Changelog:
* Noteworthy changes in release 8.25 (2016-01-20) [stable]
** Bug fixes
cp now correctly copies files with a hole at the end of the file,
and extents allocated beyond the apparent size of the file.
That combination resulted in the trailing hole not being reproduced.
[bug introduced in coreutils-8.10]
cut --fields no longer outputs extraneous characters on some uClibc configs.
[bug introduced in coreutils-6.11]
install -D again copies relative file names when absolute file names
are also specified along with an absolute destination directory name.
[bug introduced in coreutils-6.2]
ls no longer prematurely wraps lines when printing short file names.
[bug introduced in coreutils-5.1.0]
mv no longer causes data loss due to removing a source directory specified
multiple times, when that directory is also specified as the destination.
[bug introduced in coreutils-8.24]
shred again uses defined patterns for all iteration counts.
[bug introduced in coreutils-5.93]
sort --debug -b now correctly marks the matching extents for keys
that specify an offset for the first field.
[bug introduced with the --debug feature in coreutils-8.6]
tail -F now works with initially non existent files on a remote file system.
[bug introduced in coreutils-7.5]
** New commands
base32 is added to complement the existing base64 command,
and encodes and decodes printable text as per RFC 4648.
** New features
comm,cut,head,numfmt,paste,tail now have the -z,--zero-terminated option, and
tac --separator accepts an empty argument, to work with NUL delimited items.
dd now summarizes sizes in --human-readable format too, not just --si.
E.g., "3441325000 bytes (3.4 GB, 3.2 GiB) copied". It omits the summaries
if they would not provide useful information, e.g., "3 bytes copied".
Its status=progress output now uses the same format as ordinary status,
perhaps with trailing spaces to erase previous progress output.
md5sum now supports the --ignore-missing option to allow
verifying a subset of files given a larger list of checksums.
This also affects sha1sum, sha224sum, sha256sum, sha384sum and sha512sum.
printf now supports the '%q' format to print arguments in a form that
is reusable by most shells, with non-printable characters escaped
with the POSIX proposed $'...' syntax.
stty now supports the "[-]drain" setting to control whether to wait
for transmission of pending output before application of settings.
** Changes in behavior
base64 no longer supports hex or oct --wrap parameters,
thus better supporting decimals with leading zeros.
date --iso-8601 now uses +00:00 timezone format rather than +0000.
The standard states to use this "extended" format throughout a timestamp.
df now prefers sources towards the root of a device when
eliding duplicate bind mounted entries.
ls now quotes file names unambiguously and appropriate for use in a shell,
when outputting to a terminal.
join, sort, uniq with --zero-terminated, now treat '\n' as a field delimiter.
** Improvements
All utilities now quote user supplied arguments in error strings,
which avoids confusing error messages in the presence of '\r' chars etc.
Utilities that traverse directories, like chmod, cp, and rm etc., will operate
more efficiently on XFS through the use of "leaf optimization".
md5sum now ensures a single line per file for status on standard output,
by using a '\' at the start of the line, and replacing any newlines with '\n'.
This also affects sha1sum, sha224sum, sha256sum, sha384sum and sha512sum.
dircolors now supports globbing of TERM entries in its database.
For example "TERM *256color*" is now supported.
du no longer stats all mount points at startup, only doing so
upon detection of a directory cycle.
[issue introduced in coreutils-8.20]
ls -w0 is now interpreted as no limit on the length of the outputted line.
stat -f --format=%T now reports the file system type for new Linux
pseudo file systems "bpf_fs", "btrfs_test", "nsfs", "overlayfs"
and "tracefs", and remote file system "acfs".
wc now ensures a single line per file for counts on standard output,
by quoting names containing '\n' characters; appropriate for use in a shell.
* Noteworthy changes in release 8.24 (2015-07-03) [stable]
** Bug fixes
dd supports more robust SIGINFO/SIGUSR1 handling for outputting statistics.
Previously those signals may have inadvertently terminated the process.
df --local no longer hangs with inaccessible remote mounts.
[bug introduced in coreutils-8.21]
du now silently ignores all directory cycles due to bind mounts.
Previously it would issue a warning and exit with a failure status.
[bug introduced in coreutils-8.1 and partially fixed in coreutils-8.23]
chroot again calls chroot(DIR) and chdir("/"), even if DIR is "/".
This handles separate bind mounted "/" trees, and environments
depending on the implicit chdir("/").
[bugs introduced in coreutils-8.23]
cp no longer issues an incorrect warning about directory hardlinks when a
source directory is specified multiple times. Now, consistent with other
file types, a warning is issued for source directories with duplicate names,
or with -H the directory is copied again using the symlink name.
factor avoids writing partial lines, thus supporting parallel operation.
[the bug dates back to the initial implementation]
head, od, split, tac, tail, and wc no longer mishandle input from files in
/proc and /sys file systems that report somewhat-incorrect file sizes.
mkdir --parents -Z now correctly sets the context for the last component,
even if the parent directory exists and has a different default context.
[bug introduced with the -Z restorecon functionality in coreutils-8.22]
numfmt no longer outputs incorrect overflowed values seen with certain
large numbers, or with numbers with increased precision.
[bug introduced when numfmt was added in coreutils-8.21]
numfmt now handles leading zeros correctly, not counting them when
settings processing limits, and making them optional with floating point.
[bug introduced when numfmt was added in coreutils-8.21]
paste no longer truncates output for large input files. This would happen
for example with files larger than 4GiB on 32 bit systems with a '\n'
character at the 4GiB position.
[the bug dates back to the initial implementation]
rm indicates the correct number of arguments in its confirmation prompt,
on all platforms. [bug introduced in coreutils-8.22]
shuf -i with a single redundant operand, would crash instead of issuing
a diagnostic. [bug introduced in coreutils-8.22]
tail releases inotify resources when unused. Previously it could exhaust
resources with many files, or with -F if files were replaced many times.
[bug introduced in coreutils-7.5]
tail -f again follows changes to a file after it's renamed.
[bug introduced in coreutils-7.5]
tail --follow no longer misses changes to files if those files were
replaced before inotify watches were created.
[bug introduced in coreutils-7.5]
tail --follow consistently outputs all data for a truncated file.
[bug introduced in the beginning]
tail --follow=name correctly outputs headers for multiple files
when those files are being created or renamed.
[bug introduced in coreutils-7.5]
** New features
chroot accepts the new --skip-chdir option to not change the working directory
to "/" after changing into the chroot(2) jail, thus retaining the current wor-
king directory. The new option is only permitted if the new root directory is
the old "/", and therefore is useful with the --group and --userspec options.
dd accepts a new status=progress level to print data transfer statistics
on stderr approximately every second.
numfmt can now process multiple fields with field range specifications similar
to cut, and supports setting the output precision with the --format option.
split accepts a new --separator option to select a record separator character
other than the default newline character.
stty allows setting the "extproc" option where supported, which is
a useful setting with high latency links.
sync no longer ignores arguments, and syncs each specified file, or with the
--file-system option, the file systems associated with each specified file.
tee accepts a new --output-error option to control operation with pipes
and output errors in general.
** Changes in behavior
df no longer suppresses separate exports of the same remote device, as
these are generally explicitly mounted. The --total option does still
suppress duplicate remote file systems.
[suppression was introduced in coreutils-8.21]
mv no longer supports moving a file to a hardlink, instead issuing an error.
The implementation was susceptible to races in the presence of multiple mv
instances, which could result in both hardlinks being deleted. Also on case
insensitive file systems like HFS, mv would just remove a hardlinked 'file'
if called like `mv file File`. The feature was added in coreutils-5.0.1.
numfmt --from-unit and --to-unit options now interpret suffixes as SI units,
and IEC (power of 2) units are now specified by appending 'i'.
tee will exit early if there are no more writable outputs.
tee does not treat the file operand '-' as meaning standard output any longer,
for better conformance to POSIX. This feature was added in coreutils-5.3.0.
timeout --foreground no longer sends SIGCONT to the monitored process,
which was seen to cause intermittent issues with GDB for example.
** Improvements
cp,install,mv will convert smaller runs of NULs in the input to holes,
and cp --sparse=always avoids speculative preallocation on XFS for example.
cp will read sparse files more efficiently when the destination is a
non regular file. For example when copying a disk image to a device node.
mv will try a reflink before falling back to a standard copy, which is
more efficient when moving files across BTRFS subvolume boundaries.
stat and tail now know about IBRIX. stat -f --format=%T now reports the file
system type, and tail -f uses polling for files on IBRIX file systems.
wc -l processes short lines much more efficiently.
References from --help and the man pages of utilities have been corrected
in various cases, and more direct links to the corresponding online
documentation are provided.
There have been 195 commits by 32 people in the 10 months since 8.21
Executive summary: 8.22 is mainly a bug fix and performance improvement release.
tail(1) has fixes for handling initially non existent files, and df now has
better handling of specified disk device nodes. On the performance side,
there have been many improvements to the memory usage of various tools, which
are detailed at http://www.pixelbeat.org/programming/avoiding_large_buffers.html
There have been large throughput improvements in cut(1) and base64 encoding.
Also the md5sum and sha*sum utilities have support for using architecture
specific routines to improve performance, if allowed on your distribution.
There are also new features available, including SELinux "restorecon" support
in various file creation utilities through the -Z option, and shred has a new
--repeat option to continually select random items from the input.
See the NEWS below for more details.
Thanks to everyone who has contributed!
The following people contributed changes to this release:
Andreas Mohr (1)
Anton Ovchinnikov (2)
Assaf Gordon (8)
Benno Schulenberg (2)
Bernhard Voelker (28)
Cojocaru Alexandru (4)
Colin Leitner (1)
D. Hugh Redelmeier (1)
Daniel J Walsh (1)
Enrico Scholz (1)
Eric Blake (1)
FUJIWARA Katsunori (1)
Filipus Klutiero (1)
Gian Piero Carrubba (1)
Jarkko Sakkinen (4)
Javier López (1)
Jim Meyering (6)
Joachim Schmitz (2)
John (1)
Karl Berry (1)
Ken Booth (1)
Mike Frysinger (1)
Ondrej Oprala (2)
Ondřej Vašík (2)
Paul Eggert (10)
Pádraig Brady (106)
Rasmus Villemoes (1)
Rémy Lefevre (1)
Sergio Durigan Junior (1)
Stefano Lattarini (3)
Tiger Lee (1)
Torbjörn Granlund (1)
Special thanks to Bernhard Voelker for his many considered patches and reviews,
and to Assaf Gordon for his work on new (and future) features.
Pádraig [on behalf of the coreutils maintainers]
==================================================================
Here is the GNU coreutils home page:
http://gnu.org/s/coreutils/
For a summary of changes and contributors, see:
http://git.sv.gnu.org/gitweb/?p=coreutils.git;a=shortlog;h=v8.22
or run this command from a git-cloned coreutils directory:
git shortlog v8.21..v8.22
To summarize the 255 gnulib-related changes, run these commands
From a git-cloned coreutils directory:
git checkout v8.22
git submodule summary v8.21
==================================================================
Here are the compressed sources and a GPG detached signature[*]:
http://ftp.gnu.org/gnu/coreutils/coreutils-8.22.tar.xzhttp://ftp.gnu.org/gnu/coreutils/coreutils-8.22.tar.xz.sig
Use a mirror for higher download bandwidth:
http://ftpmirror.gnu.org/coreutils/coreutils-8.22.tar.xzhttp://ftpmirror.gnu.org/coreutils/coreutils-8.22.tar.xz.sig
[*] Use a .sig file to verify that the corresponding file (without the
.sig suffix) is intact. First, be sure to download both the .sig file
and the corresponding tarball. Then, run a command like this:
gpg --verify coreutils-8.22.tar.xz.sig
If that command fails because you don't have the required public key,
then run this command to import it:
gpg --keyserver keys.gnupg.net --recv-keys DF6FD971306037D9
and rerun the 'gpg --verify' command.
This release was bootstrapped with the following tools:
Autoconf 2.69
Automake 1.13.4
Gnulib v0.1-38-g0658e50
Bison 2.7
NEWS
* Noteworthy changes in release 8.22 (2013-12-13) [stable]
** Bug fixes
df now processes the mount list correctly in the presence of unstatable
mount points. Previously it may have failed to output some mount points.
[bug introduced in coreutils-8.21]
df now processes symbolic links and relative paths to special files containing
a mounted file system correctly. Previously df displayed the statistics about
the file system the file is stored on rather than the one inside.
[This bug was present in "the beginning".]
df now processes disk device nodes correctly in the presence of bind mounts.
Now df shows the base mounted file system rather than the last one mounted.
[This bug was present in "the beginning".]
install now removes the target file if the strip program failed for any
reason. Before, that file was left behind, sometimes even with wrong
permissions.
[This bug was present in "the beginning".]
ln --relative now updates existing symlinks correctly. Previously it based
the relative link on the dereferenced path of an existing link.
[This bug was introduced when --relative was added in coreutils-8.16.]
ls --recursive will no longer exit with "serious" exit code (2), if there
is an error reading a directory not specified on the command line.
[Bug introduced in coreutils-5.3.0]
mkdir, mkfifo, and mknod now work better when creating a file in a directory
with a default ACL whose umask disagrees with the process's umask, on a
system such as GNU/Linux where directory ACL umasks override process umasks.
[bug introduced in coreutils-6.0]
mv will now replace empty directories in the destination with directories
from the source, when copying across file systems.
[This bug was present in "the beginning".]
od -wN with N larger than 64K on a system with 32-bit size_t would
print approximately 2*N bytes of extraneous padding.
[Bug introduced in coreutils-7.0]
rm -I now prompts for confirmation before removing a write protected file.
[Bug introduced in coreutils-6.8]
shred once again uses direct I/O on systems requiring aligned buffers.
Also direct I/O failures for odd sized writes at end of file are now handled.
[The "last write" bug was introduced in coreutils-5.3.0 but masked
by the alignment bug introduced in coreutils-6.0]
tail --retry -f now waits for the files specified to appear. Before, tail
would immediately exit when such a file is initially inaccessible.
[This bug was introduced when inotify support was added in coreutils-7.5]
tail -F has improved handling of symlinks. Previously tail didn't respond
to the symlink target (re)appearing after being (re)created.
[This bug was introduced when inotify support was added in coreutils-7.5]
** New features
cp, install, mkdir, mknod, mkfifo and mv now support "restorecon"
functionality through the -Z option, to set the SELinux context
appropriate for the new item location in the file system.
csplit accepts a new option: --suppressed-matched, to elide the lines
used to identify the split points.
df --output now accepts a 'file' field, to propagate a specified
command line argument through to the output.
du accepts a new option: --inodes to show the number of inodes instead
of the blocks used.
id accepts a new option: --zero (-z) to delimit the output entries by
a NUL instead of a white space character.
id and ls with -Z report the SMACK security context where available.
mkdir, mkfifo and mknod with -Z set the SMACK context where available.
id can now lookup by user ID, in addition to the existing name lookup.
join accepts a new option: --zero-terminated (-z). As with the sort,uniq
option of the same name, this makes join consume and produce NUL-terminated
lines rather than newline-terminated lines.
uniq accepts a new option: --group to print all items, while separating
unique groups with empty lines.
shred accepts new parameters to the --remove option to give greater
control over that operation, which can greatly reduce sync overhead.
shuf accepts a new option: --repeat (-r), which can repeat items in
the output.
** Changes in behavior
cp --link now dereferences a symbolic link as source before creating the
hard link in the destination unless the -P,--no-deref option is specified.
Previously, it would create a hard link of the symbolic link, even when
the dereferencing options -L or -H were specified.
cp, install, mkdir, mknod and mkfifo no longer accept an argument to the
short -Z option. The --context equivalent still takes an optional argument.
dd status=none now suppresses all non fatal diagnostic messages,
not just the transfer counts.
df no longer accepts the long-obsolescent --megabytes option.
stdbuf now requires at least one buffering mode option to be specified,
as per the documented interface.
** Improvements
base64 encoding throughput for bulk data is increased by about 60%.
md5sum can use libcrypto hash routines where allowed to potentially
get better performance through using more system specific logic.
sha1sum for example has improved throughput by 40% on an i3-2310M.
This also affects sha1sum, sha224sum, sha256sum, sha384sum and sha512sum.
stat and tail work better with EFIVARFS, EXOFS, F2FS, HOSTFS, SMACKFS, SNFS
and UBIFS. stat -f --format=%T now reports the file system type, and tail -f
now uses inotify for files on all those except SNFS, rather than the default
(for unknown file system types) of issuing a warning and reverting to polling.
shuf outputs subsets of large inputs much more efficiently.
Reservoir sampling is used to limit memory usage based on the number of
outputs, rather than the number of inputs.
shred increases the default write block size from 12KiB to 64KiB
to align with other utilities and reduce the system call overhead.
split --line-bytes=SIZE, now only allocates memory as needed rather
than allocating SIZE bytes at program start.
stty now supports configuring "stick" (mark/space) parity where available.
** Build-related
factor now builds on aarch64 based systems [bug introduced in coreutils-8.20]
ryoon
richard
joerg