GitLab

* NEWS: Record release date.

* tests/du/threshold.sh: use `cut` rather than
sed to avoid using the non portable \t which
fails on sed on openbsd 5 at least.
Also remove a redundant call to `tr` and avoid
explicit setting of LANG=C which is done globally.

On linkers that don't remove unused functions,
there will be a reference to a missing dev_debug symbol
in the devmsg() function.  So for now ...

* src/system.h: ... move devmsg() from here ...
* src/numfmt.c: ... to here, and document future cleanup.
* src/factor.c: Likewise.

src/numfmt.c (parse_format_string): On some systems, strtol() returns
EINVAL if no conversion was performed.  So only handle ERANGE here,
and handle other format errors directly.

* tests/local.mk (check-root): Restrict to SUBDIRS=. as traversing
into gnulib-tests induces a false failure.

* src/numfmt.c (usage): Keep a single space between the "K = 1000",
so that it's not displayed on a separate line.  Also place ','
between each unit entry to improve readability.

* src/numfmt.c (usage): Correct synopsis and make command description
clearer.  Start option descriptions with lowercase letter; use
semicolon instead of period where needed; indent continuation lines;
gettextize single options for ease of translation and maintenance;
sort options alphabetically.
* doc/coreutils.texi (numfmt invocation): Sort numfmt options
alphabetically.  Enforce double-blank-after-period style.

This addresses http://bugs.gnu.org/13681.
Improved-by: Bernhard Voelker

Both factor and numfmt recently introduced debug messages
for developers, enabled by --verbose and ---devdebug respectively.
There were a few issues though:
 1. They used different mechanisms to enable these messages.
 2. factor used --verbose which might be needed for something else
 3. They used different methods to output the messages,
    and numfmt used error() which added an unwanted newline
 4. numfmt marked all these messages for translation and factor
    marked a couple.  We really don't need these translated.
So we fix the above issues here while renaming the enabling
option for both commands to ---debug (still undocumented).

* src/factor.c (verbose): Rename to dev_debug and change from int to
bool as it's just a toggle flag.
(long_options): Rename --verbose to ---debug.
* src/system.h (devmsg): A new inline function to output a message
if enabled by a global dev_debug variable in the compilation unit.
* src/numfmt.c: Use devmsg() rather than error().
Also remove the translation tags from these messages.
Also change debug flag to bool from int.
* tests/misc/numfmt.pl: Adjust for the ---devdebug to ---debug change.
* cfg.mk (sc_marked_devdiagnostics): Add a syntax check to ensure
translations are not added to devmsg calls.

Reported by Göran Uddeborg in http://bugs.gnu.org/13665

* tests/tail-2/inotify-rotate.sh: Avoid a subshell with bash,
which in turn causes the `kill` to be ineffective to the tail
processes (as the SIGTERM is sent to the subshell which doesn't
propagate the signal on to its children).  On NFS the test
cleanup will then fail as there will be .nfs files maintained
in the directory for the files still opened by the tail processes.
Reported by Bernhard Voelker.

* tests/misc/numfmt.pl: When the system locale grouping doesn't
match our expected format for grouping 1234 in the fr_FR locale,
reset the locale to 'C' so as to skip all locale tests.

* tests/cp/fail-perm.sh: Adjust expected diagnostic to match
just-changed cp diagnostic.
* tests/ln/hard-to-sym.sh: Likewise.
* .mailmap: Also map my new address.

Notes tests/init.sh is still in sync with gnulib

* bootstrap: update to latest
* gnulib: update avoiding secure_getenv and subsequent patches
as these are reported to fail on FreeBSD at least.

Note we use "failed to {access,close}" for those single operations,
and "error {read,writ}ing" for those partial operation failures.

* src/copy.c: Improve error messages for failing read, write and close.
* src/cp.c: Improve error messages for failing access.
* src/dd.c: Improve error messages for failing read, write and open.
* src/head.c: Improve error message for failing close.
* src/install.c: Improve error messages for failing access.
* src/ln.c: Likewise.
* src/mv.c: Likewise.
* src/touch.c: Improve error message for failing close.
* src/truncate.c: Likewise.

* src/numfmt.c (usage): Change erroneous "G" to "M".

Originally requested in Red Hat bugzilla #445213.

* src/stty.c (mode_info): Add support for DTR/DSR hardware flow control,
if available.
* doc/coreutils.texi: Document it.
* tests/misc/stty.sh: Add it to the list of serial options to avoid.
* NEWS: Mention the improvement.

Prompted by the continuous integration build failure at:
http://hydra.nixos.org/build/4010493

* src/numfmt.c (parse_format_string): Correct both sign and size of
a printf format, which only gives a warning on 32 bit builds.

* AUTHORS: Add my name.
* NEWS: Mention the new program.
* README: Reference the new program.
* src/numfmt.c: New file.
* src/.gitignore: Ignore the new binary.
* build-aux/gen-lists-of-programs.sh: Update.
* scripts/git-hooks/commit-msg: Allow numfmt: commit prefix.
* po/POTFILES.in: Add new c file.
* tests/misc/numfmt.pl: A new test file giving >93% coverage.
* tests/local.mk: Reference the new test.
* man/.gitignore: Ignore the new man page.
* man/local.mk: Reference the new man page.
* man/numfmt.x: A new template.
* doc/coreutils.texi: Document the new command.

Fixes the issue introduced in unreleased commit v8.20-60-gec48bead.

* src/cut.c (set_fields): Don't access the bit array if
we've an open ended range that's outside any finite range.
* tests/misc/cut.pl: Add tests for this case.
Reported by Marcel Böhme in http://bugs.gnu.org/13627

This fixes Bug#12115, reported by Reuben Thomas.
* doc/coreutils.texi (tac invocation): Document how to reverse a
file character by character.  Break out MS-DOS into a separate
section, like 'cat' does.

Like any other pseudo file system, df should show rootfs only
when the -a option is specified, i.e. specifying -trootfs alone
is not sufficient.  As the rootfs entry is now elided by the
general deduplication in filter_mount_list (commit v8.20-103-gbb116d35),
all other references to rootfs can be removed again.

* src/df.c (show_rootfs): Remove global variable.
(ROOTFS): Remove constant.
(filter_mount_list): Remove case to handle rootfs specially.
(main): In the case for handling the -t option, remove setting
of the show_rootfs variable.
* tests/df/skip-rootfs.sh: Adapt the test case "df -t rootfs":
the rootfs file system must not be printed (because no -a).
* doc/coreutils.texi (df invocation): Correct the documentation
about eliding mount entries: it is not the first occurrence of
the the device which wins, but now rather the entry with the
shortest mount point name.  Also adapt the description about
eliding pseudo file system types like rootfs.
* NEWS (Changes in behavior): Adapt entry.

* src/df.c (struct devlist): Add a new element for storing
pointers to mount_entry structures.
(devlist_head, dev_examined): Remove.
(filter_mount_list): Add new function to filter out the rootfs
entry (unless -trootfs is specified), and duplicities. The
function favors entries with a '/' character in me_devname
or those with the shortest me_mountdir string, if multiple
entries fulfill the first condition.
Use struct devlist to build up a list of entries already known,
and finally rebuild the global mount_list.
(get_all_entries): Call the above new function unless the -a
option is specified.
(get_dev): Remove the code for skipping rootfs and duplicities.
* tests/df/skip-duplicates.sh: Add test cases.
Co-authored-by: Bernhard Voelker <mail@bernhard-voelker.de>

* src/timeout.c (unblock_signal): A new function to unblock a
specified signal, or warn if not possible.
(set_timeout): Ensure SIGALRM is unblocked before we setup the timer.
* tests/misc/timeout-blocked.pl: A new test for the issue.
* tests/local.mk: Reference the new test.
* NEWS: Mention the fix.
Fixes: http://bugs.gnu.org/13535

* src/seq.c (main): With 3 positive integer args we were
checking the end value was == "1", rather than the step value.
* tests/misc/seq.pl: Add tests for this case.
Reported by Marcel Böhme in http://bugs.gnu.org/13525

* src/seq.c (get_default_format): Also account for the case where '.'
is auto added to the start value, which is significant when the
number sequence narrows.
* tests/misc/seq.pl: Add two new tests for the failing cases.
* NEWS: Mention the fix.
Fixes http://bugs.gnu.org/13394

* src/cut.c (cut_fields): Handle the edge case where '\n' is
the delimiter, which could be used for example to suppress
the last line if it doesn't contain a '\n'.
* test/misc/cut.pl: Add tests for this edge case.

Previously line N+1 was inspected before line N was fully output,
which causes output ordering issues at the terminal or delays
from intermittent sources like tail -f.

* src/cut.c (cut_fields): Adjust so that we record the
previous output character so we can use that info to
determine wether to output a '\n' or not.
* tests/misc/cut.pl: Add tests to ensure existing
functionality isn't broken.
* NEWS: Mention the fix.
Fixes bug http://bugs.gnu.org/13498

* bootstrap.conf (gnulib_modules): Add statat.
The fstatat module was split in two, and we need both halves.

* man/od.x: s/-w 16/-w16/ as -w takes an optional
parameter and so the space is significant.

* src/du.c (usage): Bring options into alphabetical order.
* doc/coreutils.texi (du invocation): Likewise.
Furthermore, use the @itemx macro for the long options
--max-depth and --threshold instead of @item.

Each program with at least one long option which is marked as
'required_argument' and which has also a short option for that
option, should print a note about mandatory arguments.
Define that well-known note centrally and use it rather than
literal printf/fputs, and add it where it was missing.

* src/system.h (emit_mandatory_arg_note): Add new function.

* src/cp.c (usage): Use it rather than literal printf/fputs.
* src/csplit.c, src/cut.c, src/date.c, src/df.c, src/du.c:
* src/expand.c, src/fmt.c, src/fold.c, src/head.c, src/install.c:
* src/kill.c, src/ln.c, src/ls.c, src/mkdir.c, src/mkfifo.c:
* src/mknod.c, src/mv.c, src/nl.c, src/od.c, src/paste.c:
* src/pr.c, src/ptx.c, src/shred.c, src/shuf.c, src/sort.c:
* src/split.c, src/stdbuf.c, src/tac.c, src/tail.c, src/timeout.c:
* src/touch.c, src/truncate.c, src/unexpand.c, src/uniq.c:
Likewise.

* src/base64.c (usage): Add call of the above new function
because at least one long option has a required argument.
* src/basename.c, src/chcon.c, src/date.c, src/env.c:
* src/nice.c, src/runcon.c, src/seq.c, src/stat.c, src/stty.c:
Likewise.

* src/du.c (opt_threshold): Add variable to hold the value of
the --threshold option specified by the user.
(long_options): Add a required_argument entry for the new
--threshold option.
(usage): Add --threshold option.
(process_file): Elide printing the entry if its size does not
meet the value specified by the --threshold option.
(main): In the argument parsing loop, add a case for the new
-t option. Convert the given argument by permitting the
well-known suffixes for megabyte, gigabytes, etc.
Handle the special case "-0": give an error as this value is
not permitted.
* doc/coreutils.texi (du invocation): Add documentation for the
above new option.
* tests/du/threshold.sh: Add new test to exercise the new option.
* tests/local.mk (all_tests): Mention the above test.
Co-authored-by: Bernhard Voelker <mail@bernhard-voelker.de>

This test tried to ensure that not all symlinks (across all
file system types) have Zero size and refers to a change
in system.h from 2002-08-31 (commit SH-UTILS-2_0_15-55-g62808a7f).
The test used to do this by working on symlinks to long file
names.  This assumption is dependant on the underlying file
system, and in some environments like XEN does not even work
on file systems known to work otherwise.

The test for dereferencing and no-dereferencing symlinks is
already covered by other tests (du/deref.sh, du/deref-args.sh,
and du/no-deref.sh).  Therefore, remove it.

* tests/du/slink.sh: Remove file.
* tests/local.mk (all_tests): Remove the above test.

Discussed in:
http://lists.gnu.org/archive/html/coreutils/2013-01/msg00053.html

Since commit v8.20-67-g0f525b6f, .gitignore sometimes
showed up as changed because the entries "*.gcda" and
"*.gcno" had not been in alphabetical order.

* .gitignore: Exchange the entries "*.gcda" and "*.gcno".

* src/uptime.c: Add calls to gettext() and select_plural().

* src/ln.c (usage): Move a newline to the next text fragment, so
the preceding fragment about backup methods becomes the same as
the ones for cp, mv, and install.  A bit easier for translators.

In addition to the previous 64 bit guards we've placed in longlong.h
there are additional _LP64 guards required for mips with -mcpu >= 3,
to avoid a build failure (http://bugs.gnu.org/13353) and on sparc
with -mcpu >= v9 in 32 bit mode where for example,
`factor 2123123123123123123123` would go into an infinite loop.

Since factor.c currently operates on uintmax_t, we restrict the use
of the assembly in longlong.h to when 'long' has the same width, to
provide a more general guard for this code.

* src/factor.c: Restrict the use of longlong.h assembly code,
to when the width of intmax_t == long.
* src/longlong.h: Remove the previous _LP64 guards to avoid
divergence from GMP's longlong.h
* NEWS: Adjust the info on build and runtime fixes.

* doc/coreutils.texi (Top): Sync from gnulib parse-datetime.texi menu.

* doc/coreutils.texi: avoid @sc and use explicit capitals.
* doc/local.mk (sc-use-small-caps-NUL): Remove, as no longer applicable.

* coreutils.texi (pr invocation): remove list of ancient news
items; the main documentation already covers what is needed.

Run "make update-copyright", but then also run this,
  perl -pi -e 's/2\d\d\d-//' tests/sample-test
to make that one script use the single most recent year number.