]> git.ipfire.org Git - thirdparty/coreutils.git/log
thirdparty/coreutils.git
5 weeks agobuild: fold: fix build failure with C99
Pádraig Brady [Wed, 27 Aug 2025 10:48:13 +0000 (11:48 +0100)] 
build: fold: fix build failure with C99

GCC 10.2 gave the following error:
"error: label at end of compound statement"

* src/fold.c (fold_file): Add a ";" to avoid C2X specific syntax.

5 weeks agofold: don't truncate multibyte characters at the end of the buffer
Collin Funk [Tue, 26 Aug 2025 06:15:21 +0000 (23:15 -0700)] 
fold: don't truncate multibyte characters at the end of the buffer

* src/fold.c (fold_file): Replace invalid characters with the original
byte read. Copy multibyte sequences that may not yet be read to the
start of the buffer before reading more bytes.
* tests/fold/fold-characters.sh: Add a test case.

5 weeks agotests: fold: consolidate all fold tests in tests/fold
Pádraig Brady [Mon, 25 Aug 2025 10:39:56 +0000 (11:39 +0100)] 
tests: fold: consolidate all fold tests in tests/fold

* tests/misc/fold.pl: Move from here to ...
* tests/fold/fold.pl: ... here.
* tests/local.mk: Adjust accordingly.

5 weeks agotests: fold: add a memory constraint test
Pádraig Brady [Mon, 25 Aug 2025 10:38:42 +0000 (11:38 +0100)] 
tests: fold: add a memory constraint test

Enforcing this interface behavior is worthwhile
irrespective of our current implementation,
to ensure future or other implementations conform.

* tests/fold/fold-characters.sh: Ensure the fold implementation
uses bounded memory.

6 weeks agofold: use fread instead of getline
Collin Funk [Sun, 24 Aug 2025 19:05:41 +0000 (12:05 -0700)] 
fold: use fread instead of getline

* src/fold.c: Include ioblksize.h.
(fold_file): Use two IO_BUFSIZE-sized buffers. Use fread instead of
getline. Check for if we reached the end of file.

6 weeks agotests: nproc: fix false failure on some systems
Pádraig Brady [Sun, 24 Aug 2025 10:41:01 +0000 (11:41 +0100)] 
tests: nproc: fix false failure on some systems

* tests/nproc/nproc-quota.sh: Also simulate sched_getscheduler()
as this will not be called on older or non linux, or
may return ENOSYS on Alpine.
Fixes https://bugs.gnu.org/79299

6 weeks agomaint: prefer STRUCT_UTMP to struct gl_utmp
Collin Funk [Sat, 23 Aug 2025 21:26:45 +0000 (14:26 -0700)] 
maint: prefer STRUCT_UTMP to struct gl_utmp

* cfg.mk (sc_prohibit-struct-gl_utmp): New rule for 'make syntax-check'.
* src/pinky.c (time_string, print_entry, scan_entries, short_pinky):
Use STRUCT_UTMP instead of struct gl_utmp.
* src/uptime.c (print_uptime, print_uptime, uptime): Likewise.
* src/users.c (list_entries_users, users): Likewise.
* src/who.c (time_string, print_user, print_boottime, print_deadprocs)
(print_login, print_initspawn, print_clockchange, print_runlevel)
list_entries_who, scan_entries, who): Likewise.

6 weeks agotests: cp: ensure copy offload is not disabled for sparse files
Pádraig Brady [Sat, 23 Aug 2025 17:53:17 +0000 (18:53 +0100)] 
tests: cp: ensure copy offload is not disabled for sparse files

Related to commits v9.1-109-g879d2180d and v9.7-248-g306de6c26

* tests/cp/sparse-perf.sh: This edge case was missed a couple of times,
so add a test to ensure we attempt copy offload.

6 weeks agofold: add the --characters option
Collin Funk [Thu, 21 Aug 2025 04:13:52 +0000 (21:13 -0700)] 
fold: add the --characters option

* src/fold.c: Include mcel.h.
(count_bytes): Remove variable.
(counting_mode, last_character_width): New variables.
(shortopts, long_options): Add the option.
(adjust_column): If --characters is in used account for number of
characters instead of their width.
(fold_file): Use getline and iterate over the result with mcel functions
to handle multibyte characters.
(main): Check for the option.
* src/local.mk (src_fold_LDADD): Add $(LIBC32CONV), $(LIBUNISTRING), and
$(MBRTOWC_LIB).
* tests/fold/fold-characters.sh: New file.
* tests/fold/fold-spaces.sh: New file.
* tests/fold/fold-nbsp.sh: New file.
* tests/local.mk (all_tests): Add the tests.
* NEWS: Mention the new option.
* doc/coreutils.texi (fold invocation): Likewise.

6 weeks agocp: improve hole handling on squashfs
Paul Eggert [Sat, 23 Aug 2025 00:34:04 +0000 (17:34 -0700)] 
cp: improve hole handling on squashfs

Better fix for problem reported by Jeremy Allison
<https://bugs.gnu.org/79267>.
* src/copy.c (struct scan_inference): New type, replacing
union scan_inference.  All uses changed.  This is so
infer_scantype can report the first hole's offset when known.
(lseek_copy): 5th arg is now struct scan_inference const *,
not just off_t.  All uses changed.
(infer_scantype): If SEEK_SET+SEEK_HOLE do not find a hole,
fall back on ZERO_SCANTYPE.

6 weeks agocp: go back to copy_file_range optimization
Paul Eggert [Fri, 22 Aug 2025 17:37:50 +0000 (10:37 -0700)] 
cp: go back to copy_file_range optimization

This reverts part of the previous change.
* src/copy.c (lseek_copy): When calling sparse_copy, do not
ask it to scan for zeros unless --sparse=always, so that it
can use copy_file_range which can be far more efficient.

6 weeks agocp: always punch holes that we make
Paul Eggert [Thu, 21 Aug 2025 22:02:10 +0000 (16:02 -0600)] 
cp: always punch holes that we make

Problem reported by Jeremy Allison <https://bugs.gnu.org/79267>.
* src/copy.c (create_hole, sparse_copy): Omit arg PUNCH_HOLES,
as we always punch holes now.  All uses changed.
(lseek_copy): When calling sparse_copy, scan for holes when
sparse_mode == SPARSE_AUTO, as that means we are making holes.
(copy_reg): Always punch any hole made at end.

6 weeks agotests: nproc: add a test for cgroup quotas
Pádraig Brady [Wed, 20 Aug 2025 17:35:29 +0000 (18:35 +0100)] 
tests: nproc: add a test for cgroup quotas

* tests/nproc/nproc-quota.sh: New root only test.
* tests/local.mk: Reference the new test.

6 weeks agomaint: prefer https to http
Collin Funk [Tue, 19 Aug 2025 18:39:04 +0000 (11:39 -0700)] 
maint: prefer https to http

* doc/sort-version.texi (Other version/natural sort implementations):
Use https in documentation link.
* tests/chmod/symlinks.sh: Use https in license text.

6 weeks agonproc: honor cgroup v2 CPU quotas
Pádraig Brady [Tue, 19 Aug 2025 15:49:22 +0000 (16:49 +0100)] 
nproc: honor cgroup v2 CPU quotas

* NEWS: Mention the new feature.
* doc/coreutils.texi (nproc invocation): Mention that
cgroup CPU quotas can limit the reported number.
* gnulib: Update to new nproc gnulib implementation:
https://github.com/coreutils/gnulib/commit/9b07115f4a

6 weeks agologname: list David MacKenzie as the author
Collin Funk [Tue, 19 Aug 2025 03:42:48 +0000 (20:42 -0700)] 
logname: list David MacKenzie as the author

* src/logname.c (AUTHORS): List David as the author.
* AUTHORS: Likewise.

7 weeks agorealpath: improve doc and NEWS
Paul Eggert [Fri, 15 Aug 2025 04:08:00 +0000 (21:08 -0700)] 
realpath: improve doc and NEWS

7 weeks agomaint: avoid syntax-check failure from previous commit
Collin Funk [Fri, 15 Aug 2025 02:43:52 +0000 (19:43 -0700)] 
maint: avoid syntax-check failure from previous commit

* src/tsort.c: Don't include long-options.h since the previous commit
removed the call to parse_gnu_standard_options_only. This avoids a
sc_prohibit_long_options_without_use syntax-check failure.

7 weeks agotsort: add do-nothing -w option
Paul Eggert [Thu, 14 Aug 2025 16:17:51 +0000 (09:17 -0700)] 
tsort: add do-nothing -w option

This is for conformance to POSIX.1-2024
* src/tsort.c: Include getopt.h.
(main): Accept and ignore -w.  Do not bother altering
the usage message, as the option is useless.
* tests/misc/tsort.pl (cycle-3): New test.

7 weeks agomaint: use short form bug URLs
Pádraig Brady [Tue, 12 Aug 2025 13:31:03 +0000 (14:31 +0100)] 
maint: use short form bug URLs

* cfg.mk (sc_prohibit-long-form-bug-urls): Disallow long form in code.
* scripts/git-hooks/commit-msg: Disallow long form in commit messages.
* NEWS: Shorten long urls.
* bootstrap.conf: Likewise.
* configure.ac: Likewise.
* scripts/git-hooks/commit-msg: Likewise.
* src/csplit.c: Likewise.
* src/fmt.c: Likewise.
* src/make-prime-list.c: Likewise.
* src/nohup.c: Likewise.
* tests/od/od-float.sh: Likewise.
* tests/rm/r-root.sh: Likewise.
* tests/tail/inotify-race.sh: Likewise.
* tests/tail/inotify-race2.sh: Likewise.

7 weeks agobasenc: Don't trigger undefined behaviour in mini-gmp
Bruno Haible [Tue, 12 Aug 2025 00:25:41 +0000 (02:25 +0200)] 
basenc: Don't trigger undefined behaviour in mini-gmp

* src/basenc.c (base58_encode): Avoid calling mpz_import on an empty
limb sequence.

8 weeks agorealpath: support the -E option required by POSIX
Collin Funk [Sun, 10 Aug 2025 00:53:29 +0000 (17:53 -0700)] 
realpath: support the -E option required by POSIX

* src/realpath.c (longopts): Add the option.
(main): Likewise.
(usage): Add the option to the --help message.
* tests/misc/realpath.sh: Add a simple test.
* doc/coreutils.texi (realpath invocation): Mention the new option.
* NEWS: Likewise.

8 weeks agodoc: --base58: add example usage to info
Pádraig Brady [Sun, 10 Aug 2025 15:15:37 +0000 (16:15 +0100)] 
doc: --base58: add example usage to info

* doc/coreutils.texi (basenc invocation): Add an example
using --base58 to generate a unique ID.  This also demonstrates
compound usage of the basenc command, to convert to/from binary.

8 weeks agodoc: rearrange NEWS to more standard ordering
Pádraig Brady [Fri, 8 Aug 2025 11:35:50 +0000 (12:35 +0100)] 
doc: rearrange NEWS to more standard ordering

* NEWS: Also move items to more appropriate sections.
Also use more consistent quoting.

8 weeks agobasenc: add base58 support
Pádraig Brady [Tue, 22 Jul 2025 19:30:03 +0000 (20:30 +0100)] 
basenc: add base58 support

A 58 character encoding that:
 - avoids visually ambiguous 0OIl characters
 - uses only alphanumeric characters
Described at:
 - https://tools.ietf.org/html/draft-msporny-base58-03

This implementation uses GMP (or gnulib's gmp fallback).
Performance is good in comparison to other implementations.
For example when using libgmp on an i7-5600U system,
encoding is 530 times faster, and decoding 830 times faster
than the implementation using arbitrary precision ints in cpython 3.13.

Memory use is proportional to the size of input.

Encoding benchmarks:

  $ time yes | head -c65535 | src/basenc --base58 -w0 >file.enc
  real    0m0.018s

  ./configure --quiet --without-libgmp && make -j $(nproc)
  $ time yes | head -c65535 | src/basenc --base58 -w0 >file.enc
  real    0m3.431s

  # dnf install python3-base58
  $ time yes | head -c65535 | base58 >file.enc  # cpython 3.13
  real    0m9.700s

Decoding benchmarks:

  $ time src/basenc --base58 -d <file.enc >/dev/null
  real    0m0.010s

  $ ./configure --without-libgmp && make  # gnulib gmp
  $ time src/basenc --base58 -d <file.enc >/dev/null
  real    0m0.145s

  $ time base58 -d <file.enc >/dev/null  # cpython 3.13
  real    0m8.302s

* src/basenc.c (base_decode_ctx_finalize, base_encode_ctx_init,
base_encode_ctx, base_encode_ctx_finalize): New functions to
provide more general processing functionality.
(base58_{de,en}code_ctx{_init,,_finalize}): New functions to
accumulate all input before calling ...
(base58_{de,en}code): ... the GMP based encoding/decoding routines.
(do_encode, do_decode): Call the ctx variants if enabled.
* doc/coreutils.texi (basenc invocation): Describe the new option,
and indicate the main use case being interactive user use.
* src/local.mk: Link basenc with GMP.
* tests/basenc/basenc.pl: Add test cases.
* NEWS: Mention the new feature.

8 weeks agobasenc: fix stripping of '=' chars in some encodings
Pádraig Brady [Thu, 7 Aug 2025 19:39:24 +0000 (20:39 +0100)] 
basenc: fix stripping of '=' chars in some encodings

* src/basenc.c (do_decode): With -i ensure we strip '=' chars
if there is no padding for the chosen encoding.
* tests/basenc/basenc.pl: Add a test case.
* NEWS: Mention the bug fix.

8 weeks agomaint: prefer attribute.h in .c files
Collin Funk [Sat, 9 Aug 2025 03:32:18 +0000 (20:32 -0700)] 
maint: prefer attribute.h in .c files

* src/basenc.c (base16_encode, z85_encoding, do_decode): Use
ATTRIBUTE_NONSTRING instead of ATTRIBUTE_NONSTRING.
* src/basenc.c (sc_prohibit-_gl-attributes): New rule for
'make syntax-check'.

8 weeks agocp: omit some needless lseek calls
Paul Eggert [Thu, 7 Aug 2025 22:42:23 +0000 (15:42 -0700)] 
cp: omit some needless lseek calls

The sparse code sometimes issued multiple lseeks against the
same file without doing anything in betwee.  Optimize them away
by keeping track of the last hole output, in a way that
crosses the sparse_copy function call boundary.
* src/copy.c (sparse_copy): New arg hole_size, replacing old args
scan_holes and last_write_made_hole.  All callers changed.
(sparse_copy, lseek_copy): Do not create hole at
end; let the caller deal with it.  All callers changed.
(lseek_copy): New args hole_size and total_n_read.  Caller changed.
(copy_reg): Create hole at end for both lseek_copy and sparse_copy.

8 weeks agocp: --sparse=always was missing some holes
Paul Eggert [Wed, 6 Aug 2025 23:25:01 +0000 (16:25 -0700)] 
cp: --sparse=always was missing some holes

The sparse code assumed that st_blksize was the minimum hole size.
However, st_blksize is an optimum I/O buffer size, not the file
system fundamental block size.  Use ST_NBLOCKSIZE instead;
although it may underestimate the true block size that just slows
‘cp’ down a bit, without introducing bugs.
* src/copy.c (sparse_copy): Arg scan_holes replaces
the old hole_size arg.  All callers changed.
(lseek_copy): Remove hole_size arg; no longer needed.
Caller changed.

2 months agomaint: use consistent references to standard files in messages
Collin Funk [Wed, 6 Aug 2025 01:46:04 +0000 (18:46 -0700)] 
maint: use consistent references to standard files in messages

* cfg.mk (sc_standard_outputs): Add a grep command for source files.
* src/du.c (main): Use standard input instead of stdin, standard output
instead of stdout, and standard error instead of stderr in messages.
* src/nohup.c (main): Likewise.
* src/sort.c (main): Likewise.
* src/split.c (main): Likewise.
* src/stdbuf.c (main): Likewise.
* src/wc.c (main): Likewise.
* tests/du/files0-from.pl (@Tests): Adjust test case to new messages.
* tests/sort/sort-files0-from.pl: Likewise.
* tests/wc/wc-files0-from.pl: Likewise.

2 months agomaint: remove now-unused include of 'safe-read.h'
Bernhard Voelker [Mon, 4 Aug 2025 06:36:53 +0000 (08:36 +0200)] 
maint: remove now-unused include of 'safe-read.h'

'make syntax-check' complains:
  src/tail.c
  maint.mk: the above files include safe-read.h but don't use it
  make: *** [maint.mk:737: sc_prohibit_safe_read_without_use] Error 1

The removal was missed for tail.c in recent commit d3c7072a0950.

* src/tail.c (safe-read.h): Remove include.

2 months agomaint: prefer same-inode.h
Paul Eggert [Sun, 3 Aug 2025 22:04:01 +0000 (15:04 -0700)] 
maint: prefer same-inode.h

This does not change behavior on POSIX platforms; it’s mostly to
make it clearer when we’re looking for file identity.
* src/cat.c (main):
* src/copy.c (struct dir_list, is_ancestor, copy_internal):
* src/tail.c (struct File_spec, record_open_fd, recheck)
(tail_forever_inotify, tail_file):
* src/test.c (binary_operator):
Use psame_inode, PSAME_INODE, or SAME_INODE instead of comparing
device and inode numbers by hand.

2 months agotail: refactor ‘failable’
Paul Eggert [Sun, 3 Aug 2025 15:29:52 +0000 (08:29 -0700)] 
tail: refactor ‘failable’

* src/tail.c (recheck, tail_file): Do not mark a file as tailable
merely because --retry is not in effect.  Simplify internal logic.
This should not change behavior; it’s just for clarity and to
make the code match the comments better.

2 months agotail: fix race between read and fstat
Paul Eggert [Sun, 3 Aug 2025 15:19:53 +0000 (08:19 -0700)] 
tail: fix race between read and fstat

* src/tail.c (get_file_status): Remove, since after the changes
described below it would be called in just one place and it’s a
bit clearer to inline by hand.
(tail_file): Don’t call fstat after reading the file, as that
misses changes arriving between read and fstat.  Instead, reuse
the fstat done before reading the file.

2 months agotail: track errno more accurately
Paul Eggert [Sat, 2 Aug 2025 23:55:09 +0000 (16:55 -0700)] 
tail: track errno more accurately

This matters only in some obscure cases hard to test for.
* src/tail.c (file_lines, pipe_lines, pipe_bytes, start_bytes)
(start_lines, tail_bytes, tail_lines, tail): New return
convention, which reports errno.  All callers changed.
(recheck): Don’t lose track of errno if a regular file is
replaced by a symlink.
(get_file_status): Set errno to 0 on success.
(tail_file): Be more careful about f->errnum.
It is now -1 only if the failure was not due to
a system call failing.

2 months agotail: omit redundant assignment
Paul Eggert [Sat, 2 Aug 2025 20:49:48 +0000 (13:49 -0700)] 
tail: omit redundant assignment

* src/tail.c (recheck): f->remote must be true already,
so don’t set it to true.

2 months agotail: prefer readlink to lstat+S_ISLNK
Paul Eggert [Sat, 2 Aug 2025 19:58:06 +0000 (12:58 -0700)] 
tail: prefer readlink to lstat+S_ISLNK

When not already calling lstat for some other reason,
prefer readlink to lstat+S_ISLNK,
as readlink does not suffer from EOVERFLOW issues.
* src/rmdir.c (main):
* src/tail.c (recheck, any_symlinks):
* src/test.c (unary_operator):

2 months agotail: fix unlikely races with >=2 --pids
Paul Eggert [Fri, 1 Aug 2025 20:46:33 +0000 (13:46 -0700)] 
tail: fix unlikely races with >=2 --pids

Also, fix commentary to talk about “nonexistent” rather than
“dead” processes, since the code looks for the former not the
latter and the difference matters for zombies.
* src/tail.c (some_writers_exist): Rename from writers_are_dead,
negate the sense, don’t have a special and counterintuitive case
for !nbpids, remove PIDs found not to exist, and avoid some
though not all unlikely races when kernels reuse PIDs.
(tail_forever): Optimize via blocking I/O even if --pid was used,
so long as all the writers no longer exist.
(tail_forever, tail_forever_inotify): Simplify the writers_dead
logic; there is no need to have a local var to track this, since
we can use pids and nbpids now.
(parse_options): Also free and clear pids if !forever.

2 months agotail: prefer < 0 to == -1
Paul Eggert [Fri, 1 Aug 2025 14:58:08 +0000 (07:58 -0700)] 
tail: prefer < 0 to == -1

* src/tail.c (valid_file_spec, recheck, writers_are_dead)
(tail_forever, check_fspec, tail_forever_inotify, tail_file)
(parse_options, main): Be a bit more systematic about checking
for sign, rather than for exact equality or inequality,
when the sign is enough.  Makes the code a bit clearer
now that -2 sometimes means success.

2 months agotail: record file offset more carefully
Paul Eggert [Fri, 1 Aug 2025 01:45:12 +0000 (18:45 -0700)] 
tail: record file offset more carefully

* src/tail.c (struct File_spec): New member read_pos, replacing
size, since the value was really a read position not a size.
All uses changed.
(xlseek): Move defn up.
(record_open_fd): If the read_pos (formerly) size arg is unknown,
compute it here if it is a regular file.
(file_lines): Return the resulting read pos (or -1 on failure)
instead of storing it via a pointer.  Caller changed.
Simplify by using SEEK_CUR instead of SEEK_SET when that is easy.
Avoid reading the same data twice when there are not enough
lines in the file.
(pipe_lines): Return -2 on success, -1 on failure, rather than
updating a read pos via a pointer (which was weird for pipes anyway).
Caller changed.
(pipe_bytes, tail_bytes, tail_lines, tail):
Return -1 on failure, a file offset if known, and < -1 otherwise,
instead of storing a file offset via a pointer.  Caller changed.
(pipe_bytes): Take initial file offset as an arg, or -1 if unknown.
(start_bytes, start_lines): Return -1 (not 1) on error, -2 (not
-1) on EOF, and do not accept pointer to read pos as an arg since
neither we nor our caller know the read pos.  Callers changed.
(recheck): Do not assume a newly-opened file is at offset zero,
as this is not always true on Solaris.
(tail_forever, check_fspec): Use dump_remainder result only
on regular files, to prevent (very unlikely) overflow.
(tail_file): Remove no-longer-needed TAIL_TEST_SLEEP code.

2 months agomaint: prefer 'read' to 'safe_read'
Paul Eggert [Thu, 31 Jul 2025 19:31:58 +0000 (12:31 -0700)] 
maint: prefer 'read' to 'safe_read'

In the old days, safe_read acted more like what full_read does now.
When that went away, some code that invoked safe_read should
have gone back to plain 'read' but I guess we never got around to it.
Simplify this code by going back to plain 'read'.
Use safe_read only in csplit.c, which has a signal handler
and where 'read' can therefore fail with EINTR.
Although safe_read also checks for oversize buffers,
that is better done via io_blksize.
* src/cat.c (simple_cat, cat):
* src/head.c (copy_fd, elide_tail_lines_pipe)
* src/tac.c (tac_seekable, copy_to_temp):
(elide_tail_lines_seekable, head_lines):
* src/tail.c (dump_remainder, file_lines, pipe_lines)
(pipe_bytes, start_bytes, start_lines, tail_forever_inotify):
* src/tr.c (plain_read):
Use plain 'read', not safe_read, since there is no
need to worry about signals or oversize requests.
Also, there is no longer a need to include safe-read.h.
* src/ioblksize.h: Include sys-limits.h, for SYS_BUFSIZE_MAX.
(io_blksize): Max out at SYS_BUFSIZE_MAX.

2 months agotail: prefer signed types to size_t, blksize_t
Paul Eggert [Wed, 30 Jul 2025 22:11:31 +0000 (15:11 -0700)] 
tail: prefer signed types to size_t, blksize_t

* src/tail.c (struct File_spec, xwrite_stdout, file_lines)
(pipe_lines, pipe_bytes, start_bytes, any_live_files)
(tail_forever, any_remote_file, any_non_remote_file)
(any_symlinks, any_non_regular_fifo, tailable_stdin)
(tail_forever_inotify, ignore_fifo_and_pipe, main):
Prefer a signed type to size_t, if possible.
Ordinarily this is idx_t, but use int when the value
must fit in int anyway.
(file_lines): Similarly for blksize_t, which had no business
being here anyway.
(main): Check for overflow in the oddball case where ptrdiff_t is
narrower than int.

2 months agotail: prefer intmax_t to uintmax_t
Paul Eggert [Wed, 30 Jul 2025 15:46:04 +0000 (08:46 -0700)] 
tail: prefer intmax_t to uintmax_t

Signed types let us debug better, by using -fsanitize=undefined.
* doc/local.mk (doc/constants.texi):
Adjust change from macro to enum.
* src/tail.c (COPY_TO_EOF, COPY_A_BUFFER)
(DEFAULT_MAX_N_UNCHANGED_STATS_BETWEEN_OPENS):
Now enum constants, not macros.
(COPY_TO_EOF, COPY_A_BUFFER): Now negative, not positive.
(count_t): New typedef.  Use it instead of uintmax_t.
(COUNT_MAX): New macro; use it instead of UINTMAX_MAX.
(struct File_spec, max_n_unchanged_stats_between_opens)
(dump_remainder, file_lines, pipe_lines, pipe_bytes)
(start_bytes, start_lines, tail_forever, check_fspec)
(tail_forever_inotify, tail_bytes, tail_lines, tail, tail_file)
(parse_obsolete_option, parse_options, main):
Prefer count_t to uintmax_t.

2 months agotail: don’t output more lines than requested
Paul Eggert [Wed, 30 Jul 2025 14:49:14 +0000 (07:49 -0700)] 
tail: don’t output more lines than requested

* src/tail.c (file_lines): Fix an unlikely bug where ‘tail -n N’
could output more than N lines if standard input is a largish
regular file with large initial offset that starts with (say) N-1
lines after the initial offset, but grows to N+1 lines between the
fstat and read calls.  In this case ‘tail -n N’ now outputs N-1
lines, not N+1; that is, it pretends the file grew after ‘tail’
read it.  That is better than outputting more than N lines.

2 months agotail: xlseek switch → table
Paul Eggert [Tue, 29 Jul 2025 20:23:58 +0000 (13:23 -0700)] 
tail: xlseek switch → table

* src/tail.c (xlseek): Turn a switch statement into a table.

2 months agotail: prettyname cleanup
Paul Eggert [Tue, 29 Jul 2025 20:22:41 +0000 (13:22 -0700)] 
tail: prettyname cleanup

* src/tail.c: Use ‘prettyname’ consistently as the identifier
for a prettified file name, as opposed to ‘pretty_filename’,
‘pretty_name’, and ‘name’.  This makes the code easier to follow.
(struct File_spec): New member prettyname.
(pretty_name): Remove.
All uses of pretty_name (f) replaced by f->prettyname.
(close_fd, fremote): Accept struct File_spec, not name.
All callers changed.
(main): Initialize the new prettyname member.
This is simpler/smaller than calling pretty_name everywhere.

2 months agotail: optimize tail -n +2**63
Paul Eggert [Tue, 29 Jul 2025 16:37:44 +0000 (09:37 -0700)] 
tail: optimize tail -n +2**63

* src/tail.c (tail_lines): Also optimize ‘tail -n +N’
on a seekable file, where OFF_T_MAX <= N < UINTMAX_MAX.
Of course this is very unlikely.

2 months agotail: refactor SEEK_END and lines
Paul Eggert [Tue, 29 Jul 2025 16:14:37 +0000 (09:14 -0700)] 
tail: refactor SEEK_END and lines

* src/tail.c (tail_lines): Refactor to simplify the confusing
code for using SEEK_END when counting lines.  The old code
had a ‘end_pos != 0’ expression that was always true.

2 months agotail: refactor to skip stat call on failure
Paul Eggert [Tue, 29 Jul 2025 15:24:58 +0000 (08:24 -0700)] 
tail: refactor to skip stat call on failure

* src/tail.c (tail_bytes): New function.
(tail_bytes, tail_lines, tail): Accept struct stat pointer from
caller instead of calling fstat ourselves.  All callers changed.
(tail_file): Skip a call to fstat if fstat already failed.
* tests/tail/follow-stdin.sh: Adjust to match new behavior
on failure, which omits a redundant diagnostic.

2 months agotail: speed up -c N for huge N
Paul Eggert [Tue, 29 Jul 2025 06:39:30 +0000 (23:39 -0700)] 
tail: speed up -c N for huge N

When the user specifies -c N where 2**63 <= N, don’t give
up and use the slow method (which will exhaust memory if
the file is large).  Instead, treat it as N = 2**63 - 1,
since that has equivalent effect.
* src/tail.c (tail_bytes): With -c N and large N, adjust
the code so that lseeks can still be used without
affecting correctness.  Formerly the code gave up and
did a sequential pass through the whole input, which
could easily exhaust memory.

2 months agotail: allow >=2**64 in traditional form
Paul Eggert [Mon, 28 Jul 2025 18:24:38 +0000 (11:24 -0700)] 
tail: allow >=2**64 in traditional form

This better matches the treatment of POSIX form,
e.g., ‘tail +Nc’ is now like ‘tail -c +N’ even when N is large.
* src/tail.c: Don’t include xstrtol.h.
(parse_obsolete_option): Treat numbers greater than UINTMAX_MAX as
if they are UINTMAX_MAX.  Parse the number by hand with saturating
arithmetic; nowadays that’s simpler than using xstrtoumax.  There
is no need for a diagnostic now, as the error cannot happen any more.
* tests/tail/tail.pl (obs-plus-c3): New test.

2 months agotail: check OFF_T_MAX vs COPY_A_BUFFER
Paul Eggert [Sun, 27 Jul 2025 19:36:53 +0000 (12:36 -0700)] 
tail: check OFF_T_MAX vs COPY_A_BUFFER

* src/tail.c: Document an otherwise-unstated assumption.

2 months agotail: fix SEEK_END typo
Paul Eggert [Wed, 23 Jul 2025 15:31:38 +0000 (08:31 -0700)] 
tail: fix SEEK_END typo

* src/tail.c (tail_lines): Fix embarrassing SEEK_END typo.
Luckily this matters only in never-used (though valid) invocations.

2 months agotail: simplify -c+ on regular files
Paul Eggert [Wed, 23 Jul 2025 05:05:17 +0000 (22:05 -0700)] 
tail: simplify -c+ on regular files

* src/tail.c (tail_bytes): Simplify the -c+N case by treating
regular files like other files; if the lseek fails for whatever
reason, fall back on the unoptimized version instead of reporting
a fatal error.

2 months agomaint: don’t lseek memory objects
Paul Eggert [Mon, 14 Jul 2025 23:41:05 +0000 (16:41 -0700)] 
maint: don’t lseek memory objects

This affects behavior only on memory objects, which are not in Linux.
Formerly the code would use lseek on these objects,
but POSIX says the result of lseek is unspecified on them,
and in QNX lseek has undefined behavior on typed memory objects.
* src/head.c (elide_tail_bytes_file, elide_tail_lines_file):
Omit unnecessary uses of presume_input_pipe.
Improve some out-of-date comments.
(head): Do not assume a file is seekable merely because its
st_size is usable.  Instead, seek only on regular files.
* src/od.c (skip): Do not seek on memory objects.

2 months agowc: don’t assume opening a file puts you at start
Paul Eggert [Mon, 14 Jul 2025 20:32:27 +0000 (13:32 -0700)] 
wc: don’t assume opening a file puts you at start

This is not true on Solaris when opening /dev/stdin; it could be
at a nonzero file offset.  Arguably Linux should do likewise.
* src/wc.c (wc): Omit last arg, and act as if it is always -1.
All uses changed.

2 months agohead: simplify head_bytes
Paul Eggert [Mon, 14 Jul 2025 18:40:33 +0000 (11:40 -0700)] 
head: simplify head_bytes

* src/head.c (diagnose_read_failure): New function.
Use it when possible, to simplify callers.
(head_bytes): Use copy_fd and diagnose_copy_fd_failure instead of
reinventing the wheel.

2 months agohead: omit unnecessary lseek
Paul Eggert [Mon, 14 Jul 2025 18:07:21 +0000 (11:07 -0700)] 
head: omit unnecessary lseek

* src/head.c (elseek_diagnostic): New function.
(elseek): Use it.
(head_lines): Use it to avoid an unnecessary lseek.

2 months agoreadlink: improve documentation for POSIXLY_CORRECT in --help
Collin Funk [Sun, 3 Aug 2025 19:33:55 +0000 (12:33 -0700)] 
readlink: improve documentation for POSIXLY_CORRECT in --help

* NEWS: Improve wording to not imply POSIXLY_CORRECT=1 readlink -q will
be verbose.
* src/readlink.c (usage): Mention the affect of POSIXLY_CORRECT on
-s (--silent), -q (--quiet), and -v (--verbose) in the help message.
(main): Remove spurious newline added by previous commit.
* doc/coreutils.texi (readlink invocation): Mention that -s (--silent)
and -q (--quiet) are not the default when POSIXLY_CORRECT is set.

2 months agoreadlink: emit errors when POSIXLY_CORRECT is set
Collin Funk [Sun, 3 Aug 2025 03:51:30 +0000 (20:51 -0700)] 
readlink: emit errors when POSIXLY_CORRECT is set

* src/readlink.c (main): Set verbose if the POSIXLY_CORRECT environment
variable is set.
* tests/readlink/readlink-posix.sh: New file.
* tests/local.mk (all_tests): Add it.
* NEWS: Mention the change.
* doc/coreutils.texi (readlink invocation): Document the behavior of
POSIXLY_CORRECT.

2 months agodoc: mention copy_file_range data corruption avoidance
Pádraig Brady [Sat, 2 Aug 2025 08:52:17 +0000 (09:52 +0100)] 
doc: mention copy_file_range data corruption avoidance

* NEWS: Mention this important data corruption issue.

2 months agobuild: update gnulib submodule to latest
Paul Eggert [Fri, 1 Aug 2025 23:55:31 +0000 (16:55 -0700)] 
build: update gnulib submodule to latest

2 months agomaint: date: refactor to be entirely bottom-up
Bruno Haible [Thu, 31 Jul 2025 16:45:45 +0000 (18:45 +0200)] 
maint: date: refactor to be entirely bottom-up

* src/date.c (show_date_helper): Move function. Remove forward
declaration.

2 months agodate: fix calendar-related test failures on macOS and OpenBSD
Bruno Haible [Thu, 31 Jul 2025 16:42:02 +0000 (18:42 +0200)] 
date: fix calendar-related test failures on macOS and OpenBSD

* bootstrap.conf (gnulib_modules): Add xsetenv.
* src/date.c: Include xsetenv.h.
(set_LC_TIME): New function.
(show_date_helper): Call set_LC_TIME instead of setlocale.
* po/POTFILES.in: Add lib/xsetenv.c.
Fixes https://bugs.gnu.org/79118

2 months agobuild: update gnulib submodule to latest
Collin Funk [Fri, 1 Aug 2025 04:43:55 +0000 (21:43 -0700)] 
build: update gnulib submodule to latest

2 months agobuild: fix previous commit
Collin Funk [Fri, 1 Aug 2025 00:57:59 +0000 (17:57 -0700)] 
build: fix previous commit

* src/local.mk (src/speedlist.h): Pass correct arguments $(COMPILE). Use
tabs consistently.

2 months agobuild: don't use '$<' in non-suffix rules
Collin Funk [Wed, 30 Jul 2025 23:33:24 +0000 (16:33 -0700)] 
build: don't use '$<' in non-suffix rules

* src/local.mk (src/speedlist.h): Use the file name instead of '$<'
which is not supported by OpenBSD make.

2 months agodu: improve diagnostics for --time-style
Pádraig Brady [Mon, 28 Jul 2025 22:05:49 +0000 (23:05 +0100)] 
du: improve diagnostics for --time-style

* src/system.h (x_timestyle_args): A new function refactored from ...
* src/ls.c (decode_switches): ... here.
* src/du.c: Use refactored x_timestyle_args() to output a custom error.
Addresses https://bugs.gnu.org/79113

2 months agodoc: use consistent references to standard files
Pádraig Brady [Mon, 28 Jul 2025 10:05:25 +0000 (11:05 +0100)] 
doc: use consistent references to standard files

* cfg.mk (sc_standard_outputs): A new syntax check to
enforce standard references.
* doc/coreutils.texi: s/stderr/standard error/ etc.
* src/date.c: Likewise.
* src/dd.c: Likewise.
* src/env.c: Likewise.
* src/sort.c: Likewise.
* src/split.c: Likewise.
* src/stty.c: Likewise.
* src/timeout.c: Likewise.
* src/who.c: Likewise.

2 months agodoc: update git version documented in HACKING
Collin Funk [Sun, 27 Jul 2025 18:53:42 +0000 (11:53 -0700)] 
doc: update git version documented in HACKING

* HACKING: Mention that git version 1.6.4 or later is required since
'./bootstrap' will fail otherwise.

2 months agodoc: mention 2025 speedup on large composites
Paul Eggert [Sun, 27 Jul 2025 14:30:29 +0000 (07:30 -0700)] 
doc: mention 2025 speedup on large composites

2 months agomaint: avoid :> pattern in tests
Pádraig Brady [Sat, 26 Jul 2025 19:06:42 +0000 (20:06 +0100)] 
maint: avoid :> pattern in tests

This was seen to trigger the EXIT trap on cygwin

* cfg.mk (sc_prohibit_colon_redirection): Disallow all
cases of :> in tests/
* tests/cksum/md5sum-bsd.sh: Adjust to avoid more
stringent syntax check.

2 months agotests: avoid a false failure on Cygwin
Collin Funk [Sat, 26 Jul 2025 18:29:02 +0000 (11:29 -0700)] 
tests: avoid a false failure on Cygwin

* tests/cksum/md5sum-bsd.sh: Use 'echo' instead of ':' to redirect to
the file. Otherwise 'Exit $fail' is not called on Cygwin.

2 months agomaint: avoid syntax-check failure from previous commit
Collin Funk [Sat, 26 Jul 2025 07:17:16 +0000 (00:17 -0700)] 
maint: avoid syntax-check failure from previous commit

* src/test.c (binary_operator): Place string argument on a separate line
to avoid sc_error_quotes syntax-check failure.

2 months agotest: simplify for clarity
Paul Eggert [Fri, 25 Jul 2025 18:56:00 +0000 (11:56 -0700)] 
test: simplify for clarity

This should help avoid further audit confusion,
such as was just fixed by removing a FIXME.
* src/test.c (enum binop): New type.
(get_mtime): Return a struct timespec instead of returning
a bool and storing a struct timespec.  All callers changed.
(binop): Return an int recording either success (an enum binop)
or failure (-1).  All callers changed.
(binary_operator): Accept an enum binop, so that we need
not recompute the op type.  All callers changed.  Simplify.

2 months agotest: removed dead code for unrecognised binary operators
Harry Fellowes [Fri, 25 Jul 2025 13:26:38 +0000 (14:26 +0100)] 
test: removed dead code for unrecognised binary operators

the fallback error in binary_operator() was unreachable, since invalid
binary operators are rejected in three_arguments() via binop() which
returns false and prevents entry into binary_operator().

this dead code was unreachable and safe to remove

ref: line 636 where "binary operator expected" is generated.

Signed-off-by: Harry Fellowes <harryfellowes1@gmail.com>
Copyright-paperwork-exempt: Yes

2 months agotests: date: add tests for the Ethiopian calendar
Collin Funk [Wed, 23 Jul 2025 03:30:49 +0000 (20:30 -0700)] 
tests: date: add tests for the Ethiopian calendar

* tests/date/date-ethiopia.sh: New file.
* tests/local.mk (all_tests): Add the test.

2 months agotests: date: add tests for the Solar Hijri calendar
Collin Funk [Wed, 23 Jul 2025 03:14:35 +0000 (20:14 -0700)] 
tests: date: add tests for the Solar Hijri calendar

* tests/date/date-iran.sh: New file.
* tests/local.mk (all_tests): Add the test.

2 months agoinstall: with -d, use more accurate diagnostics
Pádraig Brady [Tue, 22 Jul 2025 19:34:50 +0000 (20:34 +0100)] 
install: with -d, use more accurate diagnostics

* gnulib: Update to latest to get the fix to
propagate the appropriate error message.
* tests/install/basic-1.sh: Add a test case.
* NEWS: Mention the bug fix.
Addresses https://bugs.gnu.org/79072

2 months agotests: date: add tests for the Thai solar calendar
Collin Funk [Fri, 18 Jul 2025 02:42:30 +0000 (19:42 -0700)] 
tests: date: add tests for the Thai solar calendar

* tests/date/date-thailand.sh: New file.
* tests/local.mk (all_tests): Add the test.

2 months agodoc: NEWS: mention that 'date' now supports non-Gregorian calendars
Collin Funk [Mon, 21 Jul 2025 21:47:30 +0000 (14:47 -0700)] 
doc: NEWS: mention that 'date' now supports non-Gregorian calendars

Suggested by Bruno Haible in:
<https://lists.gnu.org/r/bug-gnulib/2025-07/msg00091.html>.

* NEWS: Mention the improvement.

2 months agobuild: update gnulib submodule to latest
Collin Funk [Mon, 21 Jul 2025 21:43:59 +0000 (14:43 -0700)] 
build: update gnulib submodule to latest

* bootstrap: Update using './bootstrap --bootstrap-sync'.

2 months agodate: update documentation regarding non-Gregorian calendars
Bruno Haible [Thu, 17 Jul 2025 14:30:57 +0000 (16:30 +0200)] 
date: update documentation regarding non-Gregorian calendars

* doc/coreutils.texi (Calendars): New section.

2 months agodate: force a Gregorian calendar for options --iso-8601 and --rfc-3339
Bruno Haible [Thu, 17 Jul 2025 14:29:05 +0000 (16:29 +0200)] 
date: force a Gregorian calendar for options --iso-8601 and --rfc-3339

* src/date.c (show_date_helper): Add a use_c_locale parameter.
(batch_convert): Add a format_in_c_locale parameter.
(main): Set format_in_c_locale to true if any of the options --rfc-3339,
--iso-8601, -R is seen.

2 months agotests: sort: add a test for equivalent floats
Pádraig Brady [Sat, 12 Jul 2025 18:58:06 +0000 (19:58 +0100)] 
tests: sort: add a test for equivalent floats

* tests/sort/sort-float.sh: Ensure equivalent floats
are treated as such.

2 months agotests: fix fraction comparison in sort-float
Paul Eggert [Fri, 11 Jul 2025 21:43:32 +0000 (14:43 -0700)] 
tests: fix fraction comparison in sort-float

Problem reported by Cosima Neidahl <https://bugs.gnu.org/78985#13>.
* tests/sort/sort-float.sh: At top level, use C locale at first.
(dbl_minima_order): Assume C locale.
Use string comparison for the fractional parts.

2025-07-10  Paul Eggert  <eggert@cs.ucla.edu>

tests: fix integer overflow in sort-float
Problem reported by Cosima Neidahl <https://bugs.gnu.org/78985>.
* tests/sort/sort-float.sh (dbl_minima_order):
Use expr instead of test, to avoid problems with integers
too large for the shell.

2 months agotests: fix integer overflow in sort-float
Paul Eggert [Thu, 10 Jul 2025 17:17:29 +0000 (10:17 -0700)] 
tests: fix integer overflow in sort-float

Problem reported by Cosima Neidahl <https://bugs.gnu.org/78985>.
* tests/sort/sort-float.sh (dbl_minima_order):
Use expr instead of test to compare fractions,
to avoid problems with integers too large for the shell.

2 months agofactor: remove SINGLE_WORKS
Paul Eggert [Thu, 10 Jul 2025 05:26:39 +0000 (22:26 -0700)] 
factor: remove SINGLE_WORKS

* src/factor.c (SINGLE_WORKS): Remove, and remove its use,
as the buggy code that it is referring to has been removed.

2 months agofactor: use 64-bit internal counters
Paul Eggert [Thu, 10 Jul 2025 04:59:50 +0000 (21:59 -0700)] 
factor: use 64-bit internal counters

* src/factor.c (factor_using_pollard_rho)
(factor_using_pollard_rho2, mp_factor_using_pollard_rho):
Use int_fast64_t for internal counters rather than int, as int
could overflow on some somewhat-practical examples.  Problem
discovered on a hypothetical platform where W_TYPE_SIZE is neither
32 nor 64, when factoring 0x7ffffffffffffeab7fffffffffff7369 ==
170141183460469225450570946617781744489, causing k to overflow in
mp_factor_using_pollard_rho.  Presumably a similar problem exists
in the previous stable coreutils 9.7, too, on 32-bit platforms
with somewhat-larger test cases, though I haven’t take the
somewhat-extensive CPU time to discover it.

2 months agofactor: fix mp_factor_using_pollard_rho aliasing
Paul Eggert [Thu, 10 Jul 2025 04:22:12 +0000 (21:22 -0700)] 
factor: fix mp_factor_using_pollard_rho aliasing

* src/factor.c (mp_factor_using_pollard_rho):
Fix recently-introduced aliasing bug by computing q
before g gets updated in place.  Problem discovered
on a hypothetical platform where W_TYPE_SIZE
is neither 32 nor 64.

2 months agomaint: fix apostrophe in old NEWS
Paul Eggert [Wed, 9 Jul 2025 22:20:17 +0000 (15:20 -0700)] 
maint: fix apostrophe in old NEWS

2 months agofactor: speed up converting strings to uuint
Paul Eggert [Wed, 9 Jul 2025 17:36:04 +0000 (10:36 -0700)] 
factor: speed up converting strings to uuint

* src/factor.c: Do not include c-ctype.h.
(strtouuint): Don’t bother generating a number on
error; just return a strtol_error value other than LONGINT_OK.
Speed up overflow checking.

2 months agofactor: simplify primes table
Paul Eggert [Wed, 9 Jul 2025 16:48:21 +0000 (09:48 -0700)] 
factor: simplify primes table

* src/factor.c (primes_ptab): New table of primes, replacing
primes_diff and primes_diff8.  All uses changed.  This is simpler
and should improve performance slightly.  Although this limits the
table’s primes to 2**15 instead of to 668221, the limit can easily
grow to 2**32 by changing the type of ‘prime’, without hurting
performance significantly compared to the primes_diff and
primes_diff8 approach.
* src/make-prime-list.c (output_primes):
For each prime p, output p instead of two differences.

2 months agofactor: prefer false for boolean 0
Paul Eggert [Wed, 9 Jul 2025 04:23:20 +0000 (21:23 -0700)] 
factor: prefer false for boolean 0

* src/factor.c: Prefer false to 0 for booleans.

2 months agomaint: port speedgen to FreeBSD 15
Paul Eggert [Mon, 7 Jul 2025 07:26:40 +0000 (00:26 -0700)] 
maint: port speedgen to FreeBSD 15

* src/speedgen: Avoid ‘trap 'command' ERR’, as ERR
does not conform to POSIX and does not work in
FreeBSD 15 sh.

2 months agofactor: speed up Pollard-rho loop counters
Paul Eggert [Mon, 7 Jul 2025 03:18:10 +0000 (20:18 -0700)] 
factor: speed up Pollard-rho loop counters

* src/factor.c (factor_using_pollard_rho)
(factor_using_pollard_rho2, mp_factor_using_pollard_rho):
Use int, not unsigned long int, for counters that won’t go above
2**31 on practical platforms.  This yields a significant speedup
on GCC 15 x86-64, and using signed values allows for automatic
checks for overflow when using gcc -fsanitize=undefined.

2 months agofactor: avoid an mpz init+clear
Paul Eggert [Sun, 6 Jul 2025 20:59:56 +0000 (13:59 -0700)] 
factor: avoid an mpz init+clear

* src/factor.c (mp_factor_insert_ui): Rename fom
mp_factor_insert_ui, and change arg type from unsigned long int to
mp_limb_t.  All uses changed.  This avoids creating and freeing a
small mpz_t.
(mp_factor_using_division): Add a static assert requiring that
that mp_limb_t be wide enough, which it should be (and is in
standard GMP).

2 months agofactor: Pollard-rho a is now mp_limb_t
Paul Eggert [Sun, 6 Jul 2025 16:34:00 +0000 (09:34 -0700)] 
factor: Pollard-rho a is now mp_limb_t

* src/factor.c (factor_using_pollard_rho)
(factor_using_pollard_rho2): Use mp_limb_t, not unsigned long int,
for a parameter.  This avoids some casts, and avoids a theoretical
bug where converting to mp_limb_t loses info.

2 months agofactor: redo ge2 in terms of lt2
Paul Eggert [Sun, 6 Jul 2025 05:12:22 +0000 (22:12 -0700)] 
factor: redo ge2 in terms of lt2

lt2 a bit more natural, given the current implementation.
* src/factor.c (lt2): New function.
(ge2): Rewrite in terms of lt2.
(gt2): Remove.  All callers changed to use lt2.

2 months agofactor: speed up umul_ppmm when !USE_LONGLONG_H
Paul Eggert [Sat, 5 Jul 2025 18:04:54 +0000 (11:04 -0700)] 
factor: speed up umul_ppmm when !USE_LONGLONG_H

* src/factor.c (umul_ppmm): When !USE_LONGLONG_H so we need to
define this, speed things up if there is an unsigned type uuroom_t
wide enough to hold two words.  Do not make a similar change for
udiv_qrnnd, as it is not performance critical and anyway on GCC 15
x86-64 that would mean a subroutine call.