* doc/coreutils.texi (dircolors invocation, Examples of expr):
(shred invocation, seq invocation): Use $(...), not `...`.
* src/mv.c (do_move): Likewise, in a comment.
* tests/dd/sparse: The last two parts of this test would fail due to
the underlying file system at least on Solaris 10 with NFS. That file
system would report that a 3MiB file was occupying <= 1KiB of space
for nearly 50 seconds after creation.
Improved-by: Bernhard Voelker
With the "--relative --symbolic" options, ln computes the relative
symbolic link for the user.
So, ln works just as cp, but creates relative symbolic links instead
of copying the file.
I miss this feature since the beginning of using ln.
$ tree ./
/
`-- usr
|-- bin
`-- lib
`-- foo
`-- foo
4 directories, 1 file
$ ln -s -v --relative usr/lib/foo/foo usr/bin/foo
‘usr/bin/foo’ -> ‘../lib/foo/foo’
$ tree ./
/
`-- usr
|-- bin
| `-- foo -> ../lib/foo/foo
`-- lib
`-- foo
`-- foo
4 directories, 2 files
$ ln -s -v --relative usr/bin/foo usr/lib/foo/link-to-foo
‘usr/lib/foo/link-to-foo’ -> ‘foo’
$ tree ./
/
`-- usr
|-- bin
| `-- foo -> ../lib/foo/foo
`-- lib
`-- foo
|-- link-to-foo -> foo
`-- foo
4 directories, 3 files
* src/Makefile.am: Reference the relpath module.
* src/ln.c (usage): Mention the new option.
(do_link): Call the relative conversion if specified.
(convert_abs_rel): Perform the relative conversion
using the relpath module.
* tests/ln/relative: Add a new test.
* tests/Makefile.am: Reference the new test.
* doc/coreutils.texi: Document the new feature.
* NEWS: Mention the new feature.
* src/relpath.c: Refactored from realpath.c and adjusted
to support returning the relative path rather than just
printing to stdout.
* src/relpath.h: Export the relpath function.
* src/Makefile.am: Reference the refactored relpath module.
* po/POTFILES.in: Likewise.
* src/realpath.c: Adjust to the refactored relpath module.
This reverts part of commit v8.12-103-g54cbe6e.
* src/system.h: Include gnulib's pathmax.h to honor
system specific limits, and then we set PATH_MAX only if needed.
Note pathmax.h no longer uses pathconf ("/", _PC_PATH_MAX).
Note I didn't reinstate the comments about limits.h inclusion
order, because pathmax.h includes limits.h anyway.
* src/tac.c (temp_stream): Use fseeko, not fseek, on principle:
use the more modern interface. In general it is better to avoid
fseek due to its ABI-imposed 4GiB limit on the "offset", here its
use was fine because the offset was always 0. Using fseeko also
has the advantage of not triggering a GNULIB_POSIXCHECK warning.
Reported by Eric Blake in
http://thread.gmane.org/gmane.comp.gnu.coreutils.general/2426/focus=2489
Since most users won't be building with GNULIB_POSIXCHECK defined in
CFLAGS, and since we can make ./configure 10% (several seconds!) faster
by omitting the framework for a posix check, this patch makes it so
that the framework is omitted by default, while still giving
instructions for maintainers to re-enable it.
It's been a while since we've used GNULIB_POSIXCHECK; see this email:
https://lists.gnu.org/archive/html/coreutils/2012-03/msg00126.html
Some of those failures are because we are intentionally avoiding
specific gnulib modules (that is, we have chosen not to use things
like fprintf-posix), but until we work with gnulib to avoid particular
warnings, wiring up an automatic GNULIB_POSIXCHECK to happen during
'make my-distcheck' is not feasible.
* configure.ac (gl_ASSERT_NO_GNULIB_POSIXCHECK): Conditionally
define, according to whether $GNULIB_POSIXCHECK is in environment.
* dist-check.mk (coreutils-path-check): Now that we set PATH in
TESTS_ENVIRONMENT, it seems like overkill to make "distcheck"
rerun all tests just to check this.
(my-distcheck): Remove sole use.
For compatibility with MacOS relpath(1), as seen here:
http://opensource.apple.com/source/bootstrap_cmds/\
bootstrap_cmds-79/relpath.tproj/relpath.c
we implemented 'realpath --relative-base=dir1 --relative-to=dir2 file'
in the same way as 'relpath -d dir1 dir2 file'. This can result
in --relative-base rendering --relative-to as a no-op if dir1 is a
child of dir2. Document this.
* doc/coreutils.texi (realpath invocation): Mention restriction.
There is no need to recompute for every path being visited whether
the base is a prefix of the relative location.
* src/realpath.c (relpath): Hoist base check...
(main): ...here.
Based on a suggestion by Pádraig Brady.
Most of the time, if someone wants to filter which paths are
relative while leaving all others absolute, they also want to
to the filtering based on the same --relative-to directory.
Make this easier to specify.
* src/realpath.c (main): Convert error to default.
* doc/coreutils.texi (realpath invocation): Document this.
* tests/misc/realpath: Adjust test to match.
* NEWS: Document it.
'realpath --relative-base --relative-to' is identical to
--relative-base=--relative-to, so the test wasn't covering what
it claimed. Expose recent fixes for handling of // on systems
where // is distinct, and for --relative-base=/. Add test that
exposes our design decision that --relative-base that is not a
prefix of --relative-to is a no-op (if we later change behavior,
we will also have to change that part of the test).
* tests/misc/realpath: Fix typo. Add some tests.
When --relative-base is /, all other paths should be treated as
relative (except for // where it matters).
Also, on platforms like Cygwin where / and // are distinct, realpath
was incorrectly collapsing // into /. http://debbugs.gnu.org/10472.
* src/realpath.c (path_prefix, path_common_prefix): Treat /
and // as having no common match.
(relpath): Allow for no match even without --relative-base.
* NEWS: Document this.
https://lists.gnu.org/archive/html/bug-gnulib/2012-02/msg00038.html
detailed a couple of bugs in gnulib's canonicalize that were visible
through coreutils' readlink, but only on systems where // is distinct
from /. This particular test assumes the POSIX fix which requires
canonicalization of a symlink containing just slashes to behave as
if slashes separating the symlink from the rest of the name are
elided (see http://austingroupbugs.net/view.php?id=541), as that is
the only useful (and current) behavior on Cygwin. That is,
ln -s / root
ls root/dev
must list the contents of /dev, not //dev.
* tests/misc/readlink-root: New test.
* tests/Makefile.am (TESTS): Run it.
* tests/Makefile.am (TESTS): Add misc/sort-discrim.
* tests/misc/sort-discrim: New file, which tests a discriminator-based
implementation of 'sort'. Coreutils doesn't use this implementation
yet, but the test is useful anyway.
Co-authored-by: Drew Kutilek <dkutilek@ucla.edu>
Co-authored-by: James Wendt <jwendt@cs.ucla.edu>
* src/dirname.c (main): Handle new -z option and manage more than one
argument.
* doc/coreutils.texi (dirname invocation): Mention it.
* NEWS (New features): Mention it.
* tests/misc/dirname: Add a two arguments test.
* src/split.c (next_file_name): If `suffix_auto' is true and the first
suffix character is 'z', generate a new file file name adding `z' to
the prefix and increasing the suffix length by one.
(set_suffix_length): Disable auto suffix width in various cases.
* tests/split/suffix-auto-length: Test it.
* doc/coreutils.texi (split invocation): Mention it.
* NEWS (Improvements): Likewise.
* NEWS: Document this.
* doc/perm.texi (Operator Numeric Modes): New section.
(Numeric Modes, Directory Setuid and Setgid): Document new behavior.
* src/chmod.c (usage): Document new behavior.
(main): Support new options -0, -1, etc.
* tests/chmod/setgid: Test these new features.
Surprise! "du -x non-DIR" would print nothing.
Note that the problem arises only when processing a non-directory
specified on the command line. Not surprisingly, "du -x" still
works as expected for any directory argument.
When performing its same-file-system check, du may skip an entry
only if it is at fts_level 1 or greater. Command-line arguments
are at fts_level == 0 (FTS_ROOTLEVEL).
* src/du.c (process_file): Don't use the top-level FTS->fts_dev
when testing for --one-file-system (-x). It happens to be valid
for directories, but it is always 0 for a non-directory.
* tests/du/one-file-system: Add tests for this.
* NEWS (Bug fixes): Mention it.
Reported by Daniel Stavrovski in http://bugs.gnu.org/10967.
Introduced by commit v8.14-95-gcfe1040.
* cfg.mk: Set XZ_OPT = -8e (determined empirically).
This sacrifices 8 KiB of compressed tarball size for a 32-MiB
decrease in the memory required during decompression. I.e.,
using -9e would shave off only 8 KiB from the tar.xz file, yet
would force every decompression process to use 32 MiB more memory.
* src/basename.c (perform_basename): New function refactored from
main() that performs the basename work on a STRING, optionally
removes a trailing SUFFIX and outputs the result.
(main): Handle new options.
* doc/coreutils.texi (basename invocation): Mention new options.
* test/misc/basename: Add new options test cases.
* NEWS (New features): Mention it.
Prompted by the continuous integration build failure at:
http://hydra.nixos.org/build/2188210 (which uses XFS).
* tests/dd/sparse (alloc_equal): Add a block allocation
comparison function that accounts for variations due
to alignment.
* src/timeout.c (usage): Document the exit status for this case,
in --help and thus in the man page. Word so that it covers
both the -s9 and -k options.
* doc/coreutils.texi (timeout invocation): Document the exit
status for this case.
* tests/ls/getxattr-speedup: Compile and link in one step with $CC.
If the shared object file is created by ld (binutils), then the
destructor print_call_count() may not run (seen on OpenSuSE 12.1).
See http://lists.opensuse.org/opensuse/2012-02/msg01342.html
Thanks to Cristian Rodríguez.
Notes:
Small seeks are not coalesced to larger ones,
like is done in cache_round() for example.
conv= is used rather then oflag= for FreeBSD compatibility.
* src/dd.c (final_op_was_seek): A new global boolean to flag
whether the final "write" was converted to a seek.
(usage): Describe the new conf=sparse option.
(iwrite): Convert a write of a NUL block to a seek if requested.
(do_copy): Initialize the output buffer to have a sentinel,
to allow for efficient testing for NUL output blocks.
If the last block in the file was converted to a seek,
then convert back to a write so the size is updated.
* NEWS: Mention the new feature.
* tests/dd/sparse: A new test for the feature.
* tests/Makefile.am: Reference the new test.
* src/sort.c (default_sort_size): Don't divide advice by 2.
Just divide the hard limits by 2. This matches the comments.
Reported by Rogier Wolff in http://bugs.gnu.org/10877
Add the --additional-suffix option, to append an
additional static suffix to output file names.
* src/split.c (next_file_name): Append suffix to output file names.
(main): Handle new --additional-suffix option.
* NEWS (New features): Mention it.
* doc/coreutils.texi (split invocation): Mention it.
* tests/split/additional-suffix: New file. Test --additional-suffix.
* tests/Makefile.am (TESTS): Add it.
Requested by Peng Yu, in bug 6554
* src/ls.c (errno_unsupported): Remove EBUSY, as this caters for
the case where ACLs can't be accessed because the _file_ is locked.
http://lists.gnu.org/archive/html/bug-coreutils/2005-06/msg00191.html
Also ENOENT is not safe to include as you get that if the _file_
is removed between the stat() and subsequent querying of xattrs.
Modern <stdio.h> provides ssize_t, but the one from Debian's
libc6-dev 2.11.2-10 apparently does not.
* tests/ls/getxattr-speedup: Also include <sys/types.h>.
Like the optimization to avoid always-failing getfilecon calls,
this change avoids always-failing queries for whether a file has
a nontrivial ACL and for whether a file has certain "capabilities".
When such a query fails for one file (indicating no support), we know it
will always fail that way for the affected device. With this change, we
have thus eliminated nearly all failing-unsupported getxattr syscalls.
* src/ls.c (has_capability) [!HAVE_CAP]: Set errno to ENOTSUP.
(errno_unsupported): Expand the list of E* errno values to match
that of lib/acl-internal.h's ACL_NOT_WELL_SUPPORTED macro.
(file_has_acl_cache, has_capability_cache): New functions.
(gobble_file): Use them in place of non-caching ones.
* NEWS (Improvements): Mention it.
Suggested by Sven Breuner in
http://thread.gmane.org/gmane.comp.gnu.coreutils.general/2187
While eliminating most getfilecon calls saved about 33%,
eliminating these other calls can save almost all of the
remaining ~67% cost, on some remote file systems.
On systems or file systems without SELinux support, all getfilecon
and lgetfilecon calls would fail due to lack of support. We can non-
invasively cache such failure (on most recently accessed device) and
avoid the vast majority of the failing underlying getxattr syscalls.
* src/ls.c (errno_unsupported): New function.
(selinux_challenged_device): New file-scoped global.
(getfilecon_cache, lgetfilecon_cache): New error-caching wrapper
functions.
(gobble_file): Use the caching wrappers, for when many *getfilecon
calls would fail with ENOTSUP or EOPNOTSUPP.
Suggested by Sven Breuner in
http://thread.gmane.org/gmane.comp.gnu.coreutils.general/2187
Improved-by: Pádraig Brady.
Allow changing the --numeric-suffixes start number
from the default of 0.
* src/split.c (next_file_name): Initialize the suffix index
and the output filename according to start value.
(main): Check that the suffix length is large enough for the
numerical suffix start value.
* doc/coreutils.texi (split invocation): Mention it.
* NEWS (New features): Mention it.
* tests/split/numeric: New file. Test --numeric-suffixes[=FROM].
* tests/Makefile.am (TESTS): Reference the new test.
