Jim Meyering [Wed, 19 Nov 2008 18:36:45 +0000 (19:36 +0100)]
cp: use far less memory in some cases
cp --link was "remembering" many name,dev,inode triples unnecessarily.
cp was doing the same, even without --link, for every directory in the
source hierarchy, while it can do its job with entries merely for the
command-line arguments. Prompted by a report from Patrick Shoenfeld.
Details <http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/15081>.
* src/copy.c (copy_internal): Refrain from remembering
name,dev,inode for most files, when invoked via cp --link.
Record an infloop-avoidance triple for each directory specified
on the command line, not for each directory in the source tree.
Don't record a dir-triple when x->hard_link is set.
* NEWS (Buf fixes): Mention it.
* tests/cp/link-heap: New file. Test for cp's lowered memory usage.
* tests/Makefile.am (TESTS): Add link-heap.
Jim Meyering [Sun, 16 Nov 2008 08:53:48 +0000 (09:53 +0100)]
tests: change misc/printf-surprise so that it doesn't segfault anymore
* tests/misc/printf-surprise: Disable MALLOC_PERTURB_, so that "make
check" no longer provokes a segfault from printf(1). Before, that
would be detected as a known problem and cause the test to be skipped.
Adjust the test so that a segfault once again results in test failure.
Jim Meyering [Wed, 12 Nov 2008 11:53:54 +0000 (12:53 +0100)]
avoid spurious test failure when df always fails
* tests/df/total-verify: Skip upon _df_ failure, not tee failure.
Reported by Ondřej Vašík. Details in
<http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/15028>.
Jim Meyering [Sat, 8 Nov 2008 18:59:52 +0000 (19:59 +0100)]
emit more precise bug-reporting instructions
Hoping that this will decrease the volume of bug reports
mistakenly directed to the coreutils mailing list.
* src/system.h (emit_bug_reporting_address): Specify the program
name in the "Report bugs to <...>." line in each program's --help
output and man page. Now, it will say "Report PROGRAM_NAME bugs
to <...>." Suggestion from Ondřej Vašík.
Jim Meyering [Wed, 5 Nov 2008 18:18:59 +0000 (19:18 +0100)]
include gnulib's strdup module explicitly
* bootstrap.conf (gnulib_modules): Include strdup explicitly, to
keep the latest version of gnulib-tool from omitting this now-marked-
as-obsolete module.
Kamil Dudka [Mon, 27 Oct 2008 14:38:23 +0000 (15:38 +0100)]
ls: --color now highlights hard linked files, too
* src/ls.c (print_color_indicator): Colorize hard linked files.
* src/dircolors.c: Add color for hard link.
* src/dircolors.hin: Add color for hard link.
* tests/ls/hardlink: Test for ls - colorize hard linked files.
* tests/Makefile.am (TESTS): Add ls/hardlink.
* NEWS: Mention the change.
Paul Eggert [Sun, 26 Oct 2008 07:45:18 +0000 (00:45 -0700)]
seq: improve quality of format-checking code
* src/seq.c (validate_format): Remove. Migrate its checks into...
(long_double_format): Report an error and exit if an error is found,
instead of returning NULL. All callers changed.
Use a more-consistent format for diagnostics.
* tests/misc/seq: Adjust to the more-consistent format for diagnostics.
Paul Eggert [Fri, 24 Oct 2008 20:43:40 +0000 (22:43 +0200)]
factor: remove --bignum and --no-bignum options
Here's a patch to remove the --bignum and --no-bignum options from
'factor'. The case for removing --bignum isn't as strong as that for
'expr', but still, it seems to me that these options are not needed and
complicate the documentation unnecessarily.
* doc/coreutils.texi (factor invocation): Remove --bignum, --no-bignum.
* src/factor.c (algorithm, ALGORITHM_CHOICE, USE_BIGNUM, NO_USE_BIGNUM):
Remove; all uses removed.
(extract_factors_multi): Remove, replacing with....
(print_factors_multi): New function, with signature similar to that
of new signature of print_factors_single.
(print_factors_single): Migrate checking code to caller.
(print_factors): Use GMP if it's available; don't bother asking user.
Improve accuracy of check for "large" numbers.
(long_options, main): Remove support for --bignum.
Paul Eggert [Sat, 25 Oct 2008 05:52:22 +0000 (22:52 -0700)]
seq: revert Solaris 8 work-around that caused x86 regression
* src/seq.c: Don't include <math.h>, <float.h>.
(abs_rel_diff): Remove.
(print_numbers): Test for equality, not for an epsilonish value.
This reverts 4827dd27b0c655a685947aaa01426a5ecba179f3, aka v6.10-185-g4827dd2, which broke 'seq' on the x86; for example, it
causes "seq 92233720368547758079223372036854775808" to incorrectly
output 3 numbers instead of 2. It's better to punish obsolescent
hosts that have incorrectly-working floating-point than to punish
correctly-working hosts.
* tests/misc/seq: Use 0.9000000000000, rather than
0.90000000000000000000, to avoid tickling a bug in Solaris 8 strtold,
which converts "0.9" and "0.9000000000000" correctly, but incorrectly
converts "0.90000000000000000000" to a smaller value.
Jim Meyering [Wed, 2 Apr 2008 20:26:45 +0000 (22:26 +0200)]
ls: use '.' (not +) as SELinux-only alt. access flag in ls -l output
* src/ls.c (gobble_file) [long_format]: Map SELinux-only to '.',
any other nonempty combination of MAC and ACL to '+', and all else
to the usual ' '. Suggested by Michael Stone.
* tests/misc/selinux: Adapt: expect '.', not '+'.
* doc/coreutils.texi (What information is listed): Document this.
* NEWS (Changes in behavior): Mention it.
Pádraig Brady [Wed, 22 Oct 2008 21:13:49 +0000 (22:13 +0100)]
Remove lzma as a specific build requirement
* bootstrap.conf: lzma was added as a build requirement
in the recently added build tools requirements check.
Remove that as it's quite new and also only required
for the make dist stage.
Pádraig Brady [Tue, 21 Oct 2008 21:40:12 +0000 (22:40 +0100)]
Add better checks and docs for build tools
Prompted by a report from Ed Avis:
<http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/14710>
* README-hacking: Organise LZMA and Valgrind as
as optional requirements rather than in their own sections.
Mention bootstrap will now check tool versions.
* README-prereq: Make a start on specific instructions
for optaining build tools. Currently we just have notes
for Fedora linux.
* bootstrap.conf: Add the list of tools and versions required.
* bootstrap: Add the logic to check for the required tools,
and list all required tools and versions if any are missing.
Jim Meyering [Wed, 22 Oct 2008 10:44:17 +0000 (12:44 +0200)]
clean up gl/modules/selinux-at
* gl/modules/selinux-at:
Ensure that LIB_SELINUX is cleared, in case it's set in the environment.
m4-quote the first two args to AC_SEARCH_LIBS.
Don't violate autoconf's ac_ namespace: s/ac_save/gl_save/
Drop the useless double quotes around a simple assignment RHS.
Jim Meyering [Tue, 21 Oct 2008 13:38:48 +0000 (15:38 +0200)]
prepare to move selinux-h module to gnulib
* gl/modules/selinux-h (Makefile.am)
[selinux/selinux.h, selinux/context.h]:
Remove temporary file and target, in case they're read-only.
Use $(MKDIR_P), not mkdir -p.
(License): Relax to LGPLv2+.
Remove vestigial comments.
Kamil Dudka [Mon, 20 Oct 2008 13:39:07 +0000 (15:39 +0200)]
ls: make it possible to disable file capabilities checking
* ls.c (print_color_indicator): Do not check for file capability
if that attribute is not being colored.
Since 84f6abfe00b4ab533145623638b417a2221f9c75, ls --color would check
each file for "capabilities". In <http://bugzilla.redhat.com/467508>,
James Antill reported that ls --color seemed slower with capabilities
detection.
Pádraig Brady [Thu, 16 Oct 2008 18:09:59 +0000 (19:09 +0100)]
expr: Fixup authors
* src/expr.c: Standardise the format of AUTHORS to
that used in other utils with multiple authors.
Also add Paul Eggert since he basically rewrote it
with his bignum fixes.
* AUTHORS (expr): Add Paul Eggert.
Jim Meyering [Wed, 8 Oct 2008 08:44:12 +0000 (10:44 +0200)]
maint: avoid warnings due to attribute warn_unused_result
Now that a (void) cast no longer suffices to ignore warnings from gcc
about uses of functions marked with the warn_unused_result attribute,
we need an alternative. For the record, here's one of the ignorable
warnings: "copy.c:233: warning: ignoring return value of 'fchown',
declared with attribute warn_unused_result"
* bootstrap.conf (gnulib_modules): Import ignore-value.
* src/copy.c: Include "ignore-value.h".
(set_owner): Use ignore_value in place of "(void)" casts,
to ignore lchown and fchown failures.
* src/cp.c (re_protect): Likewise, to ignore lchown failure.
* src/remove.c (preprocess_dir): Remove unnecessary "(void)" cast.
Jim Meyering [Thu, 16 Oct 2008 12:22:46 +0000 (14:22 +0200)]
build: when po-check fails, say why and suggest a fix
* maint.mk (po-check): Before, when this check failed, it just
spat out a diff mentioning two temporary files. Now, it tells
you what's wrong and suggests a fix with a patch using the name
of the affected file (rather than temporary file names) in the
diff output.
Eric Blake [Tue, 14 Oct 2008 04:04:51 +0000 (22:04 -0600)]
csplit: prefer sigaction over signal
* bootstrap.conf (gnulib_modules): Import sigaction.
* src/csplit.c (sigprocmask, siginterrupt) [SA_NOCLDSTOP]: Delete
workarounds.
(interrupt_handler, main): Drop use of signal. Rely on sigaction
to block fatal signal during cleanup, and to restore it to default
in case of nested signals.
Paul Eggert [Wed, 15 Oct 2008 13:29:38 +0000 (15:29 +0200)]
expr: remove --bignum and --no-bignum options
* doc/coreutils.texi (expr invocation): Remove the --bignum and
--no-bignum options. They weren't really needed, and they broke
longstanding (albeit nonportable) scripts.
* src/expr.c: Don't include <assert.h>. Include "inttostr.h",
"long-options.h", "verify.h". Check at compile-time that
size_t fits in unsigned long int, as the code assumes this in
several places.
(HAVE_GMP): Define to 0 if not defined, for convenience.
(mpz_t, mpz_clear, mpz_init_set_ui, mpz_init_set_str, mpz_add):
(mpz_sub, mpz_mul, mpz_tdiv_q, mpz_tdiv_r, mpz_get_str, mpz_sgn):
(mpz_fits_ulong_p, mpz_get_ui, mpz_out_str):
Supply substitutes when !HAVE_GMP, which work well enough for
expr's purposes.
(mp_integer): Remove. All integers are gmp, if gmp is available.
(struct valinfo): Remove 'z' member; no longer needed. The 'i'
member is always of type mpz_t.
(enum arithmetic_mode, MP_NEVER, MP_ALWAYS, MP_AUTO, mode):
Remove; no longer needed.
(usage): Remove documentation of --bignum and --no-bignum.
(integer_overflow): Abort if error misbehaves, to pacify GCC.
Restore old message on arithmetic overflow, to be conservative.
(die): Omit exit_status parameter; not needed (is always EXPR_FAILURE).
(string_too_long, USE_BIGNUM, NO_USE_BIGNUM, long_options):
Remove; no longer needed.
(main): Don't use getopt_long; this breaks old nonportable scripts.
(int_value): Arg is unsigned, in case we have strings whose length
exceeds LONG_MAX (!).
(int_value, freev, printv, null, tostring, toarith):
(eval6, eval4, eval3):
Always use mpz_ functions, to simplify the code.
(substr_value): Remove; no longer needed.
(getsize): Simplify the API: one arg rather than 3. Don't assume
unsigned long int fits in size_t.
(promote, domult, dodivide, doadd): Remove; no longer needed.
* tests/misc/expr: Don't use --bignum to test for bignum support.
Instead, use big numbers to test this.
Jim Meyering [Wed, 15 Oct 2008 07:19:29 +0000 (09:19 +0200)]
expr: don't interpret argv[0] as an operand
* src/expr.c (main): Given a first argument like -22 (negative, with
two or more digits), expr would decrement optind to 0 and then attempt
to evaluate argv[0].
Jim Meyering [Sun, 12 Oct 2008 17:53:20 +0000 (19:53 +0200)]
doc: reorder tsort/ptx nodes so body and menu ordering are consistent
* doc/coreutils.texi (tsort invocation, tsort background): Move these
nodes "down", so that they follow the ptx nodes, thus matching the
alphabetized order in the menu.
(tsort background): Make this a subsection.
Suggested by Karl Berry.
Jim Meyering [Sun, 12 Oct 2008 09:05:00 +0000 (11:05 +0200)]
tests: move check-misc targets from src/Makefile.am to maint.mk
* src/Makefile.am (check): Remove check-misc.
(check-misc): Remove target.
* maint.mk (sc_prohibit_stat_st_blocks): New target.
(sc_prohibit_S_IS_definition): New target.
* Makefile.am (EXTRA_DIST): Add .x-sc_prohibit_stat_st_blocks.
* .x-sc_prohibit_stat_st_blocks: New file.
Andreas Schwab reported that "make check" could fail to detect
violation of the no-S_IS*-definition policy.
Jim Meyering [Sun, 12 Oct 2008 08:40:22 +0000 (10:40 +0200)]
tests: df: avoid test failure when using mawk
* tests/df/total-awk: Rewrite to use $PERL, not $AWK.
mawk can't even count above 2^31. Reported by Ed Avis.
http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/14792/focus=14802
* tests/df/total-verify: Rename from total-awk.
* tests/Makefile.am (TESTS): Reflect renaming.
Ondřej Vašík [Mon, 6 Oct 2008 10:42:45 +0000 (12:42 +0200)]
chcon: remove traces of unimplemented --change (-c) option
This option was never supported in upstream coreutils.
* chcon.c (usage): Remove --change (-c) from help message.
(main): Remove 'c' from getopt string.
Remove related and now-unused enums.
Jim Meyering [Sat, 4 Oct 2008 15:12:08 +0000 (17:12 +0200)]
tests: fix the install/strip-program test
* tests/install/strip-program: Use $PREFERABLY_POSIX_SHELL,
not POSIX_SHELL. The latter may be empty, and would fail
on OpenBSD 3.9.
* tests/check.mk (TESTS_ENVIRONMENT): Propagate
PREFERABLY_POSIX_SHELL to tests.
Jim Meyering [Fri, 3 Oct 2008 16:29:39 +0000 (18:29 +0200)]
timeout.c: don't use perror; exit 125 upon failed fork
* src/timeout.c (main): Use "error", not perror.
Elbert Pol noticed a build failure on OS/2.
* src/timeout.c (main): Exit 125 (not errno) upon failed fork.
Make the failed fork diagnostic match the one from install.c.
Kamil Dudka [Fri, 3 Oct 2008 09:03:40 +0000 (11:03 +0200)]
ls and sort: use filevercmp instead of strverscmp
* src/ls.c (cmp_version): Use filevercmp instead of strverscmp.
* src/sort.c (usage): Remove mna reference to strverscmp(3).
(compare_version): Use filevercmp instead of strverscmp.
* bootstrap.conf: Add filevercmp to list of gnulib modules.
* tests/misc/sort-version: Remove conflicting string and enhance test.
* NEWS: Mention the change.
Jim Meyering [Wed, 1 Oct 2008 10:21:58 +0000 (12:21 +0200)]
remove.c: combine two helper functions, to avoid sign-extension
* src/remove.c (fs_handles_readdir_ordered_dirents_efficiently):
Remove function, so as not to have to worry about the type of
statfs.f_type and sign extension.
(dirent_inode_sort_may_be_useful): Adjust comment.
Perform the switch directly on the struct.member here, instead.
Andreas Schwab spotted the potential for a sign-extension bug,
that happens not to bite for the S_* f_type values currently used.
Jim Meyering [Wed, 1 Oct 2008 06:46:46 +0000 (08:46 +0200)]
rm: don't preprocess a directory on a file system of type tmpfs or nfs
The preprocessing phase is not necessary on tmpfs, and induces
a 20% performance decrease when removing a 2M-entry directory.
* src/remove.c (fs_handles_readdir_ordered_dirents_efficiently):
(dirent_inode_sort_may_be_useful): New functions from gnulib/fts.c.
They'll probably become a gnulib module -- eventually.
(preprocess_dir): Use dirent_inode_sort_may_be_useful.
* src/seq.c (print_numbers): Don't switch c_strtold -> strtold
in order to accommodate the locale-dependent behavior of our internal
asprintf use. Instead, simply set the locale to C before calling
asprintf, and then set it back afterwards.
Jim Meyering [Mon, 22 Sep 2008 20:42:12 +0000 (22:42 +0200)]
rm -r: avoid O(n^2) performance for a directory with very many entries
This enhancement works around a problem that is specific to at least
ext3 and ext4 file systems. With them, it would take hours to remove
a two-million-entry directory. RAM-backed file systems (tmpfs) are
not affected, since there is no seek penalty.
* remove.c (rm_malloc, rm_free, compare_ino): New functions.
(dirent_count, preprocess_dir): New function.
[struct readdir_data]: New struct.
(remove_cwd_entries): Call preprocess_dir.
* tests/rm/ext3-perf: New file. Test for the performance fix.
* NEWS: mention the new feature
Jim Meyering [Wed, 24 Sep 2008 08:27:35 +0000 (10:27 +0200)]
remove.c: don't use xmalloc; don't let obstack call exit on failure
(obstack_chunk_alloc, obstack_chunk_free): Don't define.
(top_dir): Param is no longer "const".
Use malloc, not xmalloc, and call longjmp upon failed malloc.
(obstack_init_minimal): New function.
(ds_init): Don't use xmalloc. Instead, use caller-supplied buffer.
Use obstack_specify_allocation_with_arg, not obstack_init, so
that we control what happens upon allocation failure.
Arrange for ds_free not to free uninitialized if/when
any obstack_specify_allocation_with_arg allocation fails.
(ds_free): Don't free DS, now that it's no longer malloc'd.
(rm): Allocate DS on the stack.
Arrange to handle ds_init allocation failure.
Jim Meyering [Thu, 25 Sep 2008 18:01:24 +0000 (20:01 +0200)]
seq -0.1 0.1 2: print final number when locale's decimal point is ","
* src/seq.c (print_numbers): Use strtold, not c_strtold to convert
from just-formatted-using-asprintf string back to double, since
asprintf may have used something other than "." as the decimal point.
Reported by lsof@nodata.co.uk as <http://bugzilla.redhat.com/463556>.
Thanks to Ondřej Vašík for discovering that the bug was locale-related.
* tests/check.mk (TESTS_ENVIRONMENT): Add LOCALE_FR_UTF8, for...
* tests/misc/seq [locale-dec-pt]: New test for the above.
* NEWS (bug fix): Mention it.
* src/Makefile.am (cu_install_program): New variable, set to
either ./ginstall or @INSTALL_PROGRAM@ (for cross compilation).
(INSTALL_PROGRAM): Override AC_SUBST unconditionally, to avoid
warning, and wrong initialization order.