platform/upstream/coreutils.git
15 years agotests: don't use lang-default from individual tests
Jim Meyering [Sun, 29 Mar 2009 17:54:14 +0000 (19:54 +0200)]
tests: don't use lang-default from individual tests

* tests/sample-test: Don't recommend using lang-default here.
It is now run for each test automatically, via TESTS_ENVIRONMENT.
* tests/dd/reblock: Don't source lang-default here.
* tests/misc/truncate-fail-diag: Likewise.

15 years agotests: other-fs-tmpdir: remove just-added duplicate /usr/tmp
Jim Meyering [Sun, 29 Mar 2009 17:42:34 +0000 (19:42 +0200)]
tests: other-fs-tmpdir: remove just-added duplicate /usr/tmp

* tests/other-fs-tmpdir (CANDIDATE_TMP_DIRS): Remove 2nd /usr/tmp.

15 years agotests: other-fs-tmpdir: avoid bogus "skipped **********..." diagnostic
Jim Meyering [Sun, 29 Mar 2009 07:59:23 +0000 (09:59 +0200)]
tests: other-fs-tmpdir: avoid bogus "skipped **********..." diagnostic

* tests/other-fs-tmpdir (CANDIDATE_TMP_DIRS): Add /dev/shm to the list.
Don't use a line of asterisks as the first line of the "why skipped"
diagnostic, now that that line is printed on 'make check's stderr.

15 years agobuild: use automake's --silent-rules option when possible
Jim Meyering [Sat, 28 Mar 2009 11:48:24 +0000 (12:48 +0100)]
build: use automake's --silent-rules option when possible

* bootstrap: Use automake's --silent-rules option.

15 years agotests: test-argv-iter.c: avoid declaration-after-stmt
Jim Meyering [Fri, 27 Mar 2009 16:33:26 +0000 (17:33 +0100)]
tests: test-argv-iter.c: avoid declaration-after-stmt

* gl/tests/test-argv-iter.c: Move declaration "up", so as not
to ruffle feathers of crufty compilers.  We accommodate such
compilers here solely because this module may migrate to gnulib.
Tiny patch by Matthew Woehlke.

15 years agotests: fix 'make sc_tight_scope' on cygwin
Eric Blake [Thu, 26 Mar 2009 12:39:14 +0000 (06:39 -0600)]
tests: fix 'make sc_tight_scope' on cygwin

* src/Makefile.am (sc_tight_scope): Ignore leading underscores in symbols.

15 years agotests: avoid spurious test failure in set-GID build dir
Jim Meyering [Fri, 27 Mar 2009 08:26:03 +0000 (09:26 +0100)]
tests: avoid spurious test failure in set-GID build dir

* tests/install/install-C-root: Skip this test if in set-gid directory.
Reported by Sven Joachim and C de-Avillez.

15 years agotests: migrate setgid-check into test-lib.sh
Jim Meyering [Fri, 27 Mar 2009 08:23:35 +0000 (09:23 +0100)]
tests: migrate setgid-check into test-lib.sh

* tests/test-lib.sh (skip_if_setgid_): New function.
* tests/setgid-check: Remove file.
* tests/Makefile.am (EXTRA_DIST): Remove setgid-check.
* tests/chmod/c-option: Use the new function rather than sourcing
the separate file.
* tests/cp/cp-parents: Likewise.
* tests/install/install-C: Likewise.
* tests/mkdir/parents: Likewise.
* tests/mkdir/perm: Likewise.

15 years agotests: mark the rm/ext3-perf test as "very" expensive
Jim Meyering [Thu, 26 Mar 2009 09:33:37 +0000 (10:33 +0100)]
tests: mark the rm/ext3-perf test as "very" expensive

* tests/rm/ext3-perf: Relegate this test to the very_expensive
category, since it is failing too often for me, when run in
parallel (-j4) with other disk-hammering tests.  This test fails
when it takes too long, so the fact that it fails under pressure
is an inconvenience more than a real problem.

15 years ago* gnulib: Update submodule to latest.
Jim Meyering [Thu, 26 Mar 2009 09:20:43 +0000 (10:20 +0100)]
* gnulib: Update submodule to latest.

15 years agodf: port the known-value fix to AIX as well
Paul Eggert [Thu, 26 Mar 2009 03:50:17 +0000 (20:50 -0700)]
df: port the known-value fix to AIX as well

* src/df.c (known_value): New function, which also works on AIX
/proc file systems.
(df_readable, show_dev): Use it instead of hardcoding comparison to
UINTMAX_MAX.  Suggested by Jim Meyering and Matthew Woehlke.

15 years agodf: fix a bug when totaling unknown values
Paul Eggert [Wed, 25 Mar 2009 21:16:46 +0000 (14:16 -0700)]
df: fix a bug when totaling unknown values

* src/df.c (show_dev): Don't add UINTMAX_MAX to grand totals, as that
value indicates that the true value is unknown; adding it effectively
subtracts 1 from the total, whereas we want to leave the total alone.

15 years agotests: don't fail "make syntax-check" on cygwin
Jim Meyering [Wed, 25 Mar 2009 21:49:09 +0000 (22:49 +0100)]
tests: don't fail "make syntax-check" on cygwin

* maint.mk (NEWS_hash): Remove trailing " -" or " *-".
* cfg.mk (old_NEWS_hash): Regenerate, removing " -".
Reported by Eric Blake.

15 years agopwd: support -L and -P
Eric Blake [Mon, 23 Mar 2009 20:48:19 +0000 (14:48 -0600)]
pwd: support -L and -P

* src/pwd.c (longopts): New variable.
(logical_getcwd): New function.
(main): Use it.
(usage): Document new options.
* doc/coreutils.texi (pwd invocation): Likewise.
* NEWS: Likewise.
* TODO (pwd): Mark it done.
* tests/misc/pwd-option: New file.
* tests/Makefile.am (TESTS): Add test.
* THANKS: Update.
Reported by Paul D. Smith, in savannah bug 24949.

15 years agodoc: update suffix regex in version sort
Ondřej Vašík [Wed, 25 Mar 2009 11:04:11 +0000 (12:04 +0100)]
doc: update suffix regex in version sort

* doc/coreutils.texi: Reflect current filevercmp regex.

15 years agotests: CuTmpdir.pm: avoid use of undefined variable in warning
Ralf Wildenhues [Wed, 25 Mar 2009 09:15:46 +0000 (10:15 +0100)]
tests: CuTmpdir.pm: avoid use of undefined variable in warning

* tests/CuTmpdir.pm (chmod_tree): Do not warn if $dir is undefined.

15 years agotests: ls -v: exercise yet another change in gnulib's new filevercmp
Jim Meyering [Tue, 24 Mar 2009 21:33:47 +0000 (22:33 +0100)]
tests: ls -v: exercise yet another change in gnulib's new filevercmp

* tests/misc/ls-misc: Add names with ~ and ~.1~ suffixes.

15 years ago* gnulib: Update submodule to latest.
Jim Meyering [Tue, 24 Mar 2009 20:49:03 +0000 (21:49 +0100)]
* gnulib: Update submodule to latest.

15 years agols: make it possible to disable coloring of hard links
Kamil Dudka [Tue, 24 Mar 2009 11:16:20 +0000 (12:16 +0100)]
ls: make it possible to disable coloring of hard links

* src/ls.c (print_color_indicator): Don't color hard links if disabled
(when there is an empty hl= in the LS_COLORS environment variable).
* tests/ls/hardlink: Add test case for disabled hard link highlighting.
* NEWS: Mention the change.

15 years agotests: search only files that are under version control
Jim Meyering [Sat, 21 Mar 2009 08:43:56 +0000 (09:43 +0100)]
tests: search only files that are under version control

* maint.mk (sc_error_exit_success): Search only files
that are under version control.
(sc_makefile_path_separator_check): Likewise.
Check *.mk files, too, not just Makefile.am files.

15 years agotests: follow own advice: use $(PATH_SEPARATOR), not ":"
Jim Meyering [Sat, 21 Mar 2009 08:52:41 +0000 (09:52 +0100)]
tests: follow own advice: use $(PATH_SEPARATOR), not ":"

* maint.mk (coreutils-path-check): Use the variable, not the literal.

15 years agotests: modernize: use $(var) makefile notation, not @var@
Jim Meyering [Sat, 21 Mar 2009 08:56:00 +0000 (09:56 +0100)]
tests: modernize: use $(var) makefile notation, not @var@

* maint.mk (sc_makefile_path_separator_check): Recommend
the use of $(PATH_SEPARATOR), not @PATH_SEPARATOR@.
Remove an obsolete comment.

15 years agotests: teach sc_dd_O_FLAGS about new syntax in dd.c
Jim Meyering [Fri, 20 Mar 2009 14:20:09 +0000 (15:20 +0100)]
tests: teach sc_dd_O_FLAGS about new syntax in dd.c

* cfg.mk (sc_dd_O_FLAGS): Adapt to new syntax.

15 years agodd: use a more portable definition of O_FULLBLOCK
Eric Blake [Thu, 19 Mar 2009 19:14:26 +0000 (20:14 +0100)]
dd: use a more portable definition of O_FULLBLOCK

* src/dd.c (O_FULLBLOCK): Compute its value without using a 180KB
macro.  This avoids triggering a compilation failure with HP-UX's cc.
Reported by Matthew Woehlke.

15 years agotests: avoid spurious mkdir/selinux test failure on HP-UX 11.23
Jim Meyering [Thu, 19 Mar 2009 18:04:19 +0000 (19:04 +0100)]
tests: avoid spurious mkdir/selinux test failure on HP-UX 11.23

* tests/mkdir/selinux: Accept yet another diagnostic.
Reported by Matthew Woehlke.

15 years agotests: df/total-verify: avoid test failure with older Perl
Jim Meyering [Thu, 19 Mar 2009 17:48:46 +0000 (18:48 +0100)]
tests: df/total-verify: avoid test failure with older Perl

* tests/df/total-verify: Don't use perl's -f option.
Reported by Matthew Woehlke.

15 years ago* gnulib: Update submodule to latest.
Jim Meyering [Thu, 19 Mar 2009 06:55:03 +0000 (07:55 +0100)]
* gnulib: Update submodule to latest.

15 years agotests: adjust sort-continue not to fail under valgrind
Jim Meyering [Tue, 17 Mar 2009 20:23:51 +0000 (21:23 +0100)]
tests: adjust sort-continue not to fail under valgrind

* tests/misc/sort-continue: Don't run cat inside fd-limited shell.
If sort fails to run in an fd-limited shell, skip the test.

15 years agotests: add another sort/nmerge test
Paul Eggert [Mon, 9 Mar 2009 21:56:13 +0000 (14:56 -0700)]
tests: add another sort/nmerge test

* tests/Makefile.am (TESTS): Add sort-merge-fdlimit.
* tests/misc/sort-merge-fdlimit: New file.
* doc/coreutils.texi (sort invocation): Document that we now silently
lower nmerge if necessary.
Patch by Paul Eggert, Nima Nikzad, Max Chang, Alexander Nguyen,
Sahil Amoli, and Nick Graham.

15 years agosort: handle fd exhaustion better when merging
Paul Eggert [Fri, 13 Mar 2009 22:48:30 +0000 (15:48 -0700)]
sort: handle fd exhaustion better when merging

This is an alternative to my 9 March patch labeled "Silently lower
nmerge; don't (sometimes incorrectly) range-check"
<http://lists.gnu.org/archive/html/bug-coreutils/2009-03/msg00070.html>.
It differs by not using 'dup' to probe for extra file descriptors;
instead, it simply calls 'open' (and 'pipe') to open files and pipes,
until one of these calls fails due to file descriptor exhaustion; it
then backs off by 1, does a merge with the files that it has opened,
and then retries with the (now-smaller) number of files.

This patch requires quite a few more changes to the source code than
the earlier patch, but it is in some sense "better" because it doesn't
need to call "dup" ahead of time in order to decide whether "open" or
"pipe" will fail.  Also, it's more robust in the case where "open" or
"pipe" fails with errno==EMFILE because some system-wide limit is
exhausted.

* src/sort.c (create_temp_file): New arg SURVIVE_FD_EXHAUSTION.
(stream_open): New function, containing guts of xfopen.
(xfopen): Use it.
(pipe_fork): Set errno on failure.
(maybe_create_temp): New function, containing guts of create_temp.
(create_temp): Use it.
(open_temp): Distinguish failures due to file descriptor exhaustion
from other failures, and on fd exhaustion return a notice to caller
rather than dying.  Don't test execlp's return value; when it returns,
it *always* returns -1.
(open_input_files): New function.
(mergefps): New arg FPS. It's now the caller's responsibility to open
the input and output files.  All callers changed.
(mergefiles): New function.
(avoid_trashing_input, merge): Handle the case where a single merge
can't merge as much as we wanted due to file descriptor exhaustion, by
merging as much as we can and then retrying.
* tests/Makefile.am (TESTS): Add misc/sort-continue.
* tests/misc/sort-continue: New file.
* THANKS: Add Glen Lenker and Matt Pham who coauthored this patch.

15 years agomaint: normalize leading-TAB indentation in Makefiles
Jim Meyering [Wed, 18 Mar 2009 11:20:32 +0000 (12:20 +0100)]
maint: normalize leading-TAB indentation in Makefiles

* maint.mk (sc_makefile_TAB_only_indentation): New rule.
Replace each TAB+8-space sequence with two TABs.
* man/Makefile.am: Likewise.
* build-aux/check.mk: Likewise.
I used this command (run it more than once, if needed):
t=$'\t'; git grep -l -E "$t {8}"|grep -E 'Makefile|\.mk$' \
| xargs perl -pi -e 's/\t {8}/\t\t/'

15 years agomaint: remove long-unused Makefile
Jim Meyering [Sun, 15 Mar 2009 20:40:54 +0000 (21:40 +0100)]
maint: remove long-unused Makefile

* tests/sort-time/Makefile: Remove long-unused file.

15 years agomaint: remove unused include and definitions
Jim Meyering [Sat, 14 Mar 2009 11:48:44 +0000 (12:48 +0100)]
maint: remove unused include and definitions

* gl/lib/randread.c (__attribute__): Remove unused definition.
(ATTRIBUTE_UNUSED): Likewise.
* gl/lib/selinux-at.c (_): Remove unused definition.
Don't include <gettext.h>.  No longer needed.

15 years ago* gnulib: Update submodule to latest.
Jim Meyering [Sun, 15 Mar 2009 07:09:53 +0000 (08:09 +0100)]
* gnulib: Update submodule to latest.

15 years agosystem.h: correct compilation error: MAX not defined
Jim Meyering [Sat, 14 Mar 2009 20:29:33 +0000 (21:29 +0100)]
system.h: correct compilation error: MAX not defined

* src/system.h (io_blksize): Move definition down, so it follows that
of MAX -- to avoid compilation failure on OpenSolaris.  Reported by
David Bartley.  Details in
http://lists.gnu.org/archive/html/bug-coreutils/2009-03/msg00190.html

15 years ago* gnulib: Update to pull in select/interix work-around.
Jim Meyering [Sat, 14 Mar 2009 10:49:56 +0000 (11:49 +0100)]
* gnulib: Update to pull in select/interix work-around.

15 years agodoc: update "version sort" documentation
Kamil Dudka [Wed, 25 Feb 2009 09:10:36 +0000 (10:10 +0100)]
doc: update "version sort" documentation

* doc/coreutils.texi: Reflect current filevercmp behavior.

15 years agotests: groups-version: don't require that id be built
Jim Meyering [Thu, 12 Mar 2009 13:41:38 +0000 (14:41 +0100)]
tests: groups-version: don't require that id be built

* tests/misc/groups-version: groups is now independent of id.

15 years agotests: skip the groups-dash test when "groups" is not built
Jim Meyering [Thu, 12 Mar 2009 13:40:45 +0000 (14:40 +0100)]
tests: skip the groups-dash test when "groups" is not built

* tests/misc/groups-dash: Skip if groups is not being installed.
Suggested by Mike Frysinger.
Also fail if groups exits with nonzero status.

15 years agomaint: mention Autoconf version required to build Automake
Pádraig Brady [Thu, 12 Mar 2009 14:16:54 +0000 (14:16 +0000)]
maint: mention Autoconf version required to build Automake

* README-prereq: coreutils currently requires an unreleased
version of Automake, so mention the minimum version of Autoconf
required to build that. Suggestion from James Youngman.

15 years ago* gnulib: Update submodule to latest.
Jim Meyering [Thu, 12 Mar 2009 09:46:31 +0000 (10:46 +0100)]
* gnulib: Update submodule to latest.

15 years agocp: make -a option preserve xattrs, but with reduced diagnostics
Ondřej Vašík [Wed, 11 Mar 2009 15:08:20 +0000 (16:08 +0100)]
cp: make -a option preserve xattrs, but with reduced diagnostics

* copy.c (copy_attr_by_fd): Reduce xattr diagnostics for 'cp -a'.
(copy_attr_by_name): Likewise.
* cp.c (main): Preserve xattrs with -a option, when possible.
* doc/coreutils.texi: Document that xattrs are preserved with
cp -a, with no added diagnostics.
* NEWS: Mention the change.
* tests/misc/xattr: Add tests for 'cp --preserve=all' and 'cp -a'.

15 years agomaint: Document how to best to enable gcc warnings for developers
Pádraig Brady [Wed, 11 Mar 2009 12:09:08 +0000 (12:09 +0000)]
maint: Document how to best to enable gcc warnings for developers

* README-hacking: Mention --enable-gcc-warnings configure option
* configure.ac: Don't scare developers away from using gcc warnings

15 years agomaint: remove an obsolete FIXME comment
Jim Meyering [Tue, 10 Mar 2009 09:04:23 +0000 (10:04 +0100)]
maint: remove an obsolete FIXME comment

* src/Makefile.am (../AUTHORS): Remove obsolete FIXME comment.
There are no longer any installed *.sh programs.

15 years agocat: Fix immediate output of processed data
Pádraig Brady [Wed, 11 Mar 2009 11:39:33 +0000 (11:39 +0000)]
cat: Fix immediate output of processed data

Introduced by commit 790892db, 2006-06-08 "Ensure that cat works ...".
* NEWS: Mention the bugfix.
* src/cat.c (cat): Fix the typo which stopped the writing
of processed data before a blocking read() is done.
* tests/misc/cat-buf: Add to ensure processed data is not buffered.
* tests/Makefile.am: Reference the new test.

15 years agocat,cp,mv,install,split: Set the minimum IO block size used to 32KiB
Pádraig Brady [Fri, 6 Mar 2009 22:30:55 +0000 (22:30 +0000)]
cat,cp,mv,install,split: Set the minimum IO block size used to 32KiB

This is following on from this change:
[02c3dc9d 2008-03-06 cat: use larger buffer sizes ...]
which increased the IO block size used by cat by 8 times,
but also capped it at 32KiB.
* NEWS: Mention the change in behavior.
* src/system.h: Add a new io_blksize() function that
returns the max of ST_BLKSIZE or 32KiB, as this was
seen as a good value for a minimum block size to use
to get good performance while minimizing system call overhead.
* src/cat.c: Use it.
* src/copy.c: ditto
* src/split.c: ditto

15 years agotests: add a test for newly-fixed bug in comm --check-order
Jim Meyering [Sun, 8 Mar 2009 17:37:08 +0000 (18:37 +0100)]
tests: add a test for newly-fixed bug in comm --check-order

* tests/misc/comm (ooo-prefix): Add a test for today's fix.
* NEWS (Bug fixes): Mention it.

15 years agocomm: fix a bug in its new --check-order option
Bruno Haible [Sun, 8 Mar 2009 17:25:29 +0000 (18:25 +0100)]
comm: fix a bug in its new --check-order option

* src/comm.c: Include memcmp2.h.
(check_order): Use memcmp2 instead of memcmp.
* bootstrap.conf (gnulib_modules): Add memcmp2.

15 years agotests: ls -v: exercise the bug fixed by gnulib's new filevercmp
Jim Meyering [Sat, 7 Mar 2009 19:41:00 +0000 (20:41 +0100)]
tests: ls -v: exercise the bug fixed by gnulib's new filevercmp

* tests/misc/ls-misc (version-sort): New test.
(mk_file): New function.
Reported by Josh Triplett in <http://bugs.debian.org/517558>.
Details in http://thread.gmane.org/gmane.comp.lib.gnulib.bugs/16902
* NEWS (Bug fixes): Mention it.

15 years agognulib: sync submodule to latest
Jim Meyering [Sun, 8 Mar 2009 20:16:30 +0000 (21:16 +0100)]
gnulib: sync submodule to latest

15 years agopr: fix a bug whereby --indent=N (-o) did not indent header lines
Jim Meyering [Fri, 6 Mar 2009 19:03:26 +0000 (20:03 +0100)]
pr: fix a bug whereby --indent=N (-o) did not indent header lines

* src/pr.c (print_header): Honor chars_per_margin also for the header.
Inspired by a patch from C de-Avillez.
This change also adjusts two other %*s directives to be %*.*s instead,
(for the padding on either side of "file_text") to correct a bug that
could make pr print one space of padding instead of none.
* NEWS (Bug fixes): Mention it.
* tests/pr/o3Jml24f-lm-lo: Adjust for new expected output.
* tests/pr/o3a3Sl24f-tn: Likewise.
* tests/pr/o3a3Snl24f-tn: Likewise.
* tests/pr/o3a3l24f-tn: Likewise.
* tests/pr/o3b3Sl24f-tn: Likewise.
* tests/pr/o3b3Snl24f-tn: Likewise.
* tests/pr/o3b3l24f-tn: Likewise.
* tests/pr/o3mSl24f-bl-tn: Likewise.
* tests/pr/o3mSnl24fbltn: Likewise.
* tests/pr/o3ml24f-bl-tn: Likewise.

15 years agobootstrap: add to XGETTEXT_OPTIONS
Jim Meyering [Sat, 7 Mar 2009 15:20:55 +0000 (16:20 +0100)]
bootstrap: add to XGETTEXT_OPTIONS

* bootstrap.conf (XGETTEXT_OPTIONS): Add c-format --flag options
for verror and verror_at_line; alphabetize.

15 years agomaint: use a git submodule for gnulib
Jim Meyering [Sat, 7 Mar 2009 09:19:46 +0000 (10:19 +0100)]
maint: use a git submodule for gnulib

* .gitmodules: New file, to track gnulib.
* gnulib: New file, created by running this:
git submodule add git://git.sv.gnu.org/gnulib.git gnulib

15 years agobuild: allow ./bootstrap --srcdir=... to work with a git submodule
Jim Meyering [Sat, 7 Mar 2009 15:11:54 +0000 (16:11 +0100)]
build: allow ./bootstrap --srcdir=... to work with a git submodule

* bootstrap (git_modules_config): Use GIT_CONFIG, not GIT_CONFIG_LOCAL.
The latter was removed in v1.5.6.1-109-gdc87183.

15 years agobuild: avoid "multiply defined..." warning from automake
Jim Meyering [Sat, 7 Mar 2009 12:27:29 +0000 (13:27 +0100)]
build: avoid "multiply defined..." warning from automake

* Makefile.am (AM_CFLAGS): Add $(WARN_CFLAGS) via +=, not =, to
avoid "multiply defined..." warning from automake.

15 years agocat: use larger buffer sizes to reduce read/write-syscall overhead
Jim Meyering [Fri, 6 Mar 2009 09:27:43 +0000 (10:27 +0100)]
cat: use larger buffer sizes to reduce read/write-syscall overhead

* src/cat.c (max): Remove definition.  Use MAX from system.h instead.
(compute_buffer_size): New function to compute the input and output
buffer sizes, which are now set at 8 times st_blksize with a maximum
of 32KiB. Previously the typical block sizes used were 1KiB for pipes
and 4KiB for files, and now will be 8KiB and 32KiB respectively.
(main): Use it.
This change can double throughput on modern systems.  For timings,
see http://article.gmane.org/gmane.comp.gnu.core-utils.bugs/16040
Suggestion from Tzvi Rotshtein.

15 years agomaint: avoid new unused-macro warnings from gcc
Jim Meyering [Fri, 6 Mar 2009 09:49:18 +0000 (10:49 +0100)]
maint: avoid new unused-macro warnings from gcc

* src/ls.c (ASSERT_MATCHING_DEV_INO): Comment-out unused definition,
to match commented-out use.
* src/install.c (INITIAL_HASH_MODULE, INITIAL_ENTRY_TAB_SIZE):
(READ_SIZE): Remove definitions of unused macros.
* src/mv.c (INITIAL_HASH_MODULE, INITIAL_ENTRY_TAB_SIZE): Likewise.
* src/dd.c (max): Likewise.
* src/expand.c (OUTPUT_BLOCK): Likewise.
* src/csplit.c (ALLOC_SIZE): Likewise.
* src/pr.c (NULLCOL): Likewise.
* src/ptx.c (Sword): Likewise.
* src/unexpand.c (OUTPUT_BLOCK): Likewise.
* src/factor.c (NDEBUG): Likewise.

15 years agotests: fix just-introduced typo in rm/ext3-perf
Jim Meyering [Fri, 6 Mar 2009 10:12:32 +0000 (11:12 +0100)]
tests: fix just-introduced typo in rm/ext3-perf

* tests/rm/ext3-perf: Fix typo in commit 2b1e6636bea,
2009-02-27 "tests: rm/ext3-perf: relax this test".

15 years agomaint: remove spurious case-ignoring option from a test
Jim Meyering [Fri, 6 Mar 2009 06:18:09 +0000 (07:18 +0100)]
maint: remove spurious case-ignoring option from a test

* maint.mk (sc_trailing_blank): Remove useless/harmless ignore_case=1.

15 years agopolicy: use git to help avoid trailing white space
Ralf Wildenhues [Mon, 2 Mar 2009 21:33:05 +0000 (22:33 +0100)]
policy: use git to help avoid trailing white space

* HACKING (Avoid trailing white space): Describe how git can help.

15 years agoUpdate Andreas Schwab's email address
Andreas Schwab [Tue, 3 Mar 2009 17:12:25 +0000 (18:12 +0100)]
Update Andreas Schwab's email address

* THANKS: Update Andreas Schwab's email address.

15 years agomaint: improve release-related process and documentation
Jim Meyering [Tue, 3 Mar 2009 07:24:29 +0000 (08:24 +0100)]
maint: improve release-related process and documentation

* README-release: Fix typo. it's "maintainer-clean".
Refer to HACKING for full list of dependencies.
Use cvs add's -ko option to inhibit $Id$-style keyword expansion.
* maint.mk (web-manual): Factor out definition of $(manual_title).
(vc-dist): Set XZ_OPT=-9ev for the final "make dist".
* cfg.mk (manual_title): Define it here.
Suggestions from Eric Blake.

15 years agodist: document our release procedure
Jim Meyering [Sun, 1 Mar 2009 20:29:21 +0000 (21:29 +0100)]
dist: document our release procedure

* README-release: New file.

15 years agodist: automate the post-release web-manual update process
Jim Meyering [Fri, 27 Feb 2009 12:56:49 +0000 (13:56 +0100)]
dist: automate the post-release web-manual update process

* maint.mk (web-manual): New rule, from m4's maint.mk.
* bootstrap.conf (gnulib_modules): Add gendocs.

15 years agotests: rm/ext3-perf: relax this test
Jim Meyering [Fri, 27 Feb 2009 11:54:01 +0000 (12:54 +0100)]
tests: rm/ext3-perf: relax this test

* tests/rm/ext3-perf: If set-up takes too long, use the set-up time
as the removal time limit.  Otherwise, a parallel "make check" would
usually fail when run with RUN_VERY_EXPENSIVE_TESTS=yes.

15 years agotests: don't skip setgid-dir related tests on SELinux-enabled systems
Jim Meyering [Fri, 27 Feb 2009 08:09:18 +0000 (09:09 +0100)]
tests: don't skip setgid-dir related tests on SELinux-enabled systems

* tests/setgid-check: The change in ls (use ".", not "+") caused
this check always to report the working directory as set-gid.
Now, use stat, not ls.

15 years agotests: reflect the s/+/./ SELinux indicator change
Jim Meyering [Fri, 27 Feb 2009 08:16:45 +0000 (09:16 +0100)]
tests: reflect the s/+/./ SELinux indicator change

* tests/mkdir/perm: Now that this test is no longer always skipped,
make it work independently of ls: use stat, not ls, to get perm string.
* test-lib.sh (rwx_to_mode_): Use stat, not ls to get perm string
This test was run only when RUN_VERY_EXPENSIVE_TESTS=yes was set.

15 years agotests: when skipping a test, say why on screen, not just in the log
Jim Meyering [Fri, 27 Feb 2009 08:12:14 +0000 (09:12 +0100)]
tests: when skipping a test, say why on screen, not just in the log

* tests/test-lib.sh: Redirect also to descriptor 9.
Better one-line why-skip diagnostics.
* tests/check.mk (TESTS_ENVIRONMENT): Redirect 9 to stderr.
* tests/misc/pwd-unreadable-parent: Remove redundant "skipping..." diag.

15 years agomaint: enable many gcc warnings
Jim Meyering [Sat, 14 Feb 2009 10:29:20 +0000 (11:29 +0100)]
maint: enable many gcc warnings

* bootstrap.conf (gnulib_modules): Add manywarnings.
* configure.ac: Use gl_MANYWARN_ALL_GCC, and exclude options
I don't want or that provoke too many warnings.
(WARN_CFLAGS, WERROR_CFLAGS): Define.
(lint, GNULIB_PORTCHECK): Define.
(_FORTIFY_SOURCE): Define to 2.

15 years agomaint: avoid gcc warning about missing attributes
Jim Meyering [Sat, 14 Feb 2009 10:22:22 +0000 (11:22 +0100)]
maint: avoid gcc warning about missing attributes

* src/prog-fprintf.h (prog_fprintf): Declare with printf attribute.
* src/system.h (usage): Declare with the "noreturn" attribute.

15 years agomaint: avoid warnings about potentially-counterproductive "inline"
Jim Meyering [Sat, 14 Feb 2009 10:26:14 +0000 (11:26 +0100)]
maint: avoid warnings about potentially-counterproductive "inline"

* src/dd.c (quit): Remove "inline" attribute.
* src/test.c (advance, unary_advance): Likewise.

15 years agomaint: cp: avoid gcc warning about unused macro definitions
Jim Meyering [Sat, 14 Feb 2009 09:45:53 +0000 (10:45 +0100)]
maint: cp: avoid gcc warning about unused macro definitions

* src/cp.c (INITIAL_HASH_MODULE, INITIAL_ENTRY_TAB_SIZE): Remove.

15 years agodist: do not emit MD5 or SHA1 checksums into announcement template
Jim Meyering [Mon, 2 Mar 2009 07:40:41 +0000 (08:40 +0100)]
dist: do not emit MD5 or SHA1 checksums into announcement template

* maint.mk (announcement): Use announce-gen's new --no-print-checksums.

15 years agomaint: detect unnecessary inclusion of signal.h
Jim Meyering [Tue, 24 Feb 2009 20:43:26 +0000 (21:43 +0100)]
maint: detect unnecessary inclusion of signal.h

* maint.mk (sc_prohibit_signal_without_use): New rule.

15 years agomaint: remove an unnecessary inclusion of <signal.h>
Jim Meyering [Tue, 24 Feb 2009 20:44:51 +0000 (21:44 +0100)]
maint: remove an unnecessary inclusion of <signal.h>

* src/operand2sig.c: Don't include <signal.h>.

15 years agotests: sort: Check skipping blanks in multibyte locales
Pádraig Brady [Fri, 27 Feb 2009 08:40:42 +0000 (08:40 +0000)]
tests: sort: Check skipping blanks in multibyte locales

* tests/misc/sort: On Fedora 8 at least, sort -k1b,1
mishandles blanks in multibyte locales, so add test.

15 years ago* HACKING: avoid trailing white space
Jim Meyering [Wed, 25 Feb 2009 14:37:58 +0000 (15:37 +0100)]
* HACKING: avoid trailing white space

15 years agocp: diagnose invalid "cp -rl dir dir" right away, once again
Jim Meyering [Fri, 27 Feb 2009 08:23:44 +0000 (09:23 +0100)]
cp: diagnose invalid "cp -rl dir dir" right away, once again

Running "mkdir dir; cp -rl dir dir" would create dir/dir/dir/...
rather than diagnosing the "copy-into-self" failure.

The easy fix would have been to revert this part of the change
[3ece0355 2008-11-09 cp: use far less memory in some cases]
that introduced the bug:

-         remember_copied (dst_name, dst_sb.st_ino, dst_sb.st_dev);
+         if (!x->hard_link)
+           remember_copied (dst_name, dst_sb.st_ino, dst_sb.st_dev);

However, that would have induced the failure of the new cp/link-heap
test, due to the added memory pressure of recording 10k dev/ino pairs.
And besides, I liked that improvement and wanted to keep it.

Now that it's obvious recording the just-created-directory dev/ino
needn't depend on the setting of hard_link, I realized it is necessary
to record the pair only for the first directory created for each
source command-line argument.

I made that change, then noticed the new test, cp -rl a d d, would pass
when run once, yet output the into-self diagnostic twice.  Also note
the side effect: it creates d/a and d/d.  However, running that same
command a second time, now with the modified directory, would fail.

That turned out to be due to the fact that although the first into-self
failure was detected in copy_dir, that function would continue copying
other entries regardless -- and that would make it fail (eventually)
with the unwanted recursion.

* src/copy.c (copy_internal): This function needed an indicator of
whether, for a give command line argument, it had already created its
first directory.  If so, no more need to record dev/ino pairs.  If this
is the first, then do record its pair.  Hence, the new parameter.
(copy_dir, copy): Update callers.
(copy_dir): Upon any into-self failure, break out of the loop.
* tests/cp/into-self: Test for the above.
Reported by Mikael Magnusson.

15 years agotests: stty-row-col would hang when run in the background
Jim Meyering [Thu, 26 Feb 2009 19:25:42 +0000 (20:25 +0100)]
tests: stty-row-col would hang when run in the background

* tests/misc/stty-row-col: Don't hang when run in background
at least on Solaris 10 and OpenBSD.

15 years agosort: Fix two bugs with determining the end of field
Pádraig Brady [Tue, 24 Feb 2009 08:37:18 +0000 (08:37 +0000)]
sort: Fix two bugs with determining the end of field

* src/sort.c: When no specific number of chars to skip
is specified for the end field, always skip the whole field.
Also never include leading spaces from next field.
* tests/misc/sort: Add 2 new tests for these cases.
* NEWS: Mention this bug fix.
* THANKS: Add bug reporter.
Reported by Davide Canova.

15 years agoargv-iter: avoid variadic macro
Eric Blake [Thu, 26 Feb 2009 13:54:48 +0000 (06:54 -0700)]
argv-iter: avoid variadic macro

* gl/lib/argv-iter.h (_ATTRIBUTE_NONNULL_): Don't assume C99 variadic
macros are portable yet; needed for at least IRIX/MIPSpro.
* THANKS: Update.
Reported by Stuart Shelton.

15 years agotests: Remove system specific error strings
Pádraig Brady [Thu, 26 Feb 2009 00:01:06 +0000 (00:01 +0000)]
tests: Remove system specific error strings

* tests/dd/skip-seek-past-file: Just check for
the non system specific part of the error strings.
This was causing an erroneous failure on NetBSD 1.6 at least.

15 years agotests: using printf with \e for ESC isn't portable
Jim Meyering [Wed, 25 Feb 2009 18:51:04 +0000 (19:51 +0100)]
tests: using printf with \e for ESC isn't portable

* tests/ls/color-clear-to-eol: Use \33, not \e for ascii ESC.
This test failed when using the built-in printf of /bin/sh from
NetBSD 1.6.

15 years agoEnsure comment for translators is included in po file
Pádraig Brady [Tue, 24 Feb 2009 09:47:00 +0000 (09:47 +0000)]
Ensure comment for translators is included in po file

Issue reported by Göran Uddeborg.

* src/system.h: Move the translator comment adjacent
to the translated string.
* THANKS: Update Göran's email address.

15 years agotruncate: fix typo in man page
Tobias Stoeckmann [Sun, 22 Feb 2009 15:30:48 +0000 (16:30 +0100)]
truncate: fix typo in man page

* man/truncate.x: Fix spelling.
* THANKS: Update.

15 years agodoc: describe dd's status=noxref option
Andrew Church [Sun, 22 Feb 2009 09:25:01 +0000 (10:25 +0100)]
doc: describe dd's status=noxref option

* doc/coreutils.texi (dd invocation): Describe status=noxref.

15 years agotests: reenable temporarily-disabled tests using e.g., trap '' TTOU
Jim Meyering [Sun, 22 Feb 2009 00:42:43 +0000 (01:42 +0100)]
tests: reenable temporarily-disabled tests using e.g., trap '' TTOU

Revert "tests: skip newly-modified tests unconditionally..."
This reverts commit 773a88d33f68e50a2919c8a8310c646a7e5fabc2.
* tests/misc/stty: Don't skip.
* tests/misc/stty-invalid: Likewise.
* tests/mv/i-3: Likewise.

15 years agoupdate James Youngman's email address
James Youngman [Sun, 22 Feb 2009 00:41:13 +0000 (00:41 +0000)]
update James Youngman's email address

* THANKS: Update James Youngman's email address

15 years agopost-release administrivia
Jim Meyering [Sat, 21 Feb 2009 21:49:24 +0000 (22:49 +0100)]
post-release administrivia

* NEWS: Add header line for next release.
* .prev-version: Record previous version.
* cfg.mk (old_NEWS_hash): Auto-update.

15 years agoversion 7.1 v7.1
Jim Meyering [Sat, 21 Feb 2009 21:35:33 +0000 (22:35 +0100)]
version 7.1

* NEWS: Record release date.

15 years ago* README: Mention that c99-to-c89.diff is no longer maintained.
Jim Meyering [Sat, 21 Feb 2009 20:31:34 +0000 (21:31 +0100)]
* README: Mention that c99-to-c89.diff is no longer maintained.

15 years agotests: skip new install-C test if working directory is set-gid
Jim Meyering [Sat, 21 Feb 2009 18:48:30 +0000 (19:48 +0100)]
tests: skip new install-C test if working directory is set-gid

* tests/install/install-C: Use setgid-check.

15 years agotests: skip newly-modified tests unconditionally, just to be safe
Jim Meyering [Sat, 21 Feb 2009 14:56:26 +0000 (15:56 +0100)]
tests: skip newly-modified tests unconditionally, just to be safe

* tests/misc/stty-invalid: Skip.
* tests/misc/stty: Likewise.
* tests/mv/i-3: Likewise.

15 years agotests: don't let SIGTTOU/TTIN hang a backgrounded "make check"
Jim Meyering [Sat, 21 Feb 2009 10:30:41 +0000 (11:30 +0100)]
tests: don't let SIGTTOU/TTIN hang a backgrounded "make check"

Running "make check &" would hang due to SIGTTOU and SIGTTIN signals.
* tests/misc/stty: Ignore SIGTTOU.
* tests/misc/stty-invalid: Likewise.
* tests/mv/i-3: Ignore SIGTTIN.

15 years agodoc: ls: clarify description of -s
Jim Meyering [Fri, 20 Feb 2009 18:31:31 +0000 (19:31 +0100)]
doc: ls: clarify description of -s

* src/ls.c (usage): Make the description of -s mention "allocated size",
so that it's less likely to be confused with an apparent byte-count.
Suggested by Vito Caputo.

15 years agodoc: dd: document that the default block size is 512 bytes
Jim Meyering [Thu, 19 Feb 2009 11:35:58 +0000 (12:35 +0100)]
doc: dd: document that the default block size is 512 bytes

* src/dd.c (usage): Document the default block size.
* doc/coreutils.texi (dd invocation): Document that the default
block size (bs, ibs, obs) is 512 bytes.
Reported by Petr Uzel.

15 years agomaint: changing NEWS' copyright date must not invalidate checksum
Jim Meyering [Wed, 18 Feb 2009 15:09:45 +0000 (16:09 +0100)]
maint: changing NEWS' copyright date must not invalidate checksum

* maint.mk (NEWS_hash): Don't let the Copyright date line contribute
to the checksum.
* cfg.mk (old_NEWS_hash): Update to reflect the new policy.

15 years agodoc: explain how to recover from changes committed to master
Jim Meyering [Wed, 18 Feb 2009 13:50:04 +0000 (14:50 +0100)]
doc: explain how to recover from changes committed to master

* HACKING: A beginner's commit-to-master is so common,
and causes enough confusion, that we describe how to recover.

15 years agomaint: add copyright notices
Jim Meyering [Wed, 18 Feb 2009 13:34:23 +0000 (14:34 +0100)]
maint: add copyright notices

* HACKING: Add copyright notice.
* tests/test-lib.sh: Likewise.

15 years agocp: -a now preserves SELinux context, with reduced diagnostics
Ondřej Vašík [Tue, 17 Feb 2009 14:53:39 +0000 (15:53 +0100)]
cp: -a now preserves SELinux context, with reduced diagnostics

* copy.c (copy_reg): Reduce SELinux context diagnostics for 'cp -a'.
(copy_internal): Likewise
* copy.h (cp_options): Add boolean reduce_diagnostics.
* cp.c (usage): Say that --archive (-a) behaves like -dR --preserve=all.
(cp_option_init): Initialize added reduce_diagnostics.
(main): Add reduce_diagnostics for the -a option, and preserve SELinux
context, if possible.
* mv.c (cp_options_init): Initialize new cp_options booleans.
* install.c (cp_option_init): Likewise.
* NEWS: Mention those behaviour changes.
* doc/coreutils.texi: Document --preserve=context, document that
diagnostics are not shown for failures of non-mandatory attributes
(just SELinux at the moment).
* tests/cp/cp-a-selinux: Check not only failures, but succesful use
of preserving SELinux context in cp.