7 * All fixes from 5.2.9.
11 - Added new LZMA_FILTER_LZMA1EXT for raw encoder and decoder to
12 handle raw LZMA1 streams that don't have end of payload marker
13 (EOPM) alias end of stream (EOS) marker. It can be used in
14 filter chains, for example, with the x86 BCJ filter.
16 - Added lzma_str_to_filters(), lzma_str_from_filters(), and
17 lzma_str_list_filters() to make it easier for applications
18 to get custom compression options from a user and convert
19 it to an array of lzma_filter structures.
21 - Added lzma_filters_free().
23 - lzma_filters_update() can now be used with the multi-threaded
24 encoder (lzma_stream_encoder_mt()) to change the filter chain
25 after LZMA_FULL_BARRIER or LZMA_FULL_FLUSH.
27 - In lzma_options_lzma, allow nice_len = 2 and 3 with the match
28 finders that require at least 3 or 4. Now it is internally
31 - ARM64 filter was modified. It is still experimental.
33 - Fixed LTO build with Clang if -fgnuc-version=10 or similar
34 was used to make Clang look like GCC >= 10. Now it uses
35 __has_attribute(__symver__) which should be reliable.
39 - --threads=+1 or -T+1 is now a way to put xz into multi-threaded
40 mode while using only one worker thread.
42 - In --lzma2=nice=NUMBER allow 2 and 3 with all match finders
43 now that liblzma handles it.
45 * Updated translations: Chinese (simplified), Korean, and Turkish.
48 5.3.4alpha (2022-11-15)
50 * All fixes from 5.2.7 and 5.2.8.
54 - Minor improvements to the threaded decoder.
56 - Added CRC64 implementation that uses SSSE3, SSE4.1, and CLMUL
57 instructions on 32/64-bit x86 and E2K. On 32-bit x86 it's
58 not enabled unless --disable-assembler is used but then
59 the non-CLMUL code might be slower. Processor support is
60 detected at runtime so this is built by default on x86-64
61 and E2K. On these platforms, if compiler flags indicate
62 unconditional CLMUL support (-msse4.1 -mpclmul) then the
63 generic version is not built, making liblzma 8-9 KiB smaller
64 compared to having both versions included.
66 With extremely compressible files this can make decompression
67 up to twice as fast but with typical files 5 % improvement
68 is a more realistic expectation.
70 The CLMUL version is slower than the generic version with
71 tiny inputs (especially at 1-8 bytes per call, but up to
72 16 bytes). In normal use in xz this doesn't matter at all.
74 - Added an experimental ARM64 filter. This is *not* the final
75 version! Files created with this experimental version won't
76 be supported in the future versions! The filter design is
77 a compromise where improving one use case makes some other
80 - Added decompression support for the .lz (lzip) file format
81 version 0 and the original unextended version 1. See the
82 API docs of lzma_lzip_decoder() for details. Also
83 lzma_auto_decoder() supports .lz files.
85 - Building with --disable-threads --enable-small
86 is now thread-safe if the compiler supports
87 __attribute__((__constructor__))
91 - Added support for OpenBSD's pledge(2) as a sandboxing method.
93 - Don't mention endianness for ARM and ARM-Thumb filters in
94 --long-help. The filters only work for little endian
95 instruction encoding but modern ARM processors using
96 big endian data access still use little endian
97 instruction encoding. So the help text was misleading.
98 In contrast, the PowerPC filter is only for big endian
99 32/64-bit PowerPC code. Little endian PowerPC would need
102 - Added --experimental-arm64. This will be renamed once the
103 filter is finished. Files created with this experimental
104 filter will not be supported in the future!
106 - Added new fields to the output of xz --robot --info-memory.
108 - Added decompression support for the .lz (lzip) file format
109 version 0 and the original unextended version 1. It is
110 autodetected by default. See also the option --format on
113 * Scripts now support the .lz format using xz.
117 - New #defines in config.h: HAVE_ENCODER_ARM64,
118 HAVE_DECODER_ARM64, HAVE_LZIP_DECODER, HAVE_CPUID_H,
119 HAVE_FUNC_ATTRIBUTE_CONSTRUCTOR, HAVE_USABLE_CLMUL
121 - New configure options: --disable-clmul-crc,
122 --disable-microlzma, --disable-lzip-decoder, and
123 'pledge' is now an option in --enable-sandbox (but
124 it's autodetected by default anyway).
126 - INSTALL was updated to document the new configure options.
128 - PACKAGERS now lists also --disable-microlzma and
129 --disable-lzip-decoder as configure options that must
130 not be used in builds for non-embedded use.
134 - Fix some of the tests so that they skip instead of fail if
135 certain features have been disabled with configure options.
136 It's still not perfect.
138 - Other improvements to tests.
140 * Updated translations: Croatian, Finnish, Hungarian, Polish,
141 Romanian, Spanish, Swedish, and Ukrainian.
144 5.3.3alpha (2022-08-22)
146 * All fixes from 5.2.6.
150 - Fixed 32-bit build.
152 - Added threaded .xz decompressor lzma_stream_decoder_mt().
153 It can use multiple threads with .xz files that have multiple
154 Blocks with size information in Block Headers. The threaded
155 encoder in xz has always created such files.
157 Single-threaded encoder cannot store the size information in
158 Block Headers even if one used LZMA_FULL_FLUSH to create
159 multiple Blocks, so this threaded decoder cannot use multiple
160 threads with such files.
162 If there are multiple Streams (concatenated .xz files), one
163 Stream will be decompressed completely before starting the
166 - A new decoder flag LZMA_FAIL_FAST was added. It makes the
167 threaded decompressor report errors soon instead of first
168 flushing all pending data before the error location.
172 - Using -T0 (--threads=0) will now use multi-threaded encoder
173 even on a single-core system. This is to ensure that output
174 from the same xz binary is identical on both single-core and
177 - A default soft memory usage limit is now used for compression
178 when -T0 is used and no explicit limit has been specified.
179 This soft limit is used to restrict the number of threads
180 but if the limit is exceeded with even one thread then xz
181 will continue with one thread using the multi-threaded
182 encoder and this limit is ignored. If the number of threads
183 is specified manually then no default limit will be used;
184 this affects only -T0.
186 This change helps on systems that have very many cores and
187 using all of them for xz makes no sense. Previously xz -T0
188 could run out of memory on such systems because it attempted
189 to reserve memory for too many threads.
191 This also helps with 32-bit builds which don't have a large
192 amount of address space that would be required for many
193 threads. The default limit is 1400 MiB on all 32-bit
196 Now xz -T0 should just work. It might use too few threads
197 in some cases but at least it shouldn't easily run out of
198 memory. It's possible that this will be tweaked before 5.4.0.
200 - Changes to --memlimit-compress and --no-adjust:
202 In single-threaded mode, --memlimit-compress can make xz
203 scale down the LZMA2 dictionary size to meet the memory usage
204 limit. This obviously affects the compressed output. However,
205 if xz was in threaded mode, --memlimit-compress could make xz
206 reduce the number of threads but it wouldn't make xz switch
207 from multi-threaded mode to single-threaded mode or scale
208 down the LZMA2 dictionary size. This seemed illogical.
210 Now --memlimit-compress can make xz switch to single-threaded
211 mode if one thread in multi-threaded mode uses too much
212 memory. If memory usage is still too high, then the LZMA2
213 dictionary size can be scaled down too.
215 The option --no-adjust was also changed so that it no longer
216 prevents xz from scaling down the number of threads as that
217 doesn't affect compressed output (only performance). After
218 this commit --no-adjust only prevents adjustments that affect
219 compressed output, that is, with --no-adjust xz won't switch
220 from multithreaded mode to single-threaded mode and won't
221 scale down the LZMA2 dictionary size.
223 - Added a new option --memlimit-decompress-mt=LIMIT. This is
224 used to limit the number of decompressor threads (possibly
225 falling back to single-threaded mode) but it will never make
226 xz refuse to decompress a file. This has a system-specific
227 default value because without any limit xz could end up
228 allocating memory for the whole compressed input file, the
229 whole uncompressed output file, multiple thread-specific
230 decompressor instances and so on. Basically xz could
231 attempt to use an insane amount of memory even with fairly
234 The new option works together with the existing option
235 --memlimit-decompress=LIMIT. The old option sets a hard limit
236 that must not be exceeded (xz will refuse to decompress)
237 while the new option only restricts the number of threads.
238 If the limit set with --memlimit-mt-decompress is greater
239 than the limit set with --memlimit-compress, then the latter
240 value is used also for --memlimit-mt-decompress.
244 - Added a few more tests.
246 - Added tests/code_coverage.sh to create a code coverage report
251 - Automake's parallel test harness is now used to make tests
254 - Added the CMake files to the distribution tarball. These were
255 supposed to be in 5.2.5 already.
257 - Added liblzma tests to the CMake build.
259 - Windows: Fix building of liblzma.dll with the included
260 Visual Studio project files.
263 5.3.2alpha (2021-10-28)
265 This release was made on short notice so that recent erofs-utils can
266 be built with LZMA support without needing a snapshot from xz.git.
267 Thus many pending things were not included, not even updated
268 translations (which would need to be updated for the new --list
271 * All fixes from 5.2.5.
275 - When copying metadata from the source file to the destination
276 file, don't try to set the group (GID) if it is already set
277 correctly. This avoids a failure on OpenBSD (and possibly on
278 a few other OSes) where files may get created so that their
279 group doesn't belong to the user, and fchown(2) can fail even
280 if it needs to do nothing.
282 - The --keep option now accepts symlinks, hardlinks, and
283 setuid, setgid, and sticky files. Previously this required
286 - Split the long strings used in --list and --info-memory modes
287 to make them much easier for translators.
289 - If built with sandbox support and enabling the sandbox fails,
290 xz will now immediately exit with exit status of 1. Previously
291 it would only display a warning if -vv was used.
293 - Cap --memlimit-compress to 2000 MiB on MIPS32 because on
294 MIPS32 userspace processes are limited to 2 GiB of address
299 - Added lzma_microlzma_encoder() and lzma_microlzma_decoder().
300 The API is in lzma/container.h.
302 The MicroLZMA format is a raw LZMA stream (without end marker)
303 whose first byte (always 0x00) has been replaced with
304 bitwise-negation of the LZMA properties (lc/lp/pb). It was
305 created for use in EROFS but may be used in other contexts
306 as well where it is important to avoid wasting bytes for
307 stream headers or footers. The format is also supported by
310 The MicroLZMA encoder API in liblzma can compress into a
311 fixed-sized output buffer so that as much data is compressed
312 as can be fit into the buffer while still creating a valid
313 MicroLZMA stream. This is needed for EROFS.
315 - Added fuzzing support.
317 - Support Intel Control-flow Enforcement Technology (CET) in
318 32-bit x86 assembly files.
320 - Visual Studio: Use non-standard _MSVC_LANG to detect C++
321 standard version in the lzma.h API header. It's used to
322 detect when "noexcept" can be used.
326 - Fix exit status of xzdiff/xzcmp. Exit status could be 2 when
327 the correct value is 1.
329 - Fix exit status of xzgrep.
331 - Detect corrupt .bz2 files in xzgrep.
333 - Add zstd support to xzgrep and xzdiff/xzcmp.
335 - Fix less(1) version detection in xzless. It failed if the
336 version number from "less -V" contained a dot.
338 * Fix typos and technical issues in man pages.
342 - Windows: Fix building of resource files when config.h isn't
343 used. CMake + Visual Studio can now build liblzma.dll.
345 - Various fixes to the CMake support. It might still need a few
346 more fixes even for liblzma-only builds.
349 5.3.1alpha (2018-04-29)
351 * All fixes from 5.2.4.
353 * Add lzma_file_info_decoder() into liblzma and use it in xz to
354 implement the --list feature.
356 * Capsicum sandbox support is enabled by default where available
364 - Fixed an infinite loop in LZMA encoder initialization
365 if dict_size >= 2 GiB. (The encoder only supports up
368 - Fixed two cases of invalid free() that can happen if
369 a tiny allocation fails in encoder re-initialization
370 or in lzma_filters_update(). These bugs had some
371 similarities with the bug fixed in 5.2.7.
373 - Fixed lzma_block_encoder() not allowing the use of
374 LZMA_SYNC_FLUSH with lzma_code() even though it was
375 documented to be supported. The sync-flush code in
376 the Block encoder was already used internally via
377 lzma_stream_encoder(), so this was just a missing flag
378 in the lzma_block_encoder() API function.
380 - GNU/Linux only: Don't put symbol versions into static
381 liblzma as it breaks things in some cases (and even if
382 it didn't break anything, symbol versions in static
383 libraries are useless anyway). The downside of the fix
384 is that if the configure options --with-pic or --without-pic
385 are used then it's not possible to build both shared and
386 static liblzma at the same time on GNU/Linux anymore;
387 with those options --disable-static or --disable-shared
390 * New email address for bug reports is <xz@tukaani.org> which
391 forwards messages to Lasse Collin and Jia Tan.
398 - If xz cannot remove an input file when it should, this
399 is now treated as a warning (exit status 2) instead of
400 an error (exit status 1). This matches GNU gzip and it
401 is more logical as at that point the output file has
402 already been successfully closed.
404 - Fix handling of .xz files with an unsupported check type.
405 Previously such printed a warning message but then xz
406 behaved as if an error had occurred (didn't decompress,
407 exit status 1). Now a warning is printed, decompression
408 is done anyway, and exit status is 2. This used to work
409 slightly before 5.0.0. In practice this bug matters only
410 if xz has been built with some check types disabled. As
411 instructed in PACKAGERS, such builds should be done in
412 special situations only.
414 - Fix "xz -dc --single-stream tests/files/good-0-empty.xz"
415 which failed with "Internal error (bug)". That is,
416 --single-stream was broken if the first .xz stream in
417 the input file didn't contain any uncompressed data.
419 - Fix displaying file sizes in the progress indicator when
420 working in passthru mode and there are multiple input files.
421 Just like "gzip -cdf", "xz -cdf" works like "cat" when the
422 input file isn't a supported compressed file format. In
423 this case the file size counters weren't reset between
424 files so with multiple input files the progress indicator
425 displayed an incorrect (too large) value.
429 - API docs in lzma/container.h:
430 * Update the list of decoder flags in the decoder
432 * Explain LZMA_CONCATENATED behavior with .lzma files
433 in lzma_auto_decoder() docs.
435 - OpenBSD: Use HW_NCPUONLINE to detect the number of
436 available hardware threads in lzma_physmem().
438 - Fix use of wrong macro to detect x86 SSE2 support.
439 __SSE2_MATH__ was used with GCC/Clang but the correct
440 one is __SSE2__. The first one means that SSE2 is used
441 for floating point math which is irrelevant here.
442 The affected SSE2 code isn't used on x86-64 so this affects
443 only 32-bit x86 builds that use -msse2 without -mfpmath=sse
444 (there is no runtime detection for SSE2). It improves LZMA
445 compression speed (not decompression).
447 - Fix the build with Intel C compiler 2021 (ICC, not ICX)
448 on Linux. It defines __GNUC__ to 10 but doesn't support
449 the __symver__ attribute introduced in GCC 10.
451 * Scripts: Ignore warnings from xz by using --quiet --no-warn.
452 This is needed if the input .xz files use an unsupported
457 - Updated Croatian and Turkish translations.
459 - One new translations wasn't included because it needed
460 technical fixes. It will be in upcoming 5.4.0. No new
461 translations will be added to the 5.2.x branch anymore.
463 - Renamed the French man page translation file from
464 fr_FR.po to fr.po and thus also its install directory
465 (like /usr/share/man/fr_FR -> .../fr).
467 - Man page translations for upcoming 5.4.0 are now handled
468 in the Translation Project.
470 * Update doc/faq.txt a little so it's less out-of-date.
477 - Made lzma_filters_copy() to never modify the destination
478 array if an error occurs. lzma_stream_encoder() and
479 lzma_stream_encoder_mt() already assumed this. Before this
480 change, if a tiny memory allocation in lzma_filters_copy()
481 failed it would lead to a crash (invalid free() or invalid
482 memory reads) in the cleanup paths of these two encoder
483 initialization functions.
485 - Added missing integer overflow check to lzma_index_append().
486 This affects xz --list and other applications that decode
487 the Index field from .xz files using lzma_index_decoder().
488 Normal decompression of .xz files doesn't call this code
489 and thus most applications using liblzma aren't affected
492 - Single-threaded .xz decoder (lzma_stream_decoder()): If
493 lzma_code() returns LZMA_MEMLIMIT_ERROR it is now possible
494 to use lzma_memlimit_set() to increase the limit and continue
495 decoding. This was supposed to work from the beginning
496 but there was a bug. With other decoders (.lzma or
497 threaded .xz decoder) this already worked correctly.
499 - Fixed accumulation of integrity check type statistics in
500 lzma_index_cat(). This bug made lzma_index_checks() return
501 only the type of the integrity check of the last Stream
502 when multiple lzma_indexes were concatenated. Most
503 applications don't use these APIs but in xz it made
504 xz --list not list all check types from concatenated .xz
505 files. In xz --list --verbose only the per-file "Check:"
506 lines were affected and in xz --robot --list only the "file"
509 - Added ABI compatibility with executables that were linked
510 against liblzma in RHEL/CentOS 7 or other liblzma builds
511 that had copied the problematic patch from RHEL/CentOS 7
512 (xz-5.2.2-compat-libs.patch). For the details, see the
513 comment at the top of src/liblzma/validate_map.sh.
515 WARNING: This uses __symver__ attribute with GCC >= 10.
516 In other cases the traditional __asm__(".symver ...")
517 is used. Using link-time optimization (LTO, -flto) with
518 GCC versions older than 10 can silently result in
519 broken liblzma.so.5 (incorrect symbol versions)! If you
520 want to use -flto with GCC, you must use GCC >= 10.
521 LTO with Clang seems to work even with the traditional
522 __asm__(".symver ...") method.
524 * xzgrep: Fixed compatibility with old shells that break if
525 comments inside command substitutions have apostrophes (').
526 This problem was introduced in 5.2.6.
530 - New #define in config.h: HAVE_SYMBOL_VERSIONS_LINUX
532 - Windows: Fixed liblzma.dll build with Visual Studio project
533 files. It broke in 5.2.6 due to a change that was made to
534 improve CMake support.
536 - Windows: Building liblzma with UNICODE defined should now
539 - CMake files are now actually included in the release tarball.
540 They should have been in 5.2.5 already.
542 - Minor CMake fixes and improvements.
544 * Added a new translation: Turkish
551 - The --keep option now accepts symlinks, hardlinks, and
552 setuid, setgid, and sticky files. Previously this required
555 - When copying metadata from the source file to the destination
556 file, don't try to set the group (GID) if it is already set
557 correctly. This avoids a failure on OpenBSD (and possibly on
558 a few other OSes) where files may get created so that their
559 group doesn't belong to the user, and fchown(2) can fail even
560 if it needs to do nothing.
562 - Cap --memlimit-compress to 2000 MiB instead of 4020 MiB on
563 MIPS32 because on MIPS32 userspace processes are limited
564 to 2 GiB of address space.
568 - Fixed a missing error-check in the threaded encoder. If a
569 small memory allocation fails, a .xz file with an invalid
570 Index field would be created. Decompressing such a file would
571 produce the correct output but result in an error at the end.
572 Thus this is a "mild" data corruption bug. Note that while
573 a failed memory allocation can trigger the bug, it cannot
574 cause invalid memory access.
576 - The decoder for .lzma files now supports files that have
577 uncompressed size stored in the header and still use the
578 end of payload marker (end of stream marker) at the end
579 of the LZMA stream. Such files are rare but, according to
580 the documentation in LZMA SDK, they are valid.
581 doc/lzma-file-format.txt was updated too.
583 - Improved 32-bit x86 assembly files:
584 * Support Intel Control-flow Enforcement Technology (CET)
585 * Use non-executable stack on FreeBSD.
587 - Visual Studio: Use non-standard _MSVC_LANG to detect C++
588 standard version in the lzma.h API header. It's used to
589 detect when "noexcept" can be used.
593 - Fixed arbitrary command injection via a malicious filename
594 (CVE-2022-1271, ZDI-CAN-16587). A standalone patch for
595 this was released to the public on 2022-04-07. A slight
596 robustness improvement has been made since then and, if
597 using GNU or *BSD grep, a new faster method is now used
598 that doesn't use the old sed-based construct at all. This
599 also fixes bad output with GNU grep >= 3.5 (2020-09-27)
600 when xzgrepping binary files.
602 This vulnerability was discovered by:
603 cleemy desu wayo working with Trend Micro Zero Day Initiative
605 - Fixed detection of corrupt .bz2 files.
607 - Improved error handling to fix exit status in some situations
608 and to fix handling of signals: in some situations a signal
609 didn't make xzgrep exit when it clearly should have. It's
610 possible that the signal handling still isn't quite perfect
611 but hopefully it's good enough.
613 - Documented exit statuses on the man page.
615 - xzegrep and xzfgrep now use "grep -E" and "grep -F" instead
616 of the deprecated egrep and fgrep commands.
618 - Fixed parsing of the options -E, -F, -G, -P, and -X. The
619 problem occurred when multiple options were specied in
620 a single argument, for example,
622 echo foo | xzgrep -Fe foo
624 treated foo as a filename because -Fe wasn't correctly
627 - Added zstd support.
631 - Fixed wrong exit status. Exit status could be 2 when the
634 - Documented on the man page that exit status of 2 is used
635 for decompression errors.
637 - Added zstd support.
641 - Fix less(1) version detection. It failed if the version number
642 from "less -V" contained a dot.
646 - Added new translations: Catalan, Croatian, Esperanto,
647 Korean, Portuguese, Romanian, Serbian, Spanish, Swedish,
650 - Updated the Brazilian Portuguese translation.
652 - Added French man page translation. This and the existing
653 German translation aren't complete anymore because the
654 English man pages got a few updates and the translators
655 weren't reached so that they could update their work.
659 - Windows: Fix building of resource files when config.h isn't
660 used. CMake + Visual Studio can now build liblzma.dll.
662 - Various fixes to the CMake support. Building static or shared
663 liblzma should work fine in most cases. In contrast, building
664 the command line tools with CMake is still clearly incomplete
665 and experimental and should be used for testing only.
672 - Fixed several C99/C11 conformance bugs. Now the code is clean
673 under gcc/clang -fsanitize=undefined. Some of these changes
674 might have a negative effect on performance with old GCC
675 versions or compilers other than GCC and Clang. The configure
676 option --enable-unsafe-type-punning can be used to (mostly)
677 restore the old behavior but it shouldn't normally be used.
679 - Improved API documentation of lzma_properties_decode().
681 - Added a very minor encoder speed optimization.
685 - Fixed a crash in "xz -dcfv not_an_xz_file". All four options
686 were required to trigger it. The crash occurred in the
687 progress indicator code when xz was in passthru mode where
690 - Fixed an integer overflow with 32-bit off_t. It could happen
691 when decompressing a file that has a long run of zero bytes
692 which xz would try to write as a sparse file. Since the build
693 system enables large file support by default, off_t is
694 normally 64-bit even on 32-bit systems.
696 - Fixes for --flush-timeout:
697 * Fix semi-busy-waiting.
698 * Avoid unneeded flushes when no new input has arrived
699 since the previous flush was completed.
701 - Added a special case for 32-bit xz: If --memlimit-compress is
702 used to specify a limit that exceeds 4020 MiB, the limit will
703 be set to 4020 MiB. The values "0" and "max" aren't affected
704 by this and neither is decompression. This hack can be
705 helpful when a 32-bit xz has access to 4 GiB address space
706 but the specified memlimit exceeds 4 GiB. This can happen
707 e.g. with some scripts.
709 - Capsicum sandbox is now enabled by default where available
710 (FreeBSD >= 10). The sandbox debug messages (xz -vv) were
711 removed since they seemed to be more annoying than useful.
713 - DOS build now requires DJGPP 2.05 instead of 2.04beta.
714 A workaround for a locale problem with DJGPP 2.05 was added.
716 * xzgrep and other scripts:
718 - Added a configure option --enable-path-for-scripts=PREFIX.
719 It is disabled by default except on Solaris where the default
720 is /usr/xpg4/bin. See INSTALL for details.
722 - Added a workaround for a POSIX shell detection problem on
727 - Added preliminary build instructions for z/OS. See INSTALL
730 - Experimental CMake support was added. It should work to build
731 static liblzma on a few operating systems. It may or may not
732 work to build shared liblzma. On some platforms it can build
733 xz and xzdec too but those are only for testing. See the
734 comment in the beginning of CMakeLists.txt for details.
736 - Visual Studio project files were updated.
737 WindowsTargetPlatformVersion was removed from VS2017 files
738 and set to "10.0" in the added VS2019 files. In the future
739 the VS project files will be removed when CMake support is
742 - New #defines in config.h: HAVE___BUILTIN_ASSUME_ALIGNED,
743 HAVE___BUILTIN_BSWAPXX, and TUKLIB_USE_UNSAFE_TYPE_PUNNING.
745 - autogen.sh has a new optional dependency on po4a and a new
746 option --no-po4a to skip that step. This matters only if one
747 wants to remake the build files. po4a is used to update the
748 translated man pages but as long as the man pages haven't
749 been modified, there's nothing to update and one can use
750 --no-po4a to avoid the dependency on po4a.
754 - XZ Utils translations are now handled by the Translation
755 Project: https://translationproject.org/domain/xz.html
757 - All man pages are now included in German too.
759 - New xz translations: Brazilian Portuguese, Finnish,
760 Hungarian, Chinese (simplified), Chinese (traditional),
761 and Danish (partial translation)
763 - Updated xz translations: French, German, Italian, and Polish
765 - Unfortunately a few new xz translations weren't included due
766 to technical problems like too long lines in --help output or
767 misaligned column headings in tables. In the future, many of
768 these strings will be split and e.g. the table column
769 alignment will be handled in software. This should make the
770 strings easier to translate.
777 - Allow 0 as memory usage limit instead of returning
778 LZMA_PROG_ERROR. Now 0 is treated as if 1 byte was specified,
779 which effectively is the same as 0.
781 - Use "noexcept" keyword instead of "throw()" in the public
782 headers when a C++11 (or newer standard) compiler is used.
784 - Added a portability fix for recent Intel C Compilers.
786 - Microsoft Visual Studio build files have been moved under
787 windows/vs2013 and windows/vs2017.
791 - Fix "xz --list --robot missing_or_bad_file.xz" which would
792 try to print an uninitialized string and thus produce garbage
793 output. Since the exit status is non-zero, most uses of such
794 a command won't try to interpret the garbage output.
796 - "xz --list foo.xz" could print "Internal error (bug)" in a
797 corner case where a specific memory usage limit had been set.
804 - Always close a file before trying to delete it to avoid
805 problems on some operating system and file system combinations.
807 - Fixed copying of file timestamps on Windows.
809 - Added experimental (disabled by default) sandbox support using
810 Capsicum (FreeBSD >= 10). See --enable-sandbox in INSTALL.
812 * C99/C11 conformance fixes to liblzma. The issues affected at least
813 some builds using link-time optimizations.
815 * Fixed bugs in the rarely-used function lzma_index_dup().
817 * Use of external SHA-256 code is now disabled by default.
818 It can still be enabled by passing --enable-external-sha256
819 to configure. The reasons to disable it by default (see INSTALL
822 - Some OS-specific SHA-256 implementations conflict with
823 OpenSSL and cause problems in programs that link against both
824 liblzma and libcrypto. At least FreeBSD 10 and MINIX 3.3.0
827 - The internal SHA-256 is faster than the SHA-256 code in
828 some operating systems.
830 * Changed CPU core count detection to use sched_getaffinity() on
831 GNU/Linux and GNU/kFreeBSD.
833 * Fixes to the build-system and xz to make xz buildable even when
834 encoders, decoders, or threading have been disabled from libilzma
835 using configure options. These fixes added two new #defines to
836 config.h: HAVE_ENCODERS and HAVE_DECODERS.
841 * Fixed bugs in QNX-specific code.
843 * Omitted the use of pipe2() even if it is available to avoid
844 portability issues with some old Linux and glibc combinations.
846 * Updated German translation.
848 * Added project files to build static and shared liblzma (not the
849 whole XZ Utils) with Visual Studio 2013 update 2 or later.
851 * Documented that threaded decompression hasn't been implemented
852 yet. A 5.2.0 NEWS entry describing multi-threading support had
853 incorrectly said "decompression" when it should have said
859 * Fixed a compression-ratio regression in fast mode of LZMA1 and
860 LZMA2. The bug is present in 5.1.4beta and 5.2.0 releases.
862 * Fixed a portability problem in xz that affected at least OpenBSD.
864 * Fixed xzdiff to be compatible with FreeBSD's mktemp which differs
865 from most other mktemp implementations.
867 * Changed CPU core count detection to use cpuset_getaffinity() on
875 * All fixes from 5.0.8
877 * liblzma: Fixed lzma_stream_encoder_mt_memusage() when a preset
880 * xzdiff: If mktemp isn't installed, mkdir will be used as
881 a fallback to create a temporary directory. Installing mktemp
882 is still recommended.
884 * Updated French, German, Italian, Polish, and Vietnamese
887 Summary of fixes and new features added in the 5.1.x development
892 - Added support for multi-threaded compression. See the
893 lzma_mt structure, lzma_stream_encoder_mt(), and
894 lzma_stream_encoder_mt_memusage() in <lzma/container.h>,
895 lzma_get_progress() in <lzma/base.h>, and lzma_cputhreads()
896 in <lzma/hardware.h> for details.
898 - Made the uses of lzma_allocator const correct.
900 - Added lzma_block_uncomp_encode() to create uncompressed
901 .xz Blocks using LZMA2 uncompressed chunks.
903 - Added support for LZMA_IGNORE_CHECK.
905 - A few speed optimizations were made.
907 - Added support for symbol versioning. It is enabled by default
908 on GNU/Linux, other GNU-based systems, and FreeBSD.
910 - liblzma (not the whole XZ Utils) should now be buildable
911 with MSVC 2013 update 2 or later using windows/config.h.
915 - Fixed a race condition in the signal handling. It was
916 possible that e.g. the first SIGINT didn't make xz exit
917 if reading or writing blocked and one had bad luck. The fix
918 is non-trivial, so as of writing it is unknown if it will be
919 backported to the v5.0 branch.
921 - Multi-threaded compression can be enabled with the
922 --threads (-T) option.
923 [Fixed: This originally said "decompression".]
925 - New command line options in xz: --single-stream,
926 --block-size=SIZE, --block-list=SIZES,
927 --flush-timeout=TIMEOUT, and --ignore-check.
929 - xz -lvv now shows the minimum xz version that is required to
930 decompress the file. Currently it is 5.0.0 for all supported
931 .xz files except files with empty LZMA2 streams require 5.0.2.
933 * xzdiff and xzgrep now support .lzo files if lzop is installed.
934 The .tzo suffix is also recognized as a shorthand for .tar.lzo.
937 5.1.4beta (2014-09-14)
939 * All fixes from 5.0.6
941 * liblzma: Fixed the use of presets in threaded encoder
944 * xz --block-list and --block-size can now be used together
945 in single-threaded mode. Previously the combination only
946 worked in multi-threaded mode.
948 * Added support for LZMA_IGNORE_CHECK to liblzma and made it
949 available in xz as --ignore-check.
951 * liblzma speed optimizations:
953 - Initialization of a new LZMA1 or LZMA2 encoder has been
954 optimized. (The speed of reinitializing an already-allocated
955 encoder isn't affected.) This helps when compressing many
956 small buffers with lzma_stream_buffer_encode() and other
957 similar situations where an already-allocated encoder state
958 isn't reused. This speed-up is visible in xz too if one
959 compresses many small files one at a time instead running xz
960 once and giving all files as command-line arguments.
962 - Buffer comparisons are now much faster when unaligned access
963 is allowed (configured with --enable-unaligned-access). This
964 speeds up encoding significantly. There is arch-specific code
965 for 32-bit and 64-bit x86 (32-bit needs SSE2 for the best
966 results and there's no run-time CPU detection for now).
967 For other archs there is only generic code which probably
968 isn't as optimal as arch-specific solutions could be.
970 - A few speed optimizations were made to the SHA-256 code.
971 (Note that the builtin SHA-256 code isn't used on all
974 * liblzma can now be built with MSVC 2013 update 2 or later
975 using windows/config.h.
977 * Vietnamese translation was added.
980 5.1.3alpha (2013-10-26)
982 * All fixes from 5.0.5
986 - Fixed a deadlock in the threaded encoder.
988 - Made the uses of lzma_allocator const correct.
990 - Added lzma_block_uncomp_encode() to create uncompressed
991 .xz Blocks using LZMA2 uncompressed chunks.
993 - Added support for native threads on Windows and the ability
994 to detect the number of CPU cores.
998 - Fixed a race condition in the signal handling. It was
999 possible that e.g. the first SIGINT didn't make xz exit
1000 if reading or writing blocked and one had bad luck. The fix
1001 is non-trivial, so as of writing it is unknown if it will be
1002 backported to the v5.0 branch.
1004 - Made the progress indicator work correctly in threaded mode.
1006 - Threaded encoder now works together with --block-list=SIZES.
1008 - Added preliminary support for --flush-timeout=TIMEOUT.
1009 It can be useful for (somewhat) real-time streaming. For
1010 now the decompression side has to be done with something
1011 else than the xz tool due to how xz does buffering, but this
1015 5.1.2alpha (2012-07-04)
1017 * All fixes from 5.0.3 and 5.0.4
1021 - Fixed a deadlock and an invalid free() in the threaded encoder.
1023 - Added support for symbol versioning. It is enabled by default
1024 on GNU/Linux, other GNU-based systems, and FreeBSD.
1026 - Use SHA-256 implementation from the operating system if one is
1027 available in libc, libmd, or libutil. liblzma won't use e.g.
1028 OpenSSL or libgcrypt to avoid introducing new dependencies.
1030 - Fixed liblzma.pc for static linking.
1032 - Fixed a few portability bugs.
1034 * xz --decompress --single-stream now fixes the input position after
1035 successful decompression. Now the following works:
1037 echo foo | xz > foo.xz
1038 echo bar | xz >> foo.xz
1039 ( xz -dc --single-stream ; xz -dc --single-stream ) < foo.xz
1041 Note that it doesn't work if the input is not seekable
1042 or if there is Stream Padding between the concatenated
1045 * xz -lvv now shows the minimum xz version that is required to
1046 decompress the file. Currently it is 5.0.0 for all supported .xz
1047 files except files with empty LZMA2 streams require 5.0.2.
1049 * Added an *incomplete* implementation of --block-list=SIZES to xz.
1050 It only works correctly in single-threaded mode and when
1051 --block-size isn't used at the same time. --block-list allows
1052 specifying the sizes of Blocks which can be useful e.g. when
1053 creating files for random-access reading.
1056 5.1.1alpha (2011-04-12)
1058 * All fixes from 5.0.2
1060 * liblzma fixes that will also be included in 5.0.3:
1062 - A memory leak was fixed.
1064 - lzma_stream_buffer_encode() no longer creates an empty .xz
1065 Block if encoding an empty buffer. Such an empty Block with
1066 LZMA2 data would trigger a bug in 5.0.1 and older (see the
1067 first bullet point in 5.0.2 notes). When releasing 5.0.2,
1068 I thought that no encoder creates this kind of files but
1071 - Validate function arguments better in a few functions. Most
1072 importantly, specifying an unsupported integrity check to
1073 lzma_stream_buffer_encode() no longer creates a corrupt .xz
1074 file. Probably no application tries to do that, so this
1075 shouldn't be a big problem in practice.
1077 - Document that lzma_block_buffer_encode(),
1078 lzma_easy_buffer_encode(), lzma_stream_encoder(), and
1079 lzma_stream_buffer_encode() may return LZMA_UNSUPPORTED_CHECK.
1081 - The return values of the _memusage() functions are now
1084 * Support for multithreaded compression was added using the simplest
1085 method, which splits the input data into blocks and compresses
1086 them independently. Other methods will be added in the future.
1087 The current method has room for improvement, e.g. it is possible
1088 to reduce the memory usage.
1090 * Added the options --single-stream and --block-size=SIZE to xz.
1092 * xzdiff and xzgrep now support .lzo files if lzop is installed.
1093 The .tzo suffix is also recognized as a shorthand for .tar.lzo.
1095 * Support for short 8.3 filenames under DOS was added to xz. It is
1096 experimental and may change before it gets into a stable release.
1101 * Fixed an old bug in xzgrep that affected OpenBSD and probably
1102 a few other operating systems too.
1104 * Updated French and German translations.
1106 * Added support for detecting the amount of RAM on AmigaOS/AROS.
1108 * Minor build system updates.
1113 * Fix regressions introduced in 5.0.6:
1115 - Fix building with non-GNU make.
1117 - Fix invalid Libs.private value in liblzma.pc which broke
1118 static linking against liblzma if the linker flags were
1119 taken from pkg-config.
1124 * xzgrep now exits with status 0 if at least one file matched.
1126 * A few minor portability and build system fixes
1131 * lzmadec and liblzma's lzma_alone_decoder(): Support decompressing
1132 .lzma files that have less common settings in the headers
1133 (dictionary size other than 2^n or 2^n + 2^(n-1), or uncompressed
1134 size greater than 256 GiB). The limitations existed to avoid false
1135 positives when detecting .lzma files. The lc + lp <= 4 limitation
1136 still remains since liblzma's LZMA decoder has that limitation.
1138 NOTE: xz's .lzma support or liblzma's lzma_auto_decoder() are NOT
1139 affected by this change. They still consider uncommon .lzma headers
1140 as not being in the .lzma format. Changing this would give way too
1141 many false positives.
1145 - Interaction of preset and custom filter chain options was
1146 made less illogical. This affects only certain less typical
1147 uses cases so few people are expected to notice this change.
1149 Now when a custom filter chain option (e.g. --lzma2) is
1150 specified, all preset options (-0 ... -9, -e) earlier are on
1151 the command line are completely forgotten. Similarly, when
1152 a preset option is specified, all custom filter chain options
1153 earlier on the command line are completely forgotten.
1155 Example 1: "xz -9 --lzma2=preset=5 -e" is equivalent to "xz -e"
1156 which is equivalent to "xz -6e". Earlier -e didn't put xz back
1157 into preset mode and thus the example command was equivalent
1158 to "xz --lzma2=preset=5".
1160 Example 2: "xz -9e --lzma2=preset=5 -7" is equivalent to
1161 "xz -7". Earlier a custom filter chain option didn't make
1162 xz forget the -e option so the example was equivalent to
1165 - Fixes and improvements to error handling.
1167 - Various fixes to the man page.
1169 * xzless: Fixed to work with "less" versions 448 and later.
1171 * xzgrep: Made -h an alias for --no-filename.
1173 * Include the previously missing debug/translation.bash which can
1174 be useful for translators.
1176 * Include a build script for Mac OS X. This has been in the Git
1177 repository since 2010 but due to a mistake in Makefile.am the
1178 script hasn't been included in a release tarball before.
1185 - Fix lzma_index_init(). It could crash if memory allocation
1188 - Fix the possibility of an incorrect LZMA_BUF_ERROR when a BCJ
1189 filter is used and the application only provides exactly as
1190 much output space as is the uncompressed size of the file.
1192 - Fix a bug in doc/examples_old/xz_pipe_decompress.c. It didn't
1193 check if the last call to lzma_code() really returned
1194 LZMA_STREAM_END, which made the program think that truncated
1197 - New example programs in doc/examples (old programs are now in
1198 doc/examples_old). These have more comments and more detailed
1201 * Fix "xz -lvv foo.xz". It could crash on some corrupted files.
1203 * Fix output of "xz --robot -lv" and "xz --robot -lvv" which
1204 incorrectly printed the filename also in the "foo (x/x)" format.
1206 * Fix exit status of "xzdiff foo.xz bar.xz".
1208 * Fix exit status of "xzgrep foo binary_file".
1210 * Fix portability to EBCDIC systems.
1212 * Fix a configure issue on AIX with the XL C compiler. See INSTALL
1215 * Update French, German, Italian, and Polish translations.
1222 - A memory leak was fixed.
1224 - lzma_stream_buffer_encode() no longer creates an empty .xz
1225 Block if encoding an empty buffer. Such an empty Block with
1226 LZMA2 data would trigger a bug in 5.0.1 and older (see the
1227 first bullet point in 5.0.2 notes). When releasing 5.0.2,
1228 I thought that no encoder creates this kind of files but
1231 - Validate function arguments better in a few functions. Most
1232 importantly, specifying an unsupported integrity check to
1233 lzma_stream_buffer_encode() no longer creates a corrupt .xz
1234 file. Probably no application tries to do that, so this
1235 shouldn't be a big problem in practice.
1237 - Document that lzma_block_buffer_encode(),
1238 lzma_easy_buffer_encode(), lzma_stream_encoder(), and
1239 lzma_stream_buffer_encode() may return LZMA_UNSUPPORTED_CHECK.
1241 - The return values of the _memusage() functions are now
1244 * Fix command name detection in xzgrep. xzegrep and xzfgrep now
1245 correctly use egrep and fgrep instead of grep.
1247 * French translation was added.
1252 * LZMA2 decompressor now correctly accepts LZMA2 streams with no
1253 uncompressed data. Previously it considered them corrupt. The
1254 bug can affect applications that use raw LZMA2 streams. It is
1255 very unlikely to affect .xz files because no compressor creates
1256 .xz files with empty LZMA2 streams. (Empty .xz files are a
1257 different thing than empty LZMA2 streams.)
1259 * "xz --suffix=.foo filename.foo" now refuses to compress the
1260 file due to it already having the suffix .foo. It was already
1261 documented on the man page, but the code lacked the test.
1263 * "xzgrep -l foo bar.xz" works now.
1265 * Polish translation was added.
1270 * xz --force now (de)compresses files that have setuid, setgid,
1271 or sticky bit set and files that have multiple hard links.
1272 The man page had it documented this way already, but the code
1275 * gzip and bzip2 support in xzdiff was fixed.
1279 * Minor fix to Czech translation
1284 Only the most important changes compared to 4.999.9beta are listed
1285 here. One change is especially important:
1287 * The memory usage limit is now disabled by default. Some scripts
1288 written before this change may have used --memory=max on xz command
1289 line or in XZ_OPT. THESE USES OF --memory=max SHOULD BE REMOVED
1290 NOW, because they interfere with user's ability to set the memory
1291 usage limit himself. If user-specified limit causes problems to
1292 your script, blame the user.
1294 Other significant changes:
1296 * Added support for XZ_DEFAULTS environment variable. This variable
1297 allows users to set default options for xz, e.g. default memory
1298 usage limit or default compression level. Scripts that use xz
1299 must never set or unset XZ_DEFAULTS. Scripts should use XZ_OPT
1300 instead if they need a way to pass options to xz via an
1301 environment variable.
1303 * The compression settings associated with the preset levels
1304 -0 ... -9 have been changed. --extreme was changed a little too.
1305 It is now less likely to make compression worse, but with some
1306 files the new --extreme may compress slightly worse than the old
1309 * If a preset level (-0 ... -9) is specified after a custom filter
1310 chain options have been used (e.g. --lzma2), the custom filter
1311 chain will be forgotten. Earlier the preset options were
1312 completely ignored after custom filter chain options had been
1315 * xz will create sparse files when decompressing if the uncompressed
1316 data contains long sequences of binary zeros. This is done even
1317 when writing to standard output that is connected to a regular
1318 file and certain additional conditions are met to make it safe.
1320 * Support for "xz --list" was added. Combine with --verbose or
1321 --verbose --verbose (-vv) for detailed output.
1323 * I had hoped that liblzma API would have been stable after
1324 4.999.9beta, but there have been a couple of changes in the
1325 advanced features, which don't affect most applications:
1327 - Index handling code was revised. If you were using the old
1328 API, you will get a compiler error (so it's easy to notice).
1330 - A subtle but important change was made to the Block handling
1331 API. lzma_block.version has to be initialized even for
1332 lzma_block_header_decode(). Code that doesn't do it will work
1333 for now, but might break in the future, which makes this API
1334 change easy to miss.
1336 * The major soname has been bumped to 5.0.0. liblzma API and ABI
1337 are now stable, so the need to recompile programs linking against
1338 liblzma shouldn't arise soon.