* bootstrap.conf: Reference the new "utf8" module.
* cfg.mk: Adjust checks as per mbbuf.
* gl/lib/utf8.c: Include the implementation.
* gl/lib/utf8.h: Add is_utf8_charset() and u8_count().
Note we don't give is_utf8_charset external-linkage
due to its static cache.
* gl/local.mk: Reference the new module files.
* gl/modules/utf8: Add the module definition and dependencies.
* src/cut.c: Use is_utf8_charset() and u8_count().
* src/numfmt.c: Use is_utf8_charset().
* src/system.h: Remove is_utf8_charset() definition.
These are the most common input encodings so worth optimizing for.
Performance was tested with GCC 16 on an i7-5600U with
CFLAGS='-march=native -O3 -flto' with the following setup:
$ yes | head -n10M > sl.in
$ yes $(yes eeeaae | head -n10K | paste -s -d' ') | head -n10K > llw.in
$ yes $(yes eeeaae | head -n9 | paste -s -d' ') | head -n1M > asw.in
$ yes $(yes éééááé | head -n9 | paste -s -d$'\xe2\x80\x83') |
head -n1M > mbw.in
for type in sl llw asw mbw; do
cat $type.in >/dev/null;
for imp in '-before' '-after'; do
echo ============ "${imp:-base}" $type ==============;
for d in -c; do
fields='-f1 -f10 -f100'
test "$d" = "-c" && { fields='-c1 -c10 -c100000'; d=''; }
for f in $fields; do
for loc in C.UTF-8; do
# Skip -b for UTF-8 as no different
test "$loc" = C.UTF-8 && echo "$f" | grep -q -- -b \
&& continue
# Skip multi-byte delimiter for C and not allowed
test "$loc" = C && test $(echo -n "$d" | wc -c) -ge 4 \
&& continue
bin=src/cut${imp}
LC_ALL=$loc $bin $f $d /dev/null 2>/dev/null &&
hyperfine --warmup=2 -m2 -M6 \
"LC_ALL=$loc $bin $f $d $type.in >/dev/null" ||
printf 'Benchmark 1: %s\n unsupported\n\n' \
"LC_ALL=$loc $bin $f $d $type.in >/dev/null"
done;
done;
done;
done;
done
After a little post-processing of the results, we get:
-- src/cut-before
| command | sl | llw | asw | mbw |
| --------------- | -------- | -------- | -------- | -------- |
| UTF8 -c1 | 70.7 ms | 103.5 ms | 26.9 ms | 48.6 ms |
| UTF8 -c10 | 55.1 ms | 103.9 ms | 48.6 ms | 127.3 ms |
| UTF8 -c100000 | 54.7 ms | 1.585 s | 156.9 ms | 555.9 ms |
-- src/cut-after
| command | sl | llw | asw | mbw |
| --------------- | -------- | -------- | -------- | -------- |
| UTF8 -c1 | 90.1 ms | 103.6 ms | 28.0 ms | 50.9 ms |
| UTF8 -c10 | 39.7 ms | 103.2 ms | 40.0 ms | 127.5 ms |
| UTF8 -c100000 | 39.0 ms | 104.4 ms | 18.5 ms | 30.3 ms |
* bootstrap.conf: Explicitly depend on unistr/u8-check.
* src/cut.c (cut_characters_mode): Quickly skip over unselected
ASCII/UTF-8 data.
* tests/cut/cut.pl: Add test cases.
zstd compressed sources are less than half the size
of gzip coompressed sources, so are a better secondary
distribution format in addition to xz.
* README-prereq: Remove gzip, Add zstd.
* bootstrap.conf (buildreq): Likewise.
Bump automake version check to 1.16.2 (released 2020).
* configure.ac (AM_INIT_AUTOMAKE): Disable gz, add dist-zstd.
Bump automake version requirement to 1.16.2.
* cfg.mk: Use compression level 18, and multi-threading.
* NEWS: Mention the change.
This adds an ELF note as per the UAPI Group's .note.dlopen spec:
$ readelf -p .note.dlopen src/sort
String dump of section '.note.dlopen':
[ 4] y
[ a] |@FDO
[ 10] [{"feature":"random-sort",
"description":"Support for random sorting",
"priority":"required",
"soname":["libcrypto.so.3"]}]
* bootstrap.conf: Add sd-dlopen.
* src/sort.c (DLOPEN_LIBCRYPTO && HAVE_OPENSSL_MD5): Add an ELF note
with SD_ELF_NOTE_DLOPEN() so that packagers can determine the libcrypto
dependency, needed when --random-sort is specified.
* NEWS: Mention the Build-related change.
argz is provided by glibc/gnulib,
simplifies the code a bit, and is about 15% faster,
mainly due to strlen() scanning per entry rather
than per byte scanning:
$ src/printf 'USUAL_%1$d=%1$032d\0' $(seq 30000) > env.0
$ hyperfine -N 'src/env'{,-argz}' -i --env0-from=env.0 true'
src/env-argz -i --env0-from=env.0 true ran
1.17 ± 0.27 times faster than src/env -i --env0-from=env.0 true
* .gitignore: Ignore generated lib/argz.h.
* bootstrap.conf (gnulib_modules): Add argz.
* src/env.c: Include <argz.h>.
(set_envvars_from_file): Use argz_count() and argz_next().
(set_raw_environment_from_file): Use argz_count() and argz_extract().
* bootstrap.conf: Reference the hashcode-mem module.
* src/env.c: Avoid quadratic merging using a hash of vector slots,
as the combined file entries and existing env entries may be
significant. The performance diffence become noticable
when merging 500 x 500 entries. Note we keep using system
interfaces on native windows to retain its case-insensitive behavior.
* tests/env/env0-from.sh: Add coverage for duplicate and
nonstandard inherited entries.
This avoids the "%.5119Lf" issue on Solaris 11
where an extraneous e+00 is output.
seq uses printf, while numfmt uses snprintf.
I.e. this fixes:
$ numfmt --to=si --format='%.5119f' 1 | tail -c5
$ seq -f %.5119f 1 | tail -c5
Note gnulib printf module dependencies were previously discussed at:
https://lists.gnu.org/r/coreutils/2014-11/msg00046.htmlhttps://github.com/coreutils/coreutils/commit/b817f6276
* bootstrap.conf: Depend on {printf,snprintf,sprintf}-posix modules.
* bootstrap.conf (gnulib_modules): Add execvp which as a different
prototype on Windows along with other issues mentioned in the Gnulib
manual.
* src/nohup.c (main): Check if SIGHUP is defined.
This avoids the following behavior:
$ strace -e silence=exit -e trace=unlink,rmdir \
mktemp -d > /dev/full
unlink("/tmp/tmp.ZBuPmS9ZGD") = -1 EISDIR (Is a directory)
rmdir("/tmp/tmp.ZBuPmS9ZGD") = 0
mktemp: write error: No space left on device
In the above invocation we know that we created a directory, so we
should not remove a regular file that must have been created by another
process:
$ strace -e silence=exit -e trace=unlink,rmdir \
./src/mktemp -d > /dev/full
rmdir("/tmp/tmp.hGbME1HmJr") = 0
mktemp: write error: No space left on device
* src/mktemp.c (main): Prefer rmdir and unlink depending on whether we
created a directory or regular file.
* bootstrap.conf (gnulib_modules): Remove the remove module.
* gl/lib/mbbuf.h (fill_buf): Switch from fread() to read()
as the former retries read() internally to fill the buffer.
* src/cut.c: Adjust accordingly, and avoid getc() interface entirely.
* bootstrap.h: Depend explicitly on fseterr. This is already depended
on transitively, so should not introduce new build portability issues.
* bootstrap.conf (avoided_gnulib_modules): Avoid mbiter and
mbiterf, for the same reason we avoid mbuiter and mbuiterf: these
modules are not needed because (due to mcel-prefer) we use mcel in
preference to mbiter/mbiterf/mbuiter/mbuiterf.
A good reference for the concepts used here is:
https://mazzo.li/posts/fast-pipes.html
We don't consider huge pages or busy loops here,
but use vmsplice(), and splice() to get significant speedups:
i7-5600U-laptop $ taskset 1 yes | taskset 2 pv > /dev/null
... [4.98GiB/s]
i7-5600U-laptop $ taskset 1 src/yes | taskset 2 pv > /dev/null
... [34.1GiB/s]
IBM,9043-MRX $ taskset 1 yes | taskset 2 pv > /dev/null
... [11.6GiB/s]
IBM,9043-MRX $ taskset 1 src/yes | taskset 2 pv > /dev/null
... [175GiB/s]
Also throughput to file (on BTRFS) was seen to increase significantly.
With a Fedora 43 laptop improving from 690MiB/s to 1.1GiB/s.
* bootstrap.conf: Ensure sys/uio.h is present.
This was an existing transitive dependency.
* m4/jm-macros.m4: Define HAVE_SPLICE appropriately.
We assume vmsplice() is available if splice() is as they
were introduced at the same time to Linux and glibc.
* src/yes.c (repeat_pattern): A new function to efficiently
duplicate a pattern in a buffer with memcpy calls that double in size.
This also makes the setup for the existing write() path more efficient.
(pipe_splice_size): A new function to increase the kernel pipe buffer
if possible, and use an appropriately sized buffer based on that (25%).
(splice_write): A new function to call vmplice() when outputting
to a pipe, and also splice() if outputting to a non-pipe.
* tests/misc/yes.sh: Verify the non-pipe output case,
(main): Adjust to always calling write on the minimal buffer first,
then trying vmsplice(), then falling back to write from bigger buffer.
and the vmsplice() fallback to write() case.
* NEWS: Mention the improvement.
The explicit_bzero function is a common extension, but memset_explicit
was standardized in C23. It will likely become more portable in the
future, and Gnulib provides an implementation if needed.
* bootstrap.conf (gnulib_modules): Add memset_explicit. Remove
explicit_bzero.
* gl/lib/randint.c (randint_free): Use memset_explicit instead of
explicit_bzero.
* gl/lib/randread.c (randread_free_body): Likewise.
* src/getlimits.c (main): Iterate over defined errnos,
and output shell compatible error strings.
* tests/Coreutils.pm: Adjust so shell quotes are stripped.
Fix du and ls to conform to the POSIX getenv spec,
which says you can’t modify strings returned by getenv
unless you put the string there directly, or used putenv.
This portability bug was found by strict C23 checking
using qualifier-generic functions.
* bootstrap.conf (gnulib_modules): Add xmemdup0. Sort.
* src/du.c (main):
* src/ls.c (decode_switches):
Don’t modify the string that getenv returns.
Instead, use xmemdup0 if needed, and include xmemdup0.h.
* NEWS: Remove timeout from the list of programs that use posix_spawn.
* bootstrap.conf (gnulib_modules): Remove posix_spawnattr_setsigmask.
* src/timeout.c: Don't include spawn.h.
(main): Use fork and execvp instead of posix_spawn.
This reverts commit dac96ce3e3.
* bootstrap.conf (gnulib_modules): Add findprog-in.
* src/sort.c: Include findprog.h.
(pipe_child): Look up the compress_program in $PATH and report errors
such as ENOENT or EACCES before invoking posix_spawnp.
This avoids inconsistency on systems that emulate posix_spawn through
fork/exec, as they would otherwise treat such a failure as a generic
failure and fail the sort, rather than continuing without compression.
* NEWS: Mention the improvement. Consolidate the posix_spawn
improvements into one item.
* bootstrap.conf (gnulib_modules): Add posix_spawnattr_setsigmask.
* src/timeout.c: Include spawn.h.
(main): Setup signals using a posix_spawnattr_t object. Use posix_spawn
instead of fork and execvp.
* NEWS: Mention the change.
* bootstrap.conf (gnulib_modules): Add posix_spawn,
posix_spawnattr_setsigdefault, posix_spawn_file_actions_addclose,
posix_spawn_file_actions_adddup2, and posix_spawn_file_actions_init.
* src/split.c: Include spawn.h.
(create): Use posix_spawn instead of fork and execl.
* NEWS: Mention the change.
* bootstrap.conf (gnulib_modules): Add posix_spawnattr_destroy,
posix_spawnattr_init, posix_spawnattr_setflags, and posix_spawnp.
* src/install.c (strip): Use posix_spawnp instead of fork and execlp.
* bootstrap.conf: Depend on mbsstr() to robustly search for a
multi-byte delimiter character (string) within a multi-byte string.
* src/numfmt.c (main): Accept a valid multi-byte delimiter character.
(next_field): Adjust delimiter search from single byte
to multi-byte aware. Use mbsstr to find the first match.
* tests/misc/numfmt.pl: Add test case.
* NEWS: Mention the improvement.
* src/numfmt.c (process_suffixed_number): Use gnulib's
mbs_endswith() helper, which is more robust in non UTF-8 locales.
Also always output a devmsg if a suffix is specified.
* gl/modules/mbbuf: New file.
* gl/lib/mbbuf.c: Likewise.
* gl/lib/mbbuf.h: Likewise.
* gl/local.mk (EXTRA_DIST): Add the new files.
* bootstrap.conf (gnulib_modules): Add mbbuf.
* src/fold.c: Include mbbuf.h.
(fold_file): Use the mbbuf functions instead of calling fread and
handling the input buffer ourselves.
* cfg.mk (exclude_file_name_regexp--sc_preprocessor_indentation)
(exclude_file_name_regexp--sc_GPL_version): Match gl/lib/mbbuf.c and
gl/lib/mbbuf.h.
Use new Gnulib streq function instead of rolling our own macro.
* bootstrap.conf (gnulib_modules): Add stringeq.
* src/rm.c (main): Don’t assume streq is a macro that expands to (...),
as it is now a function.
* src/system.h:
* tests/df/no-mtab-status.sh, tests/df/skip-duplicates.sh:
(STREQ): Remove. All uses replaced by streq.
This is useful to give better test coverage at least,
and may be useful for users to tune their environment.
* bootstrap.conf: Reference the cpu-supports gnulib module.
* src/cksum.c: Use cpu_supports() rather than __builtin_cpu_supports().
* src/wc.c: Likewise.
* tests/cksum/cksum.sh: Adjust to testing all implementations.
* tests/wc/wc-cpu.sh: A new test to do likewise.
* tests/local.mk: Reference the new wc test.
* src/digest.c: Include sha3.h.
(BLAKE2B_MAX_LEN): Rename to
DIGEST_MAX_LEN since it is also used for SHA-3.
(sha3_sum_stream): New function.
(enum Algorithm, algorithm_args, algorithm_args, algorithm_types)
algorithm_tags, algorithm_bits, cksumfns, cksum_output_fns): Add entries
for SHA-3.
(usage): Mention that SHA-3 is supported. Mention requirements for
--length with SHA-3.
(split_3): Use DIGEST_MAX_LEN instead of BLAKE2B_MAX_LEN. Determine the
length of the digest for SHA-3. Make sure it is 224, 256, 384, or 512.
(digest_file): Set the digest length in bytes. Use DIGEST_MAX_LEN
instead of BLAKE2B_MAX_LEN. Always append the digest length to SHA3 in
the output.
(main): Allow the use of --length with 'cksum -a sha3'. Use
DIGEST_MAX_LEN instead of BLAKE2B_MAX_LEN. Make sure it is 224, 256,
384, or 512.
* tests/cksum/cksum-base64.pl (@pairs): Add expected sha3 output.
(fmt): Modify the output to use SHA3-512 since that is the default.
(@Tests): Modify arguments for sha3 to use --length=512.
* tests/cksum/cksum-sha3.sh: New test, based on tests/cksum/b2sum.sh.
* tests/local.mk (all_tests): Add the test.
* bootstrap.conf: Add crypto/sha3.
* gnulib: Update to latest commit.
* NEWS: Mention the change.
* doc/coreutils.texi (cksum general options): Mention sha3 as a
supported argument to the -a option. Mention that 'cksum -a sha3'
supports the --length option. Mention that SHA-3 is considered secure.
* bootstrap.conf (gnulib_modules): Add issymlink and issymlinkat.
* src/copy.c: Include issymlink.h.
(copy_reg): Use issymlink instead of readlinkat.
* src/rmdir.c: Include issymlink.h.
(main): Use issymlink instead of readlink.
* src/tail.c: Include issymlink.h.
(recheck, any_symlinks): Use issymlink instead of readlink.
* src/test.c: Include issymlink.h.
(unary_operator): Use issymlink instead of readlink.
It’s long been safe to assume C99+ support for long long int.
* .gitignore: Remove m4/longlong.m4.
* bootstrap.conf (buildreq): Boost git prereq from 1.4.4 to 1.5.5,
syncing with Gnulib.
(bootstrap_post_import_hook): Remove m4/longlong.m4.
* m4/jm-macros.m4 (gl_CHECK_ALL_TYPES):
No need to require AC_TYPE_UNSIGNED_LONG_LONG_INT.
* src/factor.c (DItype, UDItype):
* src/od.c (main):
Assume HAVE_LONG_LONG_INT.
* src/od.c: (unsigned_long_long_int):
Remove. All uses replaced with unsigned long long int.
Also, change sleep and tail to not sleep less than requested.
* bootstrap.conf (gnulib_modules): Add dtimespec-bound.
* gl/lib/dtimespec-bound.c, gl/lib/dtimespec-bound.h:
* gl/modules/dtimespec-bound: New files.
* src/sleep.c, src/tail.c, src/timeout.c: Include dtimespec-bound.h.
* src/sleep.c, src/tail.c: Don’t include xstrtod.h.
* src/sleep.c (apply_suffix, main):
* src/tail.c (parse_options):
* src/timeout.c (apply_time_suffix):
Don’t sleep less than the true number of seconds.
* src/timeout.c: Don’t include ctype.h.
(is_negative): Remove; no longer needed.
(parse_duration): Use a slightly looser bound on the timeout, one
that doesn’t need -lm on GNU/Linux. Clear errno before calling
cl_strtod.
This was seen to add about 100,000 ns to the startup time,
on a 2.6 GHz i7-5600U with glibc 2.40.
* .gitignore: Remove /lib/fenv.h.
* bootstrap.conf: Remove fenv-rounding and signbit deps.
* src/local.mk: Remove fenv lib dependency.
* src/timeout.c (is_negative): A new helper function to
be equivalent of signbit in the underflow case.
(parse_duration): Remove the rounding up logic,
as a nanosecond here or there has no significance.
This handles timeouts like 16777216.000000001 correctly;
formerly the subsecond part of that timeout was ignored.
* bootstrap.conf (gnulib_modules): Add fenv-rounding, signbit.
* src/local.mk (src_timeout_LDADD): Append $(FENV_ROUNDING_LIBM).
* src/timeout.c: Include fenv.h, math.h.
Don’t include xstrtod.h, as xstrtod’s checking now gets in the way.
(parse_duration): Round up when calling cl_strtod.
Check for -1e-1000. Don’t double-round 1e-9.
* tests/timeout/timeout-parameters.sh: Test for -0.1,
-1e-1000, 1e-1000.
* who.c (scan_entries): Account for guessed tty names (e.g.
'sshd pts/1') from the readutmp module when using the systemd backend.
* bootstrap.conf (gnulib_modules): Add str_endswith.
* News: Mention the bug fix.
Addresses https://bugzilla.redhat.com/2343998
Update to latest gnulib with new copyright year.
Run "make update-copyright" and then...
* gnulib: Update included in this commit as copyright years
are the only change from the previous gnulib commit.
* tests/init.sh: Sync with gnulib to pick up copyright year.
* bootstrap: Likewise.
* tests/sample-test: Adjust to use the single most recent year.
* bootstrap.conf: Depend on crc-x86_64 rather than crc.
* gnulib: Update to latest.
* src/cksum.c (crc32b_sum_stream): Add --debug info.
* NEWS: Mention the performance improvement.
* bootstrap.conf (gnulib_modules): Add acl-permissions, which
supplies acl_errno_valid, and which we are already using
indirectly via file-has-acl.
* src/ls.c (errno_unsupported): Remove. All calls replaced
by !acl_errno_valid.
$ echo -n '123456789' | cksum --raw -a crc32b | basenc --base16
CBF43926
* bootstrap.conf: Explicitly depend on the crc module.
* doc/coreutils.texi (cksum): Add "crc32b" as an argument to -a.
* src/cksum.c (crc32b_sum_stream): A new function similar to
crc_sum_stream, but which does not include the length in
the CRC calculation.
* src/cksum.h: Add crc32b_sum_stream prototype.
* src/digest.c: Add "crc32b" as an argument to -a.
* tests/cksum/cksum.sh: Refactor to test both crc and crc32b.
* tests/cksum/cksum-a.sh: Add "crc32b" case.
* tests/cksum/cksum-base64.pl: Likewise.
* tests/misc/read-errors.sh: Likewise.
* NEWS: Mention the new feature.
Pádraig Brady