Commit | Line | Data |
---|---|---|
059ec3d9 PH |
1 | /************************************************* |
2 | * Exim - an Internet mail transport agent * | |
3 | *************************************************/ | |
4 | ||
f9ba5e22 | 5 | /* Copyright (c) University of Cambridge 1995 - 2018 */ |
059ec3d9 PH |
6 | /* See the file NOTICE for conditions of use and distribution. */ |
7 | ||
8 | /* Miscellaneous string-handling functions. Some are not required for | |
9 | utilities and tests, and are cut out by the COMPILE_UTILITY macro. */ | |
10 | ||
11 | ||
12 | #include "exim.h" | |
d12746bc | 13 | #include <assert.h> |
059ec3d9 PH |
14 | |
15 | ||
16 | #ifndef COMPILE_UTILITY | |
17 | /************************************************* | |
18 | * Test for IP address * | |
19 | *************************************************/ | |
20 | ||
21 | /* This used just to be a regular expression, but with IPv6 things are a bit | |
22 | more complicated. If the address contains a colon, it is assumed to be a v6 | |
23 | address (assuming HAVE_IPV6 is set). If a mask is permitted and one is present, | |
24 | and maskptr is not NULL, its offset is placed there. | |
25 | ||
26 | Arguments: | |
27 | s a string | |
28 | maskptr NULL if no mask is permitted to follow | |
29 | otherwise, points to an int where the offset of '/' is placed | |
1688f43b | 30 | if there is no / followed by trailing digits, *maskptr is set 0 |
059ec3d9 PH |
31 | |
32 | Returns: 0 if the string is not a textual representation of an IP address | |
33 | 4 if it is an IPv4 address | |
34 | 6 if it is an IPv6 address | |
35 | */ | |
36 | ||
37 | int | |
b1f8e4f8 | 38 | string_is_ip_address(const uschar *s, int *maskptr) |
059ec3d9 | 39 | { |
059ec3d9 PH |
40 | int yield = 4; |
41 | ||
42 | /* If an optional mask is permitted, check for it. If found, pass back the | |
43 | offset. */ | |
44 | ||
8d909960 | 45 | if (maskptr) |
059ec3d9 | 46 | { |
b1f8e4f8 | 47 | const uschar *ss = s + Ustrlen(s); |
059ec3d9 PH |
48 | *maskptr = 0; |
49 | if (s != ss && isdigit(*(--ss))) | |
50 | { | |
51 | while (ss > s && isdigit(ss[-1])) ss--; | |
52 | if (ss > s && *(--ss) == '/') *maskptr = ss - s; | |
53 | } | |
54 | } | |
55 | ||
56 | /* A colon anywhere in the string => IPv6 address */ | |
57 | ||
58 | if (Ustrchr(s, ':') != NULL) | |
59 | { | |
60 | BOOL had_double_colon = FALSE; | |
61 | BOOL v4end = FALSE; | |
059ec3d9 PH |
62 | |
63 | yield = 6; | |
64 | ||
65 | /* An IPv6 address must start with hex digit or double colon. A single | |
66 | colon is invalid. */ | |
67 | ||
68 | if (*s == ':' && *(++s) != ':') return 0; | |
69 | ||
70 | /* Now read up to 8 components consisting of up to 4 hex digits each. There | |
71 | may be one and only one appearance of double colon, which implies any number | |
72 | of binary zero bits. The number of preceding components is held in count. */ | |
73 | ||
d7978c0f | 74 | for (int count = 0; count < 8; count++) |
059ec3d9 PH |
75 | { |
76 | /* If the end of the string is reached before reading 8 components, the | |
77 | address is valid provided a double colon has been read. This also applies | |
78 | if we hit the / that introduces a mask or the % that introduces the | |
79 | interface specifier (scope id) of a link-local address. */ | |
80 | ||
8d909960 | 81 | if (*s == 0 || *s == '%' || *s == '/') return had_double_colon ? yield : 0; |
059ec3d9 PH |
82 | |
83 | /* If a component starts with an additional colon, we have hit a double | |
84 | colon. This is permitted to appear once only, and counts as at least | |
85 | one component. The final component may be of this form. */ | |
86 | ||
87 | if (*s == ':') | |
88 | { | |
89 | if (had_double_colon) return 0; | |
90 | had_double_colon = TRUE; | |
91 | s++; | |
92 | continue; | |
93 | } | |
94 | ||
95 | /* If the remainder of the string contains a dot but no colons, we | |
96 | can expect a trailing IPv4 address. This is valid if either there has | |
97 | been no double-colon and this is the 7th component (with the IPv4 address | |
98 | being the 7th & 8th components), OR if there has been a double-colon | |
99 | and fewer than 6 components. */ | |
100 | ||
101 | if (Ustrchr(s, ':') == NULL && Ustrchr(s, '.') != NULL) | |
102 | { | |
103 | if ((!had_double_colon && count != 6) || | |
104 | (had_double_colon && count > 6)) return 0; | |
105 | v4end = TRUE; | |
106 | yield = 6; | |
107 | break; | |
108 | } | |
109 | ||
110 | /* Check for at least one and not more than 4 hex digits for this | |
111 | component. */ | |
112 | ||
113 | if (!isxdigit(*s++)) return 0; | |
114 | if (isxdigit(*s) && isxdigit(*(++s)) && isxdigit(*(++s))) s++; | |
115 | ||
116 | /* If the component is terminated by colon and there is more to | |
117 | follow, skip over the colon. If there is no more to follow the address is | |
118 | invalid. */ | |
119 | ||
120 | if (*s == ':' && *(++s) == 0) return 0; | |
121 | } | |
122 | ||
123 | /* If about to handle a trailing IPv4 address, drop through. Otherwise | |
124 | all is well if we are at the end of the string or at the mask or at a percent | |
125 | sign, which introduces the interface specifier (scope id) of a link local | |
126 | address. */ | |
127 | ||
1688f43b PH |
128 | if (!v4end) |
129 | return (*s == 0 || *s == '%' || | |
130 | (*s == '/' && maskptr != NULL && *maskptr != 0))? yield : 0; | |
059ec3d9 PH |
131 | } |
132 | ||
133 | /* Test for IPv4 address, which may be the tail-end of an IPv6 address. */ | |
134 | ||
d7978c0f | 135 | for (int i = 0; i < 4; i++) |
059ec3d9 | 136 | { |
8d909960 JH |
137 | long n; |
138 | uschar * end; | |
139 | ||
059ec3d9 | 140 | if (i != 0 && *s++ != '.') return 0; |
8d909960 JH |
141 | n = strtol(CCS s, CSS &end, 10); |
142 | if (n > 255 || n < 0 || end <= s || end > s+3) return 0; | |
143 | s = end; | |
059ec3d9 PH |
144 | } |
145 | ||
8d909960 | 146 | return !*s || (*s == '/' && maskptr && *maskptr != 0) ? yield : 0; |
059ec3d9 PH |
147 | } |
148 | #endif /* COMPILE_UTILITY */ | |
149 | ||
150 | ||
151 | /************************************************* | |
152 | * Format message size * | |
153 | *************************************************/ | |
154 | ||
155 | /* Convert a message size in bytes to printing form, rounding | |
156 | according to the magnitude of the number. A value of zero causes | |
157 | a string of spaces to be returned. | |
158 | ||
159 | Arguments: | |
160 | size the message size in bytes | |
161 | buffer where to put the answer | |
162 | ||
163 | Returns: pointer to the buffer | |
164 | a string of exactly 5 characters is normally returned | |
165 | */ | |
166 | ||
167 | uschar * | |
168 | string_format_size(int size, uschar *buffer) | |
169 | { | |
45500060 | 170 | if (size == 0) Ustrcpy(buffer, " "); |
059ec3d9 PH |
171 | else if (size < 1024) sprintf(CS buffer, "%5d", size); |
172 | else if (size < 10*1024) | |
173 | sprintf(CS buffer, "%4.1fK", (double)size / 1024.0); | |
174 | else if (size < 1024*1024) | |
175 | sprintf(CS buffer, "%4dK", (size + 512)/1024); | |
176 | else if (size < 10*1024*1024) | |
177 | sprintf(CS buffer, "%4.1fM", (double)size / (1024.0 * 1024.0)); | |
178 | else | |
179 | sprintf(CS buffer, "%4dM", (size + 512 * 1024)/(1024*1024)); | |
180 | return buffer; | |
181 | } | |
182 | ||
183 | ||
184 | ||
185 | #ifndef COMPILE_UTILITY | |
186 | /************************************************* | |
187 | * Convert a number to base 62 format * | |
188 | *************************************************/ | |
189 | ||
190 | /* Convert a long integer into an ASCII base 62 string. For Cygwin the value of | |
191 | BASE_62 is actually 36. Always return exactly 6 characters plus zero, in a | |
192 | static area. | |
193 | ||
194 | Argument: a long integer | |
195 | Returns: pointer to base 62 string | |
196 | */ | |
197 | ||
198 | uschar * | |
199 | string_base62(unsigned long int value) | |
200 | { | |
201 | static uschar yield[7]; | |
202 | uschar *p = yield + sizeof(yield) - 1; | |
203 | *p = 0; | |
204 | while (p > yield) | |
205 | { | |
206 | *(--p) = base62_chars[value % BASE_62]; | |
207 | value /= BASE_62; | |
208 | } | |
209 | return yield; | |
210 | } | |
211 | #endif /* COMPILE_UTILITY */ | |
212 | ||
213 | ||
214 | ||
059ec3d9 PH |
215 | /************************************************* |
216 | * Interpret escape sequence * | |
217 | *************************************************/ | |
218 | ||
219 | /* This function is called from several places where escape sequences are to be | |
220 | interpreted in strings. | |
221 | ||
222 | Arguments: | |
223 | pp points a pointer to the initiating "\" in the string; | |
224 | the pointer gets updated to point to the final character | |
225 | Returns: the value of the character escape | |
226 | */ | |
227 | ||
228 | int | |
55414b25 | 229 | string_interpret_escape(const uschar **pp) |
059ec3d9 | 230 | { |
3fb3c68d JH |
231 | #ifdef COMPILE_UTILITY |
232 | const uschar *hex_digits= CUS"0123456789abcdef"; | |
233 | #endif | |
059ec3d9 | 234 | int ch; |
55414b25 | 235 | const uschar *p = *pp; |
059ec3d9 PH |
236 | ch = *(++p); |
237 | if (isdigit(ch) && ch != '8' && ch != '9') | |
238 | { | |
239 | ch -= '0'; | |
240 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
241 | { | |
242 | ch = ch * 8 + *(++p) - '0'; | |
243 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
244 | ch = ch * 8 + *(++p) - '0'; | |
245 | } | |
246 | } | |
247 | else switch(ch) | |
248 | { | |
c7396ac5 PP |
249 | case 'b': ch = '\b'; break; |
250 | case 'f': ch = '\f'; break; | |
059ec3d9 PH |
251 | case 'n': ch = '\n'; break; |
252 | case 'r': ch = '\r'; break; | |
253 | case 't': ch = '\t'; break; | |
c7396ac5 | 254 | case 'v': ch = '\v'; break; |
059ec3d9 PH |
255 | case 'x': |
256 | ch = 0; | |
257 | if (isxdigit(p[1])) | |
258 | { | |
259 | ch = ch * 16 + | |
260 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
261 | if (isxdigit(p[1])) ch = ch * 16 + | |
262 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
263 | } | |
264 | break; | |
265 | } | |
266 | *pp = p; | |
267 | return ch; | |
268 | } | |
059ec3d9 PH |
269 | |
270 | ||
271 | ||
272 | #ifndef COMPILE_UTILITY | |
273 | /************************************************* | |
274 | * Ensure string is printable * | |
275 | *************************************************/ | |
276 | ||
277 | /* This function is called for critical strings. It checks for any | |
278 | non-printing characters, and if any are found, it makes a new copy | |
279 | of the string with suitable escape sequences. It is most often called by the | |
280 | macro string_printing(), which sets allow_tab TRUE. | |
281 | ||
282 | Arguments: | |
283 | s the input string | |
284 | allow_tab TRUE to allow tab as a printing character | |
285 | ||
286 | Returns: string with non-printers encoded as printing sequences | |
287 | */ | |
288 | ||
55414b25 JH |
289 | const uschar * |
290 | string_printing2(const uschar *s, BOOL allow_tab) | |
059ec3d9 PH |
291 | { |
292 | int nonprintcount = 0; | |
293 | int length = 0; | |
55414b25 | 294 | const uschar *t = s; |
059ec3d9 PH |
295 | uschar *ss, *tt; |
296 | ||
297 | while (*t != 0) | |
298 | { | |
299 | int c = *t++; | |
300 | if (!mac_isprint(c) || (!allow_tab && c == '\t')) nonprintcount++; | |
301 | length++; | |
302 | } | |
303 | ||
304 | if (nonprintcount == 0) return s; | |
305 | ||
306 | /* Get a new block of store guaranteed big enough to hold the | |
307 | expanded string. */ | |
308 | ||
36719342 | 309 | ss = store_get(length + nonprintcount * 3 + 1); |
059ec3d9 | 310 | |
4c04137d | 311 | /* Copy everything, escaping non printers. */ |
059ec3d9 PH |
312 | |
313 | t = s; | |
314 | tt = ss; | |
315 | ||
316 | while (*t != 0) | |
317 | { | |
318 | int c = *t; | |
319 | if (mac_isprint(c) && (allow_tab || c != '\t')) *tt++ = *t++; else | |
320 | { | |
321 | *tt++ = '\\'; | |
322 | switch (*t) | |
323 | { | |
324 | case '\n': *tt++ = 'n'; break; | |
325 | case '\r': *tt++ = 'r'; break; | |
326 | case '\b': *tt++ = 'b'; break; | |
327 | case '\v': *tt++ = 'v'; break; | |
328 | case '\f': *tt++ = 'f'; break; | |
329 | case '\t': *tt++ = 't'; break; | |
330 | default: sprintf(CS tt, "%03o", *t); tt += 3; break; | |
331 | } | |
332 | t++; | |
333 | } | |
334 | } | |
335 | *tt = 0; | |
c7396ac5 PP |
336 | return ss; |
337 | } | |
79fe97d8 PP |
338 | #endif /* COMPILE_UTILITY */ |
339 | ||
c7396ac5 PP |
340 | /************************************************* |
341 | * Undo printing escapes in string * | |
342 | *************************************************/ | |
343 | ||
344 | /* This function is the reverse of string_printing2. It searches for | |
345 | backslash characters and if any are found, it makes a new copy of the | |
346 | string with escape sequences parsed. Otherwise it returns the original | |
347 | string. | |
348 | ||
349 | Arguments: | |
350 | s the input string | |
351 | ||
352 | Returns: string with printing escapes parsed back | |
353 | */ | |
354 | ||
355 | uschar * | |
356 | string_unprinting(uschar *s) | |
357 | { | |
358 | uschar *p, *q, *r, *ss; | |
359 | int len, off; | |
360 | ||
361 | p = Ustrchr(s, '\\'); | |
362 | if (!p) return s; | |
363 | ||
364 | len = Ustrlen(s) + 1; | |
365 | ss = store_get(len); | |
366 | ||
367 | q = ss; | |
368 | off = p - s; | |
369 | if (off) | |
370 | { | |
371 | memcpy(q, s, off); | |
372 | q += off; | |
373 | } | |
374 | ||
375 | while (*p) | |
376 | { | |
377 | if (*p == '\\') | |
378 | { | |
55414b25 | 379 | *q++ = string_interpret_escape((const uschar **)&p); |
823ad74f | 380 | p++; |
c7396ac5 PP |
381 | } |
382 | else | |
383 | { | |
384 | r = Ustrchr(p, '\\'); | |
385 | if (!r) | |
386 | { | |
387 | off = Ustrlen(p); | |
388 | memcpy(q, p, off); | |
389 | p += off; | |
390 | q += off; | |
391 | break; | |
392 | } | |
393 | else | |
394 | { | |
395 | off = r - p; | |
396 | memcpy(q, p, off); | |
397 | q += off; | |
398 | p = r; | |
399 | } | |
400 | } | |
401 | } | |
402 | *q = '\0'; | |
403 | ||
059ec3d9 PH |
404 | return ss; |
405 | } | |
059ec3d9 PH |
406 | |
407 | ||
408 | ||
409 | ||
e59797e3 | 410 | #ifdef HAVE_LOCAL_SCAN |
059ec3d9 PH |
411 | /************************************************* |
412 | * Copy and save string * | |
413 | *************************************************/ | |
414 | ||
415 | /* This function assumes that memcpy() is faster than strcpy(). | |
416 | ||
417 | Argument: string to copy | |
418 | Returns: copy of string in new store | |
419 | */ | |
420 | ||
421 | uschar * | |
e59797e3 | 422 | string_copy_function(const uschar *s) |
059ec3d9 PH |
423 | { |
424 | int len = Ustrlen(s) + 1; | |
425 | uschar *ss = store_get(len); | |
426 | memcpy(ss, s, len); | |
427 | return ss; | |
428 | } | |
429 | ||
430 | ||
059ec3d9 PH |
431 | /************************************************* |
432 | * Copy and save string, given length * | |
433 | *************************************************/ | |
434 | ||
435 | /* It is assumed the data contains no zeros. A zero is added | |
436 | onto the end. | |
437 | ||
438 | Arguments: | |
439 | s string to copy | |
440 | n number of characters | |
441 | ||
442 | Returns: copy of string in new store | |
443 | */ | |
444 | ||
445 | uschar * | |
e59797e3 | 446 | string_copyn_function(const uschar *s, int n) |
059ec3d9 PH |
447 | { |
448 | uschar *ss = store_get(n + 1); | |
449 | Ustrncpy(ss, s, n); | |
450 | ss[n] = 0; | |
451 | return ss; | |
452 | } | |
e59797e3 | 453 | #endif |
059ec3d9 PH |
454 | |
455 | ||
456 | /************************************************* | |
e59797e3 | 457 | * Copy and save string in malloc'd store * |
059ec3d9 PH |
458 | *************************************************/ |
459 | ||
e59797e3 | 460 | /* This function assumes that memcpy() is faster than strcpy(). |
059ec3d9 | 461 | |
e59797e3 JH |
462 | Argument: string to copy |
463 | Returns: copy of string in new store | |
059ec3d9 PH |
464 | */ |
465 | ||
466 | uschar * | |
e59797e3 | 467 | string_copy_malloc(const uschar *s) |
059ec3d9 | 468 | { |
e59797e3 JH |
469 | int len = Ustrlen(s) + 1; |
470 | uschar *ss = store_malloc(len); | |
471 | memcpy(ss, s, len); | |
059ec3d9 PH |
472 | return ss; |
473 | } | |
474 | ||
475 | ||
476 | ||
e28326d8 PH |
477 | /************************************************* |
478 | * Copy string if long, inserting newlines * | |
479 | *************************************************/ | |
480 | ||
481 | /* If the given string is longer than 75 characters, it is copied, and within | |
482 | the copy, certain space characters are converted into newlines. | |
483 | ||
484 | Argument: pointer to the string | |
485 | Returns: pointer to the possibly altered string | |
486 | */ | |
487 | ||
488 | uschar * | |
489 | string_split_message(uschar *msg) | |
490 | { | |
491 | uschar *s, *ss; | |
492 | ||
493 | if (msg == NULL || Ustrlen(msg) <= 75) return msg; | |
494 | s = ss = msg = string_copy(msg); | |
495 | ||
496 | for (;;) | |
497 | { | |
498 | int i = 0; | |
499 | while (i < 75 && *ss != 0 && *ss != '\n') ss++, i++; | |
500 | if (*ss == 0) break; | |
501 | if (*ss == '\n') | |
502 | s = ++ss; | |
503 | else | |
504 | { | |
505 | uschar *t = ss + 1; | |
506 | uschar *tt = NULL; | |
507 | while (--t > s + 35) | |
508 | { | |
509 | if (*t == ' ') | |
510 | { | |
511 | if (t[-1] == ':') { tt = t; break; } | |
512 | if (tt == NULL) tt = t; | |
513 | } | |
514 | } | |
515 | ||
516 | if (tt == NULL) /* Can't split behind - try ahead */ | |
517 | { | |
518 | t = ss + 1; | |
519 | while (*t != 0) | |
520 | { | |
521 | if (*t == ' ' || *t == '\n') | |
522 | { tt = t; break; } | |
523 | t++; | |
524 | } | |
525 | } | |
526 | ||
527 | if (tt == NULL) break; /* Can't find anywhere to split */ | |
528 | *tt = '\n'; | |
529 | s = ss = tt+1; | |
530 | } | |
531 | } | |
532 | ||
533 | return msg; | |
534 | } | |
535 | ||
536 | ||
537 | ||
059ec3d9 PH |
538 | /************************************************* |
539 | * Copy returned DNS domain name, de-escaping * | |
540 | *************************************************/ | |
541 | ||
542 | /* If a domain name contains top-bit characters, some resolvers return | |
543 | the fully qualified name with those characters turned into escapes. The | |
544 | convention is a backslash followed by _decimal_ digits. We convert these | |
545 | back into the original binary values. This will be relevant when | |
546 | allow_utf8_domains is set true and UTF-8 characters are used in domain | |
547 | names. Backslash can also be used to escape other characters, though we | |
548 | shouldn't come across them in domain names. | |
549 | ||
550 | Argument: the domain name string | |
551 | Returns: copy of string in new store, de-escaped | |
552 | */ | |
553 | ||
554 | uschar * | |
555 | string_copy_dnsdomain(uschar *s) | |
556 | { | |
557 | uschar *yield; | |
558 | uschar *ss = yield = store_get(Ustrlen(s) + 1); | |
559 | ||
560 | while (*s != 0) | |
561 | { | |
562 | if (*s != '\\') | |
563 | { | |
564 | *ss++ = *s++; | |
565 | } | |
566 | else if (isdigit(s[1])) | |
567 | { | |
568 | *ss++ = (s[1] - '0')*100 + (s[2] - '0')*10 + s[3] - '0'; | |
569 | s += 4; | |
570 | } | |
571 | else if (*(++s) != 0) | |
572 | { | |
573 | *ss++ = *s++; | |
574 | } | |
575 | } | |
576 | ||
577 | *ss = 0; | |
578 | return yield; | |
579 | } | |
580 | ||
581 | ||
582 | #ifndef COMPILE_UTILITY | |
583 | /************************************************* | |
584 | * Copy space-terminated or quoted string * | |
585 | *************************************************/ | |
586 | ||
587 | /* This function copies from a string until its end, or until whitespace is | |
588 | encountered, unless the string begins with a double quote, in which case the | |
589 | terminating quote is sought, and escaping within the string is done. The length | |
590 | of a de-quoted string can be no longer than the original, since escaping always | |
591 | turns n characters into 1 character. | |
592 | ||
593 | Argument: pointer to the pointer to the first character, which gets updated | |
594 | Returns: the new string | |
595 | */ | |
596 | ||
597 | uschar * | |
55414b25 | 598 | string_dequote(const uschar **sptr) |
059ec3d9 | 599 | { |
55414b25 | 600 | const uschar *s = *sptr; |
059ec3d9 PH |
601 | uschar *t, *yield; |
602 | ||
603 | /* First find the end of the string */ | |
604 | ||
605 | if (*s != '\"') | |
059ec3d9 | 606 | while (*s != 0 && !isspace(*s)) s++; |
059ec3d9 PH |
607 | else |
608 | { | |
609 | s++; | |
8c513105 | 610 | while (*s && *s != '\"') |
059ec3d9 PH |
611 | { |
612 | if (*s == '\\') (void)string_interpret_escape(&s); | |
613 | s++; | |
614 | } | |
8c513105 | 615 | if (*s) s++; |
059ec3d9 PH |
616 | } |
617 | ||
618 | /* Get enough store to copy into */ | |
619 | ||
620 | t = yield = store_get(s - *sptr + 1); | |
621 | s = *sptr; | |
622 | ||
623 | /* Do the copy */ | |
624 | ||
625 | if (*s != '\"') | |
626 | { | |
627 | while (*s != 0 && !isspace(*s)) *t++ = *s++; | |
628 | } | |
629 | else | |
630 | { | |
631 | s++; | |
632 | while (*s != 0 && *s != '\"') | |
633 | { | |
634 | if (*s == '\\') *t++ = string_interpret_escape(&s); | |
635 | else *t++ = *s; | |
636 | s++; | |
637 | } | |
638 | if (*s != 0) s++; | |
639 | } | |
640 | ||
641 | /* Update the pointer and return the terminated copy */ | |
642 | ||
643 | *sptr = s; | |
644 | *t = 0; | |
645 | return yield; | |
646 | } | |
647 | #endif /* COMPILE_UTILITY */ | |
648 | ||
649 | ||
650 | ||
651 | /************************************************* | |
652 | * Format a string and save it * | |
653 | *************************************************/ | |
654 | ||
94759fce | 655 | /* The formatting is done by string_vformat, which checks the length of |
059ec3d9 PH |
656 | everything. |
657 | ||
658 | Arguments: | |
659 | format a printf() format - deliberately char * rather than uschar * | |
660 | because it will most usually be a literal string | |
661 | ... arguments for format | |
662 | ||
663 | Returns: pointer to fresh piece of store containing sprintf'ed string | |
664 | */ | |
665 | ||
666 | uschar * | |
1ba28e2b | 667 | string_sprintf(const char *format, ...) |
059ec3d9 | 668 | { |
d12746bc | 669 | #ifdef COMPILE_UTILITY |
059ec3d9 | 670 | uschar buffer[STRING_SPRINTF_BUFFER_SIZE]; |
d12746bc JH |
671 | gstring g = { .size = STRING_SPRINTF_BUFFER_SIZE, .ptr = 0, .s = buffer }; |
672 | gstring * gp = &g; | |
673 | #else | |
674 | gstring * gp = string_get(STRING_SPRINTF_BUFFER_SIZE); | |
675 | #endif | |
676 | gstring * gp2; | |
677 | va_list ap; | |
678 | ||
059ec3d9 | 679 | va_start(ap, format); |
d12746bc JH |
680 | gp2 = string_vformat(gp, FALSE, format, ap); |
681 | gp->s[gp->ptr] = '\0'; | |
059ec3d9 | 682 | va_end(ap); |
d12746bc JH |
683 | |
684 | if (!gp2) | |
685 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, | |
686 | "string_sprintf expansion was longer than %d; format string was (%s)\n" | |
687 | "expansion started '%.32s'", | |
688 | gp->size, format, gp->s); | |
689 | ||
690 | #ifdef COMPILE_UTILITY | |
691 | return string_copy(gp->s); | |
692 | #else | |
e59797e3 | 693 | gstring_release_unused(gp); |
d12746bc JH |
694 | return gp->s; |
695 | #endif | |
059ec3d9 PH |
696 | } |
697 | ||
698 | ||
699 | ||
700 | /************************************************* | |
701 | * Case-independent strncmp() function * | |
702 | *************************************************/ | |
703 | ||
704 | /* | |
705 | Arguments: | |
706 | s first string | |
707 | t second string | |
708 | n number of characters to compare | |
709 | ||
710 | Returns: < 0, = 0, or > 0, according to the comparison | |
711 | */ | |
712 | ||
713 | int | |
1ba28e2b | 714 | strncmpic(const uschar *s, const uschar *t, int n) |
059ec3d9 PH |
715 | { |
716 | while (n--) | |
717 | { | |
718 | int c = tolower(*s++) - tolower(*t++); | |
719 | if (c) return c; | |
720 | } | |
721 | return 0; | |
722 | } | |
723 | ||
724 | ||
725 | /************************************************* | |
726 | * Case-independent strcmp() function * | |
727 | *************************************************/ | |
728 | ||
729 | /* | |
730 | Arguments: | |
731 | s first string | |
732 | t second string | |
733 | ||
734 | Returns: < 0, = 0, or > 0, according to the comparison | |
735 | */ | |
736 | ||
737 | int | |
1ba28e2b | 738 | strcmpic(const uschar *s, const uschar *t) |
059ec3d9 PH |
739 | { |
740 | while (*s != 0) | |
741 | { | |
742 | int c = tolower(*s++) - tolower(*t++); | |
743 | if (c != 0) return c; | |
744 | } | |
745 | return *t; | |
746 | } | |
747 | ||
748 | ||
749 | /************************************************* | |
750 | * Case-independent strstr() function * | |
751 | *************************************************/ | |
752 | ||
753 | /* The third argument specifies whether whitespace is required | |
754 | to follow the matched string. | |
755 | ||
756 | Arguments: | |
757 | s string to search | |
758 | t substring to search for | |
759 | space_follows if TRUE, match only if whitespace follows | |
760 | ||
761 | Returns: pointer to substring in string, or NULL if not found | |
762 | */ | |
763 | ||
764 | uschar * | |
765 | strstric(uschar *s, uschar *t, BOOL space_follows) | |
766 | { | |
767 | uschar *p = t; | |
768 | uschar *yield = NULL; | |
769 | int cl = tolower(*p); | |
770 | int cu = toupper(*p); | |
771 | ||
772 | while (*s) | |
773 | { | |
774 | if (*s == cl || *s == cu) | |
775 | { | |
776 | if (yield == NULL) yield = s; | |
777 | if (*(++p) == 0) | |
778 | { | |
779 | if (!space_follows || s[1] == ' ' || s[1] == '\n' ) return yield; | |
780 | yield = NULL; | |
781 | p = t; | |
782 | } | |
783 | cl = tolower(*p); | |
784 | cu = toupper(*p); | |
785 | s++; | |
786 | } | |
787 | else if (yield != NULL) | |
788 | { | |
789 | yield = NULL; | |
790 | p = t; | |
791 | cl = tolower(*p); | |
792 | cu = toupper(*p); | |
793 | } | |
794 | else s++; | |
795 | } | |
796 | return NULL; | |
797 | } | |
798 | ||
799 | ||
800 | ||
d12746bc JH |
801 | #ifdef COMPILE_UTILITY |
802 | /* Dummy version for this function; it should never be called */ | |
803 | static void | |
804 | gstring_grow(gstring * g, int p, int count) | |
805 | { | |
806 | assert(FALSE); | |
807 | } | |
808 | #endif | |
809 | ||
810 | ||
811 | ||
059ec3d9 PH |
812 | #ifndef COMPILE_UTILITY |
813 | /************************************************* | |
814 | * Get next string from separated list * | |
815 | *************************************************/ | |
816 | ||
817 | /* Leading and trailing space is removed from each item. The separator in the | |
818 | list is controlled by the int pointed to by the separator argument as follows: | |
819 | ||
ec95d1a6 PH |
820 | If the value is > 0 it is used as the separator. This is typically used for |
821 | sublists such as slash-separated options. The value is always a printing | |
822 | character. | |
823 | ||
824 | (If the value is actually > UCHAR_MAX there is only one item in the list. | |
059ec3d9 PH |
825 | This is used for some cases when called via functions that sometimes |
826 | plough through lists, and sometimes are given single items.) | |
059ec3d9 | 827 | |
ec95d1a6 PH |
828 | If the value is <= 0, the string is inspected for a leading <x, where x is an |
829 | ispunct() or an iscntrl() character. If found, x is used as the separator. If | |
830 | not found: | |
831 | ||
832 | (a) if separator == 0, ':' is used | |
833 | (b) if separator <0, -separator is used | |
834 | ||
835 | In all cases the value of the separator that is used is written back to the | |
836 | int so that it is used on subsequent calls as we progress through the list. | |
837 | ||
838 | A literal ispunct() separator can be represented in an item by doubling, but | |
839 | there is no way to include an iscntrl() separator as part of the data. | |
059ec3d9 PH |
840 | |
841 | Arguments: | |
842 | listptr points to a pointer to the current start of the list; the | |
843 | pointer gets updated to point after the end of the next item | |
844 | separator a pointer to the separator character in an int (see above) | |
845 | buffer where to put a copy of the next string in the list; or | |
846 | NULL if the next string is returned in new memory | |
847 | buflen when buffer is not NULL, the size of buffer; otherwise ignored | |
848 | ||
849 | Returns: pointer to buffer, containing the next substring, | |
850 | or NULL if no more substrings | |
851 | */ | |
852 | ||
853 | uschar * | |
55414b25 | 854 | string_nextinlist(const uschar **listptr, int *separator, uschar *buffer, int buflen) |
059ec3d9 | 855 | { |
55414b25 JH |
856 | int sep = *separator; |
857 | const uschar *s = *listptr; | |
ec95d1a6 | 858 | BOOL sep_is_special; |
059ec3d9 | 859 | |
8b455685 | 860 | if (!s) return NULL; |
ec95d1a6 PH |
861 | |
862 | /* This allows for a fixed specified separator to be an iscntrl() character, | |
863 | but at the time of implementation, this is never the case. However, it's best | |
864 | to be conservative. */ | |
865 | ||
866 | while (isspace(*s) && *s != sep) s++; | |
867 | ||
868 | /* A change of separator is permitted, so look for a leading '<' followed by an | |
869 | allowed character. */ | |
059ec3d9 PH |
870 | |
871 | if (sep <= 0) | |
872 | { | |
ec95d1a6 | 873 | if (*s == '<' && (ispunct(s[1]) || iscntrl(s[1]))) |
059ec3d9 PH |
874 | { |
875 | sep = s[1]; | |
b72f857f | 876 | if (*++s) ++s; |
ec95d1a6 | 877 | while (isspace(*s) && *s != sep) s++; |
059ec3d9 PH |
878 | } |
879 | else | |
8b455685 | 880 | sep = sep ? -sep : ':'; |
059ec3d9 PH |
881 | *separator = sep; |
882 | } | |
883 | ||
ec95d1a6 PH |
884 | /* An empty string has no list elements */ |
885 | ||
8b455685 | 886 | if (!*s) return NULL; |
059ec3d9 | 887 | |
ec95d1a6 PH |
888 | /* Note whether whether or not the separator is an iscntrl() character. */ |
889 | ||
890 | sep_is_special = iscntrl(sep); | |
891 | ||
059ec3d9 PH |
892 | /* Handle the case when a buffer is provided. */ |
893 | ||
617d3932 | 894 | if (buffer) |
059ec3d9 | 895 | { |
d4ff61d1 | 896 | int p = 0; |
8b455685 | 897 | for (; *s; s++) |
059ec3d9 | 898 | { |
ec95d1a6 | 899 | if (*s == sep && (*(++s) != sep || sep_is_special)) break; |
059ec3d9 PH |
900 | if (p < buflen - 1) buffer[p++] = *s; |
901 | } | |
902 | while (p > 0 && isspace(buffer[p-1])) p--; | |
8b455685 | 903 | buffer[p] = '\0'; |
059ec3d9 PH |
904 | } |
905 | ||
906 | /* Handle the case when a buffer is not provided. */ | |
907 | ||
908 | else | |
909 | { | |
acec9514 | 910 | gstring * g = NULL; |
ec95d1a6 | 911 | |
059ec3d9 | 912 | /* We know that *s != 0 at this point. However, it might be pointing to a |
ec95d1a6 PH |
913 | separator, which could indicate an empty string, or (if an ispunct() |
914 | character) could be doubled to indicate a separator character as data at the | |
915 | start of a string. Avoid getting working memory for an empty item. */ | |
059ec3d9 PH |
916 | |
917 | if (*s == sep) | |
918 | { | |
919 | s++; | |
ec95d1a6 PH |
920 | if (*s != sep || sep_is_special) |
921 | { | |
922 | *listptr = s; | |
923 | return string_copy(US""); | |
924 | } | |
059ec3d9 PH |
925 | } |
926 | ||
ec95d1a6 PH |
927 | /* Not an empty string; the first character is guaranteed to be a data |
928 | character. */ | |
929 | ||
930 | for (;;) | |
059ec3d9 | 931 | { |
d7978c0f JH |
932 | const uschar * ss; |
933 | for (ss = s + 1; *ss && *ss != sep; ) ss++; | |
acec9514 | 934 | g = string_catn(g, s, ss-s); |
ec95d1a6 | 935 | s = ss; |
8b455685 | 936 | if (!*s || *++s != sep || sep_is_special) break; |
059ec3d9 | 937 | } |
acec9514 JH |
938 | while (g->ptr > 0 && isspace(g->s[g->ptr-1])) g->ptr--; |
939 | buffer = string_from_gstring(g); | |
e59797e3 | 940 | gstring_release_unused(g); |
059ec3d9 PH |
941 | } |
942 | ||
943 | /* Update the current pointer and return the new string */ | |
944 | ||
945 | *listptr = s; | |
946 | return buffer; | |
947 | } | |
059ec3d9 PH |
948 | |
949 | ||
4226691b JH |
950 | static const uschar * |
951 | Ustrnchr(const uschar * s, int c, unsigned * len) | |
952 | { | |
953 | unsigned siz = *len; | |
954 | while (siz) | |
955 | { | |
956 | if (!*s) return NULL; | |
957 | if (*s == c) | |
958 | { | |
959 | *len = siz; | |
960 | return s; | |
961 | } | |
962 | s++; | |
963 | siz--; | |
964 | } | |
965 | return NULL; | |
966 | } | |
967 | ||
968 | ||
76146973 | 969 | /************************************************ |
1d9ddac9 | 970 | * Add element to separated list * |
76146973 | 971 | ************************************************/ |
4226691b JH |
972 | /* This function is used to build a list, returning an allocated null-terminated |
973 | growable string. The given element has any embedded separator characters | |
76146973 JH |
974 | doubled. |
975 | ||
4226691b JH |
976 | Despite having the same growable-string interface as string_cat() the list is |
977 | always returned null-terminated. | |
978 | ||
76146973 | 979 | Arguments: |
acec9514 | 980 | list expanding-string for the list that is being built, or NULL |
76146973 | 981 | if this is a new list that has no contents yet |
4c04137d JS |
982 | sep list separator character |
983 | ele new element to be appended to the list | |
76146973 JH |
984 | |
985 | Returns: pointer to the start of the list, changed if copied for expansion. | |
986 | */ | |
987 | ||
acec9514 JH |
988 | gstring * |
989 | string_append_listele(gstring * list, uschar sep, const uschar * ele) | |
76146973 | 990 | { |
76146973 JH |
991 | uschar * sp; |
992 | ||
acec9514 JH |
993 | if (list && list->ptr) |
994 | list = string_catn(list, &sep, 1); | |
76146973 | 995 | |
e3dd1d67 | 996 | while((sp = Ustrchr(ele, sep))) |
76146973 | 997 | { |
acec9514 JH |
998 | list = string_catn(list, ele, sp-ele+1); |
999 | list = string_catn(list, &sep, 1); | |
76146973 JH |
1000 | ele = sp+1; |
1001 | } | |
acec9514 JH |
1002 | list = string_cat(list, ele); |
1003 | (void) string_from_gstring(list); | |
4226691b | 1004 | return list; |
76146973 | 1005 | } |
00ba27c5 JH |
1006 | |
1007 | ||
acec9514 JH |
1008 | gstring * |
1009 | string_append_listele_n(gstring * list, uschar sep, const uschar * ele, | |
1010 | unsigned len) | |
00ba27c5 | 1011 | { |
00ba27c5 JH |
1012 | const uschar * sp; |
1013 | ||
acec9514 JH |
1014 | if (list && list->ptr) |
1015 | list = string_catn(list, &sep, 1); | |
00ba27c5 JH |
1016 | |
1017 | while((sp = Ustrnchr(ele, sep, &len))) | |
1018 | { | |
acec9514 JH |
1019 | list = string_catn(list, ele, sp-ele+1); |
1020 | list = string_catn(list, &sep, 1); | |
00ba27c5 JH |
1021 | ele = sp+1; |
1022 | len--; | |
1023 | } | |
acec9514 JH |
1024 | list = string_catn(list, ele, len); |
1025 | (void) string_from_gstring(list); | |
4226691b | 1026 | return list; |
00ba27c5 | 1027 | } |
76146973 JH |
1028 | |
1029 | ||
059ec3d9 | 1030 | |
bce15b62 JH |
1031 | /* A slightly-bogus listmaker utility; the separator is a string so |
1032 | can be multiple chars - there is no checking for the element content | |
1033 | containing any of the separator. */ | |
1034 | ||
1035 | gstring * | |
1036 | string_append2_listele_n(gstring * list, const uschar * sepstr, | |
1037 | const uschar * ele, unsigned len) | |
1038 | { | |
bce15b62 JH |
1039 | if (list && list->ptr) |
1040 | list = string_cat(list, sepstr); | |
1041 | ||
1042 | list = string_catn(list, ele, len); | |
1043 | (void) string_from_gstring(list); | |
1044 | return list; | |
1045 | } | |
1046 | ||
1047 | ||
1048 | ||
acec9514 | 1049 | /************************************************/ |
d12746bc JH |
1050 | /* Add more space to a growable-string. |
1051 | ||
1052 | Arguments: | |
1053 | g the growable-string | |
938593e9 JH |
1054 | p current end of data |
1055 | count amount to grow by | |
1056 | */ | |
1057 | ||
1058 | static void | |
acec9514 JH |
1059 | gstring_grow(gstring * g, int p, int count) |
1060 | { | |
1061 | int oldsize = g->size; | |
1062 | ||
1063 | /* Mostly, string_cat() is used to build small strings of a few hundred | |
1064 | characters at most. There are times, however, when the strings are very much | |
1065 | longer (for example, a lookup that returns a vast number of alias addresses). | |
1066 | To try to keep things reasonable, we use increments whose size depends on the | |
1067 | existing length of the string. */ | |
1068 | ||
1069 | unsigned inc = oldsize < 4096 ? 127 : 1023; | |
1070 | g->size = ((p + count + inc) & ~inc) + 1; | |
1071 | ||
1072 | /* Try to extend an existing allocation. If the result of calling | |
1073 | store_extend() is false, either there isn't room in the current memory block, | |
1074 | or this string is not the top item on the dynamic store stack. We then have | |
1075 | to get a new chunk of store and copy the old string. When building large | |
1076 | strings, it is helpful to call store_release() on the old string, to release | |
1077 | memory blocks that have become empty. (The block will be freed if the string | |
1078 | is at its start.) However, we can do this only if we know that the old string | |
1079 | was the last item on the dynamic memory stack. This is the case if it matches | |
1080 | store_last_get. */ | |
1081 | ||
1082 | if (!store_extend(g->s, oldsize, g->size)) | |
459fca58 | 1083 | g->s = store_newblock(g->s, g->size, p); |
acec9514 JH |
1084 | } |
1085 | ||
1086 | ||
1087 | ||
d12746bc JH |
1088 | /************************************************* |
1089 | * Add chars to string * | |
1090 | *************************************************/ | |
059ec3d9 PH |
1091 | /* This function is used when building up strings of unknown length. Room is |
1092 | always left for a terminating zero to be added to the string that is being | |
1093 | built. This function does not require the string that is being added to be NUL | |
1094 | terminated, because the number of characters to add is given explicitly. It is | |
1095 | sometimes called to extract parts of other strings. | |
1096 | ||
1097 | Arguments: | |
1098 | string points to the start of the string that is being built, or NULL | |
1099 | if this is a new string that has no contents yet | |
059ec3d9 PH |
1100 | s points to characters to add |
1101 | count count of characters to add; must not exceed the length of s, if s | |
42055a33 | 1102 | is a C string. |
059ec3d9 | 1103 | |
059ec3d9 PH |
1104 | Returns: pointer to the start of the string, changed if copied for expansion. |
1105 | Note that a NUL is not added, though space is left for one. This is | |
1106 | because string_cat() is often called multiple times to build up a | |
1107 | string - there's no point adding the NUL till the end. | |
a1b8a755 | 1108 | |
059ec3d9 | 1109 | */ |
96f5fe4c | 1110 | /* coverity[+alloc] */ |
059ec3d9 | 1111 | |
acec9514 JH |
1112 | gstring * |
1113 | string_catn(gstring * g, const uschar *s, int count) | |
059ec3d9 | 1114 | { |
acec9514 | 1115 | int p; |
059ec3d9 | 1116 | |
acec9514 | 1117 | if (!g) |
059ec3d9 | 1118 | { |
acec9514 JH |
1119 | unsigned inc = count < 4096 ? 127 : 1023; |
1120 | unsigned size = ((count + inc) & ~inc) + 1; | |
1121 | g = string_get(size); | |
059ec3d9 PH |
1122 | } |
1123 | ||
acec9514 JH |
1124 | p = g->ptr; |
1125 | if (p + count >= g->size) | |
1126 | gstring_grow(g, p, count); | |
1127 | ||
059ec3d9 PH |
1128 | /* Because we always specify the exact number of characters to copy, we can |
1129 | use memcpy(), which is likely to be more efficient than strncopy() because the | |
acec9514 | 1130 | latter has to check for zero bytes. */ |
059ec3d9 | 1131 | |
acec9514 JH |
1132 | memcpy(g->s + p, s, count); |
1133 | g->ptr = p + count; | |
1134 | return g; | |
059ec3d9 | 1135 | } |
acec9514 JH |
1136 | |
1137 | ||
1138 | gstring * | |
1139 | string_cat(gstring *string, const uschar *s) | |
c2f669a4 | 1140 | { |
acec9514 | 1141 | return string_catn(string, s, Ustrlen(s)); |
c2f669a4 | 1142 | } |
059ec3d9 PH |
1143 | |
1144 | ||
1145 | ||
059ec3d9 PH |
1146 | /************************************************* |
1147 | * Append strings to another string * | |
1148 | *************************************************/ | |
1149 | ||
1150 | /* This function can be used to build a string from many other strings. | |
1151 | It calls string_cat() to do the dirty work. | |
1152 | ||
1153 | Arguments: | |
acec9514 | 1154 | string expanding-string that is being built, or NULL |
059ec3d9 | 1155 | if this is a new string that has no contents yet |
059ec3d9 PH |
1156 | count the number of strings to append |
1157 | ... "count" uschar* arguments, which must be valid zero-terminated | |
1158 | C strings | |
1159 | ||
1160 | Returns: pointer to the start of the string, changed if copied for expansion. | |
1161 | The string is not zero-terminated - see string_cat() above. | |
1162 | */ | |
1163 | ||
acec9514 JH |
1164 | __inline__ gstring * |
1165 | string_append(gstring *string, int count, ...) | |
059ec3d9 PH |
1166 | { |
1167 | va_list ap; | |
059ec3d9 PH |
1168 | |
1169 | va_start(ap, count); | |
acec9514 | 1170 | while (count-- > 0) |
059ec3d9 PH |
1171 | { |
1172 | uschar *t = va_arg(ap, uschar *); | |
acec9514 | 1173 | string = string_cat(string, t); |
059ec3d9 PH |
1174 | } |
1175 | va_end(ap); | |
1176 | ||
1177 | return string; | |
1178 | } | |
1179 | #endif | |
1180 | ||
1181 | ||
1182 | ||
1183 | /************************************************* | |
1184 | * Format a string with length checks * | |
1185 | *************************************************/ | |
1186 | ||
1187 | /* This function is used to format a string with checking of the length of the | |
1188 | output for all conversions. It protects Exim from absent-mindedness when | |
1189 | calling functions like debug_printf and string_sprintf, and elsewhere. There | |
1190 | are two different entry points to what is actually the same function, depending | |
1191 | on whether the variable length list of data arguments are given explicitly or | |
1192 | as a va_list item. | |
1193 | ||
1194 | The formats are the usual printf() ones, with some omissions (never used) and | |
c0b9d3e8 | 1195 | three additions for strings: %S forces lower case, %T forces upper case, and |
acec9514 | 1196 | %#s or %#S prints nothing for a NULL string. Without the # "NULL" is printed |
c0b9d3e8 JH |
1197 | (useful in debugging). There is also the addition of %D and %M, which insert |
1198 | the date in the form used for datestamped log files. | |
059ec3d9 PH |
1199 | |
1200 | Arguments: | |
1201 | buffer a buffer in which to put the formatted string | |
1202 | buflen the length of the buffer | |
1203 | format the format string - deliberately char * and not uschar * | |
1204 | ... or ap variable list of supplementary arguments | |
1205 | ||
1206 | Returns: TRUE if the result fitted in the buffer | |
1207 | */ | |
1208 | ||
1209 | BOOL | |
d12746bc | 1210 | string_format(uschar * buffer, int buflen, const char * format, ...) |
059ec3d9 | 1211 | { |
d12746bc | 1212 | gstring g = { .size = buflen, .ptr = 0, .s = buffer }, *gp; |
059ec3d9 PH |
1213 | va_list ap; |
1214 | va_start(ap, format); | |
d12746bc | 1215 | gp = string_vformat(&g, FALSE, format, ap); |
059ec3d9 | 1216 | va_end(ap); |
d12746bc JH |
1217 | g.s[g.ptr] = '\0'; |
1218 | return !!gp; | |
059ec3d9 PH |
1219 | } |
1220 | ||
1221 | ||
d12746bc JH |
1222 | |
1223 | ||
1224 | ||
1225 | /* Bulid or append to a growing-string, sprintf-style. | |
1226 | ||
1227 | If the "extend" argument is true, the string passed in can be NULL, | |
1228 | empty, or non-empty. | |
1229 | ||
1230 | If the "extend" argument is false, the string passed in may not be NULL, | |
1231 | will not be grown, and is usable in the original place after return. | |
1232 | The return value can be NULL to signify overflow. | |
1233 | ||
1234 | Returns the possibly-new (if copy for growth was needed) string, | |
1235 | not nul-terminated. | |
1236 | */ | |
1237 | ||
1238 | gstring * | |
1239 | string_vformat(gstring * g, BOOL extend, const char *format, va_list ap) | |
059ec3d9 | 1240 | { |
d12746bc JH |
1241 | enum ltypes { L_NORMAL=1, L_SHORT=2, L_LONG=3, L_LONGLONG=4, L_LONGDOUBLE=5, L_SIZE=6 }; |
1242 | ||
1243 | int width, precision, off, lim; | |
1244 | const char * fp = format; /* Deliberately not unsigned */ | |
b1c749bb | 1245 | |
d12746bc JH |
1246 | string_datestamp_offset = -1; /* Datestamp not inserted */ |
1247 | string_datestamp_length = 0; /* Datestamp not inserted */ | |
1248 | string_datestamp_type = 0; /* Datestamp not inserted */ | |
059ec3d9 | 1249 | |
d12746bc JH |
1250 | #ifdef COMPILE_UTILITY |
1251 | assert(!extend); | |
1252 | assert(g); | |
1253 | #else | |
1254 | ||
1255 | /* Ensure we have a string, to save on checking later */ | |
1256 | if (!g) g = string_get(16); | |
1257 | #endif /*!COMPILE_UTILITY*/ | |
1258 | ||
1259 | lim = g->size - 1; /* leave one for a nul */ | |
1260 | off = g->ptr; /* remember initial offset in gstring */ | |
059ec3d9 PH |
1261 | |
1262 | /* Scan the format and handle the insertions */ | |
1263 | ||
d12746bc | 1264 | while (*fp) |
059ec3d9 | 1265 | { |
b1c749bb | 1266 | int length = L_NORMAL; |
059ec3d9 PH |
1267 | int *nptr; |
1268 | int slen; | |
d12746bc JH |
1269 | const char *null = "NULL"; /* ) These variables */ |
1270 | const char *item_start, *s; /* ) are deliberately */ | |
1271 | char newformat[16]; /* ) not unsigned */ | |
1272 | char * gp = CS g->s + g->ptr; /* ) */ | |
059ec3d9 PH |
1273 | |
1274 | /* Non-% characters just get copied verbatim */ | |
1275 | ||
1276 | if (*fp != '%') | |
1277 | { | |
d12746bc JH |
1278 | /* Avoid string_copyn() due to COMPILE_UTILITY */ |
1279 | if (g->ptr >= lim - 1) | |
1100a343 JH |
1280 | { |
1281 | if (!extend) return NULL; | |
1282 | gstring_grow(g, g->ptr, 1); | |
1283 | lim = g->size - 1; | |
1284 | } | |
d12746bc | 1285 | g->s[g->ptr++] = (uschar) *fp++; |
059ec3d9 PH |
1286 | continue; |
1287 | } | |
1288 | ||
1289 | /* Deal with % characters. Pick off the width and precision, for checking | |
1290 | strings, skipping over the flag and modifier characters. */ | |
1291 | ||
1292 | item_start = fp; | |
1293 | width = precision = -1; | |
1294 | ||
1295 | if (strchr("-+ #0", *(++fp)) != NULL) | |
1296 | { | |
1297 | if (*fp == '#') null = ""; | |
1298 | fp++; | |
1299 | } | |
1300 | ||
1301 | if (isdigit((uschar)*fp)) | |
1302 | { | |
1303 | width = *fp++ - '0'; | |
1304 | while (isdigit((uschar)*fp)) width = width * 10 + *fp++ - '0'; | |
1305 | } | |
1306 | else if (*fp == '*') | |
1307 | { | |
1308 | width = va_arg(ap, int); | |
1309 | fp++; | |
1310 | } | |
1311 | ||
1312 | if (*fp == '.') | |
059ec3d9 PH |
1313 | if (*(++fp) == '*') |
1314 | { | |
1315 | precision = va_arg(ap, int); | |
1316 | fp++; | |
1317 | } | |
1318 | else | |
d12746bc JH |
1319 | for (precision = 0; isdigit((uschar)*fp); fp++) |
1320 | precision = precision*10 + *fp - '0'; | |
059ec3d9 | 1321 | |
91a246f6 | 1322 | /* Skip over 'h', 'L', 'l', 'll' and 'z', remembering the item length */ |
b1c749bb PH |
1323 | |
1324 | if (*fp == 'h') | |
1325 | { fp++; length = L_SHORT; } | |
1326 | else if (*fp == 'L') | |
1327 | { fp++; length = L_LONGDOUBLE; } | |
1328 | else if (*fp == 'l') | |
b1c749bb | 1329 | if (fp[1] == 'l') |
d12746bc | 1330 | { fp += 2; length = L_LONGLONG; } |
b1c749bb | 1331 | else |
d12746bc | 1332 | { fp++; length = L_LONG; } |
91a246f6 PP |
1333 | else if (*fp == 'z') |
1334 | { fp++; length = L_SIZE; } | |
059ec3d9 PH |
1335 | |
1336 | /* Handle each specific format type. */ | |
1337 | ||
1338 | switch (*fp++) | |
1339 | { | |
1340 | case 'n': | |
d12746bc JH |
1341 | nptr = va_arg(ap, int *); |
1342 | *nptr = g->ptr - off; | |
1343 | break; | |
059ec3d9 PH |
1344 | |
1345 | case 'd': | |
1346 | case 'o': | |
1347 | case 'u': | |
1348 | case 'x': | |
1349 | case 'X': | |
d12746bc JH |
1350 | width = length > L_LONG ? 24 : 12; |
1351 | if (g->ptr >= lim - width) | |
1100a343 JH |
1352 | { |
1353 | if (!extend) return NULL; | |
1354 | gstring_grow(g, g->ptr, width); | |
1355 | lim = g->size - 1; | |
1356 | gp = CS g->s + g->ptr; | |
1357 | } | |
d12746bc JH |
1358 | strncpy(newformat, item_start, fp - item_start); |
1359 | newformat[fp - item_start] = 0; | |
b1c749bb | 1360 | |
d12746bc JH |
1361 | /* Short int is promoted to int when passing through ..., so we must use |
1362 | int for va_arg(). */ | |
b1c749bb | 1363 | |
d12746bc JH |
1364 | switch(length) |
1365 | { | |
1366 | case L_SHORT: | |
1367 | case L_NORMAL: | |
1368 | g->ptr += sprintf(gp, newformat, va_arg(ap, int)); break; | |
1369 | case L_LONG: | |
1370 | g->ptr += sprintf(gp, newformat, va_arg(ap, long int)); break; | |
1371 | case L_LONGLONG: | |
1372 | g->ptr += sprintf(gp, newformat, va_arg(ap, LONGLONG_T)); break; | |
1373 | case L_SIZE: | |
1374 | g->ptr += sprintf(gp, newformat, va_arg(ap, size_t)); break; | |
1375 | } | |
1376 | break; | |
059ec3d9 PH |
1377 | |
1378 | case 'p': | |
2e8db779 JH |
1379 | { |
1380 | void * ptr; | |
d12746bc | 1381 | if (g->ptr >= lim - 24) |
1100a343 JH |
1382 | { |
1383 | if (!extend) return NULL; | |
1384 | gstring_grow(g, g->ptr, 24); | |
1385 | lim = g->size - 1; | |
1386 | gp = CS g->s + g->ptr; | |
1387 | } | |
81f358da JH |
1388 | /* sprintf() saying "(nil)" for a null pointer seems unreliable. |
1389 | Handle it explicitly. */ | |
2e8db779 JH |
1390 | if ((ptr = va_arg(ap, void *))) |
1391 | { | |
1392 | strncpy(newformat, item_start, fp - item_start); | |
1393 | newformat[fp - item_start] = 0; | |
d12746bc | 1394 | g->ptr += sprintf(gp, newformat, ptr); |
2e8db779 JH |
1395 | } |
1396 | else | |
d12746bc | 1397 | g->ptr += sprintf(gp, "(nil)"); |
2e8db779 | 1398 | } |
059ec3d9 PH |
1399 | break; |
1400 | ||
1401 | /* %f format is inherently insecure if the numbers that it may be | |
870f6ba8 TF |
1402 | handed are unknown (e.g. 1e300). However, in Exim, %f is used for |
1403 | printing load averages, and these are actually stored as integers | |
1404 | (load average * 1000) so the size of the numbers is constrained. | |
1405 | It is also used for formatting sending rates, where the simplicity | |
1406 | of the format prevents overflow. */ | |
059ec3d9 PH |
1407 | |
1408 | case 'f': | |
1409 | case 'e': | |
1410 | case 'E': | |
1411 | case 'g': | |
1412 | case 'G': | |
d12746bc JH |
1413 | if (precision < 0) precision = 6; |
1414 | if (g->ptr >= lim - precision - 8) | |
1100a343 JH |
1415 | { |
1416 | if (!extend) return NULL; | |
1417 | gstring_grow(g, g->ptr, precision+8); | |
1418 | lim = g->size - 1; | |
1419 | gp = CS g->s + g->ptr; | |
1420 | } | |
d12746bc JH |
1421 | strncpy(newformat, item_start, fp - item_start); |
1422 | newformat[fp-item_start] = 0; | |
1423 | if (length == L_LONGDOUBLE) | |
1424 | g->ptr += sprintf(gp, newformat, va_arg(ap, long double)); | |
1425 | else | |
1426 | g->ptr += sprintf(gp, newformat, va_arg(ap, double)); | |
1427 | break; | |
059ec3d9 PH |
1428 | |
1429 | /* String types */ | |
1430 | ||
1431 | case '%': | |
d12746bc | 1432 | if (g->ptr >= lim - 1) |
1100a343 JH |
1433 | { |
1434 | if (!extend) return NULL; | |
1435 | gstring_grow(g, g->ptr, 1); | |
1436 | lim = g->size - 1; | |
1437 | } | |
d12746bc JH |
1438 | g->s[g->ptr++] = (uschar) '%'; |
1439 | break; | |
059ec3d9 PH |
1440 | |
1441 | case 'c': | |
d12746bc | 1442 | if (g->ptr >= lim - 1) |
1100a343 JH |
1443 | { |
1444 | if (!extend) return NULL; | |
1445 | gstring_grow(g, g->ptr, 1); | |
1446 | lim = g->size - 1; | |
1447 | } | |
d12746bc JH |
1448 | g->s[g->ptr++] = (uschar) va_arg(ap, int); |
1449 | break; | |
059ec3d9 | 1450 | |
f1e5fef5 | 1451 | case 'D': /* Insert daily datestamp for log file names */ |
d12746bc JH |
1452 | s = CS tod_stamp(tod_log_datestamp_daily); |
1453 | string_datestamp_offset = g->ptr; /* Passed back via global */ | |
1454 | string_datestamp_length = Ustrlen(s); /* Passed back via global */ | |
1455 | string_datestamp_type = tod_log_datestamp_daily; | |
1456 | slen = string_datestamp_length; | |
1457 | goto INSERT_STRING; | |
f1e5fef5 PP |
1458 | |
1459 | case 'M': /* Insert monthly datestamp for log file names */ | |
d12746bc JH |
1460 | s = CS tod_stamp(tod_log_datestamp_monthly); |
1461 | string_datestamp_offset = g->ptr; /* Passed back via global */ | |
1462 | string_datestamp_length = Ustrlen(s); /* Passed back via global */ | |
1463 | string_datestamp_type = tod_log_datestamp_monthly; | |
1464 | slen = string_datestamp_length; | |
1465 | goto INSERT_STRING; | |
059ec3d9 PH |
1466 | |
1467 | case 's': | |
1468 | case 'S': /* Forces *lower* case */ | |
c0b9d3e8 | 1469 | case 'T': /* Forces *upper* case */ |
d12746bc | 1470 | s = va_arg(ap, char *); |
059ec3d9 | 1471 | |
d12746bc JH |
1472 | if (!s) s = null; |
1473 | slen = Ustrlen(s); | |
059ec3d9 | 1474 | |
f1e5fef5 PP |
1475 | INSERT_STRING: /* Come to from %D or %M above */ |
1476 | ||
059ec3d9 | 1477 | { |
d12746bc | 1478 | BOOL truncated = FALSE; |
059ec3d9 | 1479 | |
d12746bc JH |
1480 | /* If the width is specified, check that there is a precision |
1481 | set; if not, set it to the width to prevent overruns of long | |
1482 | strings. */ | |
059ec3d9 | 1483 | |
d12746bc JH |
1484 | if (width >= 0) |
1485 | { | |
1486 | if (precision < 0) precision = width; | |
1487 | } | |
059ec3d9 | 1488 | |
d12746bc JH |
1489 | /* If a width is not specified and the precision is specified, set |
1490 | the width to the precision, or the string length if shorted. */ | |
059ec3d9 | 1491 | |
d12746bc JH |
1492 | else if (precision >= 0) |
1493 | width = precision < slen ? precision : slen; | |
059ec3d9 | 1494 | |
d12746bc | 1495 | /* If neither are specified, set them both to the string length. */ |
059ec3d9 | 1496 | |
d12746bc JH |
1497 | else |
1498 | width = precision = slen; | |
1499 | ||
1500 | if (!extend) | |
1501 | { | |
1502 | if (g->ptr == lim) return NULL; | |
1503 | if (g->ptr >= lim - width) | |
1504 | { | |
1505 | truncated = TRUE; | |
1506 | width = precision = lim - g->ptr - 1; | |
1507 | if (width < 0) width = 0; | |
1508 | if (precision < 0) precision = 0; | |
1509 | } | |
1510 | } | |
1511 | else if (g->ptr >= lim - width) | |
1100a343 JH |
1512 | { |
1513 | gstring_grow(g, g->ptr, width - (lim - g->ptr)); | |
1514 | lim = g->size - 1; | |
1515 | gp = CS g->s + g->ptr; | |
1516 | } | |
d12746bc JH |
1517 | |
1518 | g->ptr += sprintf(gp, "%*.*s", width, precision, s); | |
1519 | if (fp[-1] == 'S') | |
1520 | while (*gp) { *gp = tolower(*gp); gp++; } | |
1521 | else if (fp[-1] == 'T') | |
1522 | while (*gp) { *gp = toupper(*gp); gp++; } | |
1523 | ||
1524 | if (truncated) return NULL; | |
1525 | break; | |
059ec3d9 | 1526 | } |
059ec3d9 PH |
1527 | |
1528 | /* Some things are never used in Exim; also catches junk. */ | |
1529 | ||
1530 | default: | |
d12746bc JH |
1531 | strncpy(newformat, item_start, fp - item_start); |
1532 | newformat[fp-item_start] = 0; | |
1533 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, "string_format: unsupported type " | |
1534 | "in \"%s\" in \"%s\"", newformat, format); | |
1535 | break; | |
059ec3d9 PH |
1536 | } |
1537 | } | |
1538 | ||
d12746bc JH |
1539 | return g; |
1540 | } | |
059ec3d9 | 1541 | |
059ec3d9 | 1542 | |
d12746bc JH |
1543 | |
1544 | #ifndef COMPILE_UTILITY | |
1545 | ||
1546 | gstring * | |
1547 | string_fmt_append(gstring * g, const char *format, ...) | |
1548 | { | |
1549 | va_list ap; | |
1550 | va_start(ap, format); | |
1551 | g = string_vformat(g, TRUE, format, ap); | |
1552 | va_end(ap); | |
1553 | return g; | |
059ec3d9 PH |
1554 | } |
1555 | ||
1556 | ||
1557 | ||
059ec3d9 PH |
1558 | /************************************************* |
1559 | * Generate an "open failed" message * | |
1560 | *************************************************/ | |
1561 | ||
1562 | /* This function creates a message after failure to open a file. It includes a | |
1563 | string supplied as data, adds the strerror() text, and if the failure was | |
1564 | "Permission denied", reads and includes the euid and egid. | |
1565 | ||
1566 | Arguments: | |
1567 | eno the value of errno after the failure | |
1568 | format a text format string - deliberately not uschar * | |
1569 | ... arguments for the format string | |
1570 | ||
1571 | Returns: a message, in dynamic store | |
1572 | */ | |
1573 | ||
1574 | uschar * | |
1ba28e2b | 1575 | string_open_failed(int eno, const char *format, ...) |
059ec3d9 PH |
1576 | { |
1577 | va_list ap; | |
d12746bc | 1578 | gstring * g = string_get(1024); |
059ec3d9 | 1579 | |
d12746bc | 1580 | g = string_catn(g, US"failed to open ", 15); |
059ec3d9 PH |
1581 | |
1582 | /* Use the checked formatting routine to ensure that the buffer | |
1583 | does not overflow. It should not, since this is called only for internally | |
1584 | specified messages. If it does, the message just gets truncated, and there | |
1585 | doesn't seem much we can do about that. */ | |
1586 | ||
d12746bc JH |
1587 | va_start(ap, format); |
1588 | (void) string_vformat(g, FALSE, format, ap); | |
1589 | string_from_gstring(g); | |
e59797e3 | 1590 | gstring_release_unused(g); |
cb570b5e | 1591 | va_end(ap); |
059ec3d9 | 1592 | |
d12746bc JH |
1593 | return eno == EACCES |
1594 | ? string_sprintf("%s: %s (euid=%ld egid=%ld)", g->s, strerror(eno), | |
1595 | (long int)geteuid(), (long int)getegid()) | |
1596 | : string_sprintf("%s: %s", g->s, strerror(eno)); | |
059ec3d9 PH |
1597 | } |
1598 | #endif /* COMPILE_UTILITY */ | |
1599 | ||
1600 | ||
1601 | ||
059ec3d9 PH |
1602 | |
1603 | ||
bc3c7bb7 HSHR |
1604 | #ifndef COMPILE_UTILITY |
1605 | /* qsort(3), currently used to sort the environment variables | |
1606 | for -bP environment output, needs a function to compare two pointers to string | |
1607 | pointers. Here it is. */ | |
1608 | ||
1609 | int | |
84bbb4d8 | 1610 | string_compare_by_pointer(const void *a, const void *b) |
bc3c7bb7 | 1611 | { |
35a5627d | 1612 | return Ustrcmp(* CUSS a, * CUSS b); |
bc3c7bb7 HSHR |
1613 | } |
1614 | #endif /* COMPILE_UTILITY */ | |
059ec3d9 PH |
1615 | |
1616 | ||
1617 | ||
d12746bc | 1618 | |
059ec3d9 PH |
1619 | /************************************************* |
1620 | ************************************************** | |
1621 | * Stand-alone test program * | |
1622 | ************************************************** | |
1623 | *************************************************/ | |
1624 | ||
1625 | #ifdef STAND_ALONE | |
1626 | int main(void) | |
1627 | { | |
1628 | uschar buffer[256]; | |
1629 | ||
1630 | printf("Testing is_ip_address\n"); | |
1631 | ||
1632 | while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) | |
1633 | { | |
1634 | int offset; | |
1635 | buffer[Ustrlen(buffer) - 1] = 0; | |
1636 | printf("%d\n", string_is_ip_address(buffer, NULL)); | |
1637 | printf("%d %d %s\n", string_is_ip_address(buffer, &offset), offset, buffer); | |
1638 | } | |
1639 | ||
1640 | printf("Testing string_nextinlist\n"); | |
1641 | ||
1642 | while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) | |
1643 | { | |
1644 | uschar *list = buffer; | |
1645 | uschar *lp1, *lp2; | |
1646 | uschar item[256]; | |
1647 | int sep1 = 0; | |
1648 | int sep2 = 0; | |
1649 | ||
1650 | if (*list == '<') | |
1651 | { | |
1652 | sep1 = sep2 = list[1]; | |
1653 | list += 2; | |
1654 | } | |
1655 | ||
1656 | lp1 = lp2 = list; | |
1657 | for (;;) | |
1658 | { | |
1659 | uschar *item1 = string_nextinlist(&lp1, &sep1, item, sizeof(item)); | |
1660 | uschar *item2 = string_nextinlist(&lp2, &sep2, NULL, 0); | |
1661 | ||
1662 | if (item1 == NULL && item2 == NULL) break; | |
1663 | if (item == NULL || item2 == NULL || Ustrcmp(item1, item2) != 0) | |
1664 | { | |
1665 | printf("***ERROR\nitem1=\"%s\"\nitem2=\"%s\"\n", | |
1666 | (item1 == NULL)? "NULL" : CS item1, | |
1667 | (item2 == NULL)? "NULL" : CS item2); | |
1668 | break; | |
1669 | } | |
1670 | else printf(" \"%s\"\n", CS item1); | |
1671 | } | |
1672 | } | |
1673 | ||
1674 | /* This is a horrible lash-up, but it serves its purpose. */ | |
1675 | ||
1676 | printf("Testing string_format\n"); | |
1677 | ||
1678 | while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) | |
1679 | { | |
1680 | void *args[3]; | |
ed72ace5 | 1681 | long long llargs[3]; |
059ec3d9 PH |
1682 | double dargs[3]; |
1683 | int dflag = 0; | |
ed72ace5 | 1684 | int llflag = 0; |
059ec3d9 PH |
1685 | int n = 0; |
1686 | int count; | |
1687 | int countset = 0; | |
1688 | uschar format[256]; | |
1689 | uschar outbuf[256]; | |
1690 | uschar *s; | |
1691 | buffer[Ustrlen(buffer) - 1] = 0; | |
1692 | ||
1693 | s = Ustrchr(buffer, ','); | |
1694 | if (s == NULL) s = buffer + Ustrlen(buffer); | |
1695 | ||
1696 | Ustrncpy(format, buffer, s - buffer); | |
1697 | format[s-buffer] = 0; | |
1698 | ||
1699 | if (*s == ',') s++; | |
1700 | ||
1701 | while (*s != 0) | |
1702 | { | |
1703 | uschar *ss = s; | |
1704 | s = Ustrchr(ss, ','); | |
1705 | if (s == NULL) s = ss + Ustrlen(ss); | |
1706 | ||
1707 | if (isdigit(*ss)) | |
1708 | { | |
1709 | Ustrncpy(outbuf, ss, s-ss); | |
1710 | if (Ustrchr(outbuf, '.') != NULL) | |
1711 | { | |
1712 | dflag = 1; | |
1713 | dargs[n++] = Ustrtod(outbuf, NULL); | |
1714 | } | |
ed72ace5 PH |
1715 | else if (Ustrstr(outbuf, "ll") != NULL) |
1716 | { | |
1717 | llflag = 1; | |
1718 | llargs[n++] = strtoull(CS outbuf, NULL, 10); | |
1719 | } | |
059ec3d9 PH |
1720 | else |
1721 | { | |
1722 | args[n++] = (void *)Uatoi(outbuf); | |
1723 | } | |
1724 | } | |
1725 | ||
1726 | else if (Ustrcmp(ss, "*") == 0) | |
1727 | { | |
1728 | args[n++] = (void *)(&count); | |
1729 | countset = 1; | |
1730 | } | |
1731 | ||
1732 | else | |
1733 | { | |
1734 | uschar *sss = malloc(s - ss + 1); | |
1735 | Ustrncpy(sss, ss, s-ss); | |
1736 | args[n++] = sss; | |
1737 | } | |
1738 | ||
1739 | if (*s == ',') s++; | |
1740 | } | |
1741 | ||
ed72ace5 PH |
1742 | if (!dflag && !llflag) |
1743 | printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, | |
1744 | args[0], args[1], args[2])? "True" : "False"); | |
1745 | ||
1746 | else if (dflag) | |
1747 | printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, | |
1748 | dargs[0], dargs[1], dargs[2])? "True" : "False"); | |
059ec3d9 PH |
1749 | |
1750 | else printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, | |
ed72ace5 | 1751 | llargs[0], llargs[1], llargs[2])? "True" : "False"); |
059ec3d9 PH |
1752 | |
1753 | printf("%s\n", CS outbuf); | |
1754 | if (countset) printf("count=%d\n", count); | |
1755 | } | |
1756 | ||
1757 | return 0; | |
1758 | } | |
1759 | #endif | |
1760 | ||
1761 | /* End of string.c */ |