. 317* Made maximum chunk allocable on 32bits 2/3 of a GB again limiting total ram 318used to 2GB, regardless of how much crazy ram people have put on such systems 319since the rest is unusable. 320* Now builds on Cygwin courtesy of Тулебаев Салават. 321* Check for podman on ./configure. 322* Disable md5 generation and checking on Apple till it's fixed. 323* Display full version when -V is invoked. 324* Don't restrict file permissions from unknown source, respecting umode 325instead. 326 327MARCH 2011, version 0.601 Con Kolivas 328* Fixed the symlinks for lrzuntar, lrunzip not being installed. 329* Fixed the pod-based manpages not being installed. 330* Clearly demonstrate that ASM is not supported on 64bits. 331 332MARCH 2011, version 0.600 Con Kolivas 333* Massive rewrite with new file format to accomodate new features. 334* Allocate temporary buffers of safely sized ram that can act as temporary 335storage for performing de/compression to/from STDIN/STDOUT without requiring 336temporary physical files. Files compressed on machines with much larger ram 337being decompressed on smaller ram machines may still require temporary files, 338but as much as possible is done using in-ram decompression, and minimally 339sized temporary files. Information displayed is more verbose and accurate in 340STDIN/STDOUT mode. 341* The temporary buffers created for decompressing to STDOUT are also used 342for decompressing regular files now avoiding multiple write/reads when 343re-constructing the file on decompression. This can dramatically speed up 344the rzip phase of decompression on complicated files with many small matches, 345and always makes testing faster. 346* Compress block headers as well now since we know how many bytes can be used 347to describe the length of the block, decreasing overall file size. 348* Store the rzip chunk size per chunk to make it possible to check total file 349size by summating each rzip chunk size when it's not known till the end (as 350happens when compressing from STDIN). 351* Implement password protected encryption. Import the polarssl code for 352sha512 and aes128 routines. Read password without echoing to screen by 353disabling echo via terminfo. Take the password then multiply hash it 354according to the date it was generated. Inrease the number of hashes according 355to Moore's law so it always takes approximately 1 second per password on the 356most modern hardware when first encrypted. Hash the password against 8 bytes 357of salt which is a combination of the 2 byte encoded loop counter (for how 358many times to hash the password) and 6 random bytes. Take random from 359/dev/urandom if it's available and fall back to random() if not. Encrypt each 360block of compressed data with 8 extra bytes of random salt. Once the headers 361are written, go back and encrypt the headers as well. Then encrypt the md5 362hash value as well. Anything beyond the initial lrzip magic header should 363apppear as random data and no two successive encryptions of the same data with 364the same password should generate the same data. 365* New build system should be more robust and portable. 366* Abstract out functions better into separate files and headers, and remove 367all use of global variables. This will make the generation of an lrzip 368library possible in the future. 369* Prevent testmalloc from coming up with a negative number when determining 370how big a block of memory to allocate by decreasing the number of threads to 371be used and then aborting to a minimum value should it still be too much 372apparent ram. 373* Big Endian fixes. 374* Change lrztar and lrzuntar to not use temporary files. 375* Countless other fixes, documentation and cleanups. 376 377MARCH 2011, version 0.571 Con Kolivas 378* Only retry mmaping if it's a memory error, otherwise it may give spurious 379errors. 380* Check for free space before compression/decompression and abort if there 381is inadequate free space if the -f option is not passed. 382* Fix the wrong check in rzip.c which was rounding down the page size and 383making for one extra small chunk at the end. 384* Check the correct stdout when refusing to pipe to a terminal. 385* Fix windows EOL on lzma.txt. 386* Ignore what stdout is going to when testing from stdin. 387* More verbose summary after we know whether we have stdin/out to more 388accurately reflect the window that will be used. 389* Updated gitignore 390* Unlink temporary files immediately to avoid files lying around. 391* Check free space AFTER reading magic, and not when decompressing to stdout. 392* Don't dump output to stdout when just testing a file. 393* Dump the temporary file generated on emulating stdout on decompression after 394every chunk is decompressed instead of after the whole file is decompressed. 395* Don't attempt to file check decompression when it's not meaningful (test 396only or to stdout). 397* Clean up and simplify the times displayed component. It could potentially 398show thousands of seconds. 399* Remove open coded strerror messages when they're to be printed by fatal(). 400 401MARCH 2011, Michael Blumenkrantz 402* Updated autotools/conf build system. 403 404FEBRUARY 2011, version 0.570 Con Kolivas 405* Change the lzo testing to a bool on/off instead of taking a parameter. 406* Clean up the messy help output. 407* Refuse to read from or write to terminal in stdin/stdout mode. 408* Delete temporary files generated when testing from stdin. 409* Don't call perror on failures that aren't system related. 410* Improve visual output. 411* Significant multi-threading speed-ups - as the spawning of threads is 412partially serialised, use one more thread than CPUs. 413* Don't split up the chunks into multiple threads when no back end 414compression is being used. 415* Take into account the precise amount of overhead that the memory hungry 416back-end compression modes lzma and zpaq use when determining how much ram to 417allocate, and limit total back-end compression ram to 1/3. 418* Deprecate the -M option as it is no longer of any significant utility 419compared to the -U option. 420* Strip symbols by default in compilation and add -W -Wall to CXXFLAGS to be 421consistent with CFLAGS. 422* Update benchmarks. 423* Add verbose and max verbose modes for -i information giving a breakdown of 424each rzip chunk and compressed block. 425* Modify purpose of -T threshold option to mean Disable LZO pre-compression 426 testing. Threshold level test deprecated. 427* Make failure to allocate a buffer in the compress_ functions put out an 428error message always rather than just in max_verbose mode. 429 430FEBRUARY 2011, Peter Hyman 431* Minor updates to man pages, lrzip.conf example file. 432* Update main help screen to include environment settings. 433* Update to respect $TMP environment variable for TMP files. 434* Updated control structure to include tmpdir pointer. 435* Update lrzip.conf parser to respect -U, -H, -T, k, and -c options. 436* Update lrzip.conf example to include new parameters. 437* Reorder main Switch loop in main.c for readability. 438* Have UNLIMITED and control.window be exclusive. UNLIMITED wins. 439* Corrects heuristic computation in rzip.c which would override 440 UNLIMITED if control.window set 441* Show heuristically computed control.window when computed. 442* Remove display compression level from control.window verbose output. 443* Update print_verbose format for Testing for incompressible data in stream.c 444 to omit extra \n. 445 446FEBRUARY 2011, version 0.560 Con Kolivas 447* Rewrote the locking implementation to use pthread_mutexes instead of 448semaphores. This makes multi-threading work on OSX. 449* Re-use a single malloc in unzip_match instead of multiple malloc/frees. 450* Imported the md5 code from coreutils and modified them to be used by lrzip. 451* Implemented md5 hash checking on compression and decompression by using 452buffers to pass to the existing crc checking and re-using them for the md5 453checking. 454* Sped up the hash checking on compression by not reading the whole chunk at 455once to do the hash check, and instead read one paged size chunk. 456* Store the md5 hash value at the end of the archive, maintaining backward 457compatibility with existing 0.5 versions of lrzip, and place a flag in the 458header denoting its existence. 459* Use the header flag for md5's existence to determine what hash check to 460perform on decompression. 461* Add hash value display to maximum verbose output and in information mode. 462* Add -H option to enable hash value displaying without verbose mode. 463* Display what hash checking will be used on decompression in verbose mode and 464in information displayed with -i. 465* Add -c option to perform a hash check on the file generated on decompression, 466comparing it to that stored in the lrzip archive, or that generated on 467decompression on archives that do not have it stored. 468* Delete broken or damaged files generated on interruption or when they fail 469integrity testing by default. 470* Implement the -k option to keep broken or damaged files. 471 472FEBRUARY 2011, version 0.552 Con Kolivas 473* Fix incompressible blocks causing failure with bzip2 and gzip. 474* Freebsd doesn't do mremap, use the fake one. 475* Update copyright notices. 476* Flush writes to disk before allocating ram. 477* Relax memory allocation when no back end compression will be used. 478* Clean up unnecessary semaphore usage in decompression. 479* Fix minor typos. 480* Convert mmaps to malloc in runzip as they may silently fail if not a 481multiple of page size, thereby leading to silent corruption on decompression. 482* Return the correct length by unzip_literal. 483* Use wrappers to serialise threads on OSX and bypass use of unnamed semaphores 484which aren't supported. 485 486DECEMBER 2010, version 0.551 Con Kolivas 487* Fix stdin failing due to getting sizes all wrong. 488* Fix stdin compression values not being shown at end. 489* Fix inappropriate failure when lzma doesn't compress block. 490 491DECEMBER 2010, version 0.550 Con Kolivas 492* Move the threading on compression to higher up in the code, allowing the next 493stream to start using compression threads before the previous stream has 494finished. This speeds up compression on files that take more than one pass to 495compress. 496* Limit the number of threads decompressing stream 0 to just 1 since it's always 497followed by stream 1 chunks, and it may lead to failure to decompress due to 498running out of memory by running too many threads. 499* Default compression level and window size on lzma is set to 7 which is the 500highest it goes. Scale the 9 lrzip levels into 7, thus making the default lzma 501level 5 which uses a lot less memory and is substantially faster at the cost of 502some compression. 503* Rationalise the memory testing now that the default lzma settings use a lot 504less ram by default, and make all systems use no more than 1/3 ram in one mmap. 505This allows larger windows to be used by 32 bit at last without memory 506allocation errors. 507* Revert "Make threads spawn at regular intervals along chunk size thus speeding 508up compression" as it actually slowed it down instead of speeding it up. 509* Cope with compression/decompression threads failing by waiting till the 510previous thread has finished its work, thus serialising the work and using less 511ram, making success more likely. 512* Fix some dodgy callocs which weren't really allocating enough ram. 513* Destroy semaphores used in stream_in on closing the stream. 514* Minor output improvements. 515 516DECEMBER 2010, version 0.544 Con Kolivas 517* Make multiple stream 0 entry decompression more robust by creating separate 518thread groups for stream 0 and stream 1. 519* uclibc fails to return valid values for memory size with sysconf so create 520a workaround. 521* 32 bits has memory allocation errors on large files so be ruthless with 522limiting window sizes and use more sliding mmap. 523* Split up compression streams according to proportion of hash search thus 524evenly distributing work to threads based on raw data length rather than 525rzip pre-processed data. This makes for more CPU usage on compression and 526smoother write out on decompression with small speed ups. 527* Remove unnecessary argument passing. 528* Fix -U causing a floating point error with ultra small sized files. 529* 32 bits still sometimes fails the lzma compression stage due to internal 530memory allocation issues, so use bzip2 as a fallback on blocks that fail to 531compress. 532* Minor cleanups. 533 534NOVEMBER 2010, version 0.543 Con Kolivas 535* Fix for when there are multiple stream 0 entries per chunk, which would 536cause a failure on decompression. 537* Fix lzma windows being set far too small by mistake at all times. 538* Decrease nice level of main process compared to back end threads as it tends 539to be the rate limiting step. 540 541NOVEMBER 2010, version 0.542 Con Kolivas 542* Choose sane defaults for memory usage since linux ludicriously overcommits. 543* Use sliding mmap for any compression windows greater than 2/3 ram. 544* Consolidate and simplify testing of allocatable ram. 545* Minor tweaks to output. 546* Round up the size of the high buffer in sliding mmap to one page. 547* Squeeze a little more out of 32 bit compression windows. 548* Fix sliding mmap from not working on 2nd pass onwards. 549 550NOVEMBER 2010, version 0.541 Con Kolivas 551* Fix wrong number of passes reported. 552* Re-fix the off-by-one that wasn't off-by-one. 553* Limit lzma compression windows to 300MB as per reports of failures with larger 554windows. 555 556NOVEMBER 2010, version 0.540 Con Kolivas 557* Massive rewrite of backend decompression phse, implementing multithreading. 558This is done by taking each stream of data on read in into separate buffers for 559up to as many threads as CPUs. As each thread's data becomes available, feed it 560into runzip once it is requests more of the stream. Provided there are enough 561chunks in the originally compressed data, this provides a massive speedup 562potentially proportional to the number of CPUs. The slower the backend 563compression, the better the speed up (i.e. zpaq is the best sped up). 564* Fix the output of zpaq compress and decompress from trampling on itself and 565racing and consuming a lot of CPU time printing to the console. 566* When limiting cwindow to 6 on 32 bits, ensure that control.window is also set. 567* When testing for the maximum size of testmalloc, the multiple used was out by 568one, so increase it. 569* Minor output tweaks. 570* Build warning fixes. 571* Updated benchmarks. 572 573NOVEMBER 2010, version 0.530 Con Kolivas 574* Massive rewrite of backend compression phase. Now the stream is split up 575into as many chunks as there are CPUs, of at least 10MB in size, that are 576still mallocable. Once the stream has reached a chunk of this size, its buffer 577is handed to a new backend compression thread which works while the rzip stream 578continues processing. This has the effect of parallelising workloads almost 579linearly up to the number of CPUs on the slower compression backends. ZPAQ, 580in particular, is effectively 4x faster on quad core now. Decompression is 581unchanged. 582* Added the -p option to allow the number of processors to be specified to 583override the detected number. 584* Changed the default level back to 7 as 9 wasn't offering significanly more 585compression but was adding time. 586* Increased the size of all the buffers to other backends now as well, since 587each block adds overhead with its header. 588* Numerous alterations to screen output to cope with new threaded compression 589phase. 590* Deprecated the -P option since not setting the file permissions only 591generates a warning now, not a failure. 592* Updated docs and benchmarks. 593 594NOVEMBER 2010, version 0.520 Con Kolivas 595* Distros don't like 3 point version numbering so just repackaged as 0.520. 596 597NOVEMBER 2010, version 0.5.2 Con Kolivas 598* Fixed the Darwin build, again. 599* Fixed cases of extreme ram usage on 32 bit failing by limiting zpaq to 600MB 600windows as well. 601* Check page size if we can instead of assuming it's always 4k. 602* Improve the progress output. 603* Change failure to chmod and failure to set nice level to warnings only. 604* Standardise what's a stderr message and what's output. 605 606NOVEMBER 2010, version 0.5.1 Con Kolivas 607* Fix Darwin build - Darwin doesn't support mremap so introduce a fake wrapper 608for it. 609* Fix the memopen routines, a wrongly implemented wrapper for Darwin equivalents 610was also using the faked versions on all builds. 611* Fix dodgy ordered includes. 612* Clean up excessive use of #ifdefs 613* Huge rewrite of buffer reading in rzip.c. We use a wrapper instead of 614accessing the buffer directly, thus allowing us to have window sizes larger than 615available ram. This is implemented through the use of a "sliding mmap" 616implementation. Sliding mmap uses two mmapped buffers, one large one as 617previously, and one smaller one. When an attempt is made to read beyond the end 618of the large buffer, the small buffer is remapped to the file area that's being 619accessed, while the larger one is remapped as the search progresses along the 620file. While this implementation is potentially much slower than direct mmapping, 621it allows us to implement unlimited sized compression windows. 622* Implement the -U option with unlimited sized compression windows. 623* Rework the selection of compression windows. Instead of trying to guess how 624much ram the machine might be able to access, we try to safely buffer as much 625ram as we can, and then use that to determine the file buffer size. Do not 626choose an arbitrary upper window limit unless -w is specified. 627* Rework the -M option to try to buffer the entire file, reducing the buffer 628size until we succeed. 629* Align buffer sizes to page size. 630* Clean up lots of unneeded variables. 631* Fix lots of minor logic issues to do with window sizes accepted/passed to rzip 632and the compression backends. 633* More error handling. 634* Change -L to affect rzip compression level directly as well as backend 635compression level and use 9 by default now. 636* Fix file size reporting on compressed files generated from stdin. 637* More cleanups of information output and more information. 638* Add chunk percentage to progress update. 639* Reinstated the 2GB buffer limit on 32 bit machines during compression, though 640the -U mode can work around it now. 641* Code micro-optimisations. 642* Use 3 point release numbering in case one minor version has many subversions. 643* Numerous minor cleanups and tidying. 644* Updated docs, manpages, and benchmarks. 645 646NOVEMBER 2010, version 0.5 Con Kolivas 647* Changed offset encoding in rzip stage to use variable byte width offsets 648instead of 64 bits wide. Makes for better compression and slightly faster. 649* Write the byte width into the file before each block. 650* Shrunk match lengths to maximum of 16 bits again as per original rzip as the 651larger offsets did not achieve greater compression and made final size larger. 652* New file format not backwards compatible due to variable byte widths. 653* Rewrote memory initialisation to have a pre-allocation stage to try and 654find the maximum memory usable and defragment ram. 655* Use reduced window size if allocating memory fails at higher size. 656* Change use of malloc to mmap to make it possible to address up to 44 bit 657sized offsets even on 32 bit machines on decompression. Still unable to use 658greater than 2GB windows on 32 bit machines and unsure if this is fixable. 659* Reworked the STDIN code to use an anonymous mmap and read in stdin into this 660to make it possible to compress from STDIN without the need for temporary 661files. As the file size is not known in advance, memory allocation is set to 662large and byte width to equivalent size. 663* Reallocation of ram where possible to minimise risk of running out of memory 664in the middle of a compression phase, and flushing to disk to empty dirty ram 665for the same reason. 666* More robust fatal warnings. 667* Numerous cleanups and tidying of code and addition of comments. 668* Updated documentation to reflect changes. 669 670OCTOBER 2010, version 0.47, Con Kolivas 671* Fix the symlinking problem when DESTDIR is in use reported by a billion 672people. 673 674MAY 2010, version 0.46, Con Kolivas, Ed Avis. 675* Suppress final [OK] message with -q flag EA 676* Handle mkstemp() errors correctly EA 677* Add lrzuntar manpage 678* Update manpages 679 680APRIL 2010, version 0.45, Con Kolivas, Jon Tibble, George Makrydakis 681* Fixes the nasm program test (AC_CHECK_PROG doesn't overwrite a 682variable that is already set so do it manually) JT 683* Fix compiler flags as not all compilers accept -Wall -W (cc on 684Solaris/OpenSolaris) JT 685* Fix lrztar to not try to compress files already with the .lrz extension GM 686* Fix lrztar to decompress files where the pathname is ../* GM 687* Add lrzuntar symlink to call lrztar -d 688 689 690MAR 2010, version 0.45, Con Kolivas, Jari Aalto 691* Fixed reported window size 692* Fixed 32bit windows being attempted to be larger than contiguous amounts 693by taking into account VM kernel/userspace split of 896MB. 694* Minor code cleanups 695* Added lrztar and lrunzip docs 696* Fix minor typos 697* Added distclean and maintainer-clean make targets 698 699 700DEC 2009, version 0.44, Con Kolivas, George Makrydakis 701* Added lrztar wrapper to manage whole directories. 702* Added -i option to provide information about a compressed file. 703* Fixed "nan" showing as Compression speed on very small files. 704* Fixed build for old bz library. 705* Avoid overwriting output file if input doesn't exist. 706* Implement signal handler to delete temporary files. 707 708 709DEC 2009, version 0.43, Con Kolivas, Jukka Laurila 710* Darwin support thanks to Jukka Laurila. 711* Finally added stdin/stdout support due to popular demand. This is done 712by basically using temporary files so is a low performance way of using 713lrzip. 714* Added test function. This just uses a temporary file during decompression. 715* Config files should now accept zpaq options. 716* Minor code style cleanups. 717* Updated benchmarks in docs. 718* Add a warning when attempting to decompress a file from a newer lrzip 719version. 720 721 722NOV 2009, version 0.42, Con Kolivas 723* Changed progress update to show which of 2 chunks are being compressed 724in zpaq. 725* Fixed progress update in ZPAQ to not update with each byte which was 726wasting heaps of CPU time. 727 728 729NOV 2009, version 0.41, Con Kolivas 730* Added zpaq compression backend for extremely good but extremely slow 731compression (incompatible with previous versions if used). 732* Limited chunk size passed to LZMA to 4GB to avoid library overflows. 733* Minor changes to the formatting output 734* Changed lower limit of -T threshhold to 0 to allow disabling it. 735* Added lzo_compresses check into zpaq and bzip2 as well since they're 736slow. 737 738 739NOV 2009, version 0.40, Con Kolivas 740* Massive core code rewrite. 741* All code moved to be 64bit based for compression block addressing and length 742allowing compression windows to be limited by ram only. 743* 64bit userspace should now have no restriction on compression window size, 74432bit is still limited to 2GB windows due to userspace limitations. 745* New file format using the new addressing and data types, incompatible with 746versions prior to 0.40. 747* Support for reading and decompressing older formats. 748* Minor speedups in read/write routines. 749* Countless minor code fixes throughout. 750* Code style cleanups and consistency changes in core code. 751* Configure script improvements. 752 753 754NOV 2009, version 0.31, Con Kolivas 755* Updated to be in sync with lzma SDK 9.07beta. 756* Cleanups and fixes of the configure scripts to use the correct package version 757name. 758* Massive fixes to the memory management code fixing lots of 32bit overflow 759errors. The window size limit is now 2GB on both 32bit and 64bit. While it 760appears to be smaller than the old windows, only 900MB was being used on .30 761even though it claimed to use more. This can cause huge improvements in the 762compression of very large files. 763* The offset when mmap()ing was not being set to a multiple of page size so 764it would fail if the window size was not a multiple of it. 765* Flushing of data to disk between compression windows was implemented to 766minimise disk thrashing of read vs write. 767 768 769NOV 2009, version 0.30, Con Kolivas 770* Numerous bugfixes to try and make the most of 64bit environments with huge 771memory and to barf less on 32bit environments. 772* Executable stacks were fixed. 773* Probably other weird and wonderful bugs have been introduced. 774* -P option to not set permissions on output files allowing you to write to 775braindead filesystems (eg fat32). 776 777 778JAN 2009, version 0.24, Peter Hyman, pete@peterhyman.com 779Happy New Year! 780* Upgrade LZMA SDK to 4.63. Use new C Wrapper. Invalidates 781 LZMA archives created earlier due to new Magic property 782 bytes. 783* New LZMA logic will automatically determine allow LZMA 784 code to determine optimal lc, lp, pb, fb, and dictionary 785 size settings. stream.c will only pass level and thread 786 information. Compress function will return encoded 5 byte 787 data with compression settings. This will be stored in lrz 788 file header. 789* add error messages during LZMA compression. There are some 790 edge cases where LZMA cannot allocate memory. These errors 791 are reported and the user will be advised to use a lower 792 compression window setting. 793* type changes in rzip_fd function for correctness. 794* remove function *Realloc() since it was never used. Cleaned 795 in rzip.h and util.c. 796* apply munmap prior to closing and compressing stream in 797 function rzip_chunk in rzip.c. 798* add realloc function in close_stream_out in stream.c 799 to reclaim some ram and try and allieviate out of memory 800 conditions in LZMA compression. 801* remove file acconfig.h and include DEFINE in configure.in. 802* add lrzip.conf capability. 803* add timer for compression including elapsed time and eta. 804* add compression and decompression MB/s calculation. 805* Updated WHATS-NEW, TODO and created BUGS file. 806* Updated lrzip.1 manpage and created lrzip.conf.5 manpage. 807* Added lrzip.conf.example file in doc directory. 808 809MAR 2008, Con Kolivas, kernel@kolivas.org 810* Numerous changes all over to place restrictions on window 811size to work with 32 bit limitations. 812* Various bugfixes with respect to detecting buffer sizes and 813likelihood of compressibility. 814* Fixed the inappropriate straight copying uncompressed data for 815files larger than 4GB. 816* Re-initiated the 10MB window limits for non-lzma compression. 817I was unable to reproduce any file size savings. 818* Allow compression windows larger than ramsize if people really 819really want them. 820* Decrease thresholds for the test function to a minimum of 5% 821compressibility since the hanging in lzma compression bug has been 822fixed. 823 824JAN 2008, version 0.22, Peter Hyman, pete@peterhyman.com 825* version update 826lzma/LZMALib.cpp 827 Thanks to Lasse Collin for debugging the problem LZMA 828 had with hanging on uncompressable files. 829 Update for control parameters to both compress and 830 decompress functions. 831Makefile.in 832* use of @top_srcdir@ (Lasse Collin). Also moved away 833 more cruft. 834main.c stream.c.rzip.h LZMALib.cpp lzmalib.h 835* addition of three new control structure members. 836 control.lc -- literal context bits 837 control.lp -- literal post state bits 838 control.pb -- post state bits 839 These are needed to ensure decompression will work. 840 These will now be stored along with control.compression_level 841 in the lrz file beginning at offset 0x16 for three bytes. 842 These will be passed to the functions lzma_compresses and 843 lzma_uncompress. Currently, only compression level is 844 needed or used, but the others are stored for possible future 845 use. 846 See magic file for more information. 847stream.c 848* Change to lzo_compresses function that will reject a chunk 849 without testing it if the size of the chunk is greater 850 than the compression window * threshold. This is to avoid 851 a low probability that lzma would still be passed a chunk 852 that contains uncompressible data or barely compressible 853 data. If after rzip hashing the chunk size is still close 854 to the window size, there is hardly anything worth 855 compressing. While there is no reason lzma cannot get the 856 chunk, this will save a lot of time. 857magic.headers.txt 858* updated file to show new layout that includes lzma 859 parameters. 860README-NOT-BACKWARD-COMPATIBLE 861* added warning about using lrzip-0.22 with earlier versions. 862WHATS-NEW 863* highlight of new features. 864 865DEC 2007, version 0.21. Peter Hyman, pete@peterhyman.com 866* version update. 867* Modified to use Assembler routines from lzma SDK for CRC 868 computation when hashing streams in rzip.c and runzip.c. 869 Added files 7zCrcT8.c and 7zCrcT8u.s to lzma tree. 870 Cleaned up source tree. Moved unused files out of the way. 871 Moved non-core docs to doc directory 872configure.in 873* correct AC_INIT to set program variables. 874* modified to add check for nasm assembler. 875* modified syntax of test for errno in error.h to use 876 echo $ECHO_N/$ECHO_C instead of $ac_n/$ac_c which 877 was incorrect. 878Makefile.in, lzma/Makefile 879* modified to add compile instructions for 7zCrcT8.c 880 and 7zCrcT8U.s and Assembler. Cleaned up to remove 881 targets that don't exist or sources that don't 882 exist. 883 Modified to properly set directories. Added doc install. 884 Add link command to symlink lrunzip to lrzip. 885*main.c 886 Add CrcGenerateTable() function to init CRC tables. 887 This is needed for all crc routines including those 888 in MatchFinderMT. 889rzip.c and runzip.c 890* Updated source to change call to crc32_buffer to call 891 CrcUpdate in the assembler code. Changed parameter order 892 to conform. 893stream.c 894* Removed 10MB limit on streams for bzip, gzip, and lzo. 895 This, to improve effeciency of long range analysis. For 896 some files, this could improve results. 897Current-Benchmarks.txt 898* Added file to keep benchmarks current to version. 899 (probably need to update README too). 900README.Assembler 901* Explain how to remove default compile of Assembler 902 modules. 903config.sub config.guess 904* added files for system detection. 905 906DEC 2007, version 0.20. Peter Hyman, pete@peterhyman.com 907 908* Updated to LZMA SDK 4.57. 909* Updated to p7zip POSIX version. (www.p7zip.org) 910* Added multi-threading support (up to 2x speed with LZMA). 911* Edited LZMADecompress.cpp for backward compatibility 912 with decompress function. Needed SetPropertiesRaw function. 913* Repopulated source tree for distribution. 914* Updated Makefile.in to reflect new source files. 915 Updated to include command to link lrunzip to lrzip because 916 lrzip will test if lrunzip was used on command line. 917* Updated Makefile.in for new compile time and linking options. 918* Updated LZMALibs.cpp to include new property members for 919 LZMAEncoders as well as changed default dictionaries to 920 level+16. This would make the default compression level 921 of 7 translate to a dictionary number of 23. 922* Added output to show Nice Level when verbose mode set 923 Initial add of support for zlib which seems to give quite 924 excellent performance. 925* configure.in added AC_CHECK for libz and libm. 926 Added AC_PROG_LN_S for Makefile symlink section. 927* lrzip.1 updated man page for -g option 928* main.c added option test for gzip 929 Added sysconf(_SC_NPROCESSORS_CONF) for CPU detection 930 for threading. 931 Updated verbose output to show whether or not 932 Threading will be used. 933 Added Timer for each file compressed. 934* rzip.h added flags for GZIP compression. 935 Added control member for threads. Arg passed to 936 lzma_conpress. 937* stream.c update to accomodate gzip compress and decompress 938 functions. Cleaned up file by rearranging functions into 939 groups. 940 Removed include of lzmalib.h since it was causing a 941 compile time warning with zlib.h. Prototyped functions 942 manually. 943 Cleanup output from lzo_compresses function so that 944 unnecessary linefeeds are eliminated. 945 lzma_compress function call now uses threads as argument. 946* Added README.benchmarks file to explain a method of 947 comparing results between different methods. 948* LZMALib.cpp, lzmalib.h. Adjust function lzma_compress 949 prototype and function to include new argument threads. 950 This parameter is now placed in properties. 951* lzma/Makefile. Updated to reflect new API library. 952 Updated to include Threading option. 953 954DEC 2007, version 0.19. Con Kolivas. 955* Added nice support, defaulting to nice 19. 956 957DEC 2007, version 0.19. Peter Hyman, pete@peterhyman.com 958 959* Major goal was to stop LZMA from hanging on some files. 960 Accomplished this with a threhold setting that is used by 961 the lzo_compresses function to better analyze chunk data. 962 Threshold makes it less likely that uncompressible data 963 will be passed to the LZMA compressor. 964 965main.c 966* Added Threshold option 1-10 to control LZMA compression attempt. 967 Default value=2. This means that anything over 10% compression 968 as reported by lzo_compresses will return a true value to 969 the LZMA compression function. 970* Added verbosity option and more verbosity option (-v[v]). 971* Added -O option to specify output directory. 972* Updated compress_file and decompress_file functions to handle. 973 output directories and better handle multi files and filename 974 extensions. Optimized some string handling routines. 975 Improved flexibility in determining location of output files 976 when using -O. Added fflush(stdout) to improve printf reliability. 977* decompress_file will accept any filename and will automatically 978 append .lrz if not present. Won't automatically fail. 979* Added logic to protect against conflicting options such as 980 -q and -v, -o and -O. 981* Added printout to screen of options selected. Will display 982 only when -v or -vv used. 983* Adjusted several printf statements to avoid compiler 984 warnings (use %ll for long long int types). 985 986runzip.c 987* Added decompression progress indicator. 988 Will show percent decompressed along with bytes decompressed 989 and total to be decompressed. Will show if -q option NOT used. 990 991rzip.h 992* Version incremented to 0.19. 993* Added flag DEFINESs for verbosity and more verbosity. 994* Updated control struct to include output directory and 995 threshold value. Removed verbosity member. 996 997rzip.c 998* Minor changes to handle display when verbosity set. Changed 999 number format in some printf statements to properly handle 1000 unsigned data. 1001 1002stream.c 1003* major overhaul of lzo_compresses function to use a threshold 1004 value when testing a data chunk to see if it is suitable for 1005 LZMA compression. Optimized test loop to improve performance 1006 and reduce number of passes. Improved output reporting depending 1007 on verbosity setting. 1008* Added print controls for verbosity option. 1009* Corrected if statements that tested for error condition of 1010 some lzo functions that only return a true value regardless. 1011 1012lrzip.1 1013* updated man page to show new options and explain -T threshold. 1014 1015README 1016* updated README to explain -T threshold option. 1017 1018README.lzo_compresses.test.txt 1019* Added this file to help explain the theory behind the rewrite 1020 of the lzo_compresses function and how to use the -T option. 1021 1022TODO 1023* wish list and future enhancements. 1024 1025ChangeLog 1026* added file. 1027