2 * Copyright (C) 2004-2013 Andre Noll <maan@systemlinux.org>
4 * Licensed under the GPL v2. For licencing details see COPYING.
7 /** \file string.c Memory allocation and string handling functions. */
12 #include <sys/utsname.h> /* uname() */
26 * Paraslash's version of realloc().
28 * \param p Pointer to the memory block, may be \p NULL.
29 * \param size The desired new size.
31 * A wrapper for realloc(3). It calls \p exit(\p EXIT_FAILURE) on errors,
32 * i.e. there is no need to check the return value in the caller.
34 * \return A pointer to the newly allocated memory, which is suitably aligned
35 * for any kind of variable and may be different from \a p.
39 __must_check __malloc void *para_realloc(void *p, size_t size)
42 * No need to check for NULL pointers: If p is NULL, the call
43 * to realloc is equivalent to malloc(size)
46 if (!(p = realloc(p, size))) {
47 PARA_EMERG_LOG("realloc failed (size = %zu), aborting\n",
55 * Paraslash's version of malloc().
57 * \param size The desired new size.
59 * A wrapper for malloc(3) which exits on errors.
61 * \return A pointer to the allocated memory, which is suitably aligned for any
66 __must_check __malloc void *para_malloc(size_t size)
73 PARA_EMERG_LOG("malloc failed (size = %zu), aborting\n",
81 * Paraslash's version of calloc().
83 * \param size The desired new size.
85 * A wrapper for calloc(3) which exits on errors.
87 * \return A pointer to the allocated and zeroed-out memory, which is suitably
88 * aligned for any kind of variable.
92 __must_check __malloc void *para_calloc(size_t size)
94 void *ret = para_malloc(size);
101 * Paraslash's version of strdup().
103 * \param s The string to be duplicated.
105 * A wrapper for strdup(3). It calls \p exit(EXIT_FAILURE) on errors, i.e.
106 * there is no need to check the return value in the caller.
108 * \return A pointer to the duplicated string. If \a s was the \p NULL pointer,
109 * an pointer to an empty string is returned.
113 __must_check __malloc char *para_strdup(const char *s)
117 if ((ret = strdup(s? s: "")))
119 PARA_EMERG_LOG("strdup failed, aborting\n");
124 * Print a formated message to a dynamically allocated string.
126 * \param result The formated string is returned here.
127 * \param fmt The format string.
128 * \param ap Initialized list of arguments.
130 * This function is similar to vasprintf(), a GNU extension which is not in C
131 * or POSIX. It allocates a string large enough to hold the output including
132 * the terminating null byte. The allocated string is returned via the first
133 * argument and must be freed by the caller. However, unlike vasprintf(), this
134 * function calls exit() if insufficient memory is available, while vasprintf()
135 * returns -1 in this case.
137 * \return Number of bytes written, not including the terminating \p NULL
140 * \sa printf(3), vsnprintf(3), va_start(3), vasprintf(3), \ref xasprintf().
142 __printf_2_0 unsigned xvasprintf(char **result, const char *fmt, va_list ap)
148 *result = para_malloc(size + 1);
150 ret = vsnprintf(*result, size, fmt, aq);
153 if (ret < size) /* OK */
156 *result = para_realloc(*result, size);
158 ret = vsnprintf(*result, size, fmt, aq);
160 assert(ret >= 0 && ret < size);
165 * Print to a dynamically allocated string, variable number of arguments.
167 * \param result See \ref xvasprintf().
168 * \param fmt Usual format string.
170 * \return The return value of the underlying call to \ref xvasprintf().
172 * \sa \ref xvasprintf() and the references mentioned there.
174 __printf_2_3 unsigned xasprintf(char **result, const char *fmt, ...)
180 ret = xvasprintf(result, fmt, ap);
186 * Allocate a sufficiently large string and print into it.
188 * \param fmt A usual format string.
190 * Produce output according to \p fmt. No artificial bound on the length of the
191 * resulting string is imposed.
193 * \return This function either returns a pointer to a string that must be
194 * freed by the caller or aborts without returning.
196 * \sa printf(3), xasprintf().
198 __must_check __printf_1_2 __malloc char *make_message(const char *fmt, ...)
204 xvasprintf(&msg, fmt, ap);
210 * Free the content of a pointer and set it to \p NULL.
212 * This is equivalent to "free(*arg); *arg = NULL;".
214 * \param arg The pointer whose content should be freed.
216 void freep(void *arg)
218 void **ptr = (void **)arg;
224 * Paraslash's version of strcat().
226 * \param a String to be appended to.
227 * \param b String to append.
229 * Append \p b to \p a.
231 * \return If \a a is \p NULL, return a pointer to a copy of \a b, i.e.
232 * para_strcat(NULL, b) is equivalent to para_strdup(b). If \a b is \p NULL,
233 * return \a a without making a copy of \a a. Otherwise, construct the
234 * concatenation \a c, free \a a (but not \a b) and return \a c.
238 __must_check __malloc char *para_strcat(char *a, const char *b)
243 return para_strdup(b);
246 tmp = make_message("%s%s", a, b);
252 * Paraslash's version of dirname().
254 * \param name Pointer to the full path.
256 * Compute the directory component of \p name.
258 * \return If \a name is \p NULL or the empty string, return \p NULL.
259 * Otherwise, Make a copy of \a name and return its directory component. Caller
260 * is responsible to free the result.
262 __must_check __malloc char *para_dirname(const char *name)
268 ret = para_strdup(name);
269 p = strrchr(ret, '/');
278 * Paraslash's version of basename().
280 * \param name Pointer to the full path.
282 * Compute the filename component of \a name.
284 * \return \p NULL if (a) \a name is the empty string or \p NULL, or (b) name
285 * ends with a slash. Otherwise, a pointer within \a name is returned. Caller
286 * must not free the result.
288 __must_check char *para_basename(const char *name)
294 ret = strrchr(name, '/');
302 * Cut trailing newline.
304 * \param buf The string to be chopped.
306 * Replace the last character in \p buf by zero if it is equal to
307 * the newline character.
315 if (buf[n - 1] == '\n')
320 * Get the logname of the current user.
322 * \return A dynamically allocated string that must be freed by the caller. On
323 * errors, the string "unknown_user" is returned, i.e. this function never
328 __must_check __malloc char *para_logname(void)
330 struct passwd *pw = getpwuid(getuid());
331 return para_strdup(pw? pw->pw_name : "unknown_user");
335 * Get the home directory of the current user.
337 * \return A dynamically allocated string that must be freed by the caller. If
338 * the home directory could not be found, this function returns "/tmp".
340 __must_check __malloc char *para_homedir(void)
342 struct passwd *pw = getpwuid(getuid());
343 return para_strdup(pw? pw->pw_dir : "/tmp");
347 * Get the own hostname.
349 * \return A dynamically allocated string containing the hostname.
353 __malloc char *para_hostname(void)
358 return para_strdup(u.nodename);
362 * Call a custom function for each complete line.
364 * \param flags Any combination of flags defined in \ref for_each_line_flags.
365 * \param buf The buffer containing data separated by newlines.
366 * \param size The number of bytes in \a buf.
367 * \param line_handler The custom function.
368 * \param private_data Pointer passed to \a line_handler.
370 * For each complete line in \p buf, \p line_handler is called. The first
371 * argument to \p line_handler is (a copy of) the current line, and \p
372 * private_data is passed as the second argument. If the \p FELF_READ_ONLY
373 * flag is unset, a pointer into \a buf is passed to the line handler,
374 * otherwise a pointer to a copy of the current line is passed instead. This
375 * copy is freed immediately after the line handler returns.
377 * The function returns if \p line_handler returns a negative value or no more
378 * lines are in the buffer. The rest of the buffer (last chunk containing an
379 * incomplete line) is moved to the beginning of the buffer if FELF_READ_ONLY is
382 * \return On success this function returns the number of bytes not handled to
383 * \p line_handler. The only possible error is a negative return value from the
384 * line handler. In this case processing stops and the return value of the line
385 * handler is returned to indicate failure.
387 * \sa \ref for_each_line_flags.
389 int for_each_line(unsigned flags, char *buf, size_t size,
390 line_handler_t *line_handler, void *private_data)
392 char *start = buf, *end;
393 int ret, i, num_lines = 0;
395 // PARA_NOTICE_LOG("buf: %s\n", buf);
396 while (start < buf + size) {
400 next_cr = memchr(start, '\n', buf + size - start);
401 next_null = memchr(start, '\0', buf + size - start);
402 if (!next_cr && !next_null)
404 if (next_cr && next_null) {
405 end = next_cr < next_null? next_cr : next_null;
406 } else if (next_null) {
411 if (!(flags & FELF_DISCARD_FIRST) || start != buf) {
412 if (flags & FELF_READ_ONLY) {
413 size_t s = end - start;
414 char *b = para_malloc(s + 1);
417 ret = line_handler(b, private_data);
421 ret = line_handler(start, private_data);
428 i = buf + size - start;
429 if (i && i != size && !(flags & FELF_READ_ONLY))
430 memmove(buf, start, i);
434 /** Return the hex characters of the lower 4 bits. */
435 #define hex(a) (hexchar[(a) & 15])
437 static void write_size_header(char *buf, int n)
439 static char hexchar[] = "0123456789abcdef";
441 buf[0] = hex(n >> 12);
442 buf[1] = hex(n >> 8);
443 buf[2] = hex(n >> 4);
449 * Read a four-byte hex-number and return its value.
451 * Each status item sent by para_server is prefixed with such a hex number in
452 * ASCII which describes the size of the status item.
454 * \param buf The buffer which must be at least four bytes long.
456 * \return The value of the hex number on success, \p -E_SIZE_PREFIX if the
457 * buffer did not contain only hex digits.
459 int read_size_header(const char *buf)
463 for (i = 0; i < 4; i++) {
464 unsigned char c = buf[i];
466 if (c >= '0' && c <= '9') {
470 if (c >= 'a' && c <= 'f') {
474 return -E_SIZE_PREFIX;
477 return -E_SIZE_PREFIX;
482 * Safely print into a buffer at a given offset.
484 * \param b Determines the buffer, its size, and the offset.
485 * \param fmt The format string.
487 * This function prints into the buffer given by \a b at the offset which is
488 * also given by \a b. If there is not enough space to hold the result, the
489 * buffer size is doubled until the underlying call to vsnprintf() succeeds
490 * or the size of the buffer exceeds the maximal size specified in \a b.
492 * In the latter case the unmodified \a buf and \a offset values as well as the
493 * private_data pointer of \a b are passed to the \a max_size_handler of \a b.
494 * If this function succeeds, i.e. returns a non-negative value, the offset of
495 * \a b is reset to zero and the given data is written to the beginning of the
496 * buffer. If \a max_size_handler() returns a negative value, this value is
497 * returned by \a para_printf().
499 * Upon return, the offset of \a b is adjusted accordingly so that subsequent
500 * calls to this function append data to what is already contained in the
503 * It's OK to call this function with \p b->buf being \p NULL. In this case, an
504 * initial buffer is allocated.
506 * \return The number of bytes printed into the buffer (not including the
507 * terminating \p NULL byte) on success, negative on errors. If there is no
508 * size-bound on \a b, i.e. if \p b->max_size is zero, this function never
511 * \sa make_message(), vsnprintf(3).
513 __printf_2_3 int para_printf(struct para_buffer *b, const char *fmt, ...)
515 int ret, sz_off = (b->flags & PBF_SIZE_PREFIX)? 5 : 0;
518 b->buf = para_malloc(128);
523 char *p = b->buf + b->offset;
524 size_t size = b->size - b->offset;
529 ret = vsnprintf(p + sz_off, size - sz_off, fmt, ap);
531 if (ret > -1 && ret < size - sz_off) { /* success */
532 b->offset += ret + sz_off;
534 write_size_header(p, ret);
538 /* check if we may grow the buffer */
539 if (!b->max_size || 2 * b->size < b->max_size) { /* yes */
540 /* try again with more space */
542 b->buf = para_realloc(b->buf, b->size);
545 /* can't grow buffer */
546 if (!b->offset || !b->max_size_handler) /* message too large */
547 return -ERRNO_TO_PARA_ERROR(ENOSPC);
548 ret = b->max_size_handler(b->buf, b->offset, b->private_data);
555 /** \cond llong_minmax */
556 /* LLONG_MAX and LLONG_MIN might not be defined. */
558 #define LLONG_MAX 9223372036854775807LL
561 #define LLONG_MIN (-LLONG_MAX - 1LL)
563 /** \endcond llong_minmax */
566 * Convert a string to a 64-bit signed integer value.
568 * \param str The string to be converted.
569 * \param value Result pointer.
573 * \sa para_atoi32(), strtol(3), atoi(3).
575 int para_atoi64(const char *str, int64_t *value)
580 errno = 0; /* To distinguish success/failure after call */
581 tmp = strtoll(str, &endptr, 10);
582 if (errno == ERANGE && (tmp == LLONG_MAX || tmp == LLONG_MIN))
583 return -E_ATOI_OVERFLOW;
584 if (errno != 0 && tmp == 0) /* other error */
587 return -E_ATOI_NO_DIGITS;
588 if (*endptr != '\0') /* Further characters after number */
589 return -E_ATOI_JUNK_AT_END;
595 * Convert a string to a 32-bit signed integer value.
597 * \param str The string to be converted.
598 * \param value Result pointer.
604 int para_atoi32(const char *str, int32_t *value)
608 const int32_t max = 2147483647;
610 ret = para_atoi64(str, &tmp);
613 if (tmp > max || tmp < -max - 1)
614 return -E_ATOI_OVERFLOW;
619 static inline int loglevel_equal(const char *arg, const char * const ll)
621 return !strncasecmp(arg, ll, strlen(ll));
625 * Compute the loglevel number from its name.
627 * \param txt The name of the loglevel (debug, info, ...).
629 * \return The numeric representation of the loglevel name.
631 int get_loglevel_by_name(const char *txt)
633 if (loglevel_equal(txt, "debug"))
635 if (loglevel_equal(txt, "info"))
637 if (loglevel_equal(txt, "notice"))
639 if (loglevel_equal(txt, "warning"))
641 if (loglevel_equal(txt, "error"))
643 if (loglevel_equal(txt, "crit"))
645 if (loglevel_equal(txt, "emerg"))
650 static int get_next_word(const char *buf, const char *delim, char **word)
652 enum line_state_flags {LSF_HAVE_WORD = 1, LSF_BACKSLASH = 2,
653 LSF_SINGLE_QUOTE = 4, LSF_DOUBLE_QUOTE = 8};
658 out = para_malloc(strlen(buf) + 1);
661 for (in = buf; *in; in++) {
666 if (state & LSF_BACKSLASH) /* \\ */
668 state |= LSF_BACKSLASH;
669 state |= LSF_HAVE_WORD;
673 if (state & LSF_BACKSLASH) { /* \n or \t */
674 *out++ = (*in == 'n')? '\n' : '\t';
675 state &= ~LSF_BACKSLASH;
680 if (state & LSF_BACKSLASH) /* \" */
682 if (state & LSF_SINGLE_QUOTE) /* '" */
684 if (state & LSF_DOUBLE_QUOTE) {
685 state &= ~LSF_DOUBLE_QUOTE;
688 state |= LSF_HAVE_WORD;
689 state |= LSF_DOUBLE_QUOTE;
692 if (state & LSF_BACKSLASH) /* \' */
694 if (state & LSF_DOUBLE_QUOTE) /* "' */
696 if (state & LSF_SINGLE_QUOTE) {
697 state &= ~LSF_SINGLE_QUOTE;
700 state |= LSF_HAVE_WORD;
701 state |= LSF_SINGLE_QUOTE;
704 for (p = delim; *p; p++) {
707 if (state & LSF_BACKSLASH)
709 if (state & LSF_SINGLE_QUOTE)
711 if (state & LSF_DOUBLE_QUOTE)
713 if (state & LSF_HAVE_WORD)
717 if (*p) /* ignore delimiter at the beginning */
720 state |= LSF_HAVE_WORD;
722 state &= ~LSF_BACKSLASH;
725 if (!(state & LSF_HAVE_WORD))
727 ret = -ERRNO_TO_PARA_ERROR(EINVAL);
728 if (state & LSF_BACKSLASH) {
729 PARA_ERROR_LOG("trailing backslash\n");
732 if ((state & LSF_SINGLE_QUOTE) || (state & LSF_DOUBLE_QUOTE)) {
733 PARA_ERROR_LOG("unmatched quote character\n");
746 * Get the number of the word the cursor is on.
748 * \param buf The zero-terminated line buffer.
749 * \param delim Characters that separate words.
750 * \param point The cursor position.
752 * \return Zero-based word number.
754 int compute_word_num(const char *buf, const char *delim, int point)
760 for (p = buf, num_words = 0; ; p += ret, num_words++) {
761 ret = get_next_word(p, delim, &word);
765 if (p + ret >= buf + point)
772 * Free an array of words created by create_argv() or create_shifted_argv().
774 * \param argv A pointer previously obtained by \ref create_argv().
776 void free_argv(char **argv)
782 for (i = 0; argv[i]; i++)
787 static int create_argv_offset(int offset, const char *buf, const char *delim,
790 char *word, **argv = para_malloc((offset + 1) * sizeof(char *));
794 for (i = 0; i < offset; i++)
796 for (p = buf; p && *p; p += ret, i++) {
797 ret = get_next_word(p, delim, &word);
802 argv = para_realloc(argv, (i + 2) * sizeof(char*));
817 * Split a buffer into words.
819 * This parser honors single and double quotes, backslash-escaped characters
820 * and special characters like \p \\n. The result contains pointers to copies
821 * of the words contained in \a buf and has to be freed by using \ref
824 * \param buf The buffer to be split.
825 * \param delim Each character in this string is treated as a separator.
826 * \param result The array of words is returned here.
828 * \return Number of words in \a buf, negative on errors.
830 int create_argv(const char *buf, const char *delim, char ***result)
832 return create_argv_offset(0, buf, delim, result);
836 * Split a buffer into words, offset one.
838 * This is similar to \ref create_argv() but the returned array is one element
839 * larger, words start at index one and element zero is initialized to \p NULL.
840 * Callers must set element zero to a non-NULL value before calling free_argv()
841 * on the returned array to avoid a memory leak.
843 * \param buf See \ref create_argv().
844 * \param delim See \ref create_argv().
845 * \param result See \ref create_argv().
847 * \return Number of words plus one on success, negative on errors.
849 int create_shifted_argv(const char *buf, const char *delim, char ***result)
851 return create_argv_offset(1, buf, delim, result);
855 * Find out if the given string is contained in the arg vector.
857 * \param arg The string to look for.
858 * \param argv The array to search.
860 * \return The first index whose value equals \a arg, or \p -E_ARG_NOT_FOUND if
861 * arg was not found in \a argv.
863 int find_arg(const char *arg, char **argv)
868 return -E_ARG_NOT_FOUND;
869 for (i = 0; argv[i]; i++)
870 if (strcmp(arg, argv[i]) == 0)
872 return -E_ARG_NOT_FOUND;
876 * Compile a regular expression.
878 * This simple wrapper calls regcomp() and logs a message on errors.
880 * \param preg See regcomp(3).
881 * \param regex See regcomp(3).
882 * \param cflags See regcomp(3).
886 int para_regcomp(regex_t *preg, const char *regex, int cflags)
890 int ret = regcomp(preg, regex, cflags);
894 size = regerror(ret, preg, NULL, 0);
895 buf = para_malloc(size);
896 regerror(ret, preg, buf, size);
897 PARA_ERROR_LOG("%s\n", buf);
903 * strdup() for not necessarily zero-terminated strings.
905 * \param src The source buffer.
906 * \param len The number of bytes to be copied.
908 * \return A 0-terminated buffer of length \a len + 1.
910 * This is similar to strndup(), which is a GNU extension. However, one
911 * difference is that strndup() returns \p NULL if insufficient memory was
912 * available while this function aborts in this case.
914 * \sa strdup(), \ref para_strdup().
916 char *safe_strdup(const char *src, size_t len)
920 assert(len < (size_t)-1);
921 p = para_malloc(len + 1);
929 * Copy the value of a key=value pair.
931 * This checks whether the given buffer starts with "key=", ignoring case. If
932 * yes, a copy of the value is returned. The source buffer may not be
935 * \param src The source buffer.
936 * \param len The number of bytes of the tag.
937 * \param key Only copy if it is the value of this key.
939 * \return A zero-terminated buffer, or \p NULL if the key was
940 * not of the given type.
942 char *key_value_copy(const char *src, size_t len, const char *key)
944 int keylen = strlen(key);
948 if (strncasecmp(src, key, keylen))
950 if (src[keylen] != '=')
952 return safe_strdup(src + keylen + 1, len - keylen - 1);
955 static bool utf8_mode(void)
957 static bool initialized, have_utf8;
960 char *info = nl_langinfo(CODESET);
961 have_utf8 = (info && strcmp(info, "UTF-8") == 0);
963 PARA_INFO_LOG("%susing UTF-8 character encoding\n",
964 have_utf8? "" : "not ");
970 * glibc's wcswidth returns -1 if the string contains a tab character, which
971 * makes the function next to useless. The two functions below are taken from
975 #define IsWPrint(wc) (iswprint(wc) || wc >= 0xa0)
977 static int mutt_wcwidth(wchar_t wc, size_t pos)
981 if (wc == 0x09) /* tab */
982 return (pos | 7) + 1 - pos;
984 if (IsWPrint(wc) && n > 0)
993 static size_t mutt_wcswidth(const wchar_t *s, size_t n)
998 w += mutt_wcwidth(*s++, w);
1003 * Skip a given number of cells at the beginning of a string.
1005 * \param s The input string.
1006 * \param cells_to_skip Desired number of cells that should be skipped.
1007 * \param bytes_to_skip Result.
1009 * This function computes how many input bytes must be skipped to advance a
1010 * string by the given width. If the current character encoding is not UTF-8,
1011 * this is simply the given number of cells, i.e. \a cells_to_skip. Otherwise,
1012 * \a s is treated as a multibyte string and on successful return, \a s +
1013 * bytes_to_skip points to the start of a multibyte string such that the total
1014 * width of the multibyte characters that are skipped by advancing \a s that
1015 * many bytes equals at least \a cells_to_skip.
1019 int skip_cells(const char *s, size_t cells_to_skip, size_t *bytes_to_skip)
1023 size_t n, bytes_parsed, cells_skipped;
1026 if (cells_to_skip == 0)
1029 *bytes_to_skip = cells_to_skip;
1032 bytes_parsed = cells_skipped = 0;
1033 memset(&ps, 0, sizeof(ps));
1035 while (cells_to_skip > cells_skipped) {
1038 mbret = mbrtowc(&wc, s + bytes_parsed, n - bytes_parsed, &ps);
1040 if (mbret == (size_t)-1 || mbret == (size_t)-2)
1041 return -ERRNO_TO_PARA_ERROR(EILSEQ);
1042 bytes_parsed += mbret;
1043 cells_skipped += mutt_wcwidth(wc, cells_skipped);
1045 *bytes_to_skip = bytes_parsed;
1050 * Compute the width of an UTF-8 string.
1052 * \param s The string.
1053 * \param result The width of \a s is returned here.
1055 * If not in UTF8-mode. this function is just a wrapper for strlen(3).
1056 * Otherwise \a s is treated as an UTF-8 string and its display width is
1057 * computed. Note that this function may fail if the underlying call to
1058 * mbsrtowcs(3) fails, so the caller must check the return value.
1060 * \sa nl_langinfo(3), wcswidth(3).
1064 __must_check int strwidth(const char *s, size_t *result)
1066 const char *src = s;
1068 static wchar_t *dest;
1072 * Never call any log function here. This may result in an endless loop
1073 * as para_gui's para_log() calls this function.
1077 *result = strlen(s);
1080 memset(&state, 0, sizeof(state));
1082 num_wchars = mbsrtowcs(NULL, &src, 0, &state);
1083 if (num_wchars == (size_t)-1)
1084 return -ERRNO_TO_PARA_ERROR(errno);
1085 if (num_wchars == 0)
1087 dest = para_malloc(num_wchars * sizeof(*dest));
1089 memset(&state, 0, sizeof(state));
1090 num_wchars = mbsrtowcs(dest, &src, num_wchars, &state);
1091 assert(num_wchars > 0 && num_wchars != (size_t)-1);
1092 *result = mutt_wcswidth(dest, num_wchars);