Age | Commit message (Collapse) | Author |
|
* src/system.h (emit_ancillary_info): Take the invariant PROGRAM_NAME
as a parameter, so that consistent references are made to online docs
and texinfo nodes, when a --program-prefix is in place. Note the
man pages don't need this fix as they're generated before the program
prefix is used.
* NEWS: Mention the improvements in references to online documentation.
|
|
* build-aux/gen-single-binary.sh: Don't use ATTRIBUTE_NORETURN
for main functions.
* src/base64.c, src/basename.c, src/cat.c, src/chcon.c, src/chgrp.c:
* src/chmod.c, src/chown.c, src/chroot.c, src/cksum.c, src/comm.c:
* src/cp.c, src/csplit.c, src/cut.c, src/date.c, src/dd.c, src/df.c:
* src/dircolors.c, src/dirname.c, src/du.c, src/echo.c, src/env.c:
* src/expand.c, src/expr.c, src/factor.c, src/fmt.c, src/fold.c:
* src/getlimits.c, src/groups.c, src/head.c, src/hostid.c:
* src/hostname.c, src/id.c, src/install.c, src/join.c, src/kill.c:
* src/link.c, src/ln.c, src/logname.c, src/ls.c, src/make-prime-list.c:
* src/md5sum.c, src/mkdir.c, src/mkfifo.c, src/mknod.c, src/mktemp.c:
* src/mv.c, src/nice.c, src/nl.c, src/nohup.c, src/nproc.c:
* src/numfmt.c, src/od.c, src/paste.c, src/pathchk.c, src/pinky.c:
* src/pr.c, src/printenv.c, src/printf.c, src/ptx.c, src/pwd.c:
* src/readlink.c, src/realpath.c, src/rm.c, src/rmdir.c, src/runcon.c:
* src/seq.c, src/shred.c, src/shuf.c, src/sleep.c, src/sort.c:
* src/split.c, src/stat.c, src/stdbuf.c, src/stty.c, src/sum.c:
* src/sync.c, src/tac.c, src/tail.c, src/tee.c, src/timeout.c:
* src/touch.c, src/tr.c, src/true.c, src/truncate.c, src/tsort.c:
* src/tty.c, src/uname.c, src/unexpand.c, src/uniq.c, src/unlink.c:
* src/uptime.c, src/users.c, src/wc.c, src/who.c, src/whoami.c:
In 'main' functions, Prefer 'return status;' to 'exit (status);'.
* src/coreutils-arch.c (_single_binary_main_uname)
(_single_binary_main_arch):
* src/coreutils-dir.c, src/coreutils-vdir.c (_single_binary_main_ls)
(_single_binary_main_dir, _single_binary_main_vdir):
Omit ATTRIBUTE_NORETURN. Return a value.
* src/coreutils.c (SINGLE_BINARY_PROGRAM): Omit ATTRIBUTE_NORETURN.
(launch_program): Now static.
* src/dd.c (finish_up): New function.
(quit, main): Use it.
* src/getlimits.c (main): Return a proper exit status.
* src/test.c (test_main_return): New macro.
(main): Use it.
* src/logname.c, src/nohup.c, src/whoami.c:
Use 'error' to simplify exit status in 'main' function.
* src/yes.c (main): Use 'return' rather than 'error' to exit,
so that GCC doesn't suggest ATTRIBUTE_NORETURN.
|
|
This didn't seem to cause any invalid operation on GNU/Linux at least,
but depending on the implementation, mutex deadlocks could occur.
For example this might be the cause of lockups seen on Solaris:
http://lists.gnu.org/archive/html/coreutils/2013-03/msg00048.html
This was identified with valgrind 3.9.0 with this setup:
seq 200000 > file.sort
valgrind --tool=drd src/sort file.sort -o file.sort
With that, valgrind would _intermittently_ report the following:
Destroying locked mutex: mutex 0x5419548, recursion count 1, owner 2.
at 0x4C2E3F0: pthread_mutex_destroy(in vgpreload_drd-amd64-linux.so)
by 0x409FA2: sortlines (sort.c:3649)
by 0x409E26: sortlines (sort.c:3621)
by 0x40AA9E: sort (sort.c:3955)
by 0x40C5D9: main (sort.c:4739)
mutex 0x5419548 was first observed at:
at 0x4C2DE82: pthread_mutex_init(in vgpreload_drd-amd64-linux.so)
by 0x409266: init_node (sort.c:3276)
by 0x4092F4: init_node (sort.c:3286)
by 0x4090DD: merge_tree_init (sort.c:3234)
by 0x40AA5A: sort (sort.c:3951)
by 0x40C5D9: main (sort.c:4739)
Thread 2:
The object at address 0x5419548 is not a mutex.
at 0x4C2F4A4: pthread_mutex_unlock(in vgpreload_drd-amd64-linux.so)
by 0x4093CA: unlock_node (sort.c:3323)
by 0x409C85: merge_loop (sort.c:3531)
by 0x409F8F: sortlines (sort.c:3644)
by 0x409CE3: sortlines_thread (sort.c:3574)
by 0x4E44F32: start_thread (in /usr/lib64/libpthread-2.18.so)
by 0x514EEAC: clone (in /usr/lib64/libc-2.18.so)
* src/sort.c (sortlines): Move pthread_mutex_destroy() out to
merge_tree_destroy(), so that we don't overlap mutex destruction
with threads still operating on the nodes.
(sort): Call the destructors only with "lint" defined, as the
memory used will be deallocated implicitly at process end.
* NEWS: Mention the bug fix.
|
|
Neither issue impacts on the correct operation of sort.
The issues were detected by both valgrind 3.8.1 and 3.9.0 using:
seq 200000 > file.sort
valgrind --tool=drd src/sort file.sort -o file.sort
For tool usage and error details see:
http://valgrind.org/docs/manual/drd-manual.html
* src/sort.c (queue_insert): Unlock mutex _after_ signalling the
associated condition variable. Valgrind flags this with:
"Probably a race condition: condition variable 0xffeffffb0 has been
signaled but the associated mutex 0xffeffff88 is not locked by the
signalling thread."
The explanation at the above URL is:
"Sending a signal to a condition variable while no lock is held on
the mutex associated with the condition variable. This is a common
programming error which can cause subtle race conditions and
unpredictable behavior."
This should at least give more defined scheduling behavior.
(merge_tree_destroy): Make symmetrical with merge_tree_init() thus
destroying the correct mutex. Valgrind flags this with:
"The object at address 0x5476cf8 is not a mutex."
|
|
* src/join.c (usage): Reword to avoid implication that
the NUL byte is only generated as the output delimeter.
* src/sort.c (usage): Likewise.
* src/shuf.c (usage): Likewise. Also since we're changing the
translation string take the opportunity to separate out
the description to a separate string to reduce translation overhead.
* src/uniq.c (usage): Likewise.
* src/stty.c (usage): s/null/NUL/ for consistency.
* src/basename.c (usage): Reword for accuracy/consistency.
* src/dirname.c (usage): Likewise.
* src/du.c (usage): Likewise.
* src/env.c (usage): Likewise.
* src/printenv.c (usage): Likewise.
* src/readlink.c (usage): Likewise.
* src/realpath.c (usage): Likewise.
* doc/coreutils.texi: Consolidate/share the descriptions of
--null, --zero and --zero-terminated.
|
|
Run "make update-copyright", but then also run this,
perl -pi -e 's/2\d\d\d-//' tests/sample-test
to make that one script use the single most recent year number.
|
|
* src/sort.c: (async_safe_die): A new limited version of error(),
that outputs fixed strings and unconverted errnos to stderr.
This is safe to call in the limited context of a signal handler,
or in this particular case, between the fork() and exec() of
a multithreaded process.
(move_fd_or_die): Use the async_safe_die() rather than error().
(maybe_create_temp): Likewise.
(open_temp): Likewise.
Fixes http://bugs.gnu.org/15970
|
|
* src/copy.c (copy_internal): Change mention of the removed --reply=no
option, to the similar in this context --no-clobber.
* src/sort.c: SI and IEC suffixes can now be mixed when --human-numeric.
|
|
* configure.ac: If clang, add -Wno-format-extra-args and
-Wno-tautological-constant-out-of-range-compare.
* gl/lib/rand-isaac.c (ind):
* gl/lib/randread.c (readisaac):
* src/ls.c (dev_ino_push, dev_ino_pop):
* src/sort.c (buffer_linelim):
* src/system.h (is_nul):
* src/tail.c (tail_forever_inotify):
Rewrite to avoid casts that clang dislikes.
It's good to avoid casts anyway.
* src/expr.c (integer_overflow): Declare only if it exists.
(die): Remove; unused.
* src/ls.c (dev_ino_push): New function, replacing ...
(DEV_INO_PUSH): ... this removed macro. All uses changed.
(decode_switches): Rewrite "str"+i to &str[i].
|
|
Each program with at least one long option which is marked as
'required_argument' and which has also a short option for that
option, should print a note about mandatory arguments.
Define that well-known note centrally and use it rather than
literal printf/fputs, and add it where it was missing.
* src/system.h (emit_mandatory_arg_note): Add new function.
* src/cp.c (usage): Use it rather than literal printf/fputs.
* src/csplit.c, src/cut.c, src/date.c, src/df.c, src/du.c:
* src/expand.c, src/fmt.c, src/fold.c, src/head.c, src/install.c:
* src/kill.c, src/ln.c, src/ls.c, src/mkdir.c, src/mkfifo.c:
* src/mknod.c, src/mv.c, src/nl.c, src/od.c, src/paste.c:
* src/pr.c, src/ptx.c, src/shred.c, src/shuf.c, src/sort.c:
* src/split.c, src/stdbuf.c, src/tac.c, src/tail.c, src/timeout.c:
* src/touch.c, src/truncate.c, src/unexpand.c, src/uniq.c:
Likewise.
* src/base64.c (usage): Add call of the above new function
because at least one long option has a required argument.
* src/basename.c, src/chcon.c, src/date.c, src/env.c:
* src/nice.c, src/runcon.c, src/seq.c, src/stat.c, src/stty.c:
Likewise.
|
|
Run "make update-copyright", but then also run this,
perl -pi -e 's/2\d\d\d-//' tests/sample-test
to make that one script use the single most recent year number.
|
|
This also fixes a free-memory-read (FMR) bug: when fillbuf's realloc
of buf->buf frees the buffer into which saved_line.text points,
the processing of that just-read longer line includes comparison
against the saved line in freed memory.
* src/sort.c (overlap): Remove.
(fillbuf): Do not try to copy saved lines, as that is too risky
in the presence of parallelism, reallocated buffers, etc.
(sort): Invalidate any saved line before sorting a new batch.
|
|
sort -u could omit one or more lines of expected output.
This bug arose because sort recorded the most recently printed line via
reference, and if you were unlucky, the storage for that line would be
reused (overwritten) as additional input was read into memory. If you
were doubly unlucky, the new value of the "saved" line would not only
match the very next line, but if that next line were also the first in
a series of identical, not-yet-printed lines, then the corrupted "saved"
line value would result in the omission of all matching lines.
* src/sort.c (saved_line): New static/global, renamed and moved from...
(write_unique): ...here. Old name was "saved", which was too generic
for its new role as file-scoped global.
(fillbuf): With --unique, when we're about to read into a buffer that
overlaps the saved "preceding" line (saved_line), copy the line's .text
member to a realloc'd-as-needed temporary buffer and adjust the line's
key-defining members if they're set.
(overlap): New function.
* tests/misc/sort: New tests.
* NEWS (Bug fixes): Mention it.
* THANKS.in: Update.
Bug introduced via commit v8.5-89-g9face83.
Reported by Rasmus Borup Hansen in
http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/23173/focus=24647
|
|
* src/sort.c (fillbuf): Fix comment typo. x2nrealloc no longer
doubles the size of its input buffer.
|
|
* src/sort.c (default_sort_size): Do not exceed 3/4 of total memory.
See Jeff Janes's bug report in
<http://lists.gnu.org/archive/html/coreutils/2012-06/msg00018.html>.
|
|
* src/sort.c (stream_open): EXIT_FAILURE -> SORT_FAILURE.
Suggested by Pádraig Brady in <http://bugs.gnu.org/11816#34>.
|
|
* src/sort.c (outfd): Remove. All uses replaced by STDOUT_FILENO.
(stream_open): When writing, use stdout rather than fdopen.
(move_fd_or_die): Renamed from dup2_or_die, with the added functionality
of closing its first argument. All uses changed.
(avoid_trashing_input): Special case for !outfile no longer needed.
(check_output): Arrange for standard output to go to the file,
rather than storing the fd in outfd.
|
|
* src/sort.c (check_inputs): A new function to verify all inputs
are accessible before further processing.
(check_output): A new function to open or create a specified
output file, before futher processing.
(stream_open): Adjust to truncating the previously opened
output file rather than opening directly.
(avoid_trashing_input): Optimize to stat the output file
descriptor, rather than the file name.
(main): Call the new functions to check accessibility of
inputs and output, before processing starts.
* tests/misc/sort: Adjust to the changed error message.
* tests/misc/sort-merge-fdlimit: Account for the earlier opened
file descriptor of the specified output file.
* tests/misc/sort-exit-early: A new test to exercise the improvements.
* tests/Makefile.am: Reference the new test.
* NEWS: Mention the improvement.
Suggested-by: Bernhard Voelker
|
|
* src/sort.c (default_sort_size): Move physmem code "down" to first use.
|
|
A static analysis tool (http://labs.oracle.com/projects/parfait/)
produced some false positive diagnostics. Add assertions to help
it understand that the code is correct.
* src/stty.c: Include <assert.h>.
(display_changed): Add an assertion to placate parfait.
(display_all): Likewise.
* src/sort.c: Include <assert.h>.
(main): Add an assertion to placate parfait.
* src/fmt.c: Include <assert.h>.
(get_paragraph): Add an assertion to placate parfait.
|
|
* bootstrap.conf (gnulib_modules): Add sys_resource.
* src/sort.c: Remove #if HAVE_SYS_RESOURCE_H guard around inclusion
of <sys/resource.h> and move the inclusion "up" into the alphabetized
list of its peers. This also avoids a failure of the
sc_prohibit_always_true_header_tests syntax-check rule.
* m4/jm-macros.m4 (gl_CHECK_ALL_HEADERS): Remove sys/resource.h.
|
|
* src/sort.c (default_sort_size): Don't divide advice by 2.
Just divide the hard limits by 2. This matches the comments.
Reported by Rogier Wolff in http://bugs.gnu.org/10877
|
|
* src/sort.c (key_warnings): Use quote (quote_n, since there are two)
rather than literal single quotes ('%s') in diagnostic.
|
|
Run "make update-copyright".
|
|
* doc/coreutils.texi (Formatting the file names):
coreutils now quotes 'like this'.
* man/help2man:
* src/timeout.c (usage): Quote 'like this' in diagnostics.
* HACKING, Makefile.am, NEWS, README, README-hacking, TODO, cfg.mk:
* doc/Makefile.am, doc/coreutils.texi, m4/jm-macros.m4:
* man/Makefile.am, man/help2man, src/Makefile.am, src/copy.h:
* src/extract-magic, src/ls.c, src/pinky.c, src/pr.c, src/sort.c:
* src/split.c, src/timeout.c, src/who.c, tests/dd/skip-seek-past-file:
* tests/pr/pr-tests: Quote 'like this' in commentary.
* cfg.mk (old_NEWS_hash): Update due to changed old NEWS.
|
|
Run this (twice):
git grep -E -l '`.+'\' src/*.[ch] \
|xargs perl -pi -e 's/`(.+?'\'')/'\''$1/'
|
|
Run this (twice):
git grep -E -l '`[^ ]+'\' src/*.c \
|xargs perl -pi -e 's/`([^ ]+'\'')/'\''$1/'
|
|
All affected lines end with \ or \n\, so run this command
until it produces no new changes (4 times):
git grep -E -l '`[^ ]+'\''.*\\' src \
|xargs perl -pi -e 's/`([^ ]+'\''.*\\)/'\''$1/'
|
|
* src/csplit.c (parse_repeat_count, extract_regexp): As above.
* src/date.c (main): Likewise.
* src/ls.c (decode_switches): Likewise.
* src/od.c (decode_one_format, main): Likewise.
* src/pathchk.c (no_leading_hyphen): Likewise.
* src/pr.c (main, getoptarg): Likewise.
* src/rm.c (diagnose_leading_hyphen): Likewise.
* src/sort.c (key_warnings, incompatible_options, main): Likewise.
* src/stat.c (print_esc_char): Print '\x', not `\x' in diagnostic.
* src/test.c (main): Likewise.
* src/touch.c (main): Likewise.
* src/tr.c (build_spec_list, validate, append_range): Likewise.
* tests/misc/mktemp: This is an unusual case, since the affected
string contains only the ` of an `...' string. So we change
the long ` to a lone '.
* tests/pr/pr-tests: Manual quote adapting fix-up.
* tests/ln/hard-to-sym: Likewise.
* tests/split/suffix-length: Likewise.
* tests/mv/part-fail: Likewise.
* tests/misc/chcon: Likewise.
* tests/misc/stat-printf: Likewise.
|
|
Run this command:
perl -0777 -pi -e \
's/fprintf \(stderr, _\("Try `%s --help.*\n.*;/emit_try_help ();/m'\
src/*.c
|
|
Run "make update-copyright".
|
|
* src/sort.c (struct thread_args) [is_lo_child]: Remove member.
(sortlines): Remove unused parameter, "is_lo_child". Update callers.
* src/stat.c (out_epoch_sec): Mark unused parameter.
(do_statfs, do_stat): Remove unused parameter, "terse". Update callers.
|
|
* src/sort.c (usage): Use KEYDEF instead of POS, and call out the
specific OPTS that can occur in KEYDEF.
Based on a report by Lars Noodén, http://bugs.gnu.org/10019
|
|
These commands would fail to terminate:
yes -- -nan | head -156903 | sort -g > /dev/null
echo nan > F; sort -m -g F F
That can happen with any strtold implementation that includes
uninitialized data in its return value. The problem arises in the
mergefps function when bubble-sorting the two or more lines, each
from one of the input streams being merged: compare(a,b) returns 64,
yet compare(b,a) also returns a positive value. With a broken
comparison function like that, the bubble sort never terminates.
Why do the long-double bit strings corresponding to two identical
"nan" strings not compare equal? Because some parts of the result
are uninitialized and thus depend on the state of the stack.
For more details, see http://bugs.gnu.org/9612.
* src/sort.c (nan_compare): New function.
(general_numcompare): Use it rather than bare memcmp.
Reported by Aaron Denney in http://bugs.debian.org/642557.
* NEWS (Bug fixes): Mention it.
* tests/misc/sort-NaN-infloop: New file.
* tests/Makefile.am (TESTS): Add it.
|
|
* configure.ac (WARN_CFLAGS): Add -Wsuggest-attribute=const,
-Wsuggest-attribute=pure and -Wsuggest-attribute=noreturn.
(GNULIB_WARN_CFLAGS): But do not add them here... yet.
* src/chown-core.h (chopt_free, uid_to_name): Add function attribute(s).
* src/copy.c (is_ancestor, valid_options): Likewise.
* src/copy.h (chown_failure_ok): Likewise.
* src/dd.c (operand_matches, operand_is): Likewise.
* src/df.c (selected_fstype, excluded_fstype): Likewise.
* src/expr.c (null looks_like_integer): Likewise.
* src/md5sum.c (hex_digits): Likewise.
* src/od.c (get_lcm): Likewise.
* src/pathchk.c (component_start, component_len): Likewise.
* src/pinky.c (count_ampersands): Likewise.
* src/pr.c (cols_ready_to_print): Likewise.
* src/ptx.c (search_table): Likewise.
* src/sort.c (find_unit_order): Likewise.
* src/stty.c (mode_type_flag, string_to_baud, baud_to_value): Likewise.
* src/system.h (gcd, lcm): Likewise.
* src/tr.c (is_char_class_member, look_up_char_class): Likewise.
(star_digits_closebracket): Likewise.
* src/uniq.c (find_field): Likewise.
* src/wc.c (compute_number_width): Likewise.
* lib/xfts.h (cycle_warning_required): Likewise.
* gl/lib/randint.h (randint_get_source): Likewise.
* gl/lib/randperm.c (ceil_lg): Likewise.
* gl/lib/randperm.h (randperm_bound): Likewise.
* lib/strnumcmp.h (strintcmp): Likewise.
|
|
* NEWS: "an misleading"
* src/expr.c: "a integer
* src/ptx.c (find_occurs_in_text): "a end"
* src/shred.c (do_wipefd): "a infinite"
* src/sort.c (SUBTHREAD_LINES_HEURISTIC): "an dual-core"
(compare_random): "an checksum"
* cfg.mk (old_NEWS_hash): Update, since the typo was in old news.
|
|
* src/sort.c (key_warn): `sort -k2,1n --debug` would output
warnings about being both "zero width" and "spanning multiple fields".
Suppress the latter one.
* tests/misc/sort-debug-warn: Add a couple of test cases.
|
|
* src/sort.c (INPUT_FILE_SIZE_GUESS): Decrease initial allocation
factor used to size buffer used when reading a non-regular file.
For motivation, see discussion here:
http://thread.gmane.org/gmane.comp.gnu.coreutils.general/878/focus=887
|
|
* src/sort.c (SUBTHREAD_LINES_HEURISTIC): Do not spawn a new thread
for every 4 lines. Increase this from 4 to 128K. 128K lines seems
appropriate for a 5-year-old dual-core laptop, but it is too low for
some common combinations of short lines and/or newer systems.
* NEWS (Bug fixes): Mention it.
|
|
This case was overlooked in commit bdde34f9, 2010-08-05,
"sort: tune and refactor --debug code, and fix minor underlining bug"
* src/sort.c (debug_key): Don't adjust the key end when
it's before the key start.
* tests/misc/sort-debug-keys: Add a test case.
|
|
* src/pr.c (char_to_clump): Remove a dead store.
* src/remove.c (fts_skip_tree): Likewise.
* src/sort.c (key_warnings): Likewise.
(sort): Suppress an uninitialized pointer warning.
|
|
Run "make update-copyright".
|
|
* cfg.mk (LINE_LEN_MAX, FILTER_LONG_LINES): New macros.
(sc_long_lines): New rule.
* HACKING: Use shorter URLs to the same material.
* doc/Makefile.am, doc/coreutils.texi, m4/boottime.m4:
* man/help2man, man/stdbuf.x, src/Makefile.am, src/cat.c, src/copy.c:
* src/cp.c, src/dd.c, src/df.c, src/du.c, src/groups.c, src/install.c:
* src/ls.c, src/md5sum.c, src/mv.c, src/od.c, src/pinky.c, src/ptx.c:
* src/readlink.c, src/remove.c, src/rmdir.c, src/setuidgid.c:
* src/sort.c, src/tail.c, src/touch.c, tests/Coreutils.pm:
* tests/cp/existing-perm-race, tests/cp/perm, tests/cp/preserve-gid:
* tests/du/2g, tests/du/long-from-unreadable, tests/init.sh:
* tests/install/basic-1, tests/ls/nameless-uid:
* tests/ls/readdir-mountpoint-inode, tests/misc/chroot-credentials:
* tests/misc/cut, tests/misc/date, tests/misc/join, tests/misc/md5sum:
* tests/misc/sha1sum, tests/misc/sha224sum, tests/misc/sort:
* tests/misc/sort-continue, tests/misc/sort-files0-from:
* tests/misc/sort-rand, tests/misc/stdbuf, tests/misc/tr:
* tests/misc/uniq, tests/mv/atomic, tests/mv/part-fail:
* tests/mv/part-symlink, tests/mv/sticky-to-xpart, tests/pr/pr-tests:
* tests/rm/fail-2eperm, tests/rm/interactive-always:
Reformat to fit within 80 columns.
* doc/Makefile.am (BAD_POSIX_PERL): New macro.
* doc/coreutils.texi: Reword slightly, to make menus and
index lines shorter.
* src/md5sum.c: Redo --help output so that it fits within 79
columns, since that's a bit more portable and all the other --help
strings fit in 79 columns.
|
|
* src/sort.c (main): Don't invoke num_processors twice.
|
|
Also fix up Chen Guo's contacts
* src/sort.c (usage): Add a missing "of"
* THANKS: Add Chen Guo
* .mailmap: Add Chen Guo's UCLA address
|
|
* src/sort.c (main): If --parallel isn't specified,
restrict the number of threads to 8 by default.
If the --parallel option is specified, then
allow any number of threads to be set, independent
of the number of processors on the system.
* doc/coreutils.texi (sort invocation): Document the changes
to determining the number of threads to use.
Mention the memory overhead when using multiple threads.
* tests/misc/sort-spinlock-abuse: Allow single core
systems that support pthreads.
* tests/misc/sort-stale-thread-mem: Likewise.
* tests/misc/sort-unique-segv: Likewise.
* NEWS: Mention the change in behaviour.
|
|
Without this change, tests/misc/sort-compress-hang would consume
more than 10,000 process slots on my RHEL 5.5 x86-64 server,
making it likely for other applications to fail due to lack of
process slots. With this change, the same benchmark causes 'sort'
to consume at most 19 process slots. The change also improved
wall-clock time by 2% and user+system time by 14% on that benchmark.
* NEWS: Document this.
* src/sort.c (MAX_PROCS_BEFORE_REAP): Remove.
(reap_exited): Renamed from reap_some; this is a more accurate name,
since "some" incorrectly implies that it reaps at least one process.
All uses changed.
(reap_some): New function: it *does* reap at least one process.
(pipe_fork): Do not allow more than NMERGE + 2 subprocesses.
(mergefps, sort): Omit check for exited processes: no longer needed,
and anyway the code consumed too much CPU per line when 2 < nprocs.
|
|
* NEWS: Document this.
* src/sort.c (UNCOMPRESSED, UNREAPED, REAPED): New constants.
(struct tempnode): New member 'state', to hold these constants.
The pid member is now undefined if state == UNCOMPRESSED.
(struct sortfile): Replace member 'pid' with member 'temp'.
(uintptr): Remove.
(proctab_hasher, proctab_comparator, register_proc, delete_proc):
Proctab entries are now struct tempnode *, not pid_t, to handle
the case where multiple tempnode objects correspond to the same
pid. This avoids a race condition that can cause a hang.
(register_proc): Arg is now struct tempnode *, not pid_t. All
callers changed.
(delete_proc): Set tempnode state to REAPED.
(create_temp_file): No need to set pid member here; it's now
done when the pid is known.
(maybe_create_temp, create_temp): Remove PPID arg. Return struct
tempnode *, not char *. All callers changed.
(maybe_create_temp): Set node state to UNCOMPRESSED or UNREAPED.
No need to set node->pid to 0.
(open_temp): Replace NAME and PID args with a single TEMP arg.
All callers changed. Wait only for unreaped children.
(zaptemp): Wait for decompressor to finish before removing its
temporary-file input. This avoids .nfsXXXX hassles with NFS
and fixes a race (leading to a hang) regardless of NFS.
(open_input_files): Adjust to new way of dealing with temp files
and their subprocesses.
* tests/Makefile.am (TESTS): Add misc/sort-compress-hang.
* tests/misc/sort-compress-hang: New file.
|
|
* NEWS: Document this.
* src/sort.c (avoid_trashing_input): The previous fix to this
function didn't fix all the problems with this code. Replace it
with something simpler: just copy the input file. This doesn't
change the number of files, so return void instead of the updated
file count. Caller changed.
* tests/misc/sort-merge-fdlimit: Test for the bug.
|
|
* src/sort.c (avoid_trashing_input): Fix a typo that could cause a
buffer overrun in theory. In practice this is extremely unlikely,
as it requires running out of file descriptors in a small merge,
presumably because some other process is hogging all the OS's file
descriptors.
|