]> git.ipfire.org Git - thirdparty/coreutils.git/log
thirdparty/coreutils.git
47 hours agobuild: avoid make distcheck error master
Pádraig Brady [Sat, 5 Jul 2025 16:45:00 +0000 (17:45 +0100)] 
build: avoid make distcheck error

* src/local.mk: Similarly to commit v8.22-156-g09937e9d0
track speedlist.h with nodist_src_stty_SOURCES and DISTCLEANFILES
to ensure the make distcheck manifest comparison passes.
Addresses https://bug.gnu.org/78960

2 days agodoc: fix a dead link
Collin Funk [Sat, 5 Jul 2025 04:07:23 +0000 (21:07 -0700)] 
doc: fix a dead link

* doc/coreutils.texi (Block Size): Use an updated link to SI prefixes.

2 days agobuild: fix non parallel VPATH builds
Pádraig Brady [Sat, 5 Jul 2025 11:49:02 +0000 (12:49 +0100)] 
build: fix non parallel VPATH builds

* src/local.mk: Use the coarser BUILT_SOURCES mechanism
to generate speedlist.h, rather than a specific dependency
(which did seem to work for parallel builds).
Fixes https://bugs.gnu.org/78960

2 days agomaint: prefer endian.h macros to WORDS_BIGENDIAN, part 2
Collin Funk [Sat, 5 Jul 2025 03:52:37 +0000 (20:52 -0700)] 
maint: prefer endian.h macros to WORDS_BIGENDIAN, part 2

* src/blake2/blake2-impl.h: Include endian.h. Use BYTE_ORDER and
LITTLE_ENDIAN instead of WORDS_BIGENDIAN.

3 days agomaint: prefer endian.h macros to WORDS_BIGENDIAN
Collin Funk [Fri, 4 Jul 2025 17:59:39 +0000 (10:59 -0700)] 
maint: prefer endian.h macros to WORDS_BIGENDIAN

* src/od.c: Include endian.h.
(WORDS_BIGENDIAN): Remove definition.
(main): Use BYTE_ORDER, BIG_ENDIAN, LITTLE_ENDIAN instead of
WORDS_BIGENDIAN. Fix formatting.

4 days agomaint: add files generated by 'make ps' to .gitignore
Collin Funk [Thu, 3 Jul 2025 02:01:06 +0000 (19:01 -0700)] 
maint: add files generated by 'make ps' to .gitignore

* doc/.gitignore (/coreutils.ps, /coreutils.t2d/): Add entries.

4 days agodoc: remove outdated TODO/FIXME items
Collin Funk [Thu, 3 Jul 2025 01:41:53 +0000 (18:41 -0700)] 
doc: remove outdated TODO/FIXME items

* TODO: Remove note about missing documentation for 'pinky' and '['.
* doc/coreutils.texi: Likewise.

5 days agotests: cat: add test for O_APPEND bug fix
Pádraig Brady [Tue, 1 Jul 2025 21:52:16 +0000 (22:52 +0100)] 
tests: cat: add test for O_APPEND bug fix

* tests/cat/cat-self.sh: Add a test case
for commit v9.6-19-g7386c291b

6 days agodoc: remove extraneous double quotes within @samp{}
Pádraig Brady [Tue, 1 Jul 2025 11:02:08 +0000 (12:02 +0100)] 
doc: remove extraneous double quotes within @samp{}

* doc/coreutils.texi (df invocation): @samp{} items
are suitably quoted already.

6 days agodoc: use ``...'' instead of "..."
Collin Funk [Tue, 1 Jul 2025 02:50:58 +0000 (19:50 -0700)] 
doc: use ``...'' instead of "..."

* doc/coreutils.texi (split invocation): Use ``...'' for quoting text.
(cksum common options): Likewise.
(Control): Likewise.

6 days agotests: avoid false failure due to varied ERANGE errors
Pádraig Brady [Mon, 30 Jun 2025 22:28:32 +0000 (23:28 +0100)] 
tests: avoid false failure due to varied ERANGE errors

* tests/od/od.pl: Use the system error string for ERANGE,
rather than hardcoding a particular one.
Fixes https://bugs.gnu.org/78934

6 days agood: port to Apple clang 14
Paul Eggert [Mon, 30 Jun 2025 22:40:57 +0000 (15:40 -0700)] 
od: port to Apple clang 14

* src/od.c (print_function_type): New type.  Use it for convenience.
(width_bytes): Omit duplicate entries, such as ‘double’ vs ‘long
double’ on macOS.  Problem reported by Bruno Haible
<https://bugs.gnu.org/78933>.
(decode_one_format): Cast null pointer to print_function_type
to pacify Apple clang-1400.0.29.202.

7 days agobuild: fix VPATH builds with --enable-single-binary
Pádraig Brady [Mon, 30 Jun 2025 14:34:22 +0000 (15:34 +0100)] 
build: fix VPATH builds with --enable-single-binary

* src/local.mk: Adjust the dependency so that speedlist.h
is built irrespective of the object file name.
Note we could use BUILT_SOURCES for this,
but it's better to have this more accurate dependency.

7 days agood: reinstate half float validation check
Pádraig Brady [Mon, 30 Jun 2025 13:25:56 +0000 (14:25 +0100)] 
od: reinstate half float validation check

Reinstate check removed in commit 56aa549a0 so that we
disallow -f2 when configured with utils_cv_ieee_16_bit_supported=no.
Otherwise the output routines will consume floats,
i.e. 4 bytes at a time.  Without this extra check
the tests/od/od-endian.sh will fail with this configuration.

* src/od.c (decode_one_format): Reinstate the explicit check
for this configuration edge case.

7 days agomaint: avoid sc_prohibit-quotearg failure
Pádraig Brady [Mon, 30 Jun 2025 12:10:15 +0000 (13:10 +0100)] 
maint: avoid sc_prohibit-quotearg failure

* src/od.c (parse_old_offset): s/quotearg_colon/quotef/.

7 days agood: add test cases for recent fix
Paul Eggert [Mon, 30 Jun 2025 05:31:01 +0000 (22:31 -0700)] 
od: add test cases for recent fix

* tests/od/od.pl: New tests for the offset issues

7 days agood: more minor fixes for offsets
Paul Eggert [Mon, 30 Jun 2025 05:25:28 +0000 (22:25 -0700)] 
od: more minor fixes for offsets

* src/od.c (parse_old_offset): Reject invalid offsets like "++0".
Treat overflowing offsets as errors, not as file names.

7 days agood: refactor parse_old_offset
Paul Eggert [Mon, 30 Jun 2025 01:06:22 +0000 (18:06 -0700)] 
od: refactor parse_old_offset

* src/od.c (parse_old_offset): Refactor for brevity and clarity.

7 days agomaint: fix version in NEWS
Paul Eggert [Mon, 30 Jun 2025 00:13:07 +0000 (17:13 -0700)] 
maint: fix version in NEWS

* NEWS: Update version number as per Jim Meyering
<https://bugs.gnu.org/78880#46>.

7 days agood: pacify gcc -Wduplicated-cond
Paul Eggert [Mon, 30 Jun 2025 00:12:07 +0000 (17:12 -0700)] 
od: pacify gcc -Wduplicated-cond

Problem reported by Pádraig Brady <https://bugs.gnu.org/78880#43>.
This patch doesn’t fix any bugs; it merely pacifies GCC.
* src/od.c (ispec_to_format): New function, replacing
the old ISPEC_TO_FORMAT macro.  All uses changed.
This part of the change is just refactoring.
(decode_one_format): Pacify à la ispec_to_format.

8 days agood: be more consistent re sizeof
Paul Eggert [Sun, 29 Jun 2025 03:57:00 +0000 (20:57 -0700)] 
od: be more consistent re sizeof

* src/od.c (width_bytes, decode_one_format): Don’t assume a signed
type has the same size as the corresponding unsigned type.
This has no effect on practical platforms; it’s just for
consistency there.

8 days agood: fix integer overflow with large pseudos
Paul Eggert [Sun, 29 Jun 2025 00:36:04 +0000 (17:36 -0700)] 
od: fix integer overflow with large pseudos

* src/od.c (format_address_label): Diagnose overflow.

8 days agood: speed up -S
Paul Eggert [Sat, 28 Jun 2025 22:21:18 +0000 (15:21 -0700)] 
od: speed up -S

* src/od.c (read_char): Use getc, not fgetc.

8 days agood: check sign bit more often
Paul Eggert [Sat, 28 Jun 2025 22:04:30 +0000 (15:04 -0700)] 
od: check sign bit more often

* src/od.c (read_char, dump_strings, main):
Instead of testing for an exact negative number,
just look at the sign bit.  This is a very minor tweak.

8 days agood: simplify away one loop copy
Paul Eggert [Sat, 28 Jun 2025 21:01:18 +0000 (14:01 -0700)] 
od: simplify away one loop copy

* src/od.c (dump): Coalesce two loops into one.

8 days agood: simpler static initialization
Paul Eggert [Sat, 28 Jun 2025 16:49:42 +0000 (09:49 -0700)] 
od: simpler static initialization

* src/od.c (address_base, address_pad_len, format_address):
Initialize statically rather than dynamically.
(limit_bytes_to_format): Remove.  All uses replaced by
checking sign of end_offset.
(max_bytes_to_format): Remove static var.  Now local to ‘main’.
(end_offset): -1 now means no limit.  All uses changed.

8 days agood: minor lcm tuning
Paul Eggert [Sat, 28 Jun 2025 00:52:15 +0000 (17:52 -0700)] 
od: minor lcm tuning

* src/od.c (dump, main): Redo lcm calcuations to avoid a multiply.

8 days agood: omit some duplicate code
Paul Eggert [Sat, 28 Jun 2025 00:08:28 +0000 (17:08 -0700)] 
od: omit some duplicate code

On x86-64 (for example) print_long, print_long_long, and
print_intmax all behave identically, so give GCC enough info so
that it generates code for just one of these functions.
* src/od.c (enum size_spec): Arrange for enum values to
be the same if they represent types that behave the same.
(width_bytes, ISPEC_TO_FORMAT, decode_one_format):
Match the enum size_spec changes.

8 days agood: replace lookup tables with simple arithmetic
Paul Eggert [Fri, 27 Jun 2025 23:08:06 +0000 (16:08 -0700)] 
od: replace lookup tables with simple arithmetic

* src/od.c (FMT_BYTES_ALLOCATED): Use a simpler formula.
Although slightly too generous, the storage wasted is very small
and it pacifies gcc -Wformat-overflow=2.
(bytes_to_oct_digits, bytes_to_signed_dec_digits)
(bytes_to_unsigned_dec_digits, bytes_to_hex_digoits): Remove.
All uses replaced by algorithmic calculations, which are good
enough: they are valid for integers up to 2620 bits (!) and might
be slightly conservative for wider integers.  Remove related
static_asserts, which are no longer needed.

8 days agood: support uintmax_t too
Paul Eggert [Fri, 27 Jun 2025 15:45:53 +0000 (08:45 -0700)] 
od: support uintmax_t too

This has practical effect only on hypothetical platforms where
uintmax_t is wider than unsigned long long int.
* src/od.c (enum size_spec): New constant INTMAX.
(MAX_INTEGRAL_TYPE_WIDTH): Now equals UINTMAX_WIDTH.
(FMT_BYTES_ALLOCATED): Allow for the extra "l" in "%lld".
Also, fix off-by-two error in size calculation.
(width_bytes, integral_type_size): Add entries for uintmax_t.
(print_intmax): New function.
(decode_one_function): Use it.
(ISPEC_TO_FORMAT): New arg Max_fmt. All uses changed.

8 days agood: initialize type-size tables statically
Paul Eggert [Fri, 27 Jun 2025 06:51:37 +0000 (23:51 -0700)] 
od: initialize type-size tables statically

* src/od.c (NO_SIZE): Make it explicitly 0, as the
initializers now rely on this.
(MAX_INTEGRAL_TYPE_SIZE): Remove.  All uses replaced by
ARRAY_CARDINALITY (integral_type_size) - 1.
Move static assertion down to where this can be used.
(integral_type_size, fp_type_size): Make them const,
and initialize them statically.
(main): Omit no-longer-needed initialization code.

8 days agood: prefer intmax_t to uintmax_t
Paul Eggert [Thu, 26 Jun 2025 15:45:47 +0000 (08:45 -0700)] 
od: prefer intmax_t to uintmax_t

* src/od.c (MAX_ADDRESS_LENGTH, pseudo_offset, n_bytes_to_skip)
(max_bytes_to_format, end_offset, skip, format_address_none)
(format_address_std, format_address_paren, format_address_label)
(write_block, parse_old_offset, dump, dump_strings, main):
Prefer intmax_t to uintmax_t.  This makes no practical difference,
and lets -fsanitize=undefined check for signed integer overflow.
(skip, dump): Remove no-longer-needed casts.
(xstr2nonneg): New static function.  All callers of xstrtoumax
now call this function instead.
(main): Use ckd_add to detect signed integer overflow, since
the unsigned trick no longer works reliably.
Let xstrtol_fatal report the overflow, instead of doing
it by hand ourselves.

8 days agood: fix '+N.' bug
Paul Eggert [Sun, 29 Jun 2025 00:29:22 +0000 (17:29 -0700)] 
od: fix '+N.' bug

* src/od.c (parse_old_offset): First arg is now char *,
not char const *.  If a decimal number, temporarily
modify the string so that xstrtoumax does not complain
about the '.'.
* tests/od/od.pl: Test for the bug.

8 days agood: fix some unlikely integer overflows
Paul Eggert [Thu, 26 Jun 2025 06:22:37 +0000 (23:22 -0700)] 
od: fix some unlikely integer overflows

* src/od.c (print_n_spaces, pad_at, pad_at_overflow):
New static functions.
(struct tspec, PRINT_FIELDS, print_named_ascii, print_ascii)
(decode_one_format, write_block, main):
Use idx_t, not int, for counts that depend on the number
of bytes in an object.
(decode_one_format): Use print_n_spaces to output spaces.
(PRINT_FIELDS, print_named_ascii, print_ascii):
Use pad_at to avoid integer overflow.
(write_block): Do not use %*s to pad, as the total pad might
exceed INT_MAX.  Instead, pad by hand with putchar (' ').
(main): Use pad_at_overflow to report integer overflow due to
oversize -w.  Use better way to tell whether -w is used,
without needing IF_LINT.
* tests/od/big-w.sh: New test.
* tests/local.mk (all_tests): Add it.

8 days agood: prefer idx_t to size_t
Paul Eggert [Wed, 25 Jun 2025 03:34:29 +0000 (20:34 -0700)] 
od: prefer idx_t to size_t

This helps find overflow bugs when compiling with -fsanitize=undefined.
* src/od.c (struct tspec, bytes_per_block, PRINT_FIELDS)
(PRINT_TYPE, print_named_ascii, print_ascii, decode_one_format)
(skip, write_block, read_block, dump, main):
Use idx_t, not size_t.

8 days agood: don’t assume no holes in wide unsigned
Paul Eggert [Wed, 25 Jun 2025 03:01:04 +0000 (20:01 -0700)] 
od: don’t assume no holes in wide unsigned

Also, fix minor related typos.
* src/od.c (MAX_INTEGRAL_TYPE_SIZE, MAX_ADDRESS_LENGTH):
Now a constant, not a macro.
(MAX_INTEGRAL_TYPE_WIDTH): New constant.  Use it instead of
CHAR_BIT, so as not to assume that uintmax_t and unsigned long
long int are hole-free.  This doesn’t matter on practical porting
targets, though there is still a mainframe or two that have holes.
(FMT_BYTES_ALLOCATED): Fix typo by changing "jd" to "jo".
Fix off-by-one typo in static assertion.

8 days agomaint: assume long long int
Paul Eggert [Wed, 25 Jun 2025 02:13:20 +0000 (19:13 -0700)] 
maint: assume long long int

It’s long been safe to assume C99+ support for long long int.
* .gitignore: Remove m4/longlong.m4.
* bootstrap.conf (buildreq): Boost git prereq from 1.4.4 to 1.5.5,
syncing with Gnulib.
(bootstrap_post_import_hook): Remove m4/longlong.m4.
* m4/jm-macros.m4 (gl_CHECK_ALL_TYPES):
No need to require AC_TYPE_UNSIGNED_LONG_LONG_INT.
* src/factor.c (DItype, UDItype):
* src/od.c (main):
Assume HAVE_LONG_LONG_INT.
* src/od.c: (unsigned_long_long_int):
Remove.  All uses replaced with unsigned long long int.

8 days agood: fix another off-by-one issue with --strings
Paul Eggert [Sat, 28 Jun 2025 15:07:54 +0000 (08:07 -0700)] 
od: fix another off-by-one issue with --strings

* src/od.c (main): Fix off-by-one error in string_min limit.

8 days agood: fix theoretical size_t malloc overflow
Paul Eggert [Sat, 28 Jun 2025 15:15:42 +0000 (08:15 -0700)] 
od: fix theoretical size_t malloc overflow

* src/od.c (dump, dump_strings): Use idx_t allocators
rather than size_t allocators, to avoid unchecked integer
overflow on theoretical platforms where SIZE_MAX < IDX_MAX.

9 days agodoc: list signals added by POSIX.1-2024
Collin Funk [Sat, 28 Jun 2025 05:15:28 +0000 (22:15 -0700)] 
doc: list signals added by POSIX.1-2024

* doc/coreutils.texi (Signal specifications): Mention that POSIX.1-2024
added SIGWINCH.

9 days agodoc: mention that POSIX.1-2024 added dd iflag=fullblock
Collin Funk [Fri, 27 Jun 2025 01:48:51 +0000 (18:48 -0700)] 
doc: mention that POSIX.1-2024 added dd iflag=fullblock

* doc/coreutils.texi (dd invocation): Document that POSIX.1-2024 added
iflag=fullblock.

10 days agodoc: cksum: --check: improve supported formats description
Pádraig Brady [Fri, 27 Jun 2025 14:02:50 +0000 (15:02 +0100)] 
doc: cksum: --check: improve supported formats description

* doc/coreutils.texi (cksum common options): Reorder and tweak the info
to make it clearer that --check does not support the legacy crc output
from the cksum command.
Reported at https://bugs.debian.org/1108363

12 days agomaint: prefer UINTMAX_MAX to TYPE_MAXIMUM (uintmax_t)
Collin Funk [Wed, 25 Jun 2025 04:39:20 +0000 (21:39 -0700)] 
maint: prefer UINTMAX_MAX to TYPE_MAXIMUM (uintmax_t)

* src/df.c (get_dev): Use UINTMAX_MAX.
* src/shred.c (dopass): Likewise.

13 days agood: output standard diagnostics for invalid -w arguments
Pádraig Brady [Tue, 24 Jun 2025 14:47:48 +0000 (15:47 +0100)] 
od: output standard diagnostics for invalid -w arguments

* src/od.c (main): Don't pass LONGINT_OK to xstrtol_fatal(),
as otherwise it will abort().
* tests/od/od.pl: Add test cases.
* NEWS: Mention the bug fix.

Addresses https://bugs.gnu.org/78879

13 days agood: fix various off-by-one issues with --strings with -N
Pádraig Brady [Tue, 24 Jun 2025 00:17:12 +0000 (01:17 +0100)] 
od: fix various off-by-one issues with --strings with -N

* src/od.c (dump_strings): There are three related issues here
due to not accounting for the terminating NUL char appropriately.

1. Ensure BUF always has enough space for the terminating NUL.
This avoids CWE-122: Heap-based Buffer Overflow,
where we wrote a single NUL byte directly after the allocated buffer.
I.e., there should be no buffer overflow with:
  printf '%100s' | od -N100 -S1

2. Ensure we support -S == -N (END_OFFSET - STRING_MIN == ADDRESS):
I.e., there should be output with:
  printf '%100s' | od -N10 -S10

3. Ensure we always output a valid address by ensuring
the ADDRESS and I variables are kept in sync.
I.e., this should output address 0000000 not 1777777777777777777777:
  printf '%100s' | od -N10 -S1

As well as fixing these we simplify by using a single loop
to read the data, rather than two.

* doc/coreutils.texi (od invocation): Clarify that -N
implicitly NUL terminates strings.
* tests/od/od-N.sh: Add test cases.
* NEWS: Mention the bug fixes.

Fixes https://bugs.gnu.org/78880

13 days agobuild: fix compilation error on OpenBSD and Solaris
Bruno Haible [Mon, 23 Jun 2025 20:59:47 +0000 (22:59 +0200)] 
build: fix compilation error on OpenBSD and Solaris

* src/speedgen: Use 'printf', not 'sed', to emit code with newlines.

13 days agomaint: make it clear when to add speeds to the speedgen list
H. Peter Anvin [Mon, 23 Jun 2025 18:20:48 +0000 (11:20 -0700)] 
maint: make it clear when to add speeds to the speedgen list

Add a comment making it clear that it is perfectly safe to add
additional speeds to the canned list in the speedgen script, and under
what conditions it is appropriate to do so.

Signed-off-by: "H. Peter Anvin" (Intel) <hpa@zytor.com>
13 days agostty: support baud rate macro extraction on more compilers
H. Peter Anvin [Mon, 23 Jun 2025 18:20:47 +0000 (11:20 -0700)] 
stty: support baud rate macro extraction on more compilers

For generating src/speedtest.h, try a sequence of command-line options
until (or unless) one of them works:

  -dM: gcc, clang and derived compilers, icc classic
  -xdumpmacros: Sun Studio (writes to stderr!)
  -qshowmacros: IBM XL classic
  -PD: MSVC (usable with a wrapper such as cccl from the SWIG project)

Because Sun Studio -xdumpmacros unconditionally writes to stderr,
capture stderr output instead of sending it to /dev/null. This is
perfectly safe, even in the presence of stray stderr output, because:

  1. speedgen ignores input that is not of the form #define B<number>
  2. even if a line of that format would somehow spuriously appear,
     the only outcome is that the generated C code will probe for a few
     more macros.

Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
2 weeks agostdbuf: support AIX
Collin Funk [Mon, 23 Jun 2025 01:13:36 +0000 (18:13 -0700)] 
stdbuf: support AIX

* src/stdbuf.c (set_LD_PRELOAD): Use the AIX specific environment
variables LDR_PRELOAD or LDR_PRELOAD64.
* NEWS: Mention the bugfix.

2 weeks agostty: stricter floating point parsing
Pádraig Brady [Sun, 22 Jun 2025 15:40:04 +0000 (16:40 +0100)] 
stty: stricter floating point parsing

* src/stty.c (string_to_baud): Disallow extraneous characters
after floating point numbers.
* tests/stty/stty-invalid.sh: Add test cases.

2 weeks agodoc: stty: adjust description of supported speeds
Pádraig Brady [Sun, 22 Jun 2025 14:01:21 +0000 (15:01 +0100)] 
doc: stty: adjust description of supported speeds

* doc/coreutils.texi (stty invocation): Remove now imprecise
list of speeds given we may now support higher or arbitrary speeds.
Mention that we may support higher or arbitrary speeds.

2 weeks agobuild: add src/termios.c to the tarball
Collin Funk [Sun, 22 Jun 2025 05:05:19 +0000 (22:05 -0700)] 
build: add src/termios.c to the tarball

* src/local.mk (EXTRA_DIST): Add src/termios.c.

2 weeks agomaint: use _GL_CMP instead of handwriting three-valued comparisons
Collin Funk [Sun, 22 Jun 2025 04:50:27 +0000 (21:50 -0700)] 
maint: use _GL_CMP instead of handwriting three-valued comparisons

* src/comm.c (compare_files): Use _GL_CMP.
* src/join.c (keycmp): Likewise.
* src/ls.c (off_cmp): Likewise.
* src/ptx.c (compare_words, compare_occurs): Likewise.
* src/set-fields.c (compare_ranges): Likewise.
* src/sort.c (compare_random, diff_reversed, compare): Likewise.

2 weeks agodoc: NEWS: improve recent stty details
H. Peter Anvin [Sat, 21 Jun 2025 01:01:58 +0000 (18:01 -0700)] 
doc: NEWS: improve recent stty details

* src/NEWS: Fix grammar error,
and indicate baud rate probing is opportunistic.

2 weeks agotty: translate new error
Pádraig Brady [Sat, 21 Jun 2025 09:59:46 +0000 (10:59 +0100)] 
tty: translate new error

* src/tty.c (main): Translate and adjust new ttyname() error
diagnostic, to be clearer and avoid a syntax check failure.

2 weeks agotty: better fix for Bug#26371
Paul Eggert [Fri, 20 Jun 2025 18:53:21 +0000 (11:53 -0700)] 
tty: better fix for Bug#26371

* src/tty.c (TTY_USAGE): Rename from TTY_FAILURE, since this
is used only for usage failures.  All uses changed.
(TTY_TTYNAME_FAILURE): New constant.
(main): Remove no-longer-needed assignment of ENOENT to errno.
Make status-setting clearer too.
Report an error if ttyname fails even though stdin is a terminal,
instead of silently pretending that stdin is not a terminal.
* tests/tty/tty.sh: Test for this issue.  This should fix Bug#78244.

2 weeks agotests: stty: adjust tests for arbitary speeds
Pádraig Brady [Tue, 17 Jun 2025 22:32:05 +0000 (23:32 +0100)] 
tests: stty: adjust tests for arbitary speeds

* tests/stty/stty-invalid.sh: Adjust to what is now invalid.
* tests/stty/stty.sh: Add checks for valid speed variants.
* tests/stty/bad-speed.sh: New test to ensure unsupported speeds
are diagnosed.

2 weeks agostty: arbitrary or non-a priori known speed_t support
H. Peter Anvin [Mon, 16 Jun 2025 21:58:01 +0000 (14:58 -0700)] 
stty: arbitrary or non-a priori known speed_t support

Support the case where speed_t is simply a number, and in that case
assume that arbitrary values can be passed.  This is assumed to be the
case when all known speed_t macros equal their own value.

Try to probe for a variety of speed_t constants by trying to coax
$(CC) into emitting macro definitions (-E -dM).  If this is not
supported, use a fairly extensive list of constants as a
fallback.  This both improves the test for arbitrary speed support, as
well as allowing proper operation in the case where the constants are
not plain numbers and allows for handing enumerated speed constants
that were not known a priori when the source code was written.

A simple shell script (mostly using sed) is used to turn the list of
constants (probed and predefined) into a pair of conversion functions,
baud_to_value() and value_to_baud(); string_to_baud() is then
reimplemented as a wrapper around the latter.

* src/.gitignore: Add generated speedlist.h.
* src/local.mk: Generate speedlist.h.
* src/speedgen: Portable shell script to generate speedlist.h.
* src/stty.c: Adjust string_to_baud to
convert from arbitrary numeric values.
* src/termios.c: A helper used when generating speedlist.h
* NEWS: Mention the improvement.

3 weeks agodate: port test to NetBSD 10
Paul Eggert [Tue, 10 Jun 2025 19:51:05 +0000 (12:51 -0700)] 
date: port test to NetBSD 10

Problem reported by Collin Funk in:
https://lists.gnu.org/r/bug-gnulib/2025-06/msg00094.html
* tests/date/date-debug.sh: Also allow NetBSD 10 mktime behavior.
Although NetBSD contradicts POSIX, POSIX is likely wrong here and
I vaguely recall that there’s a POSIX correction in progress
that will allow the NetBSD behavior.

4 weeks agobuild: update gnulib submodule to latest
Paul Eggert [Mon, 9 Jun 2025 06:52:51 +0000 (23:52 -0700)] 
build: update gnulib submodule to latest

* tests/date/date-debug.sh: Update to match current behavior
of Gnulib's parse-datetime module.

4 weeks agodoc: cp --help: mention --reflink=auto is now the default
Pádraig Brady [Sat, 7 Jun 2025 13:27:59 +0000 (14:27 +0100)] 
doc: cp --help: mention --reflink=auto is now the default

* src/cp.c (usage): Reword the REFLINK description to
indicate that --reflink=auto is the default.
Fixes https://github.com/coreutils/coreutils/issues/103

4 weeks agodoc: make command synopsis more consistent
Pádraig Brady [Tue, 3 Jun 2025 21:58:38 +0000 (22:58 +0100)] 
doc: make command synopsis more consistent

* doc/coreutils.texi: Use @dots{} after [OPTION],
to be consistent with man pages.
* src/printenv.c (Usage): Remove unneeded "...".
* src/timeout.c (Usage): Add needed "...",
and also remove redundant [OPTION] only form.
* src/chroot.c (Usage): Likewise.
Fixes https://bugs.gnu.org/78628

5 weeks agocksum: port to Apple LLVM version 10.0.0
Paul Eggert [Thu, 29 May 2025 15:30:13 +0000 (08:30 -0700)] 
cksum: port to Apple LLVM version 10.0.0

This compiler lacks support for the x86-64 VPCLMULQDQ instruction;
compiles fail with "error: invalid cpu feature string for builtin".
Problem reported by Peter Dyballa, and fix suggested by
Pádraig Brady <https://bugs.gnu.org/78562#8>.
* configure.ac (USE_AVX512_CRC32):
Also check that __builtin_cpu_supports ("vpclmulqdq") compiles.

5 weeks agotests: sort: expand multiple output testing
Pádraig Brady [Wed, 28 May 2025 19:48:31 +0000 (20:48 +0100)] 
tests: sort: expand multiple output testing

* tests/sort/sort.pl: Ensure duplicate
output strings are supported.

5 weeks agomaint: assume O_DIRECTORY
Paul Eggert [Wed, 28 May 2025 19:35:14 +0000 (12:35 -0700)] 
maint: assume O_DIRECTORY

Recent changes in Gnulib guarantee O_DIRECTORY != 0.
* gl/lib/targetdir.c (target_directory_operand):
* src/dd.c (usage):
* src/ln.c (main):
Simplify now that O_DIRECTORY must be nonzero.

5 weeks agobuild: update gnulib submodule to latest
Paul Eggert [Wed, 28 May 2025 19:35:14 +0000 (12:35 -0700)] 
build: update gnulib submodule to latest

6 weeks agomaint: avoid a sc_system_h_headers failure
Pádraig Brady [Sat, 24 May 2025 21:01:44 +0000 (22:01 +0100)] 
maint: avoid a sc_system_h_headers failure

* cfg.mk: Also exclude libstdbuf.c from this check.

6 weeks agostdbuf: don’t allocate more than SIZE_MAX / 2
Paul Eggert [Sat, 24 May 2025 20:01:31 +0000 (13:01 -0700)] 
stdbuf: don’t allocate more than SIZE_MAX / 2

* src/libstdbuf.c (apply_mode): Be more conservative about
sizes passed to malloc, since we can’t rely on Gnulib malloc.

6 weeks agostdbuf: simplify apply_mode + getenv
Paul Eggert [Sat, 24 May 2025 19:55:13 +0000 (12:55 -0700)] 
stdbuf: simplify apply_mode + getenv

* src/libstdbuf.c (apply_mode): Take getenv arg, not result.
All callers changed.  This is simpler.

6 weeks agostdbuf: port better to Mac OS X 10.4.11
Paul Eggert [Sat, 24 May 2025 19:47:06 +0000 (12:47 -0700)] 
stdbuf: port better to Mac OS X 10.4.11

Do not link to fchownat, fchmodat, xreallocarray.
While we’re at it, depend as little as possible on problematic
functions like fileno, strtoumax.  Use only functions that
were around in C89.
Problem reported by Peter Dyballa <https://bugs.gnu.org/78509#59>.
* src/libstdbuf.c: Include stddef.h, stdlib.h, gettext.h.
Do not include stdint.h or system.h.
(_): New macro, since we no longer include system.h.
Undef strtoul, not strtoumax.
(fileno_to_name): Remove.  All uses removed.
(apply_mode): New arg STREAM_NAME, so that we no longer need
to depend on fileno which might run into a Gnulib workaround.
Don’t allocate more than ULONG_MAX - 1 bytes, as that’s
not helpful in the real world and this lessens dependency
on newer features like strtoumax.

6 weeks agotests: avoid hung tests on older macOS
Pádraig Brady [Wed, 21 May 2025 20:20:11 +0000 (21:20 +0100)] 
tests: avoid hung tests on older macOS

These tests were seen to hang on PPC Mac OS X 10.4.11

* tests/cp/existing-perm-race.sh: Protect fifo interactions
with a 1 minute timeout.
* tests/cp/file-perm-race.sh: Likewise.
Reported in https://bugs.gnu.org/78509

6 weeks agosort: fix buffer under-read (CWE-127)
Pádraig Brady [Tue, 20 May 2025 15:03:44 +0000 (16:03 +0100)] 
sort: fix buffer under-read (CWE-127)

* src/sort.c (begfield): Check pointer adjustment
to avoid Out-of-range pointer offset (CWE-823).
(limfield): Likewise.
* tests/sort/sort-field-limit.sh: Add a new test,
which triggers with ASAN or Valgrind.
* tests/local.mk: Reference the new test.
* NEWS: Mention bug fix introduced in v7.2 (2009).
Fixes https://bugs.gnu.org/78507

7 weeks agomaint: omit -Wmissing-variable-declarations from tests
Paul Eggert [Mon, 19 May 2025 06:34:28 +0000 (23:34 -0700)] 
maint: omit -Wmissing-variable-declarations from tests

* configure.ac (GNULIB_WARN_CFLAGS):
Omit -Wmissing-variable-declarations; needed for
gcc (GCC) 15.1.1 20250425 (Red Hat 15.1.1-1).

7 weeks agomaint: avoid sc_preprocessor_indentation failure
Pádraig Brady [Sun, 18 May 2025 09:37:35 +0000 (10:37 +0100)] 
maint: avoid sc_preprocessor_indentation failure

* src/factor.c: Tweak preprocessor indentation.

7 weeks agobuild: generate src/crctab.c when building from git
Collin Funk [Sun, 18 May 2025 04:43:31 +0000 (21:43 -0700)] 
build: generate src/crctab.c when building from git

* src/crctab.c: Remove file.
* src/cksum.c [!CRCTAB]: Include config.h.
* src/local.mk: Generate src/crctab.c in a temporary directory.
(EXTRA_DIST): Add src/crctab.c.
* .gitignore (/src/crctab.c): Add entry.

7 weeks agobuild: fix 'make distcheck'
Collin Funk [Sun, 18 May 2025 04:02:20 +0000 (21:02 -0700)] 
build: fix 'make distcheck'

* src/local.mk (EXTRA_DIST): Distribute src/make-prime-list.c
to satisfy the new make dependency.

7 weeks agofactor: fix bug with 128-bit uintmax_t
Paul Eggert [Sun, 18 May 2025 04:42:37 +0000 (21:42 -0700)] 
factor: fix bug with 128-bit uintmax_t

On so-far-only-theoretical platforms with 128-bit uintmax_t,
'factor' would misbehave by not factoring enough.
Work around the bug (at a performance cost)
and document the issue.  I hope someone with
more time and expertise can fix the performance cost
that this introduces.
To reproduce the correctness bug, build with
'gcc -DUSE_INT128 -DEXHIBIT_INT128_BUG';
'make check' should fail due to the new test case.
* src/factor.c (USE_INT128): New macro.
(wide_uint, wide_int, W_TYPE_SIZE, WIDE_UINT_MAX):
Define to proper values if USE_INT128.
(prime_p) [!EXHIBIT_INT128_BUG]: Work around bug
with 128-bit wide_uint, at some performance cost.
* tests/factor/factor.pl (bug-with-128-bit-uintmax_t):
New test.

7 weeks agofactor: refactor to allow non-uintmax_t ints
Paul Eggert [Sun, 18 May 2025 04:42:37 +0000 (21:42 -0700)] 
factor: refactor to allow non-uintmax_t ints

This prepares for using unsigned __int128 instead of uintmax_t.
It doesn’t change behavior.
* src/factor.c (wide_uint, wide_int):
New typedefs, replacing all uses of uintmax_t, intmax_t.
(WIDE_UINT_MAX): New macro, replacing all uses of UINTMAX_MAX.
(USE_LONGLONG_H): Don’t use LONG_MAX and INTMAX_MAX (which
should have been ULONG_MAX and UINTMAX_MAX anyway).
Instead, use W_TYPE_SIZE and ULONG_WIDTH, as that will be
easier to work with if we use unsigned __int128.
(binv): Also unroll for the 128-bit case.
* src/make-prime-list.c (print_wide_uint):
Print "wide_uint" not "uintmax_t".

7 weeks agobuild: support cross-compiling directly from git checkout
Collin Funk [Fri, 16 May 2025 05:02:26 +0000 (22:02 -0700)] 
build: support cross-compiling directly from git checkout

* src/local.mk: Compile the make-prime-list in a temporary directory
using $(BUILD_CC) so it can be run even when $(CC) is a cross-compiler.
Add some comments.
(noinst_PROGRAMS): Remove src/make-prime-list.
(src_make_prime_list_LDADD): Remove variable.
* src/make-prime-list.c: Don't include config.h and attributes.h.
Remove unnecessary #undefs
(ATTRIBUTE_CONST, ATTRIBUTE_MALLOC): Define for the host compiler.
Fixes https://bugs.gnu.org/78377

7 weeks agomaint: --help now outputs bug reporting address
Paul Eggert [Wed, 14 May 2025 20:36:40 +0000 (13:36 -0700)] 
maint: --help now outputs bug reporting address

This changes the tail of "cat --help" output from:

GNU coreutils online help: <https://www.gnu.org/software/coreutils/>
Full documentation <https://www.gnu.org/software/coreutils/cat>
or available locally via: info '(coreutils) cat invocation'

to:

Report bugs to: bug-coreutils@gnu.org
GNU coreutils home page: <https://www.gnu.org/software/coreutils/>
General help using GNU software: <https://www.gnu.org/gethelp/>
Full documentation <https://www.gnu.org/software/coreutils/cat>
or available locally via: info '(coreutils) cat invocation'

* src/system.h (emit_bug_reporting_address): Don’t #define.
(emit_ancillary_info): Call emit_bug_reporting_address
instead of doing it by hand, which omitted the bug reporting address.

7 weeks agomaint: avoid sc_gitignore_missing failure
Pádraig Brady [Mon, 12 May 2025 22:20:43 +0000 (23:20 +0100)] 
maint: avoid sc_gitignore_missing failure

* .gitignore: Add the soon to be generated lib/fts_.h
* cfg.mk (sc_gitignore_missing): Special case the mapping
of fts.in.h to fts_.h.
Fixes https://bugs.gnu.org/78398

8 weeks agodoc: NEWS: fix introduced version of a bug
Pádraig Brady [Mon, 12 May 2025 10:52:17 +0000 (11:52 +0100)] 
doc: NEWS: fix introduced version of a bug

* NEWS: The NFSv4 trivial ACL ENOTSUP issue was introduced in 9.6.

8 weeks agomaint: avoid sc_GPL_version failure
Pádraig Brady [Mon, 12 May 2025 10:50:25 +0000 (11:50 +0100)] 
maint: avoid sc_GPL_version failure

* cfg.mk (VC_LIST_ALWAYS_EXCLUDE_REGEX): Exclude longlong.h also.

8 weeks agodoc: suggest GNULIB_REFDIR instead of GNULIB_SRCDIR
Collin Funk [Sun, 11 May 2025 17:07:22 +0000 (10:07 -0700)] 
doc: suggest GNULIB_REFDIR instead of GNULIB_SRCDIR

* README-hacking: Use GNULIB_REFDIR instead of GNULIB_SRCDIR since that
is the correct way to have ./bootstrap do 'git clone --reference'.

8 weeks agofactor: check overflow with ckd_add etc
Paul Eggert [Mon, 12 May 2025 07:09:16 +0000 (00:09 -0700)] 
factor: check overflow with ckd_add etc

* src/factor.c (umul_ppmm, add_ssaaaa, sub_ddmmss, submod)
(mulredc, strto2wide_uint): Prefer ckd_add etc. to doing
overflow checking by hand.

8 weeks agomaint: pacify GCC when selinux is missing
Paul Eggert [Mon, 12 May 2025 02:26:38 +0000 (19:26 -0700)] 
maint: pacify GCC when selinux is missing

* src/selinux.h (restorecon, defaultcon): Use MAYBE_UNUSED.

8 weeks agofactor: be more cautious about C compiler versions
Paul Eggert [Mon, 12 May 2025 01:22:44 +0000 (18:22 -0700)] 
factor: be more cautious about C compiler versions

* src/factor.c (__GMP_GNUC_PREREQ): Define compatibly with GMP.

8 weeks agomaint: sync longlong.h from gmp
Paul Eggert [Mon, 12 May 2025 01:00:50 +0000 (18:00 -0700)] 
maint: sync longlong.h from gmp

* .x-update-copyright: Add src/longlong.h.
* src/longlong.h: Sync from gmp.

8 weeks agofactor: omit incorrect affirms in mulredc2
Paul Eggert [Mon, 12 May 2025 00:25:30 +0000 (17:25 -0700)] 
factor: omit incorrect affirms in mulredc2

* src/factor.c (mulredc2): Remove two ‘affirm’ calls that didn’t
match the accompanying comment, and one of which has a false
positive if UINTMAX_WIDTH == 128 and we factor 2**128 + 1.

8 weeks agobuild: update gnulib submodule to latest
Collin Funk [Sun, 11 May 2025 16:56:37 +0000 (09:56 -0700)] 
build: update gnulib submodule to latest

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

8 weeks agomaint: update NEWS from Gnulib update
Paul Eggert [Sat, 10 May 2025 05:35:34 +0000 (22:35 -0700)] 
maint: update NEWS from Gnulib update

8 weeks agobuild: update gnulib submodule to latest
Paul Eggert [Sat, 10 May 2025 01:05:48 +0000 (18:05 -0700)] 
build: update gnulib submodule to latest

8 weeks agomaint: minor followon touchup of GCC warnings
Paul Eggert [Sat, 10 May 2025 04:01:14 +0000 (21:01 -0700)] 
maint: minor followon touchup of GCC warnings

* configure.ac (GNULIB_TEST_WARN_CFLAGS):
Do not use -Wsuggest-attribute=cold, -Wsuggest-attribute=const,
-Wsuggest-attribute=format as they produce false positives with
GCC 15 x86-64.  Use -Wmissing-variable-declarations, as it no
longer seems to hurt.

8 weeks agomaint: modernize GCC warning configuration
Paul Eggert [Fri, 9 May 2025 22:55:47 +0000 (15:55 -0700)] 
maint: modernize GCC warning configuration

* configure.ac: Simplify the setup of GCC’s -W... options,
by removing options no longer needed for GCC 15 (when
configuring with --enable-gcc-warnings) or GCC 12 (without).
I hope this is good enough nowadays.
Add GCC 15’s -Wtrailing-whitespace, since that works for us.

8 weeks agomaint: use MAYBE_UNUSED more consistently
Paul Eggert [Fri, 9 May 2025 22:52:15 +0000 (15:52 -0700)] 
maint: use MAYBE_UNUSED more consistently

* src/basenc.c (no_required_padding):
* gl/lib/smack.h (smack_new_label_from_self)
(smack_set_label_for_self):
* src/cksum.c (output_crc):
* src/digest.c (md5_sum_stream, sha1_sum_stream)
(sha224_sum_stream, sha256_sum_stream, sha384_sum_stream)
(sha512_sum_stream, sm3_sum_stream, output_file):
* src/stat.c (print_statfs, print_stat):
* src/sum.c (output_bsd, output_sysv):
* src/timeout.c (chld):
Mark possibly-unused parameters with MAYBE_UNUSED.

8 weeks agomaint: fix Automake warning
Paul Eggert [Fri, 9 May 2025 22:49:23 +0000 (15:49 -0700)] 
maint: fix Automake warning

* Makefile.am (prologue): Remove macro, as \# is not portable here
and Automake complains.  Replace its only use with its definiens.

2 months agotests: avoid false failures in user namespaces
Pádraig Brady [Sat, 3 May 2025 08:59:16 +0000 (09:59 +0100)] 
tests: avoid false failures in user namespaces

* init.cfg (require_membership_in_two_groups_): Skip
overflow gids used in user namespaces, as one can't
chgrp() to these.
Fixes https://bugs.gnu.org/78225

2 months agobuild: update to latest gnulib
Pádraig Brady [Thu, 1 May 2025 13:05:59 +0000 (14:05 +0100)] 
build: update to latest gnulib

* bootstrap.conf: Adjust for hash-triple module renaming.
* src/copy.c: Likewise.
* src/ln.c: Likewise.
* gnulib: Update to latest.

2 months agotests: add a test case for negative relative date
Pádraig Brady [Wed, 23 Apr 2025 11:47:54 +0000 (12:47 +0100)] 
tests: add a test case for negative relative date

* tests/date/date.pl: To exercise option processing.

2 months agotests: fix false failure with constrained setfacl
Pádraig Brady [Mon, 21 Apr 2025 10:48:33 +0000 (11:48 +0100)] 
tests: fix false failure with constrained setfacl

* tests/ls/acl.sh: In constrained environments, like with
setxattr() disabled for example, require_acl_ does not
check for the required functionality, so use the more
stringent require_setfacl_ instead.
Reported at https://github.com/NixOS/nixpkgs/issues/394953