7 * Licensing change: The core components are now under the
8 BSD Zero Clause License (0BSD). In XZ Utils 5.4.6 and older
9 and 5.5.1alpha these components are in the public domain and
10 obviously remain so; the change affects the new releases only.
12 0BSD is an extremely permissive license which doesn't require
13 retaining or reproducing copyright or license notices when
14 distributing the code, thus in practice there is extremely
15 little difference to public domain.
19 - Significant speed optimizations to the LZMA decoder were
20 made. There are now three variants that can be chosen at
23 * Basic C version: This is a few percent faster than
24 5.4.x due to some new optimizations.
26 * Branchless C: This is currently the default on platforms
27 for which there is no assembly code. This should be a few
28 percent faster than the basic C version.
30 * x86-64 inline assembly. This works with GCC and Clang.
32 The default choice can currently be overridden by setting
33 LZMA_RANGE_DECODER_CONFIG in CPPFLAGS: 0 means the basic
34 version and 3 means that branchless C version.
36 - Optimized the CRC32 calculation on ARM64 platforms using the
37 CRC32 instructions. The instructions are optional in ARMv8.0
38 and are required in ARMv8.1 and later. Runtime detection for
39 the instruction is used on GNU/Linux, FreeBSD, Windows, and
40 macOS. If the compiler flags indicate unconditional CRC32
41 instruction support (+crc) then the generic version is not
44 * Added lz4 support to xzdiff/xzcmp and xzgrep.
46 * Man pages of xzdiff/xzcmp, xzgrep, and xzmore were rewritten
47 to simplify licensing of the man page translations.
51 - Updated Chinese (simplified), German, Korean, Polish,
52 Romanian, Spanish, Swedish, and Ukrainian translations.
54 - Updated German, Korean, Romanian, and Ukrainian man page
57 * Small improvements to the tests.
59 * Added doc/examples/11_file_info.c. It was added to the Git
60 repository in 2017 but forgotten to be added into distribution
63 * Removed doc/examples_old. These were from 2012.
65 * Removed the macos/build.sh script. It had not been updated
69 5.5.1alpha (2024-01-26)
71 * Added a new filter for RISC-V binaries. The filter can be used
72 for 32-bit and 64-bit binaries with either little or big
73 endianness. In liblzma, the Filter ID is LZMA_FILTER_RISCV (0x0B)
74 and the xz option is --riscv. liblzma filter string syntax
75 recognizes this filter as "riscv".
79 - Added lzma_mt_block_size() to recommend a Block size for
80 multithreaded encoding
82 - Added CLMUL-based CRC32 on x86-64 and E2K with runtime
83 processor detection. Similar to CRC64, on 32-bit x86 it
84 isn't available unless --disable-assembler is used.
86 - Implemented GNU indirect function (IFUNC) as a runtime
87 function dispatching method for CRC32 and CRC64 fast
88 implementations on x86. Only GNU/Linux (glibc) and FreeBSD
89 builds will use IFUNC, unless --enable-ifunc is specified to
92 - Added definitions of mask values like
93 LZMA_INDEX_CHECK_MASK_CRC32 to <lzma/index.h>.
95 - The XZ logo is now included in the Doxygen generated
96 documentation. It is licensed under Creative Commons
97 Attribution-ShareAlike 4.0.
101 - Multithreaded mode is now the default. This improves
102 compression speed and creates .xz files that can be
103 decompressed multithreaded at the cost of increased memory
104 usage and slightly worse compression ratio.
106 - Added new command line option --filters to set the filter
107 chain using liblzma filter string syntax.
109 - Added new command line options --filters1 ... --filters9 to
110 set additional filter chains using liblzma filter string
111 syntax. The --block-list option now allows specifying filter
112 chains that were set using these new options.
114 - Added support for Linux Landlock as a sandboxing method.
116 - xzdec now supports pledge(2), Capsicum, and Linux Landlock as
119 - Progress indicator time stats remain accurate after pausing
122 - Ported xz and xzdec to Windows MSVC. Visual Studio 2015 or
127 - Supports pledge(2), Capsicum, and Linux Landlock sandboxing
130 - Replacement functions for getopt_long() are used on platforms
133 * Enabled unaligned access by default on PowerPC64LE and on RISC-V
134 targets that define __riscv_misaligned_fast.
138 - Added two new fuzz targets to OSS-Fuzz.
140 - Implemented Continuous Integration (CI) testing using
143 * Changed quoting style from `...' to '...' in all messages,
144 scripts, and documentation.
146 * Added basic Codespell support to help catch typo errors.
151 * Fixed a bug involving internal function pointers in liblzma not
152 being initialized to NULL. The bug can only be triggered if
153 lzma_filters_update() is called on a LZMA1 encoder, so it does
154 not affect xz or any application known to us that uses liblzma.
158 - Fixed a regression introduced in 5.4.2 that caused encoding
159 in the raw format to unnecessarily fail if --suffix was not
160 used. For instance, the following command no longer reports
161 that --suffix must be used:
163 echo foo | xz --format=raw --lzma2 | wc -c
165 - Fixed an issue on MinGW-w64 builds that prevented reading
166 from or writing to non-terminal character devices like NUL.
175 - Use __attribute__((__no_sanitize_address__)) to avoid address
176 sanitization with CRC64 CLMUL. It uses 16-byte-aligned reads
177 which can extend past the bounds of the input buffer and
178 inherently trigger address sanitization errors. This isn't
181 - Fixed an assertion failure that could be triggered by a large
182 unpadded_size argument. It was verified that there was no
183 other bug than the assertion failure.
185 - Fixed a bug that prevented building with Windows Vista
186 threading when __attribute__((__constructor__)) is not
189 * xz now properly handles special files such as "con" or "nul" on
190 Windows. Before this fix, the following wrote "foo" to the
191 console and deleted the input file "con_xz":
193 echo foo | xz > con_xz
194 xz --suffix=_xz --decompress con_xz
198 - Allow builds with Windows win95 threading and small mode when
199 __attribute__((__constructor__)) is supported.
201 - Added a new line to liblzma.pc for MSYS2 (Windows):
203 Cflags.private: -DLZMA_API_STATIC
205 When compiling code that will link against static liblzma,
206 the LZMA_API_STATIC macro needs to be defined on Windows.
208 - CMake specific changes:
210 * Fixed a bug that allowed CLOCK_MONOTONIC to be used even
211 if the check for it failed.
213 * Fixed a bug where configuring CMake multiple times
214 resulted in HAVE_CLOCK_GETTIME and HAVE_CLOCK_MONOTONIC
217 * Fixed the build with MinGW-w64-based Clang/LLVM 17.
218 llvm-windres now has more accurate GNU windres emulation
219 so the GNU windres workaround from 5.4.1 is needed with
220 llvm-windres version 17 too.
222 * The import library on Windows is now properly named
223 "liblzma.dll.a" instead of "libliblzma.dll.a"
225 * Fixed a bug causing the Ninja Generator to fail on
226 UNIX-like systems. This bug was introduced in 5.4.0.
228 * Added a new option to disable CLMUL CRC64.
230 * A module-definition (.def) file is now created when
231 building liblzma.dll with MinGW-w64.
233 * The pkg-config liblzma.pc file is now installed on all
234 builds except when using MSVC on Windows.
236 * Added large file support by default for platforms that
237 need it to handle files larger than 2 GiB. This includes
238 MinGW-w64, even 64-bit builds.
240 * Small fixes and improvements to the tests.
242 * Updated translations: Chinese (simplified) and Esperanto.
247 * liblzma and xzdec can now build against WASI SDK when threading
248 support is disabled. xz and tests don't build yet.
252 - Fixed a bug preventing other projects from including liblzma
253 multiple times using find_package().
255 - Don't create broken symlinks in Cygwin and MSYS2 unless
256 supported by the environment. This prevented building for the
257 default MSYS2 environment. The problem was introduced in
262 - Small improvements to man pages.
264 - Small improvements and typo fixes for liblzma API
269 - Added a new section to INSTALL to describe basic test usage
270 and address recent questions about building the tests when
273 - Small fixes and improvements to the tests.
277 - Fixed a mistake that caused one of the error messages to not
278 be translated. This only affected versions 5.4.2 and 5.4.3.
280 - Updated the Chinese (simplified), Croatian, Esperanto, German,
281 Korean, Polish, Romanian, Spanish, Swedish, Ukrainian, and
282 Vietnamese translations.
284 - Updated the German, Korean, Romanian, and Ukrainian man page
290 * All fixes from 5.2.12
292 * Features in the CMake build can now be disabled as CMake cache
293 variables, similar to the Autotools build.
295 * Minor update to the Croatian translation.
300 * All fixes from 5.2.11 that were not included in 5.4.1.
302 * If xz is built with support for the Capsicum sandbox but running
303 in an environment that doesn't support Capsicum, xz now runs
304 normally without sandboxing instead of exiting with an error.
308 - Documentation was updated to improve the style, consistency,
309 and completeness of the liblzma API headers.
311 - The Doxygen-generated HTML documentation for the liblzma API
312 header files is now included in the source release and is
313 installed as part of "make install". All JavaScript is
314 removed to simplify license compliance and to reduce the
317 - Fixed a minor bug in lzma_str_from_filters() that produced
318 too many filters in the output string instead of reporting
319 an error if the input array had more than four filters. This
320 bug did not affect xz.
324 - autogen.sh now invokes the doxygen tool via the new wrapper
325 script doxygen/update-doxygen, unless the command line option
326 --no-doxygen is used.
328 - Added microlzma_encoder.c and microlzma_decoder.c to the
329 VS project files for Windows and to the CMake build. These
330 should have been included in 5.3.2alpha.
334 - Added a test to the CMake build that was forgotten in the
337 - Added and refactored a few tests.
341 - Updated the Brazilian Portuguese translation.
343 - Added Brazilian Portuguese man page translation.
350 - Fixed the return value of lzma_microlzma_encoder() if the
351 LZMA options lc/lp/pb are invalid. Invalid lc/lp/pb options
352 made the function return LZMA_STREAM_END without encoding
353 anything instead of returning LZMA_OPTIONS_ERROR.
355 - Windows / Visual Studio: Workaround a possible compiler bug
356 when targeting 32-bit x86 and compiling the CLMUL version of
357 the CRC64 code. The CLMUL code isn't enabled by the Windows
358 project files but it is in the CMake-based builds.
362 - Windows-specific CMake changes:
364 * Don't try to enable CLMUL CRC64 code if _mm_set_epi64x()
365 isn't available. This fixes CMake-based build with Visual
368 * Created a workaround for a build failure with windres
369 from GNU binutils. It is used only when the C compiler
370 is GCC (not Clang). The workaround is incompatible
371 with llvm-windres, resulting in "XZx20Utils" instead
372 of "XZ Utils" in the resource file, but without the
373 workaround llvm-windres works correctly. See the
374 comment in CMakeLists.txt for details.
376 * Included the resource files in the xz and xzdec build
377 rules. Building the command line tools is still
378 experimental but possible with MinGW-w64.
380 - Visual Studio: Added stream_decoder_mt.c to the project
381 files. Now the threaded decompressor lzma_stream_decoder_mt()
382 gets built. CMake-based build wasn't affected.
384 - Updated windows/INSTALL-MSVC.txt to mention that CMake-based
385 build is now the preferred method with Visual Studio. The
386 project files will probably be removed after 5.4.x releases.
388 - Changes to #defines in config.h:
390 * HAVE_DECL_CLOCK_MONOTONIC was replaced by
391 HAVE_CLOCK_MONOTONIC. The old macro was always defined
392 in configure-generated config.h to either 0 or 1. The
393 new macro is defined (to 1) only if the declaration of
394 CLOCK_MONOTONIC is available. This matches the way most
395 other config.h macros work and makes things simpler with
398 * HAVE_DECL_PROGRAM_INVOCATION_NAME was replaced by
399 HAVE_PROGRAM_INVOCATION_NAME for the same reason.
403 - Fixed test script compatibility with ancient /bin/sh
404 versions. Now the five test_compress_* tests should
405 no longer fail on Solaris 10.
407 - Added and refactored a few tests.
411 - Updated the Catalan and Esperanto translations.
413 - Added Korean and Ukrainian man page translations.
418 This bumps the minor version of liblzma because new features were
419 added. The API and ABI are still backward compatible with liblzma
424 * All fixes from 5.2.10.
426 * The ARM64 filter is now stable. The xz option is now --arm64.
427 Decompression requires XZ Utils 5.4.0. In the future the ARM64
428 filter will be supported by XZ for Java, XZ Embedded (including
429 the version in Linux), LZMA SDK, and 7-Zip.
433 - Updated Catalan, Croatian, German, Romanian, and Turkish
436 - Updated German man page translations.
438 - Added Romanian man page translations.
440 Summary of new features added in the 5.3.x development releases:
444 - Added threaded .xz decompressor lzma_stream_decoder_mt().
445 It can use multiple threads with .xz files that have multiple
446 Blocks with size information in Block Headers. The threaded
447 encoder in xz has always created such files.
449 Single-threaded encoder cannot store the size information in
450 Block Headers even if one used LZMA_FULL_FLUSH to create
451 multiple Blocks, so this threaded decoder cannot use multiple
452 threads with such files.
454 If there are multiple Streams (concatenated .xz files), one
455 Stream will be decompressed completely before starting the
458 - A new decoder flag LZMA_FAIL_FAST was added. It makes the
459 threaded decompressor report errors soon instead of first
460 flushing all pending data before the error location.
463 * LZMA_FILTER_ARM64 is for ARM64 binaries.
464 * LZMA_FILTER_LZMA1EXT is for raw LZMA1 streams that don't
465 necessarily use the end marker.
467 - Added lzma_str_to_filters(), lzma_str_from_filters(), and
468 lzma_str_list_filters() to convert a preset or a filter chain
469 string to a lzma_filter[] and vice versa. These should make
470 it easier to write applications that allow users to specify
471 custom compression options.
473 - Added lzma_filters_free() which can be convenient for freeing
474 the filter options in a filter chain (an array of lzma_filter
477 - lzma_file_info_decoder() to makes it a little easier to get
478 the Index field from .xz files. This helps in getting the
479 uncompressed file size but an easy-to-use random access
480 API is still missing which has existed in XZ for Java for
483 - Added lzma_microlzma_encoder() and lzma_microlzma_decoder().
484 It is used by erofs-utils and may be used by others too.
486 The MicroLZMA format is a raw LZMA stream (without end marker)
487 whose first byte (always 0x00) has been replaced with
488 bitwise-negation of the LZMA properties (lc/lp/pb). It was
489 created for use in EROFS but may be used in other contexts
490 as well where it is important to avoid wasting bytes for
491 stream headers or footers. The format is also supported by
492 XZ Embedded (the XZ Embedded version in Linux got MicroLZMA
493 support in Linux 5.16).
495 The MicroLZMA encoder API in liblzma can compress into a
496 fixed-sized output buffer so that as much data is compressed
497 as can be fit into the buffer while still creating a valid
498 MicroLZMA stream. This is needed for EROFS.
500 - Added lzma_lzip_decoder() to decompress the .lz (lzip) file
501 format version 0 and the original unextended version 1 files.
502 Also lzma_auto_decoder() supports .lz files.
504 - lzma_filters_update() can now be used with the multi-threaded
505 encoder (lzma_stream_encoder_mt()) to change the filter chain
506 after LZMA_FULL_BARRIER or LZMA_FULL_FLUSH.
508 - In lzma_options_lzma, allow nice_len = 2 and 3 with the match
509 finders that require at least 3 or 4. Now it is internally
510 rounded up if needed.
512 - CLMUL-based CRC64 on x86-64 and E2K with runtime processor
513 detection. On 32-bit x86 it currently isn't available unless
514 --disable-assembler is used which can make the non-CLMUL
515 CRC64 slower; this might be fixed in the future.
517 - Building with --disable-threads --enable-small
518 is now thread-safe if the compiler supports
519 __attribute__((__constructor__)).
523 - Using -T0 (--threads=0) will now use multi-threaded encoder
524 even on a single-core system. This is to ensure that output
525 from the same xz binary is identical on both single-core and
528 - --threads=+1 or -T+1 is now a way to put xz into
529 multi-threaded mode while using only one worker thread.
530 The + is ignored if the number is not 1.
532 - A default soft memory usage limit is now used for compression
533 when -T0 is used and no explicit limit has been specified.
534 This soft limit is used to restrict the number of threads
535 but if the limit is exceeded with even one thread then xz
536 will continue with one thread using the multi-threaded
537 encoder and this limit is ignored. If the number of threads
538 is specified manually then no default limit will be used;
539 this affects only -T0.
541 This change helps on systems that have very many cores and
542 using all of them for xz makes no sense. Previously xz -T0
543 could run out of memory on such systems because it attempted
544 to reserve memory for too many threads.
546 This also helps with 32-bit builds which don't have a large
547 amount of address space that would be required for many
548 threads. The default soft limit for -T0 is at most 1400 MiB
549 on all 32-bit platforms.
551 - Previously a low value in --memlimit-compress wouldn't cause
552 xz to switch from multi-threaded mode to single-threaded mode
553 if the limit cannot otherwise be met; xz failed instead. Now
554 xz can switch to single-threaded mode and then, if needed,
555 scale down the LZMA2 dictionary size too just like it already
556 did when it was started in single-threaded mode.
558 - The option --no-adjust no longer prevents xz from scaling down
559 the number of threads as that doesn't affect the compressed
560 output (only performance). Now --no-adjust only prevents
561 adjustments that affect compressed output, that is, with
562 --no-adjust xz won't switch from multi-threaded mode to
563 single-threaded mode and won't scale down the LZMA2
566 - Added a new option --memlimit-mt-decompress=LIMIT. This is
567 used to limit the number of decompressor threads (possibly
568 falling back to single-threaded mode) but it will never make
569 xz refuse to decompress a file. This has a system-specific
570 default value because without any limit xz could end up
571 allocating memory for the whole compressed input file, the
572 whole uncompressed output file, multiple thread-specific
573 decompressor instances and so on. Basically xz could
574 attempt to use an insane amount of memory even with fairly
575 common files. The system-specific default value is currently
576 the same as the one used for compression with -T0.
578 The new option works together with the existing option
579 --memlimit-decompress=LIMIT. The old option sets a hard limit
580 that must not be exceeded (xz will refuse to decompress)
581 while the new option only restricts the number of threads.
582 If the limit set with --memlimit-mt-decompress is greater
583 than the limit set with --memlimit-compress, then the latter
584 value is used also for --memlimit-mt-decompress.
586 - Added new information to the output of xz --info-memory and
587 new fields to the output of xz --robot --info-memory.
589 - In --lzma2=nice=NUMBER allow 2 and 3 with all match finders
590 now that liblzma handles it.
592 - Don't mention endianness for ARM and ARM-Thumb filters in
593 --long-help. The filters only work for little endian
594 instruction encoding but modern ARM processors using
595 big endian data access still use little endian
596 instruction encoding. So the help text was misleading.
597 In contrast, the PowerPC filter is only for big endian
598 32/64-bit PowerPC code. Little endian PowerPC would need
601 - Added decompression support for the .lz (lzip) file format
602 version 0 and the original unextended version 1. It is
603 autodetected by default. See also the option --format on
606 - Sandboxing enabled by default:
608 * pledge(2) (OpenBSD)
610 * Scripts now support the .lz format using xz.
612 * A few new tests were added.
614 * The liblzma-specific tests are now supported in CMake-based
615 builds too ("make test").
618 5.3.5beta (2022-12-01)
620 * All fixes from 5.2.9.
624 - Added new LZMA_FILTER_LZMA1EXT for raw encoder and decoder to
625 handle raw LZMA1 streams that don't have end of payload marker
626 (EOPM) alias end of stream (EOS) marker. It can be used in
627 filter chains, for example, with the x86 BCJ filter.
629 - Added lzma_str_to_filters(), lzma_str_from_filters(), and
630 lzma_str_list_filters() to make it easier for applications
631 to get custom compression options from a user and convert
632 it to an array of lzma_filter structures.
634 - Added lzma_filters_free().
636 - lzma_filters_update() can now be used with the multi-threaded
637 encoder (lzma_stream_encoder_mt()) to change the filter chain
638 after LZMA_FULL_BARRIER or LZMA_FULL_FLUSH.
640 - In lzma_options_lzma, allow nice_len = 2 and 3 with the match
641 finders that require at least 3 or 4. Now it is internally
642 rounded up if needed.
644 - ARM64 filter was modified. It is still experimental.
646 - Fixed LTO build with Clang if -fgnuc-version=10 or similar
647 was used to make Clang look like GCC >= 10. Now it uses
648 __has_attribute(__symver__) which should be reliable.
652 - --threads=+1 or -T+1 is now a way to put xz into multi-threaded
653 mode while using only one worker thread.
655 - In --lzma2=nice=NUMBER allow 2 and 3 with all match finders
656 now that liblzma handles it.
658 * Updated translations: Chinese (simplified), Korean, and Turkish.
661 5.3.4alpha (2022-11-15)
663 * All fixes from 5.2.7 and 5.2.8.
667 - Minor improvements to the threaded decoder.
669 - Added CRC64 implementation that uses SSSE3, SSE4.1, and CLMUL
670 instructions on 32/64-bit x86 and E2K. On 32-bit x86 it's
671 not enabled unless --disable-assembler is used but then
672 the non-CLMUL code might be slower. Processor support is
673 detected at runtime so this is built by default on x86-64
674 and E2K. On these platforms, if compiler flags indicate
675 unconditional CLMUL support (-msse4.1 -mpclmul) then the
676 generic version is not built, making liblzma 8-9 KiB smaller
677 compared to having both versions included.
679 With extremely compressible files this can make decompression
680 up to twice as fast but with typical files 5 % improvement
681 is a more realistic expectation.
683 The CLMUL version is slower than the generic version with
684 tiny inputs (especially at 1-8 bytes per call, but up to
685 16 bytes). In normal use in xz this doesn't matter at all.
687 - Added an experimental ARM64 filter. This is *not* the final
688 version! Files created with this experimental version won't
689 be supported in the future versions! The filter design is
690 a compromise where improving one use case makes some other
693 - Added decompression support for the .lz (lzip) file format
694 version 0 and the original unextended version 1. See the
695 API docs of lzma_lzip_decoder() for details. Also
696 lzma_auto_decoder() supports .lz files.
698 - Building with --disable-threads --enable-small
699 is now thread-safe if the compiler supports
700 __attribute__((__constructor__))
704 - Added support for OpenBSD's pledge(2) as a sandboxing method.
706 - Don't mention endianness for ARM and ARM-Thumb filters in
707 --long-help. The filters only work for little endian
708 instruction encoding but modern ARM processors using
709 big endian data access still use little endian
710 instruction encoding. So the help text was misleading.
711 In contrast, the PowerPC filter is only for big endian
712 32/64-bit PowerPC code. Little endian PowerPC would need
715 - Added --experimental-arm64. This will be renamed once the
716 filter is finished. Files created with this experimental
717 filter will not be supported in the future!
719 - Added new fields to the output of xz --robot --info-memory.
721 - Added decompression support for the .lz (lzip) file format
722 version 0 and the original unextended version 1. It is
723 autodetected by default. See also the option --format on
726 * Scripts now support the .lz format using xz.
730 - New #defines in config.h: HAVE_ENCODER_ARM64,
731 HAVE_DECODER_ARM64, HAVE_LZIP_DECODER, HAVE_CPUID_H,
732 HAVE_FUNC_ATTRIBUTE_CONSTRUCTOR, HAVE_USABLE_CLMUL
734 - New configure options: --disable-clmul-crc,
735 --disable-microlzma, --disable-lzip-decoder, and
736 'pledge' is now an option in --enable-sandbox (but
737 it's autodetected by default anyway).
739 - INSTALL was updated to document the new configure options.
741 - PACKAGERS now lists also --disable-microlzma and
742 --disable-lzip-decoder as configure options that must
743 not be used in builds for non-embedded use.
747 - Fix some of the tests so that they skip instead of fail if
748 certain features have been disabled with configure options.
749 It's still not perfect.
751 - Other improvements to tests.
753 * Updated translations: Croatian, Finnish, Hungarian, Polish,
754 Romanian, Spanish, Swedish, and Ukrainian.
757 5.3.3alpha (2022-08-22)
759 * All fixes from 5.2.6.
763 - Fixed 32-bit build.
765 - Added threaded .xz decompressor lzma_stream_decoder_mt().
766 It can use multiple threads with .xz files that have multiple
767 Blocks with size information in Block Headers. The threaded
768 encoder in xz has always created such files.
770 Single-threaded encoder cannot store the size information in
771 Block Headers even if one used LZMA_FULL_FLUSH to create
772 multiple Blocks, so this threaded decoder cannot use multiple
773 threads with such files.
775 If there are multiple Streams (concatenated .xz files), one
776 Stream will be decompressed completely before starting the
779 - A new decoder flag LZMA_FAIL_FAST was added. It makes the
780 threaded decompressor report errors soon instead of first
781 flushing all pending data before the error location.
785 - Using -T0 (--threads=0) will now use multi-threaded encoder
786 even on a single-core system. This is to ensure that output
787 from the same xz binary is identical on both single-core and
790 - A default soft memory usage limit is now used for compression
791 when -T0 is used and no explicit limit has been specified.
792 This soft limit is used to restrict the number of threads
793 but if the limit is exceeded with even one thread then xz
794 will continue with one thread using the multi-threaded
795 encoder and this limit is ignored. If the number of threads
796 is specified manually then no default limit will be used;
797 this affects only -T0.
799 This change helps on systems that have very many cores and
800 using all of them for xz makes no sense. Previously xz -T0
801 could run out of memory on such systems because it attempted
802 to reserve memory for too many threads.
804 This also helps with 32-bit builds which don't have a large
805 amount of address space that would be required for many
806 threads. The default limit is 1400 MiB on all 32-bit
809 Now xz -T0 should just work. It might use too few threads
810 in some cases but at least it shouldn't easily run out of
811 memory. It's possible that this will be tweaked before 5.4.0.
813 - Changes to --memlimit-compress and --no-adjust:
815 In single-threaded mode, --memlimit-compress can make xz
816 scale down the LZMA2 dictionary size to meet the memory usage
817 limit. This obviously affects the compressed output. However,
818 if xz was in threaded mode, --memlimit-compress could make xz
819 reduce the number of threads but it wouldn't make xz switch
820 from multi-threaded mode to single-threaded mode or scale
821 down the LZMA2 dictionary size. This seemed illogical.
823 Now --memlimit-compress can make xz switch to single-threaded
824 mode if one thread in multi-threaded mode uses too much
825 memory. If memory usage is still too high, then the LZMA2
826 dictionary size can be scaled down too.
828 The option --no-adjust was also changed so that it no longer
829 prevents xz from scaling down the number of threads as that
830 doesn't affect compressed output (only performance). After
831 this commit --no-adjust only prevents adjustments that affect
832 compressed output, that is, with --no-adjust xz won't switch
833 from multithreaded mode to single-threaded mode and won't
834 scale down the LZMA2 dictionary size.
836 - Added a new option --memlimit-mt-decompress=LIMIT. This is
837 used to limit the number of decompressor threads (possibly
838 falling back to single-threaded mode) but it will never make
839 xz refuse to decompress a file. This has a system-specific
840 default value because without any limit xz could end up
841 allocating memory for the whole compressed input file, the
842 whole uncompressed output file, multiple thread-specific
843 decompressor instances and so on. Basically xz could
844 attempt to use an insane amount of memory even with fairly
847 The new option works together with the existing option
848 --memlimit-decompress=LIMIT. The old option sets a hard limit
849 that must not be exceeded (xz will refuse to decompress)
850 while the new option only restricts the number of threads.
851 If the limit set with --memlimit-mt-decompress is greater
852 than the limit set with --memlimit-compress, then the latter
853 value is used also for --memlimit-mt-decompress.
857 - Added a few more tests.
859 - Added tests/code_coverage.sh to create a code coverage report
864 - Automake's parallel test harness is now used to make tests
867 - Added the CMake files to the distribution tarball. These were
868 supposed to be in 5.2.5 already.
870 - Added liblzma tests to the CMake build.
872 - Windows: Fix building of liblzma.dll with the included
873 Visual Studio project files.
876 5.3.2alpha (2021-10-28)
878 This release was made on short notice so that recent erofs-utils can
879 be built with LZMA support without needing a snapshot from xz.git.
880 Thus many pending things were not included, not even updated
881 translations (which would need to be updated for the new --list
884 * All fixes from 5.2.5.
888 - When copying metadata from the source file to the destination
889 file, don't try to set the group (GID) if it is already set
890 correctly. This avoids a failure on OpenBSD (and possibly on
891 a few other OSes) where files may get created so that their
892 group doesn't belong to the user, and fchown(2) can fail even
893 if it needs to do nothing.
895 - The --keep option now accepts symlinks, hardlinks, and
896 setuid, setgid, and sticky files. Previously this required
899 - Split the long strings used in --list and --info-memory modes
900 to make them much easier for translators.
902 - If built with sandbox support and enabling the sandbox fails,
903 xz will now immediately exit with exit status of 1. Previously
904 it would only display a warning if -vv was used.
906 - Cap --memlimit-compress to 2000 MiB on MIPS32 because on
907 MIPS32 userspace processes are limited to 2 GiB of address
912 - Added lzma_microlzma_encoder() and lzma_microlzma_decoder().
913 The API is in lzma/container.h.
915 The MicroLZMA format is a raw LZMA stream (without end marker)
916 whose first byte (always 0x00) has been replaced with
917 bitwise-negation of the LZMA properties (lc/lp/pb). It was
918 created for use in EROFS but may be used in other contexts
919 as well where it is important to avoid wasting bytes for
920 stream headers or footers. The format is also supported by
923 The MicroLZMA encoder API in liblzma can compress into a
924 fixed-sized output buffer so that as much data is compressed
925 as can be fit into the buffer while still creating a valid
926 MicroLZMA stream. This is needed for EROFS.
928 - Added fuzzing support.
930 - Support Intel Control-flow Enforcement Technology (CET) in
931 32-bit x86 assembly files.
933 - Visual Studio: Use non-standard _MSVC_LANG to detect C++
934 standard version in the lzma.h API header. It's used to
935 detect when "noexcept" can be used.
939 - Fix exit status of xzdiff/xzcmp. Exit status could be 2 when
940 the correct value is 1.
942 - Fix exit status of xzgrep.
944 - Detect corrupt .bz2 files in xzgrep.
946 - Add zstd support to xzgrep and xzdiff/xzcmp.
948 - Fix less(1) version detection in xzless. It failed if the
949 version number from "less -V" contained a dot.
951 * Fix typos and technical issues in man pages.
955 - Windows: Fix building of resource files when config.h isn't
956 used. CMake + Visual Studio can now build liblzma.dll.
958 - Various fixes to the CMake support. It might still need a few
959 more fixes even for liblzma-only builds.
962 5.3.1alpha (2018-04-29)
964 * All fixes from 5.2.4.
966 * Add lzma_file_info_decoder() into liblzma and use it in xz to
967 implement the --list feature.
969 * Capsicum sandbox support is enabled by default where available
975 * Fixed a build system bug that prevented building liblzma as a
976 shared library when configured with --disable-threads. This bug
977 affected releases 5.2.6 to 5.2.11 and 5.4.0 to 5.4.2.
979 * Include <intrin.h> for Windows intrinsic functions where they are
980 needed. This fixed a bug that prevented building liblzma using
983 * Minor update to the Croatian translation. The small change
984 applies to a string in both 5.2 and 5.4 branches.
989 * Removed all possible cases of null pointer + 0. It is undefined
990 behavior in C99 and C17. This was detected by a sanitizer and had
991 not caused any known issues.
995 - Added a workaround for building with GCC on MicroBlaze Linux.
996 GCC 12 on MicroBlaze doesn't support the __symver__ attribute
997 even though __has_attribute(__symver__) returns true. The
998 build is now done without the extra RHEL/CentOS 7 symbols
999 that were added in XZ Utils 5.2.7. The workaround only
1000 applies to the Autotools build (not CMake).
1002 - CMake: Ensure that the C compiler language is set to C99 or
1005 - CMake changes from XZ Utils 5.4.1:
1007 * Added a workaround for a build failure with
1008 windres from GNU binutils.
1010 * Included the Windows resource files in the xz
1011 and xzdec build rules.
1016 * xz: Don't modify argv[] when parsing the --memlimit* and
1017 --block-list command line options. This fixes confusing
1018 arguments in process listing (like "ps auxf").
1020 * GNU/Linux only: Use __has_attribute(__symver__) to detect if
1021 that attribute is supported. This fixes build on Mandriva where
1022 Clang is patched to define __GNUC__ to 11 by default (instead
1023 of 4 as used by Clang upstream).
1030 - Fixed an infinite loop in LZMA encoder initialization
1031 if dict_size >= 2 GiB. (The encoder only supports up
1034 - Fixed two cases of invalid free() that can happen if
1035 a tiny allocation fails in encoder re-initialization
1036 or in lzma_filters_update(). These bugs had some
1037 similarities with the bug fixed in 5.2.7.
1039 - Fixed lzma_block_encoder() not allowing the use of
1040 LZMA_SYNC_FLUSH with lzma_code() even though it was
1041 documented to be supported. The sync-flush code in
1042 the Block encoder was already used internally via
1043 lzma_stream_encoder(), so this was just a missing flag
1044 in the lzma_block_encoder() API function.
1046 - GNU/Linux only: Don't put symbol versions into static
1047 liblzma as it breaks things in some cases (and even if
1048 it didn't break anything, symbol versions in static
1049 libraries are useless anyway). The downside of the fix
1050 is that if the configure options --with-pic or --without-pic
1051 are used then it's not possible to build both shared and
1052 static liblzma at the same time on GNU/Linux anymore;
1053 with those options --disable-static or --disable-shared
1056 * New email address for bug reports is <xz@tukaani.org> which
1057 forwards messages to Lasse Collin and Jia Tan.
1064 - If xz cannot remove an input file when it should, this
1065 is now treated as a warning (exit status 2) instead of
1066 an error (exit status 1). This matches GNU gzip and it
1067 is more logical as at that point the output file has
1068 already been successfully closed.
1070 - Fix handling of .xz files with an unsupported check type.
1071 Previously such printed a warning message but then xz
1072 behaved as if an error had occurred (didn't decompress,
1073 exit status 1). Now a warning is printed, decompression
1074 is done anyway, and exit status is 2. This used to work
1075 slightly before 5.0.0. In practice this bug matters only
1076 if xz has been built with some check types disabled. As
1077 instructed in PACKAGERS, such builds should be done in
1078 special situations only.
1080 - Fix "xz -dc --single-stream tests/files/good-0-empty.xz"
1081 which failed with "Internal error (bug)". That is,
1082 --single-stream was broken if the first .xz stream in
1083 the input file didn't contain any uncompressed data.
1085 - Fix displaying file sizes in the progress indicator when
1086 working in passthru mode and there are multiple input files.
1087 Just like "gzip -cdf", "xz -cdf" works like "cat" when the
1088 input file isn't a supported compressed file format. In
1089 this case the file size counters weren't reset between
1090 files so with multiple input files the progress indicator
1091 displayed an incorrect (too large) value.
1095 - API docs in lzma/container.h:
1096 * Update the list of decoder flags in the decoder
1098 * Explain LZMA_CONCATENATED behavior with .lzma files
1099 in lzma_auto_decoder() docs.
1101 - OpenBSD: Use HW_NCPUONLINE to detect the number of
1102 available hardware threads in lzma_physmem().
1104 - Fix use of wrong macro to detect x86 SSE2 support.
1105 __SSE2_MATH__ was used with GCC/Clang but the correct
1106 one is __SSE2__. The first one means that SSE2 is used
1107 for floating point math which is irrelevant here.
1108 The affected SSE2 code isn't used on x86-64 so this affects
1109 only 32-bit x86 builds that use -msse2 without -mfpmath=sse
1110 (there is no runtime detection for SSE2). It improves LZMA
1111 compression speed (not decompression).
1113 - Fix the build with Intel C compiler 2021 (ICC, not ICX)
1114 on Linux. It defines __GNUC__ to 10 but doesn't support
1115 the __symver__ attribute introduced in GCC 10.
1117 * Scripts: Ignore warnings from xz by using --quiet --no-warn.
1118 This is needed if the input .xz files use an unsupported
1123 - Updated Croatian and Turkish translations.
1125 - One new translations wasn't included because it needed
1126 technical fixes. It will be in upcoming 5.4.0. No new
1127 translations will be added to the 5.2.x branch anymore.
1129 - Renamed the French man page translation file from
1130 fr_FR.po to fr.po and thus also its install directory
1131 (like /usr/share/man/fr_FR -> .../fr).
1133 - Man page translations for upcoming 5.4.0 are now handled
1134 in the Translation Project.
1136 * Update doc/faq.txt a little so it's less out-of-date.
1143 - Made lzma_filters_copy() to never modify the destination
1144 array if an error occurs. lzma_stream_encoder() and
1145 lzma_stream_encoder_mt() already assumed this. Before this
1146 change, if a tiny memory allocation in lzma_filters_copy()
1147 failed it would lead to a crash (invalid free() or invalid
1148 memory reads) in the cleanup paths of these two encoder
1149 initialization functions.
1151 - Added missing integer overflow check to lzma_index_append().
1152 This affects xz --list and other applications that decode
1153 the Index field from .xz files using lzma_index_decoder().
1154 Normal decompression of .xz files doesn't call this code
1155 and thus most applications using liblzma aren't affected
1158 - Single-threaded .xz decoder (lzma_stream_decoder()): If
1159 lzma_code() returns LZMA_MEMLIMIT_ERROR it is now possible
1160 to use lzma_memlimit_set() to increase the limit and continue
1161 decoding. This was supposed to work from the beginning
1162 but there was a bug. With other decoders (.lzma or
1163 threaded .xz decoder) this already worked correctly.
1165 - Fixed accumulation of integrity check type statistics in
1166 lzma_index_cat(). This bug made lzma_index_checks() return
1167 only the type of the integrity check of the last Stream
1168 when multiple lzma_indexes were concatenated. Most
1169 applications don't use these APIs but in xz it made
1170 xz --list not list all check types from concatenated .xz
1171 files. In xz --list --verbose only the per-file "Check:"
1172 lines were affected and in xz --robot --list only the "file"
1175 - Added ABI compatibility with executables that were linked
1176 against liblzma in RHEL/CentOS 7 or other liblzma builds
1177 that had copied the problematic patch from RHEL/CentOS 7
1178 (xz-5.2.2-compat-libs.patch). For the details, see the
1179 comment at the top of src/liblzma/validate_map.sh.
1181 WARNING: This uses __symver__ attribute with GCC >= 10.
1182 In other cases the traditional __asm__(".symver ...")
1183 is used. Using link-time optimization (LTO, -flto) with
1184 GCC versions older than 10 can silently result in
1185 broken liblzma.so.5 (incorrect symbol versions)! If you
1186 want to use -flto with GCC, you must use GCC >= 10.
1187 LTO with Clang seems to work even with the traditional
1188 __asm__(".symver ...") method.
1190 * xzgrep: Fixed compatibility with old shells that break if
1191 comments inside command substitutions have apostrophes (').
1192 This problem was introduced in 5.2.6.
1196 - New #define in config.h: HAVE_SYMBOL_VERSIONS_LINUX
1198 - Windows: Fixed liblzma.dll build with Visual Studio project
1199 files. It broke in 5.2.6 due to a change that was made to
1200 improve CMake support.
1202 - Windows: Building liblzma with UNICODE defined should now
1205 - CMake files are now actually included in the release tarball.
1206 They should have been in 5.2.5 already.
1208 - Minor CMake fixes and improvements.
1210 * Added a new translation: Turkish
1217 - The --keep option now accepts symlinks, hardlinks, and
1218 setuid, setgid, and sticky files. Previously this required
1221 - When copying metadata from the source file to the destination
1222 file, don't try to set the group (GID) if it is already set
1223 correctly. This avoids a failure on OpenBSD (and possibly on
1224 a few other OSes) where files may get created so that their
1225 group doesn't belong to the user, and fchown(2) can fail even
1226 if it needs to do nothing.
1228 - Cap --memlimit-compress to 2000 MiB instead of 4020 MiB on
1229 MIPS32 because on MIPS32 userspace processes are limited
1230 to 2 GiB of address space.
1234 - Fixed a missing error-check in the threaded encoder. If a
1235 small memory allocation fails, a .xz file with an invalid
1236 Index field would be created. Decompressing such a file would
1237 produce the correct output but result in an error at the end.
1238 Thus this is a "mild" data corruption bug. Note that while
1239 a failed memory allocation can trigger the bug, it cannot
1240 cause invalid memory access.
1242 - The decoder for .lzma files now supports files that have
1243 uncompressed size stored in the header and still use the
1244 end of payload marker (end of stream marker) at the end
1245 of the LZMA stream. Such files are rare but, according to
1246 the documentation in LZMA SDK, they are valid.
1247 doc/lzma-file-format.txt was updated too.
1249 - Improved 32-bit x86 assembly files:
1250 * Support Intel Control-flow Enforcement Technology (CET)
1251 * Use non-executable stack on FreeBSD.
1253 - Visual Studio: Use non-standard _MSVC_LANG to detect C++
1254 standard version in the lzma.h API header. It's used to
1255 detect when "noexcept" can be used.
1259 - Fixed arbitrary command injection via a malicious filename
1260 (CVE-2022-1271, ZDI-CAN-16587). A standalone patch for
1261 this was released to the public on 2022-04-07. A slight
1262 robustness improvement has been made since then and, if
1263 using GNU or *BSD grep, a new faster method is now used
1264 that doesn't use the old sed-based construct at all. This
1265 also fixes bad output with GNU grep >= 3.5 (2020-09-27)
1266 when xzgrepping binary files.
1268 This vulnerability was discovered by:
1269 cleemy desu wayo working with Trend Micro Zero Day Initiative
1271 - Fixed detection of corrupt .bz2 files.
1273 - Improved error handling to fix exit status in some situations
1274 and to fix handling of signals: in some situations a signal
1275 didn't make xzgrep exit when it clearly should have. It's
1276 possible that the signal handling still isn't quite perfect
1277 but hopefully it's good enough.
1279 - Documented exit statuses on the man page.
1281 - xzegrep and xzfgrep now use "grep -E" and "grep -F" instead
1282 of the deprecated egrep and fgrep commands.
1284 - Fixed parsing of the options -E, -F, -G, -P, and -X. The
1285 problem occurred when multiple options were specified in
1286 a single argument, for example,
1288 echo foo | xzgrep -Fe foo
1290 treated foo as a filename because -Fe wasn't correctly
1293 - Added zstd support.
1297 - Fixed wrong exit status. Exit status could be 2 when the
1300 - Documented on the man page that exit status of 2 is used
1301 for decompression errors.
1303 - Added zstd support.
1307 - Fix less(1) version detection. It failed if the version number
1308 from "less -V" contained a dot.
1312 - Added new translations: Catalan, Croatian, Esperanto,
1313 Korean, Portuguese, Romanian, Serbian, Spanish, Swedish,
1316 - Updated the Brazilian Portuguese translation.
1318 - Added French man page translation. This and the existing
1319 German translation aren't complete anymore because the
1320 English man pages got a few updates and the translators
1321 weren't reached so that they could update their work.
1325 - Windows: Fix building of resource files when config.h isn't
1326 used. CMake + Visual Studio can now build liblzma.dll.
1328 - Various fixes to the CMake support. Building static or shared
1329 liblzma should work fine in most cases. In contrast, building
1330 the command line tools with CMake is still clearly incomplete
1331 and experimental and should be used for testing only.
1338 - Fixed several C99/C11 conformance bugs. Now the code is clean
1339 under gcc/clang -fsanitize=undefined. Some of these changes
1340 might have a negative effect on performance with old GCC
1341 versions or compilers other than GCC and Clang. The configure
1342 option --enable-unsafe-type-punning can be used to (mostly)
1343 restore the old behavior but it shouldn't normally be used.
1345 - Improved API documentation of lzma_properties_decode().
1347 - Added a very minor encoder speed optimization.
1351 - Fixed a crash in "xz -dcfv not_an_xz_file". All four options
1352 were required to trigger it. The crash occurred in the
1353 progress indicator code when xz was in passthru mode where
1354 xz works like "cat".
1356 - Fixed an integer overflow with 32-bit off_t. It could happen
1357 when decompressing a file that has a long run of zero bytes
1358 which xz would try to write as a sparse file. Since the build
1359 system enables large file support by default, off_t is
1360 normally 64-bit even on 32-bit systems.
1362 - Fixes for --flush-timeout:
1363 * Fix semi-busy-waiting.
1364 * Avoid unneeded flushes when no new input has arrived
1365 since the previous flush was completed.
1367 - Added a special case for 32-bit xz: If --memlimit-compress is
1368 used to specify a limit that exceeds 4020 MiB, the limit will
1369 be set to 4020 MiB. The values "0" and "max" aren't affected
1370 by this and neither is decompression. This hack can be
1371 helpful when a 32-bit xz has access to 4 GiB address space
1372 but the specified memlimit exceeds 4 GiB. This can happen
1373 e.g. with some scripts.
1375 - Capsicum sandbox is now enabled by default where available
1376 (FreeBSD >= 10). The sandbox debug messages (xz -vv) were
1377 removed since they seemed to be more annoying than useful.
1379 - DOS build now requires DJGPP 2.05 instead of 2.04beta.
1380 A workaround for a locale problem with DJGPP 2.05 was added.
1382 * xzgrep and other scripts:
1384 - Added a configure option --enable-path-for-scripts=PREFIX.
1385 It is disabled by default except on Solaris where the default
1386 is /usr/xpg4/bin. See INSTALL for details.
1388 - Added a workaround for a POSIX shell detection problem on
1393 - Added preliminary build instructions for z/OS. See INSTALL
1396 - Experimental CMake support was added. It should work to build
1397 static liblzma on a few operating systems. It may or may not
1398 work to build shared liblzma. On some platforms it can build
1399 xz and xzdec too but those are only for testing. See the
1400 comment in the beginning of CMakeLists.txt for details.
1402 - Visual Studio project files were updated.
1403 WindowsTargetPlatformVersion was removed from VS2017 files
1404 and set to "10.0" in the added VS2019 files. In the future
1405 the VS project files will be removed when CMake support is
1408 - New #defines in config.h: HAVE___BUILTIN_ASSUME_ALIGNED,
1409 HAVE___BUILTIN_BSWAPXX, and TUKLIB_USE_UNSAFE_TYPE_PUNNING.
1411 - autogen.sh has a new optional dependency on po4a and a new
1412 option --no-po4a to skip that step. This matters only if one
1413 wants to remake the build files. po4a is used to update the
1414 translated man pages but as long as the man pages haven't
1415 been modified, there's nothing to update and one can use
1416 --no-po4a to avoid the dependency on po4a.
1420 - XZ Utils translations are now handled by the Translation
1421 Project: https://translationproject.org/domain/xz.html
1423 - All man pages are now included in German too.
1425 - New xz translations: Brazilian Portuguese, Finnish,
1426 Hungarian, Chinese (simplified), Chinese (traditional),
1427 and Danish (partial translation)
1429 - Updated xz translations: French, German, Italian, and Polish
1431 - Unfortunately a few new xz translations weren't included due
1432 to technical problems like too long lines in --help output or
1433 misaligned column headings in tables. In the future, many of
1434 these strings will be split and e.g. the table column
1435 alignment will be handled in software. This should make the
1436 strings easier to translate.
1443 - Allow 0 as memory usage limit instead of returning
1444 LZMA_PROG_ERROR. Now 0 is treated as if 1 byte was specified,
1445 which effectively is the same as 0.
1447 - Use "noexcept" keyword instead of "throw()" in the public
1448 headers when a C++11 (or newer standard) compiler is used.
1450 - Added a portability fix for recent Intel C Compilers.
1452 - Microsoft Visual Studio build files have been moved under
1453 windows/vs2013 and windows/vs2017.
1457 - Fix "xz --list --robot missing_or_bad_file.xz" which would
1458 try to print an uninitialized string and thus produce garbage
1459 output. Since the exit status is non-zero, most uses of such
1460 a command won't try to interpret the garbage output.
1462 - "xz --list foo.xz" could print "Internal error (bug)" in a
1463 corner case where a specific memory usage limit had been set.
1470 - Always close a file before trying to delete it to avoid
1471 problems on some operating system and file system combinations.
1473 - Fixed copying of file timestamps on Windows.
1475 - Added experimental (disabled by default) sandbox support using
1476 Capsicum (FreeBSD >= 10). See --enable-sandbox in INSTALL.
1478 * C99/C11 conformance fixes to liblzma. The issues affected at least
1479 some builds using link-time optimizations.
1481 * Fixed bugs in the rarely-used function lzma_index_dup().
1483 * Use of external SHA-256 code is now disabled by default.
1484 It can still be enabled by passing --enable-external-sha256
1485 to configure. The reasons to disable it by default (see INSTALL
1488 - Some OS-specific SHA-256 implementations conflict with
1489 OpenSSL and cause problems in programs that link against both
1490 liblzma and libcrypto. At least FreeBSD 10 and MINIX 3.3.0
1493 - The internal SHA-256 is faster than the SHA-256 code in
1494 some operating systems.
1496 * Changed CPU core count detection to use sched_getaffinity() on
1497 GNU/Linux and GNU/kFreeBSD.
1499 * Fixes to the build-system and xz to make xz buildable even when
1500 encoders, decoders, or threading have been disabled from libilzma
1501 using configure options. These fixes added two new #defines to
1502 config.h: HAVE_ENCODERS and HAVE_DECODERS.
1507 * Fixed bugs in QNX-specific code.
1509 * Omitted the use of pipe2() even if it is available to avoid
1510 portability issues with some old Linux and glibc combinations.
1512 * Updated German translation.
1514 * Added project files to build static and shared liblzma (not the
1515 whole XZ Utils) with Visual Studio 2013 update 2 or later.
1517 * Documented that threaded decompression hasn't been implemented
1518 yet. A 5.2.0 NEWS entry describing multi-threading support had
1519 incorrectly said "decompression" when it should have said
1525 * Fixed a compression-ratio regression in fast mode of LZMA1 and
1526 LZMA2. The bug is present in 5.1.4beta and 5.2.0 releases.
1528 * Fixed a portability problem in xz that affected at least OpenBSD.
1530 * Fixed xzdiff to be compatible with FreeBSD's mktemp which differs
1531 from most other mktemp implementations.
1533 * Changed CPU core count detection to use cpuset_getaffinity() on
1541 * All fixes from 5.0.8
1543 * liblzma: Fixed lzma_stream_encoder_mt_memusage() when a preset
1546 * xzdiff: If mktemp isn't installed, mkdir will be used as
1547 a fallback to create a temporary directory. Installing mktemp
1548 is still recommended.
1550 * Updated French, German, Italian, Polish, and Vietnamese
1553 Summary of fixes and new features added in the 5.1.x development
1558 - Added support for multi-threaded compression. See the
1559 lzma_mt structure, lzma_stream_encoder_mt(), and
1560 lzma_stream_encoder_mt_memusage() in <lzma/container.h>,
1561 lzma_get_progress() in <lzma/base.h>, and lzma_cputhreads()
1562 in <lzma/hardware.h> for details.
1564 - Made the uses of lzma_allocator const correct.
1566 - Added lzma_block_uncomp_encode() to create uncompressed
1567 .xz Blocks using LZMA2 uncompressed chunks.
1569 - Added support for LZMA_IGNORE_CHECK.
1571 - A few speed optimizations were made.
1573 - Added support for symbol versioning. It is enabled by default
1574 on GNU/Linux, other GNU-based systems, and FreeBSD.
1576 - liblzma (not the whole XZ Utils) should now be buildable
1577 with MSVC 2013 update 2 or later using windows/config.h.
1581 - Fixed a race condition in the signal handling. It was
1582 possible that e.g. the first SIGINT didn't make xz exit
1583 if reading or writing blocked and one had bad luck. The fix
1584 is non-trivial, so as of writing it is unknown if it will be
1585 backported to the v5.0 branch.
1587 - Multi-threaded compression can be enabled with the
1588 --threads (-T) option.
1589 [Fixed: This originally said "decompression".]
1591 - New command line options in xz: --single-stream,
1592 --block-size=SIZE, --block-list=SIZES,
1593 --flush-timeout=TIMEOUT, and --ignore-check.
1595 - xz -lvv now shows the minimum xz version that is required to
1596 decompress the file. Currently it is 5.0.0 for all supported
1597 .xz files except files with empty LZMA2 streams require 5.0.2.
1599 * xzdiff and xzgrep now support .lzo files if lzop is installed.
1600 The .tzo suffix is also recognized as a shorthand for .tar.lzo.
1603 5.1.4beta (2014-09-14)
1605 * All fixes from 5.0.6
1607 * liblzma: Fixed the use of presets in threaded encoder
1610 * xz --block-list and --block-size can now be used together
1611 in single-threaded mode. Previously the combination only
1612 worked in multi-threaded mode.
1614 * Added support for LZMA_IGNORE_CHECK to liblzma and made it
1615 available in xz as --ignore-check.
1617 * liblzma speed optimizations:
1619 - Initialization of a new LZMA1 or LZMA2 encoder has been
1620 optimized. (The speed of reinitializing an already-allocated
1621 encoder isn't affected.) This helps when compressing many
1622 small buffers with lzma_stream_buffer_encode() and other
1623 similar situations where an already-allocated encoder state
1624 isn't reused. This speed-up is visible in xz too if one
1625 compresses many small files one at a time instead running xz
1626 once and giving all files as command-line arguments.
1628 - Buffer comparisons are now much faster when unaligned access
1629 is allowed (configured with --enable-unaligned-access). This
1630 speeds up encoding significantly. There is arch-specific code
1631 for 32-bit and 64-bit x86 (32-bit needs SSE2 for the best
1632 results and there's no run-time CPU detection for now).
1633 For other archs there is only generic code which probably
1634 isn't as optimal as arch-specific solutions could be.
1636 - A few speed optimizations were made to the SHA-256 code.
1637 (Note that the builtin SHA-256 code isn't used on all
1640 * liblzma can now be built with MSVC 2013 update 2 or later
1641 using windows/config.h.
1643 * Vietnamese translation was added.
1646 5.1.3alpha (2013-10-26)
1648 * All fixes from 5.0.5
1652 - Fixed a deadlock in the threaded encoder.
1654 - Made the uses of lzma_allocator const correct.
1656 - Added lzma_block_uncomp_encode() to create uncompressed
1657 .xz Blocks using LZMA2 uncompressed chunks.
1659 - Added support for native threads on Windows and the ability
1660 to detect the number of CPU cores.
1664 - Fixed a race condition in the signal handling. It was
1665 possible that e.g. the first SIGINT didn't make xz exit
1666 if reading or writing blocked and one had bad luck. The fix
1667 is non-trivial, so as of writing it is unknown if it will be
1668 backported to the v5.0 branch.
1670 - Made the progress indicator work correctly in threaded mode.
1672 - Threaded encoder now works together with --block-list=SIZES.
1674 - Added preliminary support for --flush-timeout=TIMEOUT.
1675 It can be useful for (somewhat) real-time streaming. For
1676 now the decompression side has to be done with something
1677 else than the xz tool due to how xz does buffering, but this
1681 5.1.2alpha (2012-07-04)
1683 * All fixes from 5.0.3 and 5.0.4
1687 - Fixed a deadlock and an invalid free() in the threaded encoder.
1689 - Added support for symbol versioning. It is enabled by default
1690 on GNU/Linux, other GNU-based systems, and FreeBSD.
1692 - Use SHA-256 implementation from the operating system if one is
1693 available in libc, libmd, or libutil. liblzma won't use e.g.
1694 OpenSSL or libgcrypt to avoid introducing new dependencies.
1696 - Fixed liblzma.pc for static linking.
1698 - Fixed a few portability bugs.
1700 * xz --decompress --single-stream now fixes the input position after
1701 successful decompression. Now the following works:
1703 echo foo | xz > foo.xz
1704 echo bar | xz >> foo.xz
1705 ( xz -dc --single-stream ; xz -dc --single-stream ) < foo.xz
1707 Note that it doesn't work if the input is not seekable
1708 or if there is Stream Padding between the concatenated
1711 * xz -lvv now shows the minimum xz version that is required to
1712 decompress the file. Currently it is 5.0.0 for all supported .xz
1713 files except files with empty LZMA2 streams require 5.0.2.
1715 * Added an *incomplete* implementation of --block-list=SIZES to xz.
1716 It only works correctly in single-threaded mode and when
1717 --block-size isn't used at the same time. --block-list allows
1718 specifying the sizes of Blocks which can be useful e.g. when
1719 creating files for random-access reading.
1722 5.1.1alpha (2011-04-12)
1724 * All fixes from 5.0.2
1726 * liblzma fixes that will also be included in 5.0.3:
1728 - A memory leak was fixed.
1730 - lzma_stream_buffer_encode() no longer creates an empty .xz
1731 Block if encoding an empty buffer. Such an empty Block with
1732 LZMA2 data would trigger a bug in 5.0.1 and older (see the
1733 first bullet point in 5.0.2 notes). When releasing 5.0.2,
1734 I thought that no encoder creates this kind of files but
1737 - Validate function arguments better in a few functions. Most
1738 importantly, specifying an unsupported integrity check to
1739 lzma_stream_buffer_encode() no longer creates a corrupt .xz
1740 file. Probably no application tries to do that, so this
1741 shouldn't be a big problem in practice.
1743 - Document that lzma_block_buffer_encode(),
1744 lzma_easy_buffer_encode(), lzma_stream_encoder(), and
1745 lzma_stream_buffer_encode() may return LZMA_UNSUPPORTED_CHECK.
1747 - The return values of the _memusage() functions are now
1750 * Support for multithreaded compression was added using the simplest
1751 method, which splits the input data into blocks and compresses
1752 them independently. Other methods will be added in the future.
1753 The current method has room for improvement, e.g. it is possible
1754 to reduce the memory usage.
1756 * Added the options --single-stream and --block-size=SIZE to xz.
1758 * xzdiff and xzgrep now support .lzo files if lzop is installed.
1759 The .tzo suffix is also recognized as a shorthand for .tar.lzo.
1761 * Support for short 8.3 filenames under DOS was added to xz. It is
1762 experimental and may change before it gets into a stable release.
1767 * Fixed an old bug in xzgrep that affected OpenBSD and probably
1768 a few other operating systems too.
1770 * Updated French and German translations.
1772 * Added support for detecting the amount of RAM on AmigaOS/AROS.
1774 * Minor build system updates.
1779 * Fix regressions introduced in 5.0.6:
1781 - Fix building with non-GNU make.
1783 - Fix invalid Libs.private value in liblzma.pc which broke
1784 static linking against liblzma if the linker flags were
1785 taken from pkg-config.
1790 * xzgrep now exits with status 0 if at least one file matched.
1792 * A few minor portability and build system fixes
1797 * lzmadec and liblzma's lzma_alone_decoder(): Support decompressing
1798 .lzma files that have less common settings in the headers
1799 (dictionary size other than 2^n or 2^n + 2^(n-1), or uncompressed
1800 size greater than 256 GiB). The limitations existed to avoid false
1801 positives when detecting .lzma files. The lc + lp <= 4 limitation
1802 still remains since liblzma's LZMA decoder has that limitation.
1804 NOTE: xz's .lzma support or liblzma's lzma_auto_decoder() are NOT
1805 affected by this change. They still consider uncommon .lzma headers
1806 as not being in the .lzma format. Changing this would give way too
1807 many false positives.
1811 - Interaction of preset and custom filter chain options was
1812 made less illogical. This affects only certain less typical
1813 uses cases so few people are expected to notice this change.
1815 Now when a custom filter chain option (e.g. --lzma2) is
1816 specified, all preset options (-0 ... -9, -e) earlier are on
1817 the command line are completely forgotten. Similarly, when
1818 a preset option is specified, all custom filter chain options
1819 earlier on the command line are completely forgotten.
1821 Example 1: "xz -9 --lzma2=preset=5 -e" is equivalent to "xz -e"
1822 which is equivalent to "xz -6e". Earlier -e didn't put xz back
1823 into preset mode and thus the example command was equivalent
1824 to "xz --lzma2=preset=5".
1826 Example 2: "xz -9e --lzma2=preset=5 -7" is equivalent to
1827 "xz -7". Earlier a custom filter chain option didn't make
1828 xz forget the -e option so the example was equivalent to
1831 - Fixes and improvements to error handling.
1833 - Various fixes to the man page.
1835 * xzless: Fixed to work with "less" versions 448 and later.
1837 * xzgrep: Made -h an alias for --no-filename.
1839 * Include the previously missing debug/translation.bash which can
1840 be useful for translators.
1842 * Include a build script for Mac OS X. This has been in the Git
1843 repository since 2010 but due to a mistake in Makefile.am the
1844 script hasn't been included in a release tarball before.
1851 - Fix lzma_index_init(). It could crash if memory allocation
1854 - Fix the possibility of an incorrect LZMA_BUF_ERROR when a BCJ
1855 filter is used and the application only provides exactly as
1856 much output space as is the uncompressed size of the file.
1858 - Fix a bug in doc/examples_old/xz_pipe_decompress.c. It didn't
1859 check if the last call to lzma_code() really returned
1860 LZMA_STREAM_END, which made the program think that truncated
1863 - New example programs in doc/examples (old programs are now in
1864 doc/examples_old). These have more comments and more detailed
1867 * Fix "xz -lvv foo.xz". It could crash on some corrupted files.
1869 * Fix output of "xz --robot -lv" and "xz --robot -lvv" which
1870 incorrectly printed the filename also in the "foo (x/x)" format.
1872 * Fix exit status of "xzdiff foo.xz bar.xz".
1874 * Fix exit status of "xzgrep foo binary_file".
1876 * Fix portability to EBCDIC systems.
1878 * Fix a configure issue on AIX with the XL C compiler. See INSTALL
1881 * Update French, German, Italian, and Polish translations.
1888 - A memory leak was fixed.
1890 - lzma_stream_buffer_encode() no longer creates an empty .xz
1891 Block if encoding an empty buffer. Such an empty Block with
1892 LZMA2 data would trigger a bug in 5.0.1 and older (see the
1893 first bullet point in 5.0.2 notes). When releasing 5.0.2,
1894 I thought that no encoder creates this kind of files but
1897 - Validate function arguments better in a few functions. Most
1898 importantly, specifying an unsupported integrity check to
1899 lzma_stream_buffer_encode() no longer creates a corrupt .xz
1900 file. Probably no application tries to do that, so this
1901 shouldn't be a big problem in practice.
1903 - Document that lzma_block_buffer_encode(),
1904 lzma_easy_buffer_encode(), lzma_stream_encoder(), and
1905 lzma_stream_buffer_encode() may return LZMA_UNSUPPORTED_CHECK.
1907 - The return values of the _memusage() functions are now
1910 * Fix command name detection in xzgrep. xzegrep and xzfgrep now
1911 correctly use egrep and fgrep instead of grep.
1913 * French translation was added.
1918 * LZMA2 decompressor now correctly accepts LZMA2 streams with no
1919 uncompressed data. Previously it considered them corrupt. The
1920 bug can affect applications that use raw LZMA2 streams. It is
1921 very unlikely to affect .xz files because no compressor creates
1922 .xz files with empty LZMA2 streams. (Empty .xz files are a
1923 different thing than empty LZMA2 streams.)
1925 * "xz --suffix=.foo filename.foo" now refuses to compress the
1926 file due to it already having the suffix .foo. It was already
1927 documented on the man page, but the code lacked the test.
1929 * "xzgrep -l foo bar.xz" works now.
1931 * Polish translation was added.
1936 * xz --force now (de)compresses files that have setuid, setgid,
1937 or sticky bit set and files that have multiple hard links.
1938 The man page had it documented this way already, but the code
1941 * gzip and bzip2 support in xzdiff was fixed.
1945 * Minor fix to Czech translation
1950 Only the most important changes compared to 4.999.9beta are listed
1951 here. One change is especially important:
1953 * The memory usage limit is now disabled by default. Some scripts
1954 written before this change may have used --memory=max on xz command
1955 line or in XZ_OPT. THESE USES OF --memory=max SHOULD BE REMOVED
1956 NOW, because they interfere with user's ability to set the memory
1957 usage limit himself. If user-specified limit causes problems to
1958 your script, blame the user.
1960 Other significant changes:
1962 * Added support for XZ_DEFAULTS environment variable. This variable
1963 allows users to set default options for xz, e.g. default memory
1964 usage limit or default compression level. Scripts that use xz
1965 must never set or unset XZ_DEFAULTS. Scripts should use XZ_OPT
1966 instead if they need a way to pass options to xz via an
1967 environment variable.
1969 * The compression settings associated with the preset levels
1970 -0 ... -9 have been changed. --extreme was changed a little too.
1971 It is now less likely to make compression worse, but with some
1972 files the new --extreme may compress slightly worse than the old
1975 * If a preset level (-0 ... -9) is specified after a custom filter
1976 chain options have been used (e.g. --lzma2), the custom filter
1977 chain will be forgotten. Earlier the preset options were
1978 completely ignored after custom filter chain options had been
1981 * xz will create sparse files when decompressing if the uncompressed
1982 data contains long sequences of binary zeros. This is done even
1983 when writing to standard output that is connected to a regular
1984 file and certain additional conditions are met to make it safe.
1986 * Support for "xz --list" was added. Combine with --verbose or
1987 --verbose --verbose (-vv) for detailed output.
1989 * I had hoped that liblzma API would have been stable after
1990 4.999.9beta, but there have been a couple of changes in the
1991 advanced features, which don't affect most applications:
1993 - Index handling code was revised. If you were using the old
1994 API, you will get a compiler error (so it's easy to notice).
1996 - A subtle but important change was made to the Block handling
1997 API. lzma_block.version has to be initialized even for
1998 lzma_block_header_decode(). Code that doesn't do it will work
1999 for now, but might break in the future, which makes this API
2000 change easy to miss.
2002 * The major soname has been bumped to 5.0.0. liblzma API and ABI
2003 are now stable, so the need to recompile programs linking against
2004 liblzma shouldn't arise soon.
projects / thirdparty / xz.git / blob