The bug was in the single-threaded .xz Stream encoder
in the code that is used for both re-initialization and for
lzma_filters_update(). To trigger it, an application had
to either re-initialize an existing encoder instance with
lzma_stream_encoder() or use lzma_filters_update(), and
then one of the 1-4 tiny allocations in lzma_filters_copy()
(called from stream_encoder_update()) must fail. An error
was correctly reported but the encoder state was corrupted.
This is related to the recent fix in
f8ee61e74e which is good but
it wasn't enough to fix the main problem in stream_encoder.c.
The encoder doesn't support dictionary sizes larger than 1536 MiB.
This is validated, for example, when calculating the memory usage
via lzma_raw_encoder_memusage(). It is also enforced by the LZ
part of the encoder initialization. However, LZMA encoder with
LZMA_MODE_NORMAL did an unsafe calculation with dict_size before
such validation and that results in an infinite loop if dict_size
was 2 << 30 or greater.
This reverts commit 177bdc922c
and also does equivalent change to arm64.c.
Now that ARM64 filter will use lzma_options_bcj, this change
is not needed anymore.
It also works on E2K as it supports these intrinsics.
On x86-64 runtime detection is used so the code keeps working on
older processors too. A CLMUL-only build can be done by using
-msse4.1 -mpclmul in CFLAGS and this will reduce the library
size since the generic implementation and its 8 KiB lookup table
will be omitted.
On 32-bit x86 this isn't used by default for now because by default
on 32-bit x86 the separate assembly file crc64_x86.S is used.
If --disable-assembler is used then this new CLMUL code is used
the same way as on 64-bit x86. However, a CLMUL-only build
(-msse4.1 -mpclmul) won't omit the 8 KiB lookup table on
32-bit x86 due to a currently-missing check for disabled
assembler usage.
The configure.ac check should be such that the code won't be
built if something in the toolchain doesn't support it but
--disable-clmul-crc option can be used to unconditionally
disable this feature.
CLMUL speeds up decompression of files that have compressed very
well (assuming CRC64 is used as a check type). It is know that
the CLMUL code is significantly slower than the generic code for
tiny inputs (especially 1-8 bytes but up to 16 bytes). If that
is a real-world problem then there is already a commented-out
variant that uses the generic version for small inputs.
Thanks to Ilya Kurdyukov for the original patch which was
derived from a white paper from Intel [1] (published in 2009)
and public domain code from [2] (released in 2016).
[1] https://www.intel.com/content/dam/www/public/us/en/documents/white-papers/fast-crc-computation-generic-polynomials-pclmulqdq-paper.pdf
[2] https://github.com/rawrunprotected/crc
This uses it for CRC table initializations when using --disable-small.
It avoids mythread_once() overhead. It also means that then
--disable-small --disable-threads is thread-safe if this attribute
is supported.
__SSE2__ is the correct macro for SSE2 support with GCC, Clang,
and ICC. __SSE2_MATH__ means doing floating point math with SSE2
instead of 387. Often the latter macro is defined if the first
one is but it was still a bug.
Support for format version 0 was removed from lzip 1.18 for some
reason. .lz format version 0 files are rare (and old) but some
source packages were released in this format, and some people might
have personal files in this format too. It's very little extra code
to support it along side format version 1 so this commits adds
support for both.
The Sync Flush marker extentension to the original .lz format
version 1 isn't supported. It would require changes to the
LZMA decoder itself. Such files are very rare anyway.
See the API doc for lzma_lzip_decoder() for more details about
the .lz format support.
Thanks to Michał Górny for the original patch.
This affects lzma_memusage() and lzma_memlimit_set() when used
with the threaded decompressor. Now all allocations are reported
by lzma_memusage() (so it's not misleading) and lzma_memlimit_set()
cannot lower the limit below that value.
The alternative would have been to allow lowering the limit if
doing so is possible by freeing the cached memory but since
the primary use case of lzma_memlimit_set() is to increase
memlimit after LZMA_MEMLIMIT_ERROR this simple approach
was selected.
The cached memory was always included when enforcing
the memory usage limit while decoding.
Thanks to Jia Tan.
When encoders were disabled and threading enabled, outqueue.c and
outqueue.h were not compiled. The multi threaded decoder required
these files, so compilation failed.
The documentation states LZMA_PROG_ERROR can be returned from
lzma_index_cat. Previously, lzma_index_cat could not return
LZMA_PROG_ERROR. Now, the validation is similar to
lzma_index_append, which does a NULL check on the index
parameter.
The check type of the last Stream in dest was never copied to
dest->checks (the code tried to copy it but it was done too late).
This meant that the value returned by lzma_index_checks() would
only include the check type of the last Stream when multiple
lzma_indexes had been concatenated.
In xz --list this meant that the summary would only list the
check type of the last Stream, so in this sense this was only
a visual bug. However, it's possible that some applications
use this information for purposes other than merely showing
it to the users in an informational message. I'm not aware of
such applications though and it's quite possible that such
applications don't exist.
Regular streamed decompression in xz or any other application
doesn't use lzma_index_cat() and so this bug cannot affect them.
If lzma_code() returns LZMA_MEMLIMIT_ERROR it is now possible
to use lzma_memlimit_set() to increase the limit and continue
decoding. This was supposed to work from the beginning but
there was a bug. With other decoders (.lzma or threaded .xz)
this already worked correctly.
That is, the Filter ID will be changed once the design is final.
The current version will be removed. So files created with the
tempoary Filter ID won't be supported in the future.
lzma_stream_encoder() and lzma_stream_encoder_mt() always assumed
this. Before this patch, failing lzma_filters_copy() could result
in free(invalid_pointer) or invalid memory reads in stream_encoder.c
or stream_encoder_mt.c.
To trigger this, allocating memory for a filter options structure
has to fail. These are tiny allocations so in practice they very
rarely fail.
Certain badness in the filter chain array could also make
lzma_filters_copy() fail but both stream_encoder.c and
stream_encoder_mt.c validate the filter chain before
trying to copy it, so the crash cannot occur this way.
The documentation in src/liblzma/api/lzma/index.h suggests that
both the unpadded (compressed) size and the uncompressed size
are checked for overflow, but only the unpadded size was checked.
The uncompressed check is done first since that is more likely to
occur than the unpadded or index field size overflows.
RHEL/CentOS 7 shipped with 5.1.2alpha, including the threaded
encoder that is behind #ifdef LZMA_UNSTABLE in the API headers.
In 5.1.2alpha these symbols are under XZ_5.1.2alpha in liblzma.map.
API/ABI compatibility tracking isn't done between development
releases so newer releases didn't have XZ_5.1.2alpha anymore.
Later RHEL/CentOS 7 updated xz to 5.2.2 but they wanted to keep
the exported symbols compatible with 5.1.2alpha. After checking
the ABI changes it turned out that >= 5.2.0 ABI is backward
compatible with the threaded encoder functions from 5.1.2alpha
(but not vice versa as fixes and extensions to these functions
were made between 5.1.2alpha and 5.2.0).
In RHEL/CentOS 7, XZ Utils 5.2.2 was patched with
xz-5.2.2-compat-libs.patch to modify liblzma.map:
- XZ_5.1.2alpha was added with lzma_stream_encoder_mt and
lzma_stream_encoder_mt_memusage. This matched XZ Utils 5.1.2alpha.
- XZ_5.2 was replaced with XZ_5.2.2. It is clear that this was
an error; the intention was to keep using XZ_5.2 (XZ_5.2.2
has never been used in XZ Utils). So XZ_5.2.2 lists all
symbols that were listed under XZ_5.2 before the patch.
lzma_stream_encoder_mt and _mt_memusage are included too so
they are listed both here and under XZ_5.1.2alpha.
The patch didn't add any __asm__(".symver ...") lines to the .c
files. Thus the resulting liblzma.so exports the threaded encoder
functions under XZ_5.1.2alpha only. Listing the two functions
also under XZ_5.2.2 in liblzma.map has no effect without
matching .symver lines.
The lack of XZ_5.2 in RHEL/CentOS 7 means that binaries linked
against unpatched XZ Utils 5.2.x won't run on RHEL/CentOS 7.
This is unfortunate but this alone isn't too bad as the problem
is contained within RHEL/CentOS 7 and doesn't affect users
of other distributions. It could also be fixed internally in
RHEL/CentOS 7.
The second problem is more serious: In XZ Utils 5.2.2 the API
headers don't have #ifdef LZMA_UNSTABLE for obvious reasons.
This is true in RHEL/CentOS 7 version too. Thus now programs
using new APIs can be compiled without an extra #define. However,
the programs end up depending on symbol version XZ_5.1.2alpha
(and possibly also XZ_5.2.2) instead of XZ_5.2 as they would
with an unpatched XZ Utils 5.2.2. This means that such binaries
won't run on other distributions shipping XZ Utils >= 5.2.0 as
they don't provide XZ_5.1.2alpha or XZ_5.2.2; they only provide
XZ_5.2 (and XZ_5.0). (This includes RHEL/CentOS 8 as the patch
luckily isn't included there anymore with XZ Utils 5.2.4.)
Binaries built by RHEL/CentOS 7 users get distributed and then
people wonder why they don't run on some other distribution.
Seems that people have found out about the patch and been copying
it to some build scripts, seemingly curing the symptoms but
actually spreading the illness further and outside RHEL/CentOS 7.
The ill patch seems to be from late 2016 (RHEL 7.3) and in 2017 it
had spread at least to EasyBuild. I heard about the events only
recently. :-(
This commit splits liblzma.map into two versions: one for
GNU/Linux and another for other OSes that can use symbol versioning
(FreeBSD, Solaris, maybe others). The Linux-specific file and the
matching additions to .c files add full compatibility with binaries
that have been built against a RHEL/CentOS-patched liblzma. Builds
for OSes other than GNU/Linux won't get the vaccine as they should
be immune to the problem (I really hope that no build script uses
the RHEL/CentOS 7 patch outside GNU/Linux).
The RHEL/CentOS compatibility symbols XZ_5.1.2alpha and XZ_5.2.2
are intentionally put *after* XZ_5.2 in liblzma_linux.map. This way
if one forgets to #define HAVE_SYMBOL_VERSIONS_LINUX when building,
the resulting liblzma.so.5 will have lzma_stream_encoder_mt@@XZ_5.2
since XZ_5.2 {...} is the first one that lists that function.
Without HAVE_SYMBOL_VERSIONS_LINUX @XZ_5.1.2alpha and @XZ_5.2.2
will be missing but that's still a minor problem compared to
only having lzma_stream_encoder_mt@@XZ_5.1.2alpha!
The "local: *;" line was moved to XZ_5.0 so that it doesn't need
to be moved around. It doesn't matter where it is put.
Having two similar liblzma_*.map files is a bit silly as it is,
at least for now, easily possible to generate the generic one
from the Linux-specific file. But that adds extra steps and
increases the risk of mistakes when supporting more than one
build system. So I rather maintain two files in parallel and let
validate_map.sh check that they are in sync when "make mydist"
is run.
This adds .symver lines for lzma_stream_encoder_mt@XZ_5.2.2 and
lzma_stream_encoder_mt_memusage@XZ_5.2.2 even though these
weren't exported by RHEL/CentOS 7 (only @@XZ_5.1.2alpha was
for these two). I added these anyway because someone might
misunderstand the RHEL/CentOS 7 patch and think that @XZ_5.2.2
(@@XZ_5.2.2) versions were exported too.
At glance one could suggest using __typeof__ to copy the function
prototypes when making aliases. However, this doesn't work trivially
because __typeof__ won't copy attributes (lzma_nothrow, lzma_pure)
and it won't change symbol visibility from hidden to default (done
by LZMA_API()). Attributes could be copied with __copy__ attribute
but that needs GCC 9 and a fallback method would be needed anyway.
This uses __symver__ attribute with GCC >= 10 and
__asm__(".symver ...") with everything else. The attribute method
is required for LTO (-flto) support with GCC. Using -flto with
GCC older than 10 is now broken on GNU/Linux and will not be fixed
(can silently result in a broken liblzma build that has dangerously
incorrect symbol versions). LTO builds with Clang seem to work
with the traditional __asm__(".symver ...") method.
Thanks to Boud Roukema for reporting the problem and discussing
the details and testing the fix.
It will now return LZMA_DATA_ERROR (not LZMA_OK or LZMA_BUF_ERROR)
if LZMA_FINISH is used and there isn't enough input to finish
decoding the Block Header or the Block. The use of LZMA_DATA_ERROR
is simpler and the less risky than LZMA_BUF_ERROR but this might
be changed before 5.4.0.
Turns out that this is needed for .lzma files as the spec in
LZMA SDK says that end marker may be present even if the size
is stored in the header. Such files are rare but exist in the
real world. The code in liblzma is so old that the spec didn't
exist in LZMA SDK back then and I had understood that such
files weren't possible (the lzma tool in LZMA SDK didn't
create such files).
This modifies the internal API so that LZMA decoder can be told
if EOPM is allowed even when the uncompressed size is known.
It's allowed with .lzma and not with other uses.
Thanks to Karl Beldan for reporting the problem.
In most cases if the input file is corrupt the application won't
care about the uncompressed content at all. With this new flag
the threaded decoder will return an error as soon as any thread
has detected an error; it won't wait to copy out the data before
the location of the error.
I don't plan to use this in xz to keep the behavior consistent
between single-threaded and multi-threaded modes.
This makes it possible to call lzma_code() in a loop that only
reads new input when lzma_code() didn't fill the output buffer
completely. That isn't the calling style suggested by the
liblzma example program 02_decompress.c so perhaps the usefulness
of this feature is limited.
Also, it is possible to write such a loop so that it works
with the single-threaded decoder but not with the threaded
decoder even after this commit, or so that it works only if
lzma_mt.timeout = 0.
The zlib tutorial <https://zlib.net/zlib_how.html> is a well-known
example of a loop where more input is read only when output isn't
full. Porting this as is to liblzma would work with the
single-threaded decoder (if LZMA_CONCATENATED isn't used) but it
wouldn't work with threaded decoder even after this commit because
the loop assumes that no more output is possible when it cannot
read more input ("if (strm.avail_in == 0) break;"). This cannot
be fixed at liblzma side; the loop has to be modified at least
a little.
I'm adding this in any case because the actual code is simple
and short and should have no harmful side-effects in other
situations.
If a worker thread has consumed all input so far and it's
waiting on thr->cond and then the main thread enables
partial update for that thread, the code used to deadlock.
This commit allows one dummy decoding pass to occur in this
situation which then also does the partial update.
As part of the fix, this moves thr->progress_* updates to
avoid the second thr->mutex locking.
Thanks to Jia Tan for finding, debugging, and reporting the bug.
LZMA_TIMED_OUT is not an error and thus stopping threads on
LZMA_TIMED_OUT breaks the decoder badly.
Thanks to Jia Tan for finding the bug and for the patch.
I realize that this is about a decade late.
Big thanks to Sebastian Andrzej Siewior for the original patch.
I made a bunch of smaller changes but after a while quite a few
things got rewritten. So any bugs in the commit were created by me.
Add lzma_outq_clear_cache2() which may leave one buffer allocated
in the cache.
Add lzma_outq_outbuf_memusage() to get the memory needed for
a single lzma_outbuf. This is now used internally in outqueue.c too.
Track both the total amount of memory allocated and the amount of
memory that is in active use (not in cache).
In lzma_outbuf, allow storing the current input position that
matches the current output position. This way the main thread
can notice when no more output is possible without first providing
more input.
Allow specifying return code for lzma_outq_read() in a finished
lzma_outbuf.
If lzma_index_append() failed (most likely memory allocation failure)
it could have gone unnoticed and the resulting .xz file would have
an incorrect Index. Decompressing such a file would produce the
correct uncompressed data but then an error would occur when
verifying the Index field.
Now it limits the input and output buffer sizes that are
passed to a raw decoder. This way there's no need to check
if the sizes can grow too big or overflow when updating
Compressed Size and Uncompressed Size counts. This also means
that a corrupt file cannot cause the raw decoder to process
useless extra input or output that would exceed the size info
in Block Header (and thus cause LZMA_DATA_ERROR anyway).
More importantly, now the size information is verified more
carefully in case raw decoder returns LZMA_OK. This doesn't
really matter with the current single-threaded .xz decoder
as the errors would be detected slightly later anyway. But
this helps avoiding corner cases in the upcoming threaded
decompressor, and it might help other Block decoder uses
outside liblzma too.
The test files bad-1-lzma2-{9,10,11}.xz test these conditions.
With the single-threaded .xz decoder the only difference is
that LZMA_DATA_ERROR is detected in a difference place now.
Previously lzma_lzma_props_encode() and lzma_lzma2_props_encode()
assumed that the options pointers must be non-NULL because the
with these filters the API says it must never be NULL. It is
good to do these checks anyway.
This broke 32-bit builds due to a pointer type mismatch.
This bug was introduced with the output-size-limited encoding
in 625f4c7c99.
Thanks to huangqinjin for the bug report.
When the uncompressed size is known to be exact, after decompressing
the stream exactly comp_size bytes of input must have been consumed.
This is a minor improvement to error detection.
The caller must still not specify an uncompressed size bigger
than the actual uncompressed size.
As a downside, this now needs the exact compressed size.
Right now this is just a planned extra-compact format for use
in the EROFS file system in Linux. At this point it's possible
that the format will either change or be abandoned and removed
completely.
The special thing about the encoder is that it uses the
output-size-limited encoding added in the previous commit.
EROFS uses fixed-sized blocks (e.g. 4 KiB) to hold compressed
data so the compressors must be able to create valid streams
that fill the given block size.
With this it is possible to encode LZMA1 data without EOPM so that
the encoder will encode as much input as it can without exceeding
the specified output size limit. The resulting LZMA1 stream will
be a normal LZMA1 stream without EOPM. The actual uncompressed size
will be available to the caller via the uncomp_size pointer.
One missing thing is that the LZMA layer doesn't inform the LZ layer
when the encoding is finished and thus the LZ may read more input
when it won't be used. However, this doesn't matter if encoding is
done with a single call (which is the planned use case for now).
For proper multi-call encoding this should be improved.
This commit only adds the functionality for internal use.
Nothing uses it yet.
Before this commit all output queue buffers were allocated as
a single big allocation. Now each buffer is allocated separately
when needed. Used buffers are cached to avoid reallocation
overhead but the cache will keep only one buffer size at a time.
This should make things work OK in the decompression where most
of the time the buffer sizes will be the same but with some less
common files the buffer sizes may vary.
While this should work fine, it's still a bit preliminary
and may even get reverted if it turns out to be useless for
decompression.
When Intel CET is enabled, we need to include <cet.h> in assembly codes
to mark Intel CET support and add _CET_ENDBR to indirect jump targets.
Tested on Intel Tiger Lake under CET enabled Linux.
The comment didn't match the value of RC_SYMBOLS_MAX and the value
itself was slightly larger than actually needed. The only harm
about this was that memory usage was a few bytes larger.
Using the aligned methods requires more care to ensure that
the address really is aligned, so it's nicer if the aligned
methods are prefixed. The next commit will remove the unaligned_
prefix from the unaligned methods which in liblzma are used in
more places than the aligned ones.
LZMA_TIMED_OUT is *internally* used as a value for lzma_ret
enumeration. Previously it was #defined to 32 and cast to lzma_ret.
That way it wasn't visible in the public API, but this was hackish.
Now the public API has eight LZMA_RET_INTERNALx members and
LZMA_TIMED_OUT is #defined to LZMA_RET_INTERNAL1. This way
the code is cleaner overall although the public API has a few
extra mysterious enum members.
I should have always known this but I didn't. Here is an example
as a reminder to myself:
int mycopy(void *dest, void *src, size_t n)
{
memcpy(dest, src, n);
return dest == NULL;
}
In the example, a compiler may assume that dest != NULL because
passing NULL to memcpy() would be undefined behavior. Testing
with GCC 8.2.1, mycopy(NULL, NULL, 0) returns 1 with -O0 and -O1.
With -O2 the return value is 0 because the compiler infers that
dest cannot be NULL because it was already used with memcpy()
and thus the test for NULL gets optimized out.
In liblzma, if a null-pointer was passed to memcpy(), there were
no checks for NULL *after* the memcpy() call, so I cautiously
suspect that it shouldn't have caused bad behavior in practice,
but it's hard to be sure, and the problematic cases had to be
fixed anyway.
Thanks to Jeffrey Walton.
FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION is #defined when liblzma
is being built for fuzz testing.
Most fuzzed inputs would normally get rejected because of incorrect
CRC32 and the actual header decoding code wouldn't get fuzzed.
Disabling CRC32 checks avoids this problem. The fuzzer program
must still use LZMA_IGNORE_CHECK flag to disable verification of
integrity checks of uncompressed data.
The 0 got treated specially in a buggy way and as a result
the function did nothing. The API doc said that 0 was supposed
to return LZMA_PROG_ERROR but it didn't.
Now 0 is treated as if 1 had been specified. This is done because
0 is already used to indicate an error from lzma_memlimit_get()
and lzma_memusage().
In addition, lzma_memlimit_set() no longer checks that the new
limit is at least LZMA_MEMUSAGE_BASE. It's counter-productive
for the Index decoder and was actually needed only by the
auto decoder. Auto decoder has now been modified to check for
LZMA_MEMUSAGE_BASE.
It returned LZMA_PROG_ERROR, which was done to avoid zero as
the limit (because it's a special value elsewhere), but using
LZMA_PROG_ERROR is simply inconvenient and can cause bugs.
The fix/workaround is to treat 0 as if it were 1 byte. It's
effectively the same thing. The only weird consequence is
that then lzma_memlimit_get() will return 1 even when 0 was
specified as the limit.
This fixes a very rare corner case in xz --list where a specific
memory usage limit and a multi-stream file could print the
error message "Internal error (bug)" instead of saying that
the memory usage limit is too low.
Only one definition was visible in a translation unit.
It avoided a few casts and temp variables but seems that
this hack doesn't work with link-time optimizations in compilers
as it's not C99/C11 compliant.
Fixes:
http://www.mail-archive.com/xz-devel@tukaani.org/msg00279.html
This is the sane thing to do. The conflict with OpenSSL
on some OSes and especially that the OS-provided versions
can be significantly slower makes it clear that it was
a mistake to have the external SHA-256 support enabled by
default.
Those who want it can now pass --enable-external-sha256 to
configure. INSTALL was updated with notes about OSes where
this can be a bad idea.
The SHA-256 detection code in configure.ac had some bugs that
could lead to a build failure in some situations. These were
fixed, although it doesn't matter that much now that the
external SHA-256 is disabled by default.
MINIX >= 3.2.0 uses NetBSD's libc and thus has SHA256_Init
in libc instead of libutil. Support for the libutil version
was removed.
When optimizing, GCC can reorder code so that an uninitialized
value gets used in a comparison, which makes Valgrind unhappy.
It doesn't happen when compiled with -O0, which I tend to use
when running Valgrind.
Thanks to Rich Prohaska. I remember this being mentioned long
ago by someone else but nothing was done back then.
People shouldn't rely on the presets when decoding raw streams,
but xz uses the presets as the starting point for raw decoder
options anyway.
lzma_encocder_presets.c was renamed to lzma_presets.c to
make it clear it's not used solely by the encoder code.
lzma_index_dup() calls index_dup_stream() which, in case of
an error, calls index_stream_end() to free memory allocated
by index_stream_init(). However, it illogically didn't
actually free the memory. To make it logical, the tree
handling code was modified a bit in addition to changing
index_stream_end().
Thanks to Evan Nemerson for the bug report.
I know that soname != app version, but I skip AGE=1
in -version-info to make the soname match the liblzma
version anyway. It doesn't hurt anything as long as
it doesn't conflict with library versioning rules.
This way an invalid filter chain is detected at the Stream
encoder initialization instead of delaying it to the first
call to lzma_code() which triggers the initialization of
the actual filter encoder(s).
POSIX supports $< only in inference rules (suffix rules).
Using it elsewhere is a GNU make extension and doesn't
work e.g. with OpenBSD make.
Thanks to Christian Weisgerber for the patch.
Note that this slightly changes how lzma_block_header_decode()
has been documented. Earlier it said that the .version is set
to the lowest required value, but now it says that the .version
field is kept unchanged if possible. In practice this doesn't
affect any old code, because before this commit the only
possible .version was 0.
The Maj macro is used where multiple things are added
together, so making Maj a sum of two expressions allows
some extra freedom for the compiler to schedule the
instructions.
I learned this trick from
<http://www.hackersdelight.org/corres.txt>.
This looks weird because the rotations become sequential,
but it helps quite a bit on both 32-bit and 64-bit x86:
- It requires fewer instructions on two-operand
instruction sets like x86.
- It requires one register less which matters especially
on 32-bit x86.
I hope this doesn't hurt other archs.
I didn't invent this idea myself, but I don't remember where
I saw it first.
This way a branch isn't needed for each operation
to choose between blk0 and blk2, and still the code
doesn't grow as much as it would with full unrolling.
This commit just adds the function. Its uses will be in
separate commits.
This hasn't been tested much yet and it's perhaps a bit early
to commit it but if there are bugs they should get found quite
quickly.
Thanks to Jun I Jin from Intel for help and for pointing out
that string comparison needs to be optimized in liblzma.
This avoids a memzero() call for a newly-allocated memory,
which can be expensive when encoding small streams with
an over-sized dictionary.
To avoid using lzma_alloc_zero() for memory that doesn't
need to be zeroed, lzma_mf.son is now allocated separately,
which requires handling it separately in normalize() too.
Thanks to Vincenzo Innocente for reporting the problem.
Lasse Collin
Jia Tan
Ed Maste
H.J. Lu
Antoine Cœur
Ben Boeckel