Merge commit 'fml/master'
[paraslash.git] / string.c
1 /*
2 * Copyright (C) 2004-2008 Andre Noll <maan@systemlinux.org>
3 *
4 * Licensed under the GPL v2. For licencing details see COPYING.
5 */
6
7 /** \file string.c Memory allocation and string handling functions. */
8
9 #include "para.h"
10 #include "string.h"
11
12 #include <sys/time.h> /* gettimeofday */
13 #include <pwd.h>
14 #include <sys/utsname.h> /* uname() */
15 #include <string.h>
16
17 #include "error.h"
18
19 /**
20 * Paraslash's version of realloc().
21 *
22 * \param p Pointer to the memory block, may be \p NULL.
23 * \param size The desired new size.
24 *
25 * A wrapper for realloc(3). It calls \p exit(\p EXIT_FAILURE) on errors,
26 * i.e. there is no need to check the return value in the caller.
27 *
28 * \return A pointer to the newly allocated memory, which is suitably aligned
29 * for any kind of variable and may be different from \a p.
30 *
31 * \sa realloc(3).
32 */
33 __must_check __malloc void *para_realloc(void *p, size_t size)
34 {
35 /*
36 * No need to check for NULL pointers: If p is NULL, the call
37 * to realloc is equivalent to malloc(size)
38 */
39 assert(size);
40 if (!(p = realloc(p, size))) {
41 PARA_EMERG_LOG("realloc failed (size = %zu), aborting\n",
42 size);
43 exit(EXIT_FAILURE);
44 }
45 return p;
46 }
47
48 /**
49 * Paraslash's version of malloc().
50 *
51 * \param size The desired new size.
52 *
53 * A wrapper for malloc(3) which exits on errors.
54 *
55 * \return A pointer to the allocated memory, which is suitably aligned for any
56 * kind of variable.
57 *
58 * \sa malloc(3).
59 */
60 __must_check __malloc void *para_malloc(size_t size)
61 {
62 void *p;
63
64 assert(size);
65 p = malloc(size);
66 if (!p) {
67 PARA_EMERG_LOG("malloc failed (size = %zu), aborting\n",
68 size);
69 exit(EXIT_FAILURE);
70 }
71 return p;
72 }
73
74 /**
75 * Paraslash's version of calloc().
76 *
77 * \param size The desired new size.
78 *
79 * A wrapper for calloc(3) which exits on errors.
80 *
81 * \return A pointer to the allocated and zeroed-out memory, which is suitably
82 * aligned for any kind of variable.
83 *
84 * \sa calloc(3)
85 */
86 __must_check __malloc void *para_calloc(size_t size)
87 {
88 void *ret = para_malloc(size);
89
90 memset(ret, 0, size);
91 return ret;
92 }
93
94 /**
95 * Paraslash's version of strdup().
96 *
97 * \param s The string to be duplicated.
98 *
99 * A wrapper for strdup(3). It calls \p exit(EXIT_FAILURE) on errors, i.e.
100 * there is no need to check the return value in the caller.
101 *
102 * \return A pointer to the duplicated string. If \p s was the NULL pointer,
103 * an pointer to an empty string is returned.
104 *
105 * \sa strdup(3)
106 */
107 __must_check __malloc char *para_strdup(const char *s)
108 {
109 char *ret;
110
111 if ((ret = strdup(s? s: "")))
112 return ret;
113 PARA_EMERG_LOG("strdup failed, aborting\n");
114 exit(EXIT_FAILURE);
115 }
116
117 /**
118 * Allocate a sufficiently large string and print into it.
119 *
120 * \param fmt A usual format string.
121 *
122 * Produce output according to \p fmt. No artificial bound on the length of the
123 * resulting string is imposed.
124 *
125 * \return This function either returns a pointer to a string that must be
126 * freed by the caller or aborts without returning.
127 *
128 * \sa printf(3).
129 */
130 __must_check __printf_1_2 __malloc char *make_message(const char *fmt, ...)
131 {
132 char *msg;
133
134 PARA_VSPRINTF(fmt, msg);
135 return msg;
136 }
137
138 /**
139 * Paraslash's version of strcat().
140 *
141 * \param a String to be appended to.
142 * \param b String to append.
143 *
144 * Append \p b to \p a.
145 *
146 * \return If \a a is \p NULL, return a pointer to a copy of \a b, i.e.
147 * para_strcat(NULL, b) is equivalent to para_strdup(b). If \a b is \p NULL,
148 * return \a a without making a copy of \a a. Otherwise, construct the
149 * concatenation \a c, free \a a (but not \a b) and return \a c.
150 *
151 * \sa strcat(3)
152 */
153 __must_check __malloc char *para_strcat(char *a, const char *b)
154 {
155 char *tmp;
156
157 if (!a)
158 return para_strdup(b);
159 if (!b)
160 return a;
161 tmp = make_message("%s%s", a, b);
162 free(a);
163 return tmp;
164 }
165
166 /**
167 * Paraslash's version of dirname().
168 *
169 * \param name Pointer to the full path.
170 *
171 * Compute the directory component of \p name.
172 *
173 * \return If \a name is \p NULL or the empty string, return \p NULL.
174 * Otherwise, Make a copy of \a name and return its directory component. Caller
175 * is responsible to free the result.
176 */
177 __must_check __malloc char *para_dirname(const char *name)
178 {
179 char *p, *ret;
180
181 if (!name || !*name)
182 return NULL;
183 ret = para_strdup(name);
184 p = strrchr(ret, '/');
185 if (!p)
186 *ret = '\0';
187 else
188 *p = '\0';
189 return ret;
190 }
191
192 /**
193 * Paraslash's version of basename().
194 *
195 * \param name Pointer to the full path.
196 *
197 * Compute the filename component of \a name.
198 *
199 * \return \p NULL if (a) \a name is the empty string or \p NULL, or (b) name
200 * ends with a slash. Otherwise, a pointer within \a name is returned. Caller
201 * must not free the result.
202 */
203 __must_check const char *para_basename(const char *name)
204 {
205 const char *ret;
206
207 if (!name || !*name)
208 return NULL;
209 ret = strrchr(name, '/');
210 if (!ret)
211 return name;
212 ret++;
213 return ret;
214 }
215
216 /**
217 * Cut trailing newline.
218 *
219 * \param buf The string to be chopped.
220 *
221 * Replace the last character in \p buf by zero if it is euqal to
222 * the newline character.
223 */
224 void chop(char *buf)
225 {
226 int n = strlen(buf);
227 if (!n)
228 return;
229 if (buf[n - 1] == '\n')
230 buf[n - 1] = '\0';
231 }
232
233 /**
234 * Get a random filename.
235 *
236 * This is by no means a secure way to create temporary files in a hostile
237 * direcory like \p /tmp. However, it is OK to use for temp files, fifos,
238 * sockets that are created in ~/.paraslash. Result must be freed by the
239 * caller.
240 *
241 * \return A pointer to a random filename.
242 */
243 __must_check __malloc char *para_tmpname(void)
244 {
245 struct timeval now;
246 unsigned int seed;
247
248 gettimeofday(&now, NULL);
249 seed = now.tv_usec;
250 srand(seed);
251 return make_message("%08i", rand());
252 }
253
254 /**
255 * Get the logname of the current user.
256 *
257 * \return A dynammically allocated string that must be freed by the caller. On
258 * errors, the string "unknown user" is returned, i.e. this function never
259 * returns \p NULL.
260 *
261 * \sa getpwuid(3).
262 */
263 __must_check __malloc char *para_logname(void)
264 {
265 struct passwd *pw = getpwuid(getuid());
266 return para_strdup(pw? pw->pw_name : "unknown_user");
267 }
268
269 /**
270 * Get the home directory of the current user.
271 *
272 * \return A dynammically allocated string that must be freed by the caller. If
273 * the home directory could not be found, this function returns "/tmp".
274 */
275 __must_check __malloc char *para_homedir(void)
276 {
277 struct passwd *pw = getpwuid(getuid());
278 return para_strdup(pw? pw->pw_dir : "/tmp");
279 }
280
281 /**
282 * Split string and return pointers to its parts.
283 *
284 * \param args The string to be split.
285 * \param argv_ptr Pointer to the list of substrings.
286 * \param delim Delimiter.
287 *
288 * This function modifies \a args by replacing each occurance of \a delim by
289 * zero. A \p NULL-terminated array of pointers to char* is allocated dynamically
290 * and these pointers are initialized to point to the broken-up substrings
291 * within \a args. A pointer to this array is returned via \a argv_ptr.
292 *
293 * \return The number of substrings found in \a args.
294 */
295 __must_check unsigned split_args(char *args, char *** const argv_ptr, const char *delim)
296 {
297 char *p = args;
298 char **argv;
299 size_t n = 0, i, j;
300
301 p = args + strspn(args, delim);
302 for (;;) {
303 i = strcspn(p, delim);
304 if (!i)
305 break;
306 p += i;
307 n++;
308 p += strspn(p, delim);
309 }
310 *argv_ptr = para_malloc((n + 1) * sizeof(char *));
311 argv = *argv_ptr;
312 i = 0;
313 p = args + strspn(args, delim);
314 while (p) {
315 argv[i] = p;
316 j = strcspn(p, delim);
317 if (!j)
318 break;
319 p += strcspn(p, delim);
320 if (*p) {
321 *p = '\0';
322 p++;
323 p += strspn(p, delim);
324 }
325 i++;
326 }
327 argv[n] = NULL;
328 return n;
329 }
330
331 /**
332 * Ensure that file descriptors 0, 1, and 2 are valid.
333 *
334 * Common approach that opens /dev/null until it gets a file descriptor greater
335 * than two.
336 *
337 * \sa okir's Black Hats Manual.
338 */
339 void valid_fd_012(void)
340 {
341 while (1) {
342 int fd = open("/dev/null", O_RDWR);
343 if (fd < 0)
344 exit(EXIT_FAILURE);
345 if (fd > 2) {
346 close(fd);
347 break;
348 }
349 }
350 }
351
352 /**
353 * Get the own hostname.
354 *
355 * \return A dynammically allocated string containing the hostname.
356 *
357 * \sa uname(2).
358 */
359 __malloc char *para_hostname(void)
360 {
361 struct utsname u;
362
363 uname(&u);
364 return para_strdup(u.nodename);
365 }
366
367 /**
368 * Used to distinguish between read-only and read-write mode.
369 *
370 * \sa for_each_line(), for_each_line_ro().
371 */
372 enum for_each_line_modes{
373 /** Activate read-only mode. */
374 LINE_MODE_RO,
375 /** Activate read-write mode. */
376 LINE_MODE_RW
377 };
378
379 static int for_each_complete_line(enum for_each_line_modes mode, char *buf,
380 size_t size, line_handler_t *line_handler, void *private_data)
381 {
382 char *start = buf, *end;
383 int ret, i, num_lines = 0;
384
385 // PARA_NOTICE_LOG("buf: %s\n", buf);
386 while (start < buf + size) {
387 char *next_null;
388 char *next_cr;
389
390 next_cr = memchr(start, '\n', buf + size - start);
391 next_null = memchr(start, '\0', buf + size - start);
392 if (!next_cr && !next_null)
393 break;
394 if (next_cr && next_null) {
395 end = next_cr < next_null? next_cr : next_null;
396 } else if (next_null) {
397 end = next_null;
398 } else
399 end = next_cr;
400 num_lines++;
401 if (!line_handler) {
402 start = ++end;
403 continue;
404 }
405 if (mode == LINE_MODE_RO) {
406 size_t s = end - start;
407 char *b = para_malloc(s + 1);
408 memcpy(b, start, s);
409 b[s] = '\0';
410 // PARA_NOTICE_LOG("b: %s, start: %s\n", b, start);
411 ret = line_handler(b, private_data);
412 free(b);
413 } else {
414 *end = '\0';
415 ret = line_handler(start, private_data);
416 }
417 if (ret < 0)
418 return ret;
419 start = ++end;
420 }
421 if (!line_handler || mode == LINE_MODE_RO)
422 return num_lines;
423 i = buf + size - start;
424 if (i && i != size)
425 memmove(buf, start, i);
426 return i;
427 }
428
429 /**
430 * Call a custom function for each complete line.
431 *
432 * \param buf The buffer containing data seperated by newlines.
433 * \param size The number of bytes in \a buf.
434 * \param line_handler The custom function.
435 * \param private_data Pointer passed to \a line_handler.
436 *
437 * If \p line_handler is \p NULL, the function returns the number of complete
438 * lines in \p buf. Otherwise, \p line_handler is called for each complete
439 * line in \p buf. The first argument to \p line_handler is the current line,
440 * and \p private_data is passed as the second argument. The function returns
441 * if \p line_handler returns a negative value or no more lines are in the
442 * buffer. The rest of the buffer (last chunk containing an incomplete line)
443 * is moved to the beginning of the buffer.
444 *
445 * \return If \p line_handler is not \p NULL, this function returns the number
446 * of bytes not handled to \p line_handler on success, or the negative return
447 * value of the \p line_handler on errors.
448 *
449 * \sa for_each_line_ro().
450 */
451 int for_each_line(char *buf, size_t size, line_handler_t *line_handler,
452 void *private_data)
453 {
454 return for_each_complete_line(LINE_MODE_RW, buf, size, line_handler,
455 private_data);
456 }
457
458 /**
459 * Call a custom function for each complete line.
460 *
461 * \param buf Same meaning as in \p for_each_line().
462 * \param size Same meaning as in \p for_each_line().
463 * \param line_handler Same meaning as in \p for_each_line().
464 * \param private_data Same meaning as in \p for_each_line().
465 *
466 * This function behaves like \p for_each_line(), but \a buf is left unchanged.
467 *
468 * \return On success, the function returns the number of complete lines in \p
469 * buf, otherwise the (negative) return value of \p line_handler is returned.
470 *
471 * \sa for_each_line().
472 */
473 int for_each_line_ro(char *buf, size_t size, line_handler_t *line_handler,
474 void *private_data)
475 {
476 return for_each_complete_line(LINE_MODE_RO, buf, size, line_handler,
477 private_data);
478 }
479
480 /**
481 * Safely print into a buffer at a given offset
482 *
483 * \param b Determines the buffer, its size, and the offset.
484 * \param fmt The format string.
485 *
486 * This function prints into the buffer given by \a b at the offset which is
487 * also given by \a b. If there is not enough space to hold the result, the
488 * buffer size is doubled until the underlying call to vsnprintf() succeeds
489 * or the size of the buffer exceeds the maximal size specified in \a pb.
490 *
491 * In the latter case the unmodified \a buf and \a offset values as well as the
492 * private_data pointer of \a b are passed to the \a max_size_handler of \a b.
493 * If this function succeeds, i.e. returns a non-negative value, the offset of
494 * \a b is reset to zero and the given data is written to the beginning of the
495 * buffer.
496 *
497 * Upon return, the offset of \a b is adjusted accordingly so that subsequent
498 * calls to this function append data to what is already contained in the
499 * buffer.
500 *
501 * It's OK to call this function with \p b->buf being \p NULL. In this case, an
502 * initial buffer is allocated.
503 *
504 * \return The number of bytes printed into the buffer (not including the
505 * therminating \p NULL byte).
506 *
507 * \sa make_message(), vsnprintf(3).
508 */
509 __printf_2_3 int para_printf(struct para_buffer *b, const char *fmt, ...)
510 {
511 int ret;
512
513 if (!b->buf) {
514 b->buf = para_malloc(128);
515 b->size = 128;
516 b->offset = 0;
517 }
518 while (1) {
519 char *p = b->buf + b->offset;
520 size_t size = b->size - b->offset;
521 va_list ap;
522 if (size) {
523 va_start(ap, fmt);
524 ret = vsnprintf(p, size, fmt, ap);
525 va_end(ap);
526 if (ret > -1 && ret < size) { /* success */
527 b->offset += ret;
528 return ret;
529 }
530 }
531 /* check if we may grow the buffer */
532 if (!b->max_size || 2 * b->size < b->max_size) { /* yes */
533 /* try again with more space */
534 b->size *= 2;
535 b->buf = para_realloc(b->buf, b->size);
536 continue;
537 }
538 /* can't grow buffer */
539 if (!b->offset || !b->max_size_handler) /* message too large */
540 return -ERRNO_TO_PARA_ERROR(ENOSPC);
541 ret = b->max_size_handler(b->buf, b->offset, b->private_data);
542 if (ret < 0)
543 return ret;
544 b->offset = 0;
545 }
546 }
547
548 /** \cond LLONG_MAX and LLONG_LIN might not be defined. */
549 #ifndef LLONG_MAX
550 #define LLONG_MAX (1 << (sizeof(long) - 1))
551 #endif
552 #ifndef LLONG_MIN
553 #define LLONG_MIN (-LLONG_MAX - 1LL)
554 #endif
555 /** \endcond */
556
557 /**
558 * Convert a string to a 64-bit signed integer value.
559 *
560 * \param str The string to be converted.
561 * \param value Result pointer.
562 *
563 * \return Standard.
564 *
565 * \sa para_atoi32(), strtol(3), atoi(3).
566 */
567 int para_atoi64(const char *str, int64_t *value)
568 {
569 char *endptr;
570 long long tmp;
571
572 errno = 0; /* To distinguish success/failure after call */
573 tmp = strtoll(str, &endptr, 10);
574 if (errno == ERANGE && (tmp == LLONG_MAX || tmp == LLONG_MIN))
575 return -E_ATOI_OVERFLOW;
576 if (errno != 0 && tmp == 0) /* other error */
577 return -E_STRTOLL;
578 if (endptr == str)
579 return -E_ATOI_NO_DIGITS;
580 if (*endptr != '\0') /* Further characters after number */
581 return -E_ATOI_JUNK_AT_END;
582 *value = tmp;
583 return 1;
584 }
585
586 /**
587 * Convert a string to a 32-bit signed integer value.
588 *
589 * \param str The string to be converted.
590 * \param value Result pointer.
591 *
592 * \return Standard.
593 *
594 * \sa para_atoi64().
595 */
596 int para_atoi32(const char *str, int32_t *value)
597 {
598 int64_t tmp;
599 int ret;
600 const int32_t max = 2147483647;
601
602 ret = para_atoi64(str, &tmp);
603 if (ret < 0)
604 return ret;
605 if (tmp > max || tmp < -max - 1)
606 return -E_ATOI_OVERFLOW;
607 *value = tmp;
608 return 1;
609 }