| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
when the file's apparent size is not a multiple of its block size.
* src/copy.c (copy_reg): Don't write a NUL before calling ftruncate.
For some file sizes, writing that single byte would unnecessarily
waste a few file blocks. That write may have been necessary in the
early days of Linux, but now, removing it should be safe.
Based on a patch by Alan Curry: <http://bugs.debian.org/370792>
* tests/cp/sparse: New test for the above.
* tests/cp/Makefile.am (TESTS): Add sparse.
* tests/sparse-file: New file, essence factored out of...
* tests/du/8gb: ... here. Use the new script.
|
|
* src/stat.c (USE_STATVFS): Set to 1 if 'struct statvfs' has a field
f_fstypename.
(STATXFS_FILE_SYSTEM_TYPE_MEMBER_NAME): Define also if 'struct statvfs'
has a field f_fstypename.
This undoes the 2006-08-15 to src/stat.c.
* lib/mountlist.c [MOUNTED_GETMNTINFO2]: Include sys/statvfs.h.
(ME_DUMMY): Treat "kernfs" as a dummy.
(read_file_system_list) [MOUNTED_GETMNTINFO2]: Implement.
* m4/ls-mntd-fs.m4 (gl_LIST_MOUNTED_FILE_SYSTEMS): Also check for
sys/statvfs.h. When getmntinfo was found, check its declaration and
set either MOUNTED_GETMNTINFO or MOUNTED_GETMNTINFO2 depending on it.
* m4/stat-prog.m4 (cu_PREREQ_STAT_PROG): Test also for f_fstypename
in struct statvfs.
|
|
the CVS gettext manual now suggests 1000000.
|
|
|
|
* src/ls.c (get_type_indicator): New function. extracted from...
(print_type_indicator): ...here. Use it.
(length_of_file_name_and_frills): Use it here, too, rather than
assuming stat.st_mode is valid.
Reported by Andreas Schwab, here:
<http://article.gmane.org/gmane.comp.gnu.core-utils.bugs/7774>
FIXME: add a test for this: FYI, I did ls -CF /proc and visually
inspected the result.
|
|
to mention the now-removed cp_options.xstat member.
|
|
* src/c99-to-c89.diff: Add diffs to convert more c99-isms.
This integrates patches from Bruno Haible.
|
|
unsigned long int values.
* src/system.h (select_plural): New function.
* src/md5sum.c (digest_check): Use select_plural to avoid bug.
* src/uptime.c (print_uptime): Likewise.
* src/dd.c (print_stats): Likewise. Also, don't use ngettext to
print a floating point number, as reducing to 0 or 1 doesn't work
for some languages. Instead, just use "s" for seconds since it
doesn't need a plural form.
|
|
no output. This change was in 6.0 but inadvertently unmentioned.
* doc/coreutils.texi (df invocation): df exits nonzero if it outpus
nothing.
* src/df.c (file_systems_processed): Renamed from n_valid_args, and now
a boolean.
(show_dev): Don't set it until we actually output something.
Print the header if this is the first output.
(main): Don't print a header, as that is now show_dev's job.
* tests/misc/Makefile.am (TESTS): Add df.
* tests/misc/df: New file.
|
|
statvfs.f_type not present.
|
|
causes confusion for the Hungarian translators. Problem reported
by Egmont Koblinger.
|
|
|
|
Otherwise, help2man would misformat the output.
Reported by Adam Buchbinder in <https://launchpad.net/bugs/48917>.
|
|
(extract_regexp): Allocate fastmap separately, since otherwise
it might move due to a realloc.
|
|
ansi, color-xterm, gnome, konsole, kterm, rxvt-cygwin,
rxvt-cygwin-native, screen.linux, xterm-256color.
Sort the TERM directives.
From Mike Frysinger.
|
|
|
|
|
|
|
|
|
|
|
|
(read_input): Use it, to avoid relying on GCC-specific behavior
with void * arithmetic.
|
|
and filetype_letter have the same number of elements.
|
|
|
|
|
|
Don't include rand-isaac.c; include randint.h and randread.h instead.
(RANDOM_SOURCE_OPTION): New enum.
(long_opts, usage, main): New option --random-source.
Include md5.h, randread.h, xmemxfrm.h.
(longopts, usage, main): Remove undocumented --seed option;
it's now replaced by --random-source.
(rand_state, get_hash): Remove.
(randread_source): New static var.
(random_state, cmp_hashes, compare_random): New functions; they guarantee
no collisions in the random hash function.
(keycompare): Use compare_random for -R; don't fall back on comparing
via memcoll, since compare_random does the right thing.
|
|
Don't include rand-isaac.c; include randint.h and randread.h instead.
(RANDOM_SOURCE_OPTION): New enum.
(long_opts, usage, main): New option --random-source.
(struct irand_state, irand_init, irand32, irand_mod): Remove.
All callers changed to use randint interface.
(fillrand): Remove. All callers changed to use randread interface.
(dopass): Remove dependency on ISAAC buffer size.
(genpattern): Don't wipe the random state here.
(randint_source): New static var.
(clear_random_data): New function.
(main): Allocate random source, and arrange to wipe it on exit.
|
|
(EXTRA_DIST): Remove rand-isaac.c.
(shuf_LDADD): New macro.
|
|
file descriptor dest_desc.
|
|
option
* src/chown.c (usage): Likewise.
|
|
Correctly access SRC_SB's element ST_AUTHOR.
|
|
than having the code test for all of the other types first.
Hoist the set-uid/gid-testing code "up" into this new block.
Classify any other type of file (e.g., S_TYPEISSHM, etc.) as
C_ORPHAN, not as C_FILE.
* doc/coreutils.texi (What information is listed): Mention that missing
pieces of information are marked with "?". From Paul Eggert.
|
|
2006-07-25 Paul Eggert <eggert@cs.ucla.edu>
* src/ls.c (DT_INIT): Remove. All uses removed.
(enum filetype): Use an ordinary enum rather than trying to keep
the values in sync with DT_FIFO etc. That way, we don't have
to make special assumptions about them. All uses changed.
(whiteout): New constant member of enum filetype.
(filetype_letter): New constant, for use with enum filetype.
(FILETYPE_INDICATORS): New initializer list.
(print_dir): Add case for DT_WHT.
(gobble_file): If stat fails, don't discard information from
readdir; instead, preserve it so it can be printed.
(print_long_format): Fall back on readdir result if stat info
is not available. Use "?" to denote each unknown mode char,
instead of an overall "?", since we now know some of the mode
typically.
(print_type_indicator): Now that MODE isn't necessarily
useful, guard all uses.
Now that two blocks in the type-checking tree can set "type = C_FILE",
move the suffix-handling code out and down.
|
|
* src/ls.c [struct fileinfo] (stat_ok): Rename from stat_failed,
and adjust uses. From a patch by Paul Eggert.
|
|
|
|
print the entry name not the absolute_name -- to be consistent
with the usual case.
* tests/ls/stat-failed: Update accordingly.
|
|
|
|
|
|
* src/ls.c (print_long_format): Be careful to increment P by the
appropriate amount, even when inode_number_width and nlink_width
are zero.
* tests/ls/stat-failed: Test for the above.
|
|
have_acl member. That would happen for a directory with both a
non-stat'able entry and one with an ACL.
|
|
non-command-line file provokes an exit status of 1, not 0.
Say "cannot access" rather than "cannot stat".
* tests/ls/stat-failed: New file/test, for the above.
* tests/ls/Makefile.am (TESTS): Add stat-failed.
|
|
|
|
Replace all occurrences of "type == command_line" with the
equivalent, "command_line_arg".
|
|
coreutils-selinux.patch. From Ulrich Drepper.
This makes it so files not mentioned on the command line (e.g.,
names read from a directory that *is* mentioned on the command
line) for which stat fails are still listed. With --color,
such files are colored just like ORPHANs (aka dangling symlinks).
|
|
|
|
Reported by Tim Waugh.
Also remove the comment duplicating much of --help output.
|
|
alphabetized.
|
|
stat for a symlink, even though it wasn't always needed.
In some cases, that unnecessary stat would cause ls to fail.
* src/ls.c (gobble_file): Don't treat symlinks specially (in
requiring a stat syscall). Remove the offending exclusion.
* NEWS: Mention the fix.
* tests/ls/stat-dtype: New file/test, for the above fix.
Also exercises the new df feature, below.
|
|
file system is processed. This makes it easy to test whether
a specified directory is on a file system of a given type or types.
Otherwise, applications would have had to parse df's output.
E.g., is "." either ext3 or reiserfs: df -t ext3 -t reiserfs .
|
