Jia Tan [Tue, 21 Nov 2023 12:56:55 +0000 (20:56 +0800)]
Build: Change --enable-ifunc handling.
Some compilers support __attribute__((__ifunc__())) even though the
dynamic linker does not. The compiler is able to create the binary
but it will fail on startup. So it is not enough to just test if
the attribute is supported.
The default value for enable_ifunc is now auto, which will attempt
to compile a program using __attribute__((__ifunc__())). There are
additional checks in this program if glibc is being used or if it
is running on FreeBSD.
Setting --enable-ifunc will skip this test and always enable
__attribute__((__ifunc__())), even if is not supported.
Jia Tan [Thu, 23 Nov 2023 14:04:35 +0000 (22:04 +0800)]
xz: Create separate is_tty() function.
The new is_tty() will report if a file descriptor is a terminal or not.
On POSIX systems, it is a wrapper around isatty(). However, the native
Windows implementation of isatty() will return true for all character
devices, not just terminals. So is_tty() has a special case for Windows
so it can use alternative Windows API functions to determine if a file
descriptor is a terminal.
This fixes a bug with MSVC and MinGW-w64 builds that refused to read from
or write to non-terminal character devices because xz thought it was a
terminal. For instance:
xz foo -c > /dev/null
would fail because /dev/null was assumed to be a terminal.
Jia Tan [Fri, 17 Nov 2023 12:35:11 +0000 (20:35 +0800)]
Tests: Create test_suffix.sh.
This tests some complicated interactions with the --suffix= option.
The suffix option must be used with --format=raw, but can optionally
be used to override the default .xz suffix.
This test also verifies some recent bugs have been correctly solved
and to hopefully avoid further regressions in the future.
Jia Tan [Fri, 17 Nov 2023 12:19:26 +0000 (20:19 +0800)]
xz: Fix a bug with --files and --files0 in raw mode without a suffix.
The following command caused a segmentation fault:
xz -Fraw --lzma1 --files=foo
when foo was a valid file. The usage of --files or --files0 was not
being checked when compressing or decompressing in raw mode without a
suffix. The suffix checking code was meant to validate that all files
to be processed are "-" (if not writing to standard out), meaning the
data is only coming from standard in. In this case, there were no file
names to check since --files and --files0 store their file name in a
different place.
Later code assumed the suffix was set and caused a segmentation fault.
Now, the above command results in an error.
Jia Tan [Wed, 15 Nov 2023 15:40:13 +0000 (23:40 +0800)]
xz: Refactor suffix test with raw format.
The previous version set opt_stdout, but this caused an issue with
copying an input file to standard out when decompressing an unknown file
type. The following needs to result in an error:
echo foo | xz -df
since -c, --stdout is not used. This fixes the previous error by not
setting opt_stdout.
Jia Tan [Tue, 14 Nov 2023 12:27:46 +0000 (20:27 +0800)]
xz: Move suffix check after stdout mode is detected.
This fixes a bug introduced in cc5aa9ab138beeecaee5a1e81197591893ee9ca0
when the suffix check was initially moved. This caused a situation that
previously worked:
echo foo | xz -Fraw --lzma1 | wc -c
to fail because the old code knew that this would write to standard out
so a suffix was not needed.
Jia Tan [Tue, 14 Nov 2023 12:27:04 +0000 (20:27 +0800)]
xz: Detect when all data will be written to standard out earlier.
If the -c, --stdout argument is not used, then we can still detect when
the data will be written to standard out if all of the provided
filenames are "-" (denoting standard in) or if no filenames are
provided.
Lasse Collin [Tue, 31 Oct 2023 19:41:09 +0000 (21:41 +0200)]
liblzma: Fix compilation of fastpos_tablegen.c.
The macro lzma_attr_visibility_hidden has to be defined to make
fastpos.h usable. The visibility attribute is irrelevant to
fastpos_tablegen.c so simply #define the macro to an empty value.
fastpos_tablegen.c is never built by the included build systems
and so the problem wasn't noticed earlier. It's just a standalone
program for generating fastpos_table.c.
Fixes: https://github.com/tukaani-project/xz/pull/69
Thanks to GitHub user Jamaika1.
Lasse Collin [Sun, 22 Oct 2023 14:08:39 +0000 (17:08 +0300)]
liblzma: #define lzma_attr_visibility_hidden in common.h.
In ELF shared libs:
-fvisibility=hidden affects definitions of symbols but not
declarations.[*] This doesn't affect direct calls to functions
inside liblzma as a linker can replace a call to lzma_foo@plt
with a call directly to lzma_foo when -fvisibility=hidden is used.
[*] It has to be like this because otherwise every installed
header file would need to explictly set the symbol visibility
to default.
When accessing extern variables that aren't defined in the
same translation unit, compiler assumes that the variable has
the default visibility and thus indirection is needed. Unlike
function calls, linker cannot optimize this.
Using __attribute__((__visibility__("hidden"))) with the extern
variable declarations tells the compiler that indirection isn't
needed because the definition is in the same shared library.
About 15+ years ago, someone told me that it would be good if
the CRC tables would be defined in the same translation unit
as the C code of the CRC functions. While I understood that it
could help a tiny amount, I didn't want to change the code because
a separate translation unit for the CRC tables was needed for the
x86 assembly code anyway. But when visibility attributes are
supported, simply marking the extern declaration with the
hidden attribute will get identical result. When there are only
a few affected variables, this is trivial to do. I wish I had
understood this back then already.
Lasse Collin [Sat, 30 Sep 2023 19:54:28 +0000 (22:54 +0300)]
liblzma: Refer to MinGW-w64 instead of MinGW in the API headers.
MinGW (formely a MinGW.org Project, later the MinGW.OSDN Project
at <https://osdn.net/projects/mingw/>) has GCC 9.2.0 as the
most recent GCC package (released 2021-02-02). The project might
still be alive but majority of people have switched to MinGW-w64.
Thus it seems clearer to refer to MinGW-w64 in our API headers too.
Building with MinGW is likely to still work but I haven't tested it
in the recent years.
Lasse Collin [Fri, 29 Sep 2023 23:14:25 +0000 (02:14 +0300)]
CMake: Generate and install liblzma.pc if not using MSVC.
Autotools based build uses -pthread and thus adds it to Libs.private
in liblzma.pc. CMake doesn't use -pthread at all if pthread functions
are available in libc so Libs.private doesn't get -pthread either.
Lasse Collin [Fri, 29 Sep 2023 17:46:11 +0000 (20:46 +0300)]
liblzma: Add Cflags.private to liblzma.pc.in for MSYS2.
It properly adds -DLZMA_API_STATIC when compiling code that
will be linked against static liblzma. Having it there on
systems other than Windows does no harm.
Lasse Collin [Wed, 25 Oct 2023 16:13:25 +0000 (19:13 +0300)]
Build: Detect -fsanitize= in CFLAGS and incompatible build options.
Now configure will fail if -fsanitize= is found in CFLAGS
and sanitizer-incompatible ifunc or Landlock sandboxing
would be used. These are incompatible with one or more sanitizers.
It's simpler to reject all -fsanitize= uses instead of trying to
pass those that might not cause problems.
CMake-based build was updated similarly. It lets the configuration
finish (SEND_ERROR instead of FATAL_ERROR) so that both error
messages can be seen at once.
Jia Tan [Mon, 23 Oct 2023 16:50:08 +0000 (00:50 +0800)]
CI: Disable sandboxing in fsanitize=address,undefined job.
The sandboxing on Linux now supports Landlock, which restricts all
supported filesystem actions after xz opens the files it needs. The
sandbox is only enabled when one file is input and we are writing to
standard out. With fsanitize=address,undefined, the instrumentation
needs to read additional files after the sandbox is in place. This
forces all xz based test to fail, so the sandbox must instead be
disabled.
Lasse Collin [Wed, 11 Oct 2023 16:47:44 +0000 (19:47 +0300)]
CMake: Don't shadow the cache entry ENABLE_THREADS with a normal variable.
Using set(ENABLE_THREADS "posix") is confusing because it sets
a new normal variable and leaves the cache entry with the same
name unchanged. The intent wasn't to change the cache entry so
this switches to a different variable name.
Lasse Collin [Mon, 9 Oct 2023 17:59:24 +0000 (20:59 +0300)]
CMake: Use FATAL_ERROR if user-supplied options aren't understood.
This way typos are caught quickly and compounding error messages
are avoided (a single typo could cause more than one error).
This keeps using SEND_ERROR when the system is lacking a feature
(like threading library or sandboxing method). This way the whole
configuration log will be generated in case someone wishes to
report a problem upstream.
Lasse Collin [Mon, 9 Oct 2023 15:13:08 +0000 (18:13 +0300)]
Simplify detection of Capsicum support.
This removes support for FreeBSD 10.0 and 10.1 which used
<sys/capability.h> instead of <sys/capsicum.h>. Support for
FreeBSD 10.1 ended on 2016-12-31. So now FreeBSD >= 10.2 is
required to enable Capsicum support.
This also removes support for Capsicum on Linux (libcaprights)
which seems to have been unmaintained since 2017 and Linux 4.11:
https://github.com/google/capsicum-linux
Lasse Collin [Sat, 23 Sep 2023 00:23:32 +0000 (03:23 +0300)]
Build: Check for clock_gettime() even if not using POSIX threads.
See the new comment in the code.
This also makes the check for clock_gettime() run with MinGW-w64
with which we don't want to use clock_gettime(). The previous
commit already took care of this situation.
Lasse Collin [Sun, 24 Sep 2023 19:58:53 +0000 (22:58 +0300)]
xz/Windows: Ensure that clock_gettime() isn't used with MinGW-w64.
This commit alone doesn't change anything in the real-world:
- configure.ac currently checks for clock_gettime() only
when using pthreads.
- CMakeLists.txt doesn't check for clock_gettime() on Windows.
So clock_gettime() wasn't used with MinGW-w64 before either.
clock_gettime() provides monotonic time and it's better than
gettimeofday() in this sense. But clock_gettime() is defined
in winpthreads, and liblzma or xz needs nothing else from
winpthreads. By avoiding clock_gettime(), we avoid the dependency on
libwinpthread-1.dll or the need to link against the static version.
As a bonus, GetTickCount64() and MinGW-w64's gettimeofday() can be
faster than clock_gettime(CLOCK_MONOTONIC, &tv). The resolution
is more than good enough for the progress indicator in xz.
Jia Tan [Fri, 20 Oct 2023 11:17:46 +0000 (19:17 +0800)]
liblzma: Move is_clmul_supported() back to crc_common.h.
This partially reverts creating crc_clmul.c
(8c0f9376f58c0696d5d6719705164d35542dd891) where is_clmul_supported()
was moved, extern'ed, and renamed to lzma_is_clmul_supported(). This
caused a problem when the function call to lzma_is_clmul_supported()
results in a call through the PLT. ifunc resolvers run very early in
the dynamic loading sequence, so the PLT may not be setup properly at
this point. Whether the PLT is used or not for
lzma_is_clmul_supported() depened upon the compiler-toolchain used and
flags.
In liblzma compiled with GCC, for instance, GCC will go through the PLT
for function calls internal to liblzma if the version scripts and
symbol visibility hiding are not used. If lazy-binding is disabled,
then it would have made any program linked with liblzma fail during
dynamic loading in the ifunc resolver.
Jia Tan [Thu, 19 Oct 2023 08:23:32 +0000 (16:23 +0800)]
Build: Remove check for COND_CHECK_CRC32 in check/Makefile.inc.
Currently crc32 is always enabled, so COND_CHECK_CRC32 must always be
set. Because of this, it makes the recent change to conditionally
compile check/crc_clmul.c appear wrong since that file has CLMUL
implementations for both CRC32 and CRC64.
Jia Tan [Thu, 19 Oct 2023 08:09:01 +0000 (16:09 +0800)]
CMake: Add ALLOW_CLMUL_CRC option to enable/disable CLMUL.
The option is enabled by default, but will only be visible to a user
listing cache variables or using a CMake GUI application if the
immintrin.h header file is found.
This mirrors our Autotools build --disable-clmul-crc functionality.
Jia Tan [Wed, 18 Oct 2023 16:22:50 +0000 (00:22 +0800)]
liblzma: Fix -fsanitize=address failure with crc_clmul functions.
After forcing crc_simd_body() to always be inlined it caused
-fsanitize=address to fail for lzma_crc32_clmul() and
lzma_crc64_clmul(). The __no_sanitize_address__ attribute was added
to lzma_crc32_clmul() and lzma_crc64_clmul(), but not removed from
crc_simd_body(). ASAN and inline functions behavior has changed over
the years for GCC specifically, so while strictly required we will
keep __attribute__((__no_sanitize_address__)) on crc_simd_body() in
case this becomes a requirement in the future.
Older GCC versions refuse to inline a function with ASAN if the
caller and callee do not agree on sanitization flags
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89124#c3). If the
function was forced to be inlined, it will not compile if the callee
function has __no_sanitize_address__ but the caller doesn't.
Lasse Collin [Sat, 23 Sep 2023 20:28:48 +0000 (23:28 +0300)]
Build: Enabled unaligned access by default on PowerPC64LE and some RISC-V.
PowerPC64LE wasn't tested but it seems like a safe change.
POWER8 supports unaligned access in little endian mode. Testing
on godbolt.org shows that GCC uses unaligned access by default.
The RISC-V macro __riscv_misaligned_fast is very new and not
in any stable compiler release yet.
Documentation in INSTALL was updated to match.
Documentation about an autodetection bug when using ARM64 GCC
with -mstrict-align was added to INSTALL.
Lasse Collin [Sat, 14 Oct 2023 14:56:59 +0000 (17:56 +0300)]
tuklib_integer: Revise unaligned reads and writes on strict-align archs.
In XZ Utils context this doesn't matter much because
unaligned reads and writes aren't used in hot code
when TUKLIB_FAST_UNALIGNED_ACCESS isn't #defined.
Jia Tan [Wed, 18 Oct 2023 11:57:10 +0000 (19:57 +0800)]
liblzma: Set the MSVC optimization fix to only cover lzma_crc64_clmul().
After testing a 32-bit Release build on MSVC, only lzma_crc64_clmul()
has the bug. crc_simd_body() and lzma_crc32_clmul() do not need the
optimizations disabled.
Jia Tan [Wed, 18 Oct 2023 14:50:25 +0000 (22:50 +0800)]
liblzma: Add the crc_always_inline macro to crc_simd_body().
Forcing this to be inline has a significant speed improvement at the
cost of a few repeated instructions. The compilers tested on did not
inline this function since it is large and is used twice in the same
translation unit.
Jia Tan [Wed, 18 Oct 2023 14:48:19 +0000 (22:48 +0800)]
liblzma: Create crc_always_inline macro.
This macro must be used instead of the inline keyword. On MSVC, it is
a replacement for __forceinline which is an MSVC specific keyword that
should not be used with inline (it will issue a warning if it is).
It does not use a build system check to determine if
__attribute__((__always_inline__)) since all compilers that can use
CLMUL extensions (except the special case for MSVC) should support this
attribute. If this assumption is incorrect then it will result in a bug
report instead of silently producing slow code.
Jia Tan [Sat, 14 Oct 2023 05:23:23 +0000 (13:23 +0800)]
liblzma: Refactor CRC comments.
A detailed description of the three dispatch methods was added. Also,
duplicated comments now only appear in crc32_fast.c or were removed from
both crc32_fast.c and crc64_fast.c if they appeared in crc_clmul.c.
Jia Tan [Sat, 14 Oct 2023 04:17:57 +0000 (12:17 +0800)]
liblzma: Create crc_clmul.c.
Both crc32_clmul() and crc64_clmul() are now exported from
crc32_clmul.c as lzma_crc32_clmul() and lzma_crc64_clmul(). This
ensures that is_clmul_supported() (now lzma_is_clmul_supported()) is
not duplicated between crc32_fast.c and crc64_fast.c.
Also, it encapsulates the complexity of the CLMUL implementations into a
single file and reduces the complexity of crc32_fast.c and crc64_fast.c.
Before, CLMUL code was present in crc32_fast.c, crc64_fast.c, and
crc_common.h.
During the conversion, various cleanups were applied to code (thanks to
Lasse Collin) including:
- Require using semicolons with MASK_/L/H/LH macros.
- Variable typing and const handling improvements.
- Improvements to comments.
- Fixes to the pragmas used.
- Removed unneeded variables.
- Whitespace improvements.
- Fixed CRC_USE_GENERIC_FOR_SMALL_INPUTS handling.
- Silenced warnings and removed the need for some #pragmas
CI: Bump and ref actions by commit SHA in windows-ci.yml
Referencing actions by commit SHA in GitHub workflows guarantees you are using an immutable version. Actions referenced by tags and branches are more vulnerable to attacks, such as the tag being moved to a malicious commit or a malicious commit being pushed to the branch.
It's important to make sure the SHA's are from the original repositories and not forks.
Referencing actions by commit SHA in GitHub workflows guarantees you are using an immutable version. Actions referenced by tags and branches are more vulnerable to attacks, such as the tag being moved to a malicious commit or a malicious commit being pushed to the branch.
It's important to make sure the SHA's are from the original repositories and not forks.
Lasse Collin [Fri, 6 Oct 2023 15:55:57 +0000 (18:55 +0300)]
CMake/Windows: Fix when the windres workaround is applied.
CMake doesn't set WIN32 on CYGWIN but the workaround is
probably needed on Cygwin too. Same for MSYS and MSYS2.
The workaround must not be used with Clang that is acting in
MSVC mode. This fixes it by checking for the known environments
that need the workaround instead of using "NOT MSVC".
Jia Tan [Fri, 29 Sep 2023 14:11:54 +0000 (22:11 +0800)]
CI: Disable CLANG64 MSYS2 environment until bug is resolved.
lld 17.0.1 searches for libraries to link first in the toolchain
directories before the local directory when building. The is a problem
for us because liblzma.a is installed in MSYS2 CLANG64 by default and
xz.exe will thus use the installed library instead of the one being
built.
This causes tests to fail when they are expecting features to be
disabled. More importantly, it will compile xz.exe with an incorrect
liblzma and could cause unexpected behavior by being unable to update
liblzma code in static builds. The CLANG64 environment can be tested
again once this is fixed.
Link to bug: https://github.com/llvm/llvm-project/issues/67779.
Jia Tan [Fri, 29 Sep 2023 12:14:39 +0000 (20:14 +0800)]
CMake: Rename xz and man page symlink custom targets.
The Ninja Generator for CMake cannot have a custom target and its
BYPRODUCTS have the same name. This has prevented Ninja builds on
Unix-like systems since the xz symlinks were introduced in 80a1a8bb838842a2be343bd88ad1462c21c5e2c9.
Lasse Collin [Tue, 26 Sep 2023 18:47:13 +0000 (21:47 +0300)]
liblzma: Update a comment.
The C standards don't allow an empty translation unit which can be
avoided by declaring something, without exporting any symbols.
When I committed f644473a211394447824ea00518d0a214ff3f7f2 I had
a feeling that some specific toolchain somewhere didn't like
empty object files (assembler or maybe "ar" complained) but
I cannot find anything to confirm this now. Quite likely I
remembered nonsense. I leave this here as a note to my future self. :-)
Jia Tan [Tue, 26 Sep 2023 16:02:11 +0000 (00:02 +0800)]
liblzma: Avoid compiler warning without creating extra symbol.
When the generic fast crc64 method is used, then we omit
lzma_crc64_table[][]. Similar to d9166b52cf3458a4da3eb92224837ca8fc208d79, we can avoid compiler warnings
with -Wempty-translation-unit (Clang) or -pedantic (GCC) by creating a
never used typedef instead of an extra symbol.
Lasse Collin [Tue, 26 Sep 2023 12:00:43 +0000 (15:00 +0300)]
Build: Fix __attribute__((ifunc(...))) detection with clang -Wall.
Now if user-supplied CFLAGS contains -Wall -Wextra -Wpedantic
the two checks that need -Werror will still work.
At CMake side there is add_compile_options(-Wall -Wextra)
but it didn't affect the -Werror tests. So with both Autotools
and CMake only user-supplied CFLAGS could make the checks fail
when they shouldn't.
This is not a full fix as things like -Wunused-macros in
user-supplied CFLAGS will still cause problems with both
GCC and Clang.
Lasse Collin [Tue, 26 Sep 2023 10:14:37 +0000 (13:14 +0300)]
Build: Silence two Autoconf warnings.
There were two uses of AC_COMPILE_IFELSE that didn't use
AC_LANG_SOURCE and Autoconf warned about these. The omission
had been intentional but it turned out that this didn't do
what I thought it would.
Autoconf 2.71 manual gives an impression that AC_LANG_SOURCE
inserts all #defines that have been made with AC_DEFINE so
far (confdefs.h). The idea was that omitting AC_LANG_SOURCE
would mean that only the exact code included in the
AC_COMPILE_IFELSE call would be compiled.
With C programs this is not true: the #defines get added without
AC_LANG_SOURCE too. There seems to be no neat way to avoid this.
Thus, with the C language at least, adding AC_LANG_SOURCE makes
no other difference than silencing a warning from Autoconf. The
generated "configure" remains identical. (Docs of AC_LANG_CONFTEST
say that the #defines have been inserted since Autoconf 2.63b and
that AC_COMPILE_IFELSE uses AC_LANG_CONFTEST. So the behavior is
documented if one also reads the docs of macros that one isn't
calling directly.)
Any extra code, including #defines, can cause problems for
these two tests because these tests must use -Werror.
CC=clang CFLAGS=-Weverything is the most extreme example.
It enables -Wreserved-macro-identifier which warns about
#define __EXTENSIONS__ 1 because it begins with two underscores.
It's possible to write a test file that passes -Weverything but
it becomes impossible when Autoconf inserts confdefs.h.
So this commit adds AC_LANG_SOURCE to silence Autoconf warnings.
A different solution is needed for -Werror tests.
Jia Tan [Fri, 22 Sep 2023 19:27:00 +0000 (03:27 +0800)]
lib: Update getopt.in.h from Gnulib with modifications.
We can still avoid modifying the contents of this file during
configuration to simplify the build systems. Gnulib added replacements
for inclusions guards for Cygwin. Cygwin should not need getopt_long
replacement so this feature can be omitted.
<unistd.h> is conditionally included to avoid MSVC since it is not
available.
The definition for _GL_ARG_NONNULL was also copied into this file from
Gnulib since this stage is usually done during gnulib-tool.
Jia Tan [Fri, 22 Sep 2023 16:18:56 +0000 (00:18 +0800)]
lib: Add getopt-cdefs.h for getopt_long update.
This was modified slightly from Gnulib. In Gnulib, it expects the
@HAVE_SYS_CDEFS_H@ to be replaced. Instead, we can set HAVE_SYS_CDEFS_H
on systems that have it and avoid copying another file into the build
directory. Since we are not using gnulib-tool, copying extra files
requires extra build system updates (and special handling with CMake) so
we should avoid when possible.
Jia Tan [Fri, 22 Sep 2023 16:15:25 +0000 (00:15 +0800)]
lib: Copy new header files from Gnulib without modification.
The getopt related files have changed from Gnulib by splitting up
getopt.in.h into more modular header files. We could have kept
everything in just getopt.in.h, but this will help us continue to update
in the future.