Log message:
sysutils/coreutils | misc/gnuls : update to version 9.4
* Noteworthy changes in release 9.4 (2023-08-29) [stable]
This is a stabilization release coming about 19 weeks after the 9.3 release.
There have been 162 commits by 10 people in the 19 weeks since 9.3.
** Bug fixes
On GNU/Linux s390x and alpha, programs like 'cp' and 'ls' no longer
fail on files with inode numbers that do not fit into 32 bits.
[This bug was present in "the beginning".]
'b2sum --check' will no longer read unallocated memory when
presented with malformed checksum lines.
[bug introduced in coreutils-9.2]
'cp --parents' again succeeds when preserving mode for absolute directories.
Previously it would have failed with a "No such file or directory" error.
[bug introduced in coreutils-9.1]
'cp --sparse=never' will avoid copy-on-write (reflinking) and copy offloading,
to ensure no holes present in the destination copy.
[bug introduced in coreutils-9.0]
cksum again diagnoses read errors in its default CRC32 mode.
[bug introduced in coreutils-9.0]
'cksum --check' now ensures filenames with a leading backslash character
are escaped appropriately in the status output.
This also applies to the standalone checksumming utilities.
[bug introduced in coreutils-8.25]
dd again supports more than two multipliers for numbers.
Previously numbers of the form '1024x1024x32' gave "invalid number" \
errors.
[bug introduced in coreutils-9.1]
factor, numfmt, and tsort now diagnose read errors on the input.
[This bug was present in "the beginning".]
'install --strip' now supports installing to files with a leading hyphen.
Previously such file names would have caused the strip process to fail.
[This bug was present in "the beginning".]
ls now shows symlinks specified on the command line that can't be traversed.
Previously a "Too many levels of symbolic links" diagnostic was given.
[This bug was present in "the beginning".]
pinky, uptime, users, and who no longer misbehave on 32-bit GNU/Linux
platforms like x86 and ARM where time_t was historically 32 bits.
Also see the new --enable-systemd option mentioned below.
[bug introduced in coreutils-9.0]
'pr --length=1 --double-space' no longer enters an infinite loop.
[This bug was present in "the beginning".]
shred again operates on Solaris when built for 64 bits.
Previously it would have exited with a "getrandom: Invalid argument" \
error.
[bug introduced in coreutils-9.0]
tac now handles short reads on its input. Previously it may have exited
erroneously, especially with large input files with no separators.
[This bug was present in "the beginning".]
'uptime' no longer incorrectly prints "0 users" on OpenBSD,
and is being built again on FreeBSD and Haiku.
[bugs introduced in coreutils-9.2]
'wc -l' and 'cksum' no longer crash with an "Illegal instruction" error
on x86 Linux kernels that disable XSAVE YMM. This was seen on Xen VMs.
[bug introduced in coreutils-9.0]
** Changes in behavior
'cp -v' and 'mv -v' will no longer output a message for each file skipped
due to -i, or -u. Instead they only output this information with --debug.
I.e., 'cp -u -v' etc. will have the same verbosity as before coreutils-9.3.
'cksum -b' no longer prints base64-encoded checksums. Rather that
short option is reserved to better support emulation of the standalone
checksum utilities with cksum.
'mv dir x' now complains differently if x/dir is a nonempty directory.
Previously it said "mv: cannot move 'dir' to 'x/dir': Directory not \
empty",
where it was unclear whether 'dir' or 'x/dir' was the problem.
Now it says "mv: cannot overwrite 'x/dir': Directory not empty".
Similarly for other renames where the destination must be the problem.
[problem introduced in coreutils-6.0]
** Improvements
cp, mv, and install now avoid copy_file_range on linux kernels before 5.3
irrespective of which kernel version coreutils is built against,
reinstating that behavior from coreutils-9.0.
comm, cut, join, od, and uniq will now exit immediately upon receiving a
write error, which is significant when reading large / unbounded inputs.
split now uses more tuned access patterns for its potentially large input.
This was seen to improve throughput by 5% when reading from SSD.
split now supports a configurable $TMPDIR for handling any temporary files.
tac now falls back to '/tmp' if a configured $TMPDIR is unavailable.
'who -a' now displays the boot time on Alpine Linux, OpenBSD,
Cygwin, Haiku, and some Android distributions
'uptime' now succeeds on some Android distributions, and now counts
VM saved/sleep time on GNU (Linux, Hurd, kFreeBSD), NetBSD, OpenBSD,
Minix, and Cygwin.
On GNU/Linux platforms where utmp-format files have 32-bit timestamps,
pinky, uptime, and who can now work for times after the year 2038,
so long as systemd is installed, you configure with a new, experimental
option --enable-systemd, and you use the programs without file arguments.
(For example, with systemd 'who /var/log/wtmp' does not work because
systemd does not support the equivalent of /var/log/wtmp.)
|
Log message:
coreutils: Update to 9.2
Changelog:
* Noteworthy changes in release 9.2 (2023-03-20) [stable]
** Bug fixes
'comm --output-delimiter="" --total' now delimits columns in the total
line with the NUL character, consistent with NUL column delimiters in
the rest of the output. Previously no delimiters were used for the
total line in this case.
[bug introduced with the --total option in coreutils-8.26]
'cp -p' no longer has a security hole when cloning into a dangling
symbolic link on macOS 10.12 and later.
[bug introduced in coreutils-9.1]
'cp -rx / /mnt' no longer complains "cannot create directory /mnt/".
[bug introduced in coreutils-9.1]
cp, mv, and install avoid allocating too much memory, and possibly
triggering "memory exhausted" failures, on file systems like ZFS,
which can return varied file system I/O block size values for files.
[bug introduced in coreutils-6.0]
cp, mv, and install now immediately acknowledge transient errors
when creating copy-on-write or cloned reflink files, on supporting
file systems like XFS, BTRFS, APFS, etc.
Previously they would have tried again with other copy methods
which may have resulted in data corruption.
[bug introduced in coreutils-7.5 and enabled by default in coreutils-9.0]
cp, mv, and install now handle ENOENT failures across CIFS file systems,
falling back from copy_file_range to a better supported standard copy.
[issue introduced in coreutils-9.0]
'mv --backup=simple f d/' no longer mistakenly backs up d/f to f~.
[bug introduced in coreutils-9.1]
rm now fails gracefully when memory is exhausted.
Previously it may have aborted with a failed assertion in some cases.
[This bug was present in "the beginning".]
rm -d (--dir) now properly handles unreadable empty directories.
E.g., before, this would fail to remove d: mkdir -m0 d; src/rm -d d
[bug introduced in v8.19 with the addition of this option]
runcon --compute no longer looks up the specified command in the $PATH
so that there is no mismatch between the inspected and executed file.
[bug introduced when runcon was introduced in coreutils-6.9.90]
'sort -g' no longer infloops when given multiple NaNs on platforms
like x86_64 where 'long double' has padding bits in memory.
Although the fix alters sort -g's NaN ordering, that ordering has
long been documented to be platform-dependent.
[bug introduced 1999-05-02 and only partly fixed in coreutils-8.14]
stty ispeed and ospeed options no longer accept and silently ignore
invalid speed arguments, or give false warnings for valid speeds.
Now they're validated against both the general accepted set,
and the system supported set of valid speeds.
[This bug was present in "the beginning".]
stty now wraps output appropriately for the terminal width.
Previously it may have output 1 character too wide for certain widths.
[bug introduced in coreutils-5.3]
tail --follow=name works again with non seekable files. Previously it
exited with an "Illegal seek" error when such a file was replaced.
[bug introduced in fileutils-4.1.6]
'wc -c' will again efficiently determine the size of large files
on all systems. It no longer redundantly reads data from certain
sized files larger than SIZE_MAX.
[bug introduced in coreutils-8.24]
** Changes in behavior
Programs now support the new Ronna (R), and Quetta (Q) SI prefixes,
corresponding to 10^27 and 10^30 respectively,
along with their binary counterparts Ri (2^90) and Qi (2^100).
In some cases (e.g., 'sort -h') these new prefixes simply work;
in others, where they exceed integer width limits, they now elicit
the same integer overflow diagnostics as other large prefixes.
'cp --reflink=always A B' no longer leaves behind a newly created
empty file B merely because copy-on-write clones are not supported.
'cp -n' and 'mv -n' now exit with nonzero status if they skip their
action because the destination exists, and likewise for 'cp -i',
'ln -i', and 'mv -i' when the user declines. (POSIX specifies this
for 'cp -i' and 'mv -i'.)
cp, mv, and install again read in multiples of the reported block size,
to support unusual devices that may have this constraint.
[behavior inadvertently changed in coreutils-7.2]
du --apparent now counts apparent sizes only of regular files and
symbolic links. POSIX does not specify the meaning of apparent
sizes (i.e., st_size) for other file types, and counting those sizes
could cause confusing and unwanted size mismatches.
'ls -v' and 'sort -V' go back to sorting ".0" before ".A",
reverting to the behavior in coreutils-9.0 and earlier.
This behavior is now documented.
ls --color now matches a file extension case sensitively
if there are different sequences defined for separate cases.
printf unicode \uNNNN, \UNNNNNNNN syntax, now supports all valid
unicode code points. Previously is was restricted to the C
universal character subset, which restricted most points <= 0x9F.
runcon now exits with status 125 for internal errors. Previously upon
internal errors it would exit with status 1, which was less distinguishable
from errors from the invoked command.
'split -n N' now splits more evenly when the input size is not a
multiple of N, by creating N output files whose sizes differ by at
most 1 byte. Formerly, it did this only when the input size was
less than N.
'stat -c %s' now prints sizes as unsigned, consistent with 'ls'.
** New Features
cksum now accepts the --base64 (-b) option to print base64-encoded
checksums. It also accepts/checks such checksums.
cksum now accepts the --raw option to output a raw binary checksum.
No file name or other information is output in this mode.
cp, mv, and install now accept the --debug option to
print details on how a file is being copied.
factor now accepts the --exponents (-h) option to print factors
in the form p^e, rather than repeating the prime p, e times.
ls now supports the --time=modification option, to explicitly
select the default mtime timestamp for display and sorting.
mv now supports the --no-copy option, which causes it to fail when
asked to move a file to a different file system.
split now accepts options like '-n SIZE' that exceed machine integer
range, when they can be implemented as if they were infinity.
split -n now accepts piped input even when not in round-robin mode,
by first copying input to a temporary file to determine its size.
wc now accepts the --total={auto,never,always,only} option
to give explicit control over when the total is output.
** Improvements
cp --sparse=auto (the default), mv, and install,
will use the copy_file_range syscall now also with sparse files.
This may be more efficient, by avoiding user space copies,
and possibly employing copy offloading or reflinking,
for the non sparse portion of such sparse files.
On macOS, cp creates a copy-on-write clone in more cases.
Previously cp would only do this when preserving mode and timestamps.
date --debug now diagnoses if multiple --date or --set options are
specified, as only the last specified is significant in that case.
rm outputs more accurate diagnostics in the presence of errors
when removing directories. For example EIO will be faithfully
diagnosed, rather than being conflated with ENOTEMPTY.
tail --follow=name now works with single non regular files even
when their modification time doesn't change when new data is available.
Previously tail would not show any new data in this case.
tee -p detects when all remaining outputs have become broken pipes, and
exits, rather than waiting for more input to induce an exit when written.
tee now handles non blocking outputs, which can be seen for example with
telnet or mpirun piping through tee to a terminal.
Previously tee could truncate data written to such an output and fail,
and also potentially output a "Resource temporarily unavailable" error.
|