Paul Eggert [Fri, 30 Jul 2004 07:57:27 +0000 (07:57 +0000)]
(struct dir_attr, flag_path, remove_trailing_slashes,
re_protect, make_path_private, target_directory_operand, do_copy,
cp_option_init, decode_preserve_arg, main): Use bool when appropriate.
(target_directory_operand): Do not clear *NEW_DST if stat
succeeds. It's not necessary in that case, as *NEW_DST is always
false already.
(do_copy): Rewrite slightly to avoid need for "unreachable" comment.
(main): Use EXIT_SUCCESS, EXIT_FAILURE instead of 0, 1.
Paul Eggert [Fri, 30 Jul 2004 00:54:44 +0000 (00:54 +0000)]
(OPENOPTS, have_read_stdin, status_only, warn,
bsd_split_3, split_3, hex_digits, digest_file, digest_check, main):
Use bool when appropriate.
(digest_check): Increase limit of number of input lines to
UINTMAX_MAX from INT_MAX. Diagnose any overflows of this counter.
Use ngettext instead of hard-to-i18nize hardcoded stuff for plurals.
Paul Eggert [Fri, 30 Jul 2004 00:53:49 +0000 (00:53 +0000)]
Don't include config.h, sys/types.h, stdio.h: not needed.
(ALG_UNSPECIFIED): Remove.
(ALG_MDT): Don't make it equal to CHAR_MAX + 1; this isn't necessary.
Paul Eggert [Thu, 29 Jul 2004 06:12:27 +0000 (06:12 +0000)]
(exit_status): Remove. Now done by passing a boolean
'ok' flag around.
(simple_cat, cat): Return true if successful. All callers changed.
(simple_cat, cat, main): Use bool for booleans.
(simple_cat): Use size_t for sizes.
(cat, main): Use the same names for parameters that we use for
long options, to avoid confusion. This inverts the sense of the
show_tabs (formerly output_tabs) and number_nonblank
(formerly numbers_at_empty_lines) variables.
(main): Don't mess up (due to integer overflow) if we are given
INT_MAX - INT_MIN + 1 options.
[O_BINARY]: Don't invoke isatty unless the other options require it.
(main): When deciding whether to use simple_cat, don't worry
about binary option; it's irrelevant.
Paul Eggert [Wed, 28 Jul 2004 23:59:11 +0000 (23:59 +0000)]
Remove comments, trailing white space, and empty
lines from the output strings, to save space.
Use a narrower type like 'unsigned char' for line lengths, if
that will do.
Make the output variables static, not extern.
Paul Eggert [Wed, 28 Jul 2004 23:37:49 +0000 (23:37 +0000)]
Include inttostr.h.
(UINT_MAX_DECIMAL_DIGITS, uint_to_string): Remove.
(gid_to_name, uid_to_name): Use imaxtostr/umaxtostr
instead of uint_to_string).
(describe_change): Instead of an int flag, use a char *
auxiliary; this avoids the need for casts.
Assume free (NULL) works.
(change_file_owner): Return true/false, not 0/-1, since
we don't set errno. All callers changed.
Use bool when appropriate.
(chown_files): Likewise.
Paul Eggert [Wed, 28 Jul 2004 23:06:11 +0000 (23:06 +0000)]
Include <stdbool.h>, "inttostr.h".
(V_STRDUP): Don't assume the string's length fits in int.
(ISDIGIT): unsigned -> unsigned int
(is_number): Define only ifdef __DJGPP__; not needed elsewhere.
Use bool instead of int where appropriate.
Do not allow empty strings.
(parse_user_spec): Parse numbers as decimal integers, even if
they have a leading 0. Don't assume uids and gids fit in int.
Paul Eggert [Wed, 28 Jul 2004 22:21:18 +0000 (22:21 +0000)]
Include <stddef.h>, not <stdlib.h> and <sys/types.h>.
(LONG_MAX_32_BITS): Remove.
Include <inttypes.h> and <stdint.h> if available.
(alignof, UNALIGNEDP): New macro, portable to all C89 hosts.
(__memrchr): Don't assume unsigned long int is either 4 or 8 bytes;
let it be any number of bytes greater than or equal to 4.
Paul Eggert [Wed, 28 Jul 2004 22:20:56 +0000 (22:20 +0000)]
Include <stddef.h>, not <stdlib.h> and <sys/types.h>.
(LONG_MAX_32_BITS): Remove.
Include <inttypes.h> and <stdint.h> if available.
(alignof, UNALIGNEDP): New macro, portable to all C89 hosts.
(__memchr): Don't assume unsigned long int is either 4 or 8 bytes;
let it be any number of bytes greater than or equal to 4.
* memrchr.c: Likewise, with __memrchr.
Paul Eggert [Wed, 28 Jul 2004 20:09:39 +0000 (20:09 +0000)]
Don't include <sys/types.h> or <stdlib.h>; <stddef.h>
suffices with C89 or better.
(alignof): New macro, portable to all C89 hosts.
(UNALIGNED): Use it. Use uintptr_t if available, and assume
everything is unaligned otherwise; this is more portable than
assuming 'unsigned long int' will always work.
Paul Eggert [Wed, 28 Jul 2004 19:23:07 +0000 (19:23 +0000)]
Don't assume that Perl's getpwd agrees with our
pwd when there are multiple names for the working directory
(which can happen with an automounter, sigh).
Paul Eggert [Wed, 28 Jul 2004 18:53:22 +0000 (18:53 +0000)]
this causes Solaris 8 'make' to refuse to build "groups".
(localedir.h): Don't depend on Makefile: this causes Solaris
8 'make' to build localedir.h unnecessarily. The dependence
on Makefile is ineffective anyway, since $(localedir) might
change even if Makefile hasn't.
Paul Eggert [Wed, 28 Jul 2004 14:50:06 +0000 (14:50 +0000)]
(remove_dir): If we can't save the state of the
working directory, pretend we started from "/", not ".".
This avoids a bug on hosts like Solaris that don't let you
remove the working directory.
Paul Eggert [Wed, 28 Jul 2004 06:58:25 +0000 (06:58 +0000)]
(strtiomax, strtoumax): Declare if not already
declared: this fixes a portability bug with Solaris 8 + GCC.
(STRTOX): Parenthesize use of macro arg as expression.
(vstrtoimax, vstrtoumax, vstrtold): Remove now-unnecessary
parentheses.
Paul Eggert [Wed, 28 Jul 2004 05:00:10 +0000 (05:00 +0000)]
(gl_PREREQ_EUIDACCESS): Don't bother checking for
unistd.h, as autoconf does this for us. Check for libgen.h.
Also look for eaccess within -lgen, which is where it is in
Solaris. Set LIB_EACCESS accordingly.
Paul Eggert [Wed, 28 Jul 2004 00:31:14 +0000 (00:31 +0000)]
Remove now-inaccurate comment about the files
you need to include first. You don't need to include any files
other than the usual config.h.
Include <inttypes.h> and <stdint.h> if available, for uintmax_t.
Remove 'struct stat;' not needed since we know sys/stat.h has
been included by dev-ino.h.
(struct cycle_check_state): Change chdir_counter to uintmax_t,
not size_t, since it isn't limited by object sizes.
Change magic from long unsigned int to int; that's good enough
for our use.
Paul Eggert [Wed, 28 Jul 2004 00:30:45 +0000 (00:30 +0000)]
(is_zero_or_power_of_two): Renamed from
is_power_of_two, to reflect better what it really does.
All uses changed. Arg is now uintmax_t, not unsigned int
(it should have been unsigned long int -- that was a bug).
(cycle_check): Check for integer overflow in cycle count,
and report a cycle if that happens, as it must be a cycle
by this point.
Paul Eggert [Tue, 27 Jul 2004 23:12:56 +0000 (23:12 +0000)]
backupfile.c rewrite to avoid arbitrary limits on lengths of
numeric backup extensions.
Include "backupfile.h" first.
Include errno.h, stdbool.h, limits.h, unistd.h, xalloc.h.
(CLOSEDIR, INT_STRLEN_BOUND): Remove.
(pathconf) [! (HAVE_PATHCONF && defined _PC_NAME_MAX)]: New macro.
(_POSIX_NAME_MAX) [!defined _POSIX_NAME_MAX]: New macro.
(NAME_MAX_MAXIMUM): New macro. Unlike the old addext.c, we
also look at _XOPEN_NAME_MAX, for better performance on modern
hosts that support only file names of length 255 or more.
(ISDIGIT): unsigned -> unsigned int
(max_backup_version, version_number): Remove.
(check_extension): New function. Similar to the old addext, but
static, assumes that the extension has already been added,
and a bit more careful on DOS hosts.
(numbered_backup): New function. It does what max_backup_version
and version_number used to do, but it doesn't use integer arithmetic
to calculate extensions so it doesn't overflow.
(find_backup_file_name): Rewrite to use these new functions.
This has a new optimization: we needn't call pathconf if the
new numbered backup name has the same length as the old.
Also, use xmalloc rather than malloc, so that the caller
needn't worry about memory exhaustion.
Paul Eggert [Mon, 26 Jul 2004 06:08:50 +0000 (06:08 +0000)]
(GET_NICE_VALUE): Renamed from GET_PRIORITY.
All uses changed.
(NZERO): New macro, if system doesn't define it already.
(usage): Distinguish priorities from nice values.
Don't assume NZERO is 20.
(main): Use bool instead of int where appropriate.
If user specifies an adjustment out of range, always truncate it
to an inrange value instead of sometimes giving an error message
and sometimes not.
Do not assume that -1 is an error return from "nice" or
"getpriority", as it might be the current nice value minus NZERO.
If nice/setpriority fails with errno == EPERM, go ahead and run
the command anyway; POSIX requires this.
Paul Eggert [Mon, 26 Jul 2004 06:08:18 +0000 (06:08 +0000)]
(nice invocation): Document the "nice value", and
how it affects the scheduling priority. (The old documentation
implied that the nice value equaled the scheduling priority, which
isn't accurate.) Document that the range of nice values might
exceed -20..19. Specify what happens when you give a nice value
that is out of range, or when you don't have permissions to lower
the nice value. Bash doesn't have a builtin 'nice', so don't say
"most shells" have one.
Paul Eggert [Sun, 25 Jul 2004 07:37:55 +0000 (07:37 +0000)]
(gl_FUNC_NONREENTRANT_EUIDACCESS): New macro.
(gl_FUNC_EUIDACCESS): Use AC_CHECK_DECLS_ONCE, not AC_CHECK_DECLS.
(gl_PREREQ_EUIDACCESS): Check for eaccess and setregid decls.
Require AC_HEADER_STAT.
Paul Eggert [Sun, 25 Jul 2004 07:37:22 +0000 (07:37 +0000)]
[!defined LIBC]: Included group-member.h, stat-macros.h.
(S_IXUSR, S_IXGRP, S_IXOTH, S_IROTH, S_IWOTH, S_IXOTH):
Remove; now done by stat-macros.h.
(NGROUPS_MAX, group_member): Remove; now down by group-member.h.
No need to include <limits.h>.
(errno): Remove decl; we now assume C89 or better.
(access, getuid, getgid, geteuid, getegid, stat) [defined _LIBC]:
New macros.
(uid, gid, have_ids): Remove these static variables.
They weren't accurate for programs that also invoked setreuid etc.
(euidaccess) [defined EFF_ONLY_OK || defined ACC_SELF ||
HAVE_DECL_EACCSS]: Use builtin substitutes.
[defined _LIBC]: Ignore __libc_enable_secure; it's not a
correct optimization for programs run as root that later
invoke setreuid.
[no builtin substitutes && HAVE_DECL_SETREGID &&
PREFER_NONREENTRANT_EUIDACCESS]:
Use setreuid+setregid to get the correct answer.
[no builtin substitutes && ! (HAVE_DECL_SETREGID &&
PREFER_NONREENTRANT_EUIDACCESS)]:
Don't assume that the stat macros have their historical values,
as POSIX doesn't require this.
[defined TEST]: Include <stdlib.h>; don't include errno.h
twice; include <error.h> rather than "error.h".
Paul Eggert [Fri, 23 Jul 2004 22:33:51 +0000 (22:33 +0000)]
Include fcntl-safer.h.
(COPY_TO_EOF): Set to UINTMAX_MAX, not OFF_T_MAX (which was wrong).
(COPY_A_BUFFER): New macro.
(struct File_spec): New members mtime, mode, blocking.
Remove member n_consecutive_size_changes.
(DEFAULT_MAX_N_CONSECUTIVE_SIZE_CHANGES,
max_n_consecutive_size_changes_between_opens,
MAX_CONSECUTIVE_SIZE_CHANGES_OPTION): Remove.
(long_options, tail_forever, parse_options):
Remove (non-)support for --max-consecutive-size-changes.
(record_open_fd): New function.
(recheck, tail_file): Use it. Don't assume that stdin is open.
(dump_remainder): Add support for new COPY_A_BUFFER special value.
Treat errno==EAGAIN like EOF, since it might be a nonblocking read.
(recheck): New arg BLOCKING, specifying whether to use blocking reads.
All uses changed.
(n_live_files): Remove, replacing with...
(any_live_files): New function. All uses changed.
(tail_forever): Use nonblocking I/O unless we know that blocking I/O
is safe; this avoids some hangs when reading from a fifo.
Avoid invoking fstat or sleep when using blocking I/O.
Do not check for changes to size if the file is not a regular file,
as the size is undefined in that case.
Check for changes to mtime or mode, too; this works for non-regular
files.
(tail_forever, main): Redo fflush strategy to work even when input
is nonblocking. Don't use unbuffered output; just flush when needed.