tests: mktemp: ensure mktemp does not depend on getrandom and ASLR

* tests/mktemp/mktemp-misc.sh: Add new test.
* tests/local.mk: Reference new test.
https://github.com/coreutils/coreutils/pull/206

maint: tests: decouple debug output determination

* tests/misc/warning-errors.sh: Simply check there is output
to stderr before checking that output induces an error.

tests: avoid false test failure when using address sanitizer

* tests/misc/warning-errors.sh: Skip commands which have been built with
sanitizers, since standard error will not be closed and checked for
errors.
Reported by Bruno Haible.

tests: avoid failure on systems without an optimized 'cksum' or 'wc -l'

* tests/misc/warning-errors.sh: Expect 'wc' and 'cksum' to exit
successfully if there is not an optimized 'wc -l' implementation or
CRC32 implementation.
Reported by Bruno Haible.

tests: shuf: ensure memory exhaustion is handled gracefully

* tests/shuf/shuf.sh: Ensure we exit 1 upon failure
to allocate memory.
https://github.com/uutils/coreutils/issues/11170
https://github.com/coreutils/coreutils/pull/209

test: cp: add test for non-UTF8 directory names

Missing test identified here:
 https://github.com/uutils/coreutils/pull/11148

* tests/cp/non-utf8-name.sh: Add a new test to cover this case.
https://github.com/coreutils/coreutils/pull/207

du: fflush after outputting a line

* src/du.c (print_size): Resurrect the fflush call, since there
can be significant delay between output lines.

tests: wc,du: add additional --files0-from test cases

* tests/wc/wc-files0-from.pl ($limits): New variable.
(@Tests): Prefer the error strings from getlimits over writing them by
hand. Add test cases for --files0-from listing missing files and
duplicate files.
* tests/du/files0-from.pl ($limits): New variable.
(@Tests): Prefer the error strings from getlimits over writing them by
hand. Add test cases for --files0-from listing missing files. Add tests
for --files0-from listing duplicate files with and without the -l option
also in use.

build: update gnulib submodule to latest

* po/POTFILES.in: Remove recently added lib/cygpath.c dependency after
gnulib commit 2a893de047 (filesystem-remote: New module., 2026-02-28).

tests: ls: treat invalid UTF-8 paths starting with a dot as hidden

* tests/ls/non-utf8-hidden.sh: Add the test case.
https://github.com/uutils/coreutils/pull/11135
https://github.com/coreutils/coreutils/pull/202

tests: ln: verify that -f and -i override each other

Identified here:
<https://github.com/uutils/coreutils/pull/11129>

* tests/ln/misc.sh: Add the check.
 https://github.com/coreutils/coreutils/pull/199

test: ln: verify backup suffix path traversal prevention

missing test detected thanks to:
https://github.com/uutils/coreutils/pull/11149

* tests/ln/backup-suffix-traversal.sh: Add a test.
https://github.com/coreutils/coreutils/pull/208

maint: fix typo in previous test

* tests/shuf/shuf.sh: Use non varying $ret rather than $?

tests: shuf: ensure we handle unsupported getrandom syscall gracefully

* tests/shuf/shuf.sh: Check we fail normally or succeed where
the getrandom syscall is not available.
https://github.com/coreutils/coreutils/pull/205

build: update gnulib to latest

* NEWS: Mention the more encompassing remoteness check for df.
* po/POTFILES.in: Add new lib/cygpath.c dependency.

du: avoid locking and flushing standard output

This results in a noticeable increase in performance:

    $ yes /dev/null | head -n 10000000 | tr '\n' '\0' \
        | time --format=%E ./src/du-prev -l --files0-from=- > /dev/null
    0:20.40
    $ yes /dev/null | head -n 10000000 | tr '\n' '\0' \
        | time --format=%E ./src/du -l --files0-from=- > /dev/null
    0:16.57

* src/du.c (print_size): Prefer putchar and fputs which may be unlocked
unlike printf. Prefer ferror to fflush.

stat: handle %%%N too

* src/stat.c (main): Fix incorrect counting of '%'s before 'N'.
* tests/stat/stat-fmt.sh: Test for the bug.

id: avoid unnecessary buffer flushing

* src/groups.c (main):
* src/id.c (main, print_stuff):
Don’t flush stdout before testing for write error.
Do the test only when in a loop, as a one-shot will
test for write error soon anyway.

cksum: prefer signed int

* src/cksum.c (min_digest_line_length, digest_hex_bytes)
(digest_length, md5_sum_stream, sha1_sum_stream)
(sha224_sum_stream, sha256_sum_stream, sha384_sum_stream)
(sha512_sum_stream, sha2_sum_stream, sha3_sum_stream)
(blake2b_sum_stream, sm3_sum_stream, problematic_chars)
(filename_unescape, valid_digits, bsd_split_3)
(algorithm_from_tag, split_3, digest_file, output_file)
(b64_equal, hex_equal, digest_check, main):
* src/cksum_avx2.c (cksum_avx2):
* src/cksum_avx512.c (cksum_avx512):
* src/cksum_crc.c (cksum_fp_t, cksum_slice8, crc_sum_stream)
(crc32b_sum_stream, output_crc):
* src/cksum_pclmul.c (cksum_pclmul):
* src/cksum_vmull.c (cksum_vmull):
* src/sum.c (bsd_sum_stream, sysv_sum_stream, output_bsd, output_sysv):
Prefer signed to unsigned int where either will do.
This allows better checking with -fsanitize=undefined.
It should also help simplify future patches, so that they
needn’t worry whether comparisons like ‘i < len - 2’ will misbehave.

stat: don't check QUOTING_STYLE when --printf %%N is used

* NEWS: Mention the fix.
* src/stat.c (main): Only check QUOTING_STYLE if there is a %N that is
not preceded by a percentage sign.
* tests/stat/stat-fmt.sh: Add some test cases.

id: promptly diagnose write errors

* NEWS: Mention the improvement.
* src/id.c (print_stuff): Call fflush for each listed user to check for
write errors.
* tests/misc/io-errors.sh: Add an invocation of 'id'.

groups: promptly diagnose write errors

* NEWS: Mention the improvement.
* src/groups.c (main): Call fflush for each listed user to check for
write errors.
* tests/misc/io-errors.sh: Add an invocation of 'groups'.

tests: ensure failure to write warnings is handled gracefully

* tests/misc/warning-errors.sh: Add a new test to ensure
failure to write warnings is diagnosed in the exit status.
* tests/local.mk: Reference the new test.

tests: shuf: ensure randomization doesn't depend solely on ASLR

* tests/shuf/shuf.sh: Use setarch --addr-no-randomize to disable
ASLR, and show the output is still random.
https://github.com/coreutils/coreutils/pull/198

maint: fix description of tests/misc/io-errors.sh

* tests/misc/io-errors.sh: Promptness is checked in
write-errors.sh, not this test.

tests: nice: ensure a particular adjustment is disallowed

* tests/nice/nice-fail.sh: Ensure "1+2-3" is disallowed.
https://github.com/coreutils/coreutils/pull/197

tests: factor,numfmt: verify embedded NUL handling

* tests/factor/factor.pl: Verify that embedded NULs
on stdin terminate the _number_.
* tests/numfmt/numfmt.p: Verify that embedded NULs
on stdin terminate the _line_.
https://github.com/coreutils/coreutils/pull/196

tests: fix "Hangup" termination of non-interactive runs

This avoids the test harness being terminated like:
  make[1]: *** [Makefile:24419: check-recursive] Hangup
  make[3]: *** [Makefile:24668: check-TESTS] Hangup
  make: *** [Makefile:24922: check] Hangup
  make[2]: *** [Makefile:24920: check-am] Hangup
  make[4]: *** [Makefile:24685: tests/misc/usage_vs_refs.log] Error 129
  ...

This happened sometimes when the tests were being run non interactively.
For example when run like:

  setsid make TESTS="tests/timeout/timeout.sh \
   tests/tail/overlay-headers.sh" SUBDIRS=. -j2 check

Note the race window can be made bigger by adding a sleep
after tail is stopped in overlay-headers.sh

The race can trigger the kernel to induce its job control
mechanism to prevent stuck processes.
I.e. where it sends SIGHUP + SIGCONT to a process group
when it determines that group may become orphaned,
and there are stopped processes in that group.

* tests/tail/overlay-headers.sh: Use setsid(1) to keep the stopped
tail process in a separate process group, thus avoiding any kernel
job control protection mechanism.
* tests/timeout/timeout.sh: Use setsid(1) to avoid the kernel
checking the main process group when sleep(1) is reparented.
Fixes https://bugs.gnu.org/80477

doc: tee: avoid the use of gpg cleartext signatures in an example

Cleartext signatures have many gotchas. Therefore, the use of detached
signatures is recommended where possible. See:
<https://gnupg.org/blog/20251226-cleartext-signatures.html>.

* doc/coreutils.texi (tee invocation): Adjust gpg invocation to produce
a detached signature.

tests: whoami, logname: verify error handling

* tests/df/no-mtab-status-masked-proc.sh: Tweak unshare check.
* tests/local.mk: Reference new test.
* tests/misc/user.sh: Add new test using unshare -U, to verify
that whoami and logname diagnose failure correctly.
https://github.com/coreutils/coreutils/pull/195

doc: stty: mention the -g does not save the terminal window size

* doc/coreutils.texi (stty invocation): Mention that 'stty -g' does not
save the terminal window size as allowed by POSIX.1-2024.

tests: shuf: check that write errors are promptly diagnosed

* tests/misc/write-errors.sh: Add a shuf invocation that would run
forever if write errors weren't diagnosed.

shuf: avoid locking standard output when using --input-range

Here is the throughput before this patch:

    # write_permuted_numbers
    $ ./src/shuf-prev -i 0-100000000 | pv -r > /dev/null
    [ 153MiB/s]

    # write_random_numbers
    $ timeout 10 ./src/shuf-prev -i 0-100000 -r | pv -r > /dev/null
    [78.6MiB/s]

Here is the throughput after this patch:

    # write_permuted_numbers
    $ timeout 10 ./src/shuf -i 0-100000000 | pv -r > /dev/null
    [ 308MiB/s]

    # write_random_numbers
    $ timeout 10 ./src/shuf -i 0-100000 -r | pv -r > /dev/null
    [ 196MiB/s]

* NEWS: Mention the performance improvement.
* src/shuf.c (write_permuted_numbers, write_random_numbers): Prefer
fputs and fputc which may be unlocked over printf which locks standard
output.

tests: chmod: test that chmod(2) is always called

* tests/chmod/only-op.sh: New file.
* tests/local.mk (all_test): Add the new test.

maint: printf: prefer static initialization

* src/printf.c (main): Remove unnecessary initialization.

maint: fmt: prefer static initialization

* src/fmt.c (prefix, max_width): Initialize variables.
(main): Remove unnecessary initializations.

maint: sort: prefer static initialization

* src/sort.c (main): Remove unnecessary initialization.

maint: df: prefer static initialization

* src/df.c (human_output_opts, grand_fsu): Initialize variables.
(main): Remove unnecessary initializations.

maint: split: prefer static initialization

* src/split.c (outbase, infile): Initialize variables.
(main): Remove unnecessary initializations.

maint: tail: prefer static initialization

* src/tail.c (count_lines, line_end): Initialize variables.
(main): Remove unnecessary initializations.

maint: csplit: prefer static initialization

* src/csplit.c (prefix, remove_files): Initialize variables.
(main): Remove unnecessary initializations.

maint: comm: prefer static initialization

* src/comm.c (only_file_1, only_file_2, both): Initialize variables.
(main): Remove unnecessary initializations.

maint: wc: prefer static initialization

* src/wc.c (main): Remove unnecessary initializations.

maint: tac: prefer static initialization

* src/tac.c (separator, separator_ends_record, sentinel_length):
Initialize variables.
(main): Remove unnecessary initializations.

maint: join: prefer static initialization

* src/join.c (print_pairables): Initialize variable.
(main): Remove unnecessary initializations.

maint: cut: prefer static initialization

* src/cut.c (main): Remove unnecessary initializations.

maint: seq: prefer static initialization

* src/seq.c (separator): Initialize variable.
(main): Remove unnecessary initializations.

maint: paste: prefer static initialization

* src/paste.c (main): Remove unnecessary initializations.

maint: install: prefer static initialization

* src/install.c (main): Remove unnecessary initializations.

maint: chmod: prefer static initialization

* src/chmod.c (main): Remove unnecessary initializations.

maint: head: prefer static initialization

* src/head.c (line_end): Initialize variable.
(main): Remove unnecessary initializations.

maint: ln: prefer static initialization

* src/ln.c (main): Remove unnecessary initializations.

maint: nl: prefer static initialization

* src/nl.c (main): Remove unnecessary initialization.

maint: touch: prefer static initialization

* src/touch.c (main): Remove unnecessary initializations.

maint: fold: prefer static initialization

* src/fold.c (main): Remove unnecessary initializations.

maint: tee: prefer static initialization

* src/tee.c (main): Remove unnecessary initializations.

maint: rmdir: prefer static initialization

* src/rmdir.c (main): Remove unnecessary initialization.

maint: tty: prefer static initialization

* src/tty.c (main): Remove unnecessary initialization.

maint: tests: DRY a recent change

* tests/pwd/pwd-long.sh: Adjust so complicated command is not repeated.

wc: add aarch64 Neon optimization for wc -l

Here is an example of the performance improvement:

    $ yes abcdefghijklmnopqrstuvwxyz | head -n 100000000 > input
    $ time ./src/wc-prev -l < input
    100000000

    real 0m0.793s
    user 0m0.630s
    sys 0m0.162s
    $ time ./src/wc -l < input
    100000000

    real 0m0.230s
    user 0m0.065s
    sys 0m0.164s

* NEWS: Mention the performance improvement.
* gnulib: Update to the latest commit.
* configure.ac: Check the the necessary intrinsics and functions.
* src/local.mk (noinst_LIBRARIES) [USE_NEON_WC_LINECOUNT]: Add
src/libwc_neon.a.
(src_libwc_neon_a_SOURCES, wc_neon_ldadd, src_libwc_neon_a_CFLAGS)
[USE_NEON_WC_LINECOUNT]: New variables.
(src_wc_LDADD) [USE_NEON_WC_LINECOUNT]: Add $(wc_neon_ldadd).
* src/wc.c [USE_NEON_WC_LINECOUNT]: Include sys/auxv.h and asm/hwcap.h.
(neon_supported) [USE_NEON_WC_LINECOUNT]: New function.
(wc_lines) [USE_NEON_WC_LINECOUNT]: Use neon_supported and
wc_lines_neon.
* src/wc.h (wc_lines_neon): Add declaration.
* src/wc_neon.c: New file.
* doc/coreutils.texi (Hardware Acceleration): Document the "-ASIMD"
hwcap and the variable used in ./configure to override detection of Neon
instructions.
* tests/wc/wc-cpu.sh: Also add "-ASIMD" to disable the use of Neon
instructions.

tests: use system ENOTEMPTY message

* tests/mv/dir2dir.sh: Map system specific messages.
* tests/rm/d-2.sh: Likewise.

maint: tests: avoid ERANGE specific handling

* tests/fmt/base.pl: Verify ERANGE is diagnosed.
* tests/od/od.pl: Adjust determination of ERANGE.

maint: tests: avoid ELOOP specific handling

* tests/du/long-sloop.sh: Adjust determination of ELOOP.

maint: tests: avoid ENOSPC specific handling

* tests/csplit/csplit-io-err.sh: Adjust determination of ENOSPC.
* tests/csplit/split-io-err.sh: Likewise.
* tests/fold/fold-zero-width.sh: Verify full ENOSPC message.
* tests/tac/tac-continue.sh: Likewise.
* tests/misc/io-errors.sh: Likewise.
* tests/misc/write-errors.sh: Likewise.

tests: use system ENOSPC message

* tests/head/head-write-error.sh: Verify ENOSPC is output.
* tests/misc/yes.sh: Likewise.

tests: use system ENOTSUP message

* tests/cp/cp-mv-enotsup-xattr.sh: Don't hardcode "Operation not
supported".
* tests/cp/cp-a-selinux.sh: Likewise.

maint: tests: avoid ENOTSUP specific handling

* tests/mv/mv-exchange.sh: Use ENOTSUP rather than
transforming variations.

tests: use system EACCES message

* tests/chgrp/no-x.sh: Don't harcode "Permission denied".
* tests/chmod/no-x.sh: Likewise.
* tests/du/inacc-dest.sh: Likewise.
* tests/du/no-x.sh: Likewise.
* tests/misc/nohup.sh: Likewise.
* tests/mv/part-fail.sh: Likewise.
* tests/mv/perm-1.sh: Likewise.
* tests/rm/fail-eacces.sh: Likewise.
* tests/rm/rm1.sh: Likewise.
* tests/rm/rm2.sh: Likewise.
* tests/rm/unread2.sh: Likewise.
* tests/rm/unreadable.pl: Likewise.

maint: tests: avoid AIX specific EACCES handling

* tests/rm/inaccessible.sh: Remove AIX specific handling.
* tests/cp/fail-perm.sh: Likeise.  Also remove confusing
EPERM/HPUX mentions.
* tests/misc/sync.sh: Don't hardcode EISDIR and EACCES.

maint: tests: avoid EPERM/EACCES specific handling

* tests/Coreutils.pm: Remove associated comment.
* tests/mv/sticky-to-xpart.sh: Use $EACCES and $EPERM.
* tests/rm/fail-2eperm.sh: Likewise.
* tests/touch/not-owner.sh: Likewise.

maint: tests: avoid EPERM specific handling

* tests/dd/fail-ftruncate-fstat.sh: Use $EPERM rather than
mapping specific strings to the Linux $EPERM.
* tests/mkdir/smack-no-root.sh: Use $EPERM rather than hardcoding.

maint: expr: reduce variable scope

* src/expr.c (mbs_logical_cspn, main, trace, docolon, eval7, eval6)
(eval5, eval4, eval3, eval2, eval1, eval): Declare variables where they
are used instead of at the start of the function.

tests: pwd: ensure our getcwd fallback is tested

* tests/pwd/pwd-long.sh: Avoid the getcwd syscall,
to ensure our fallback getcwd code is exercised.
* NEWS: Mention the recent bug fix in pwd.

pwd: fix heap buffer overflow in file_name_prepend

file_name_prepend works by right-aligning path data in a growing buffer.
When the buffer is too small, it then allocates a new buffer via
xpalloc() and copies existing data to the end of the new buffer.

Unfortunately, the memcpy destination is computed as buf + p->n_alloc -
n_free, but xpalloc has already updated p->n_alloc to the new (larger)
allocation size while n_free still reflects the old state. This places
the data at too high an offset, writing past the end of the buffer.

Update to properly calculate the destination offset.

Fixes: v9.5-171-g61ab25c35 ("pwd: prefer xpalloc to xnrealloc")

build: update gnulib submodule to latest

maint: seq,who: prefer xasprintf

* src/seq.c (main): Use xasprintf instead of asprintf followed by a
check if the allocation fails.
* src/who.c (print_line): Likewise.

maint: seq: reduce variable scope

* src/seq.c (long_double_format, print_numbers, main): Declare variables
where they are used instead of at the start of the function.

tests: date: check that options can follow the format string

* tests/date/date.pl (@Tests): Add the test case.

doc: relax github issue/pull-request instructions

Be more accepting of github issues and pull-requests.
Be a bit less verbose in info provided.

tests: sort: ensure --batch-size=(rlimit+1) fails

* tests/sort/sort-merge-fdlimit.sh: Add test case.
https://github.com/coreutils/coreutils/pull/182

tests: coreutils: remove redundant validity check

* tests/misc/coreutils.sh: Remove the check for non-empty coreutils.h,
as the existing `print_ver_ coreutils` is sufficient to determine
support for single binary mode.
https://github.com/coreutils/coreutils/pull/189

tests: cp: ensure --debug output failure is diagnosed

* tests/cp/debug.sh: cp diagnostics are written to stdout,
so ensure output errors are diagnosed.
https://github.com/coreutils/coreutils/pull/192

tests: arch: add option parsing check

* tests/misc/arch.sh: Ensure -- to indicate end of options is accepted.
https://github.com/coreutils/coreutils/pull/191

tests: date: check dotted EU dates in combination

* tests/date/date.pl: Add other date elements
to ensure abbreviated dotted dates are parsed appropriately.
Corresponds to gnulib commit 451042cb76

build: update gnulib to latest

tests: determine errno string more efficiently

* tests/misc/read-errors.sh: Use getlimits_ determined EIO error string,
rather than inferring the string from bash's output.
See https://bugs.gnu.org/80353

tests: getlimits: output error strings

* src/getlimits.c (main): Iterate over defined errnos,
and output shell compatible error strings.
* tests/Coreutils.pm: Adjust so shell quotes are stripped.

maint: pinky: reduce variable scope

* src/pinky.c (create_fullname, idle_string, print_entry)
(print_long_entry, main): Declare variables where they are used instead
of at the start of the function. Initialize variables where they are
declared.

maint: remove extraneous code from test

* tests/misc/usage_vs_refs.sh: Remove copy and paste slop.
* tests/local.mk: Fix indentation.

nl: support multi-byte section delimiters

* NEWS: Mention the improvement.
* src/nl.c: Include mcel.h.
(DEFAULT_SECTION_DELIMITERS): Resize to fit 2 multi-byte characters.
(section_del_len): New variable.
(check_section): Compare against section_del_len instead of 2.
(main): Support multi-byte characters for the -d option.
* tests/nl/multibyte.sh: New file.
* tests/nl/nl.sh: New file, moved from tests/misc/nl.sh.
* tests/local.mk (all_tests): Add the new test. Adjust the existing
tests file name.
* cfg.mk (exclude_file_name_regexp--sc_space_tab): Adjust Adjust the
existing tests file name.

tests: fix false failure with dash or ksh

* tests/misc/read-errors.sh: Avoid dash or ksh as they
do _not_ print an error message upon EIO from read.
Fixes https://bugs.gnu.org/80353

date: add tests and NEWS for dd.mm.yy support

* NEWS: Mention the new feature.
* tests/date/date.pl: Add test cases.

build: update to latest gnulib

* gnulib: Update to pick up support for dd.mm.yy format
in parse-datetime.

build: fix --enable-single-binary=hardlinks with dash

With dash as /bin/sh, you get the error
```
checking for sys/capability.h... yes
./configure: 95775: test: no: unexpected operator
checking for working fork... yes
```

* configure.ac: Use '=' in test for equality, not '==',
for POSIX shell compatibility.
* NEWS: Mention the build fix.

tests: ptx: ensure bad unicode is supported

* tests/ptx/ptx-overrun.sh: Ensure bad_unicode() is processed.

tests: cp: ensure read-only dir perms preserved

* tests/cp/readonly-dir.sh: Add new test.
* tests/local.mk: Reference new test.
Identified here https://github.com/uutils/coreutils/issues/7961

doc: NEWS: mention the recent kill documentation fix

* NEWS: Mention that the kill documentation now has the anchors expected
by 'kill --help'.

tests: fold: check that all files are processed

* tests/fold/multiple-files.sh: New file.
* tests/local.mk (all_tests): Add the test.

tests: nl: check that all files are processed

* tests/misc/nl.sh: Add a test case.

maint: unexpand: remove dead store

The assign to next_tab_column in the backspace branch is never
read; in fact it's always written and then read solely inside the
blank branch.

Move the declaration of 'next_tab_column' down into the blank
branch, and remove its assignment in the backspace branch.

Spotted by the 'infer' static checker.

* src/unexpand.c (unexpand): Remove dead store of next_tab_column.