Commit | Line | Data |
---|---|---|
184e8823 | 1 | /* $Cambridge: exim/src/src/parse.c,v 1.11 2007/01/08 10:50:18 ph10 Exp $ */ |
059ec3d9 PH |
2 | |
3 | /************************************************* | |
4 | * Exim - an Internet mail transport agent * | |
5 | *************************************************/ | |
6 | ||
184e8823 | 7 | /* Copyright (c) University of Cambridge 1995 - 2007 */ |
059ec3d9 PH |
8 | /* See the file NOTICE for conditions of use and distribution. */ |
9 | ||
10 | /* Functions for parsing addresses */ | |
11 | ||
12 | ||
13 | #include "exim.h" | |
14 | ||
15 | ||
16 | static uschar *last_comment_position; | |
17 | ||
18 | ||
19 | ||
20 | /* In stand-alone mode, provide a replacement for deliver_make_addr() | |
21 | and rewrite_address[_qualify]() so as to avoid having to drag in too much | |
22 | redundant apparatus. */ | |
23 | ||
24 | #ifdef STAND_ALONE | |
25 | ||
26 | address_item *deliver_make_addr(uschar *address, BOOL copy) | |
27 | { | |
28 | address_item *addr = store_get(sizeof(address_item)); | |
29 | addr->next = NULL; | |
30 | addr->parent = NULL; | |
31 | addr->address = address; | |
32 | return addr; | |
33 | } | |
34 | ||
35 | uschar *rewrite_address(uschar *recipient, BOOL dummy1, BOOL dummy2, rewrite_rule | |
36 | *dummy3, int dummy4) | |
37 | { | |
38 | return recipient; | |
39 | } | |
40 | ||
41 | uschar *rewrite_address_qualify(uschar *recipient, BOOL dummy1) | |
42 | { | |
43 | return recipient; | |
44 | } | |
45 | ||
46 | #endif | |
47 | ||
48 | ||
49 | ||
50 | ||
51 | /************************************************* | |
52 | * Find the end of an address * | |
53 | *************************************************/ | |
54 | ||
55 | /* Scan over a string looking for the termination of an address at a comma, | |
56 | or end of the string. It's the source-routed addresses which cause much pain | |
57 | here. Although Exim ignores source routes, it must recognize such addresses, so | |
58 | we cannot get rid of this logic. | |
59 | ||
60 | Argument: | |
61 | s pointer to the start of an address | |
62 | nl_ends if TRUE, '\n' terminates an address | |
63 | ||
64 | Returns: pointer past the end of the address | |
65 | (i.e. points to null or comma) | |
66 | */ | |
67 | ||
68 | uschar * | |
69 | parse_find_address_end(uschar *s, BOOL nl_ends) | |
70 | { | |
71 | BOOL source_routing = *s == '@'; | |
72 | int no_term = source_routing? 1 : 0; | |
73 | ||
74 | while (*s != 0 && (*s != ',' || no_term > 0) && (*s != '\n' || !nl_ends)) | |
75 | { | |
76 | /* Skip single quoted characters. Strictly these should not occur outside | |
77 | quoted strings in RFC 822 addresses, but they can in RFC 821 addresses. Pity | |
78 | about the lack of consistency, isn't it? */ | |
79 | ||
80 | if (*s == '\\' && s[1] != 0) s += 2; | |
81 | ||
82 | /* Skip quoted items that are not inside brackets. Note that | |
83 | quoted pairs are allowed inside quoted strings. */ | |
84 | ||
85 | else if (*s == '\"') | |
86 | { | |
87 | while (*(++s) != 0 && (*s != '\n' || !nl_ends)) | |
88 | { | |
89 | if (*s == '\\' && s[1] != 0) s++; | |
90 | else if (*s == '\"') { s++; break; } | |
91 | } | |
92 | } | |
93 | ||
94 | /* Skip comments, which may include nested brackets, but quotes | |
95 | are not recognized inside comments, though quoted pairs are. */ | |
96 | ||
97 | else if (*s == '(') | |
98 | { | |
99 | int level = 1; | |
100 | while (*(++s) != 0 && (*s != '\n' || !nl_ends)) | |
101 | { | |
102 | if (*s == '\\' && s[1] != 0) s++; | |
103 | else if (*s == '(') level++; | |
104 | else if (*s == ')' && --level <= 0) { s++; break; } | |
105 | } | |
106 | } | |
107 | ||
108 | /* Non-special character; just advance. Passing the colon in a source | |
109 | routed address means that any subsequent comma or colon may terminate unless | |
110 | inside angle brackets. */ | |
111 | ||
112 | else | |
113 | { | |
114 | if (*s == '<') | |
115 | { | |
116 | source_routing = s[1] == '@'; | |
117 | no_term = source_routing? 2 : 1; | |
118 | } | |
119 | else if (*s == '>') no_term--; | |
120 | else if (source_routing && *s == ':') no_term--; | |
121 | s++; | |
122 | } | |
123 | } | |
124 | ||
125 | return s; | |
126 | } | |
127 | ||
128 | ||
129 | ||
130 | /************************************************* | |
131 | * Find last @ in an address * | |
132 | *************************************************/ | |
133 | ||
134 | /* This function is used when we have something that may not qualified. If we | |
135 | know it's qualified, searching for the rightmost '@' is sufficient. Here we | |
136 | have to be a bit more clever than just a plain search, in order to handle | |
137 | unqualified local parts like "thing@thong" correctly. Since quotes may not | |
138 | legally be part of a domain name, we can give up on hitting the first quote | |
139 | when searching from the right. Now that the parsing also permits the RFC 821 | |
140 | form of address, where quoted-pairs are allowed in unquoted local parts, we | |
141 | must take care to handle that too. | |
142 | ||
143 | Argument: pointer to an address, possibly unqualified | |
144 | Returns: pointer to the last @ in an address, or NULL if none | |
145 | */ | |
146 | ||
147 | uschar * | |
148 | parse_find_at(uschar *s) | |
149 | { | |
150 | uschar *t = s + Ustrlen(s); | |
151 | while (--t >= s) | |
152 | { | |
153 | if (*t == '@') | |
154 | { | |
155 | int backslash_count = 0; | |
156 | uschar *tt = t - 1; | |
157 | while (tt > s && *tt-- == '\\') backslash_count++; | |
158 | if ((backslash_count & 1) == 0) return t; | |
159 | } | |
160 | else if (*t == '\"') return NULL; | |
161 | } | |
162 | return NULL; | |
163 | } | |
164 | ||
165 | ||
166 | ||
167 | ||
168 | /*************************************************************************** | |
169 | * In all the functions below that read a particular object type from * | |
170 | * the input, return the new value of the pointer s (the first argument), * | |
171 | * and put the object into the store pointed to by t (the second argument), * | |
172 | * adding a terminating zero. If no object is found, t will point to zero * | |
173 | * on return. * | |
174 | ***************************************************************************/ | |
175 | ||
176 | ||
177 | /************************************************* | |
178 | * Skip white space and comment * | |
179 | *************************************************/ | |
180 | ||
181 | /* Algorithm: | |
182 | (1) Skip spaces. | |
183 | (2) If uschar not '(', return. | |
184 | (3) Skip till matching ')', not counting any characters | |
185 | escaped with '\'. | |
186 | (4) Move past ')' and goto (1). | |
187 | ||
188 | The start of the last potential comment position is remembered to | |
189 | make it possible to ignore comments at the end of compound items. | |
190 | ||
191 | Argument: current character pointer | |
192 | Regurns: new character pointer | |
193 | */ | |
194 | ||
195 | static uschar * | |
196 | skip_comment(uschar *s) | |
197 | { | |
198 | last_comment_position = s; | |
199 | while (*s) | |
200 | { | |
201 | int c, level; | |
202 | while (isspace(*s)) s++; | |
203 | if (*s != '(') break; | |
204 | level = 1; | |
205 | while((c = *(++s)) != 0) | |
206 | { | |
207 | if (c == '(') level++; | |
208 | else if (c == ')') { if (--level <= 0) { s++; break; } } | |
209 | else if (c == '\\' && s[1] != 0) s++; | |
210 | } | |
211 | } | |
212 | return s; | |
213 | } | |
214 | ||
215 | ||
216 | ||
217 | /************************************************* | |
218 | * Read a domain * | |
219 | *************************************************/ | |
220 | ||
221 | /* A domain is a sequence of subdomains, separated by dots. See comments below | |
222 | for detailed syntax of the subdomains. | |
223 | ||
224 | If allow_domain_literals is TRUE, a "domain" may also be an IP address enclosed | |
225 | in []. Make sure the output is set to the null string if there is a syntax | |
226 | error as well as if there is no domain at all. | |
227 | ||
228 | Arguments: | |
229 | s current character pointer | |
230 | t where to put the domain | |
231 | errorptr put error message here on failure (*t will be 0 on exit) | |
232 | ||
233 | Returns: new character pointer | |
234 | */ | |
235 | ||
236 | static uschar * | |
237 | read_domain(uschar *s, uschar *t, uschar **errorptr) | |
238 | { | |
239 | uschar *tt = t; | |
240 | s = skip_comment(s); | |
241 | ||
242 | /* Handle domain literals if permitted. An RFC 822 domain literal may contain | |
243 | any character except [ ] \, including linear white space, and may contain | |
244 | quoted characters. However, RFC 821 restricts literals to being dot-separated | |
245 | 3-digit numbers, and we make the obvious extension for IPv6. Go for a sequence | |
309bd837 PH |
246 | of digits, dots, hex digits, and colons here; later this will be checked for |
247 | being a syntactically valid IP address if it ever gets to a router. | |
059ec3d9 | 248 | |
309bd837 PH |
249 | Allow both the formal IPv6 form, with IPV6: at the start, and the informal form |
250 | without it, and accept IPV4: as well, 'cause someone will use it sooner or | |
251 | later. */ | |
059ec3d9 PH |
252 | |
253 | if (*s == '[') | |
254 | { | |
255 | *t++ = *s++; | |
256 | ||
059ec3d9 PH |
257 | if (strncmpic(s, US"IPv6:", 5) == 0 || strncmpic(s, US"IPv4:", 5) == 0) |
258 | { | |
259 | memcpy(t, s, 5); | |
260 | t += 5; | |
261 | s += 5; | |
262 | } | |
263 | while (*s == '.' || *s == ':' || isxdigit(*s)) *t++ = *s++; | |
264 | ||
059ec3d9 PH |
265 | if (*s == ']') *t++ = *s++; else |
266 | { | |
267 | *errorptr = US"malformed domain literal"; | |
268 | *tt = 0; | |
269 | } | |
270 | ||
271 | if (!allow_domain_literals) | |
272 | { | |
273 | *errorptr = US"domain literals not allowed"; | |
274 | *tt = 0; | |
275 | } | |
276 | *t = 0; | |
277 | return skip_comment(s); | |
278 | } | |
279 | ||
280 | /* Handle a proper domain, which is a sequence of dot-separated atoms. Remove | |
281 | trailing dots if strip_trailing_dot is set. A subdomain is an atom. | |
282 | ||
283 | An atom is a sequence of any characters except specials, space, and controls. | |
284 | The specials are ( ) < > @ , ; : \ " . [ and ]. This is the rule for RFC 822 | |
285 | and its successor (RFC 2822). However, RFC 821 and its successor (RFC 2821) is | |
286 | tighter, allowing only letters, digits, and hyphens, not starting with a | |
287 | hyphen. | |
288 | ||
289 | There used to be a global flag that got set when checking addresses that came | |
290 | in over SMTP and which should therefore should be checked according to the | |
291 | stricter rule. However, it seems silly to make the distinction, because I don't | |
292 | suppose anybody ever uses local domains that are 822-compliant and not | |
293 | 821-compliant. Furthermore, Exim now has additional data on the spool file line | |
294 | after an address (after "one_time" processing), and it makes use of a # | |
295 | character to delimit it. When I wrote that code, I forgot about this 822-domain | |
296 | stuff, and assumed # could never appear in a domain. | |
297 | ||
298 | So the old code is now cut out for Release 4.11 onwards, on 09-Aug-02. In a few | |
299 | years, when we are sure this isn't actually causing trouble, throw it away. | |
300 | ||
301 | March 2003: the story continues: There is a camp that is arguing for the use of | |
302 | UTF-8 in domain names as the way to internationalization, and other MTAs | |
303 | support this. Therefore, we now have a flag that permits the use of characters | |
304 | with values greater than 127, encoded in UTF-8, in subdomains, so that Exim can | |
305 | be used experimentally in this way. */ | |
306 | ||
307 | for (;;) | |
308 | { | |
309 | uschar *tsave = t; | |
310 | ||
311 | /********************* | |
312 | if (rfc821_domains) | |
313 | { | |
314 | if (*s != '-') while (isalnum(*s) || *s == '-') *t++ = *s++; | |
315 | } | |
316 | else | |
317 | while (!mac_iscntrl_or_special(*s)) *t++ = *s++; | |
318 | *********************/ | |
319 | ||
320 | if (*s != '-') | |
321 | { | |
322 | /* Only letters, digits, and hyphens */ | |
323 | ||
324 | if (!allow_utf8_domains) | |
325 | { | |
326 | while (isalnum(*s) || *s == '-') *t++ = *s++; | |
327 | } | |
328 | ||
329 | /* Permit legal UTF-8 characters to be included */ | |
330 | ||
331 | else for(;;) | |
332 | { | |
333 | int i, d; | |
334 | if (isalnum(*s) || *s == '-') /* legal ascii characters */ | |
335 | { | |
336 | *t++ = *s++; | |
337 | continue; | |
338 | } | |
339 | if ((*s & 0xc0) != 0xc0) break; /* not start of UTF-8 character */ | |
340 | d = *s << 2; | |
341 | for (i = 1; i < 6; i++) /* i is the number of additional bytes */ | |
342 | { | |
343 | if ((d & 0x80) == 0) break; | |
344 | d <<= 1; | |
345 | } | |
346 | if (i == 6) goto BAD_UTF8; /* invalid UTF-8 */ | |
347 | *t++ = *s++; /* leading UTF-8 byte */ | |
348 | while (i-- > 0) /* copy and check remainder */ | |
349 | { | |
350 | if ((*s & 0xc0) != 0x80) | |
351 | { | |
352 | BAD_UTF8: | |
353 | *errorptr = US"invalid UTF-8 byte sequence"; | |
354 | *tt = 0; | |
355 | return s; | |
356 | } | |
357 | *t++ = *s++; | |
358 | } | |
359 | } /* End of loop for UTF-8 character */ | |
360 | } /* End of subdomain */ | |
361 | ||
362 | s = skip_comment(s); | |
363 | *t = 0; | |
364 | ||
365 | if (t == tsave) /* empty component */ | |
366 | { | |
367 | if (strip_trailing_dot && t > tt && *s != '.') t[-1] = 0; else | |
368 | { | |
369 | *errorptr = US"domain missing or malformed"; | |
370 | *tt = 0; | |
371 | } | |
372 | return s; | |
373 | } | |
374 | ||
375 | if (*s != '.') break; | |
376 | *t++ = *s++; | |
377 | s = skip_comment(s); | |
378 | } | |
379 | ||
380 | return s; | |
381 | } | |
382 | ||
383 | ||
384 | ||
385 | /************************************************* | |
386 | * Read a local-part * | |
387 | *************************************************/ | |
388 | ||
389 | /* A local-part is a sequence of words, separated by periods. A null word | |
390 | between dots is not strictly allowed but apparently many mailers permit it, | |
391 | so, sigh, better be compatible. Even accept a trailing dot... | |
392 | ||
393 | A <word> is either a quoted string, or an <atom>, which is a sequence | |
394 | of any characters except specials, space, and controls. The specials are | |
395 | ( ) < > @ , ; : \ " . [ and ]. In RFC 822, a single quoted character, (a | |
396 | quoted-pair) is not allowed in a word. However, in RFC 821, it is permitted in | |
397 | the local part of an address. Rather than have separate parsing functions for | |
398 | the different cases, take the liberal attitude always. At least one MUA is | |
399 | happy to recognize this case; I don't know how many other programs do. | |
400 | ||
401 | Arguments: | |
402 | s current character pointer | |
403 | t where to put the local part | |
404 | error where to point error text | |
405 | allow_null TRUE if an empty local part is not an error | |
406 | ||
407 | Returns: new character pointer | |
408 | */ | |
409 | ||
410 | static uschar * | |
411 | read_local_part(uschar *s, uschar *t, uschar **error, BOOL allow_null) | |
412 | { | |
413 | uschar *tt = t; | |
414 | *error = NULL; | |
415 | for (;;) | |
416 | { | |
417 | int c; | |
418 | uschar *tsave = t; | |
419 | s = skip_comment(s); | |
420 | ||
421 | /* Handle a quoted string */ | |
422 | ||
423 | if (*s == '\"') | |
424 | { | |
425 | *t++ = '\"'; | |
426 | while ((c = *(++s)) != 0 && c != '\"') | |
427 | { | |
428 | *t++ = c; | |
429 | if (c == '\\' && s[1] != 0) *t++ = *(++s); | |
430 | } | |
431 | if (c == '\"') | |
432 | { | |
433 | s++; | |
434 | *t++ = '\"'; | |
435 | } | |
436 | else | |
437 | { | |
438 | *error = US"unmatched doublequote in local part"; | |
439 | return s; | |
440 | } | |
441 | } | |
442 | ||
443 | /* Handle an atom, but allow quoted pairs within it. */ | |
444 | ||
445 | else while (!mac_iscntrl_or_special(*s) || *s == '\\') | |
446 | { | |
447 | c = *t++ = *s++; | |
448 | if (c == '\\' && *s != 0) *t++ = *s++; | |
449 | } | |
450 | ||
451 | /* Terminate the word and skip subsequent comment */ | |
452 | ||
453 | *t = 0; | |
454 | s = skip_comment(s); | |
455 | ||
456 | /* If we have read a null component at this point, give an error unless it is | |
457 | terminated by a dot - an extension to RFC 822 - or if it is the first | |
458 | component of the local part and an empty local part is permitted, in which | |
459 | case just return normally. */ | |
460 | ||
461 | if (t == tsave && *s != '.') | |
462 | { | |
463 | if (t == tt && !allow_null) | |
464 | *error = US"missing or malformed local part"; | |
465 | return s; | |
466 | } | |
467 | ||
468 | /* Anything other than a dot terminates the local part. Treat multiple dots | |
469 | as a single dot, as this seems to be a common extension. */ | |
470 | ||
471 | if (*s != '.') break; | |
472 | do { *t++ = *s++; } while (*s == '.'); | |
473 | } | |
474 | ||
475 | return s; | |
476 | } | |
477 | ||
478 | ||
479 | /************************************************* | |
480 | * Read route part of route-addr * | |
481 | *************************************************/ | |
482 | ||
483 | /* The pointer is at the initial "@" on entry. Return it following the | |
484 | terminating colon. Exim no longer supports the use of source routes, but it is | |
485 | required to accept the syntax. | |
486 | ||
487 | Arguments: | |
488 | s current character pointer | |
489 | t where to put the route | |
490 | errorptr where to put an error message | |
491 | ||
492 | Returns: new character pointer | |
493 | */ | |
494 | ||
495 | static uschar * | |
496 | read_route(uschar *s, uschar *t, uschar **errorptr) | |
497 | { | |
498 | BOOL commas = FALSE; | |
499 | *errorptr = NULL; | |
500 | ||
501 | while (*s == '@') | |
502 | { | |
503 | *t++ = '@'; | |
504 | s = read_domain(s+1, t, errorptr); | |
505 | if (*t == 0) return s; | |
506 | t += Ustrlen((const uschar *)t); | |
507 | if (*s != ',') break; | |
508 | *t++ = *s++; | |
509 | commas = TRUE; | |
510 | s = skip_comment(s); | |
511 | } | |
512 | ||
513 | if (*s == ':') *t++ = *s++; | |
514 | ||
515 | /* If there is no colon, and there were no commas, the most likely error | |
516 | is in fact a missing local part in the address rather than a missing colon | |
517 | after the route. */ | |
518 | ||
519 | else *errorptr = commas? | |
520 | US"colon expected after route list" : | |
521 | US"no local part"; | |
522 | ||
523 | /* Terminate the route and return */ | |
524 | ||
525 | *t = 0; | |
526 | return skip_comment(s); | |
527 | } | |
528 | ||
529 | ||
530 | ||
531 | /************************************************* | |
532 | * Read addr-spec * | |
533 | *************************************************/ | |
534 | ||
535 | /* Addr-spec is local-part@domain. We make the domain optional - | |
536 | the expected terminator for the whole thing is passed to check this. | |
537 | This function is called only when we know we have a route-addr. | |
538 | ||
539 | Arguments: | |
540 | s current character pointer | |
541 | t where to put the addr-spec | |
542 | term expected terminator (0 or >) | |
543 | errorptr where to put an error message | |
544 | domainptr set to point to the start of the domain | |
545 | ||
546 | Returns: new character pointer | |
547 | */ | |
548 | ||
549 | static uschar * | |
550 | read_addr_spec(uschar *s, uschar *t, int term, uschar **errorptr, | |
551 | uschar **domainptr) | |
552 | { | |
553 | s = read_local_part(s, t, errorptr, FALSE); | |
554 | if (*errorptr == NULL) | |
555 | { | |
556 | if (*s != term) | |
557 | { | |
558 | if (*s != '@') | |
559 | *errorptr = string_sprintf("\"@\" or \".\" expected after \"%s\"", t); | |
560 | else | |
561 | { | |
562 | t += Ustrlen((const uschar *)t); | |
563 | *t++ = *s++; | |
564 | *domainptr = t; | |
565 | s = read_domain(s, t, errorptr); | |
566 | } | |
567 | } | |
568 | } | |
569 | return s; | |
570 | } | |
571 | ||
572 | ||
573 | ||
574 | /************************************************* | |
575 | * Extract operative address * | |
576 | *************************************************/ | |
577 | ||
578 | /* This function extracts an operative address from a full RFC822 mailbox and | |
579 | returns it in a piece of dynamic store. We take the easy way and get a piece | |
580 | of store the same size as the input, and then copy into it whatever is | |
581 | necessary. If we cannot find a valid address (syntax error), return NULL, and | |
582 | point the error pointer to the reason. The arguments "start" and "end" are used | |
583 | to return the offsets of the first and one past the last characters in the | |
584 | original mailbox of the address that has been extracted, to aid in re-writing. | |
585 | The argument "domain" is set to point to the first character after "@" in the | |
586 | final part of the returned address, or zero if there is no @. | |
587 | ||
588 | Exim no longer supports the use of source routed addresses (those of the form | |
589 | @domain,...:route_addr). It recognizes the syntax, but collapses such addresses | |
590 | down to their final components. Formerly, collapse_source_routes had to be set | |
591 | to achieve this effect. RFC 1123 allows collapsing with MAY, while the revision | |
592 | of RFC 821 had increased this to SHOULD, so I've gone for it, because it makes | |
593 | a lot of code elsewhere in Exim much simpler. | |
594 | ||
595 | There are some special fudges here for handling RFC 822 group address notation | |
596 | which may appear in certain headers. If the flag parse_allow_group is set | |
597 | TRUE and parse_found_group is FALSE when this function is called, an address | |
598 | which is the start of a group (i.e. preceded by a phrase and a colon) is | |
599 | recognized; the phrase is ignored and the flag parse_found_group is set. If | |
1eccaa59 PH |
600 | this flag is TRUE at the end of an address, and if an extraneous semicolon is |
601 | found, it is ignored and the flag is cleared. | |
602 | ||
603 | This logic is used only when scanning through addresses in headers, either to | |
604 | fulfil the -t option, or for rewriting, or for checking header syntax. Because | |
605 | the group "state" has to be remembered between multiple calls of this function, | |
606 | the variables parse_{allow,found}_group are global. It is important to ensure | |
607 | that they are reset to FALSE at the end of scanning a header's list of | |
608 | addresses. | |
059ec3d9 PH |
609 | |
610 | Arguments: | |
611 | mailbox points to the RFC822 mailbox | |
612 | errorptr where to point an error message | |
613 | start set to start offset in mailbox | |
614 | end set to end offset in mailbox | |
615 | domain set to domain offset in result, or 0 if no domain present | |
616 | allow_null allow <> if TRUE | |
617 | ||
618 | Returns: points to the extracted address, or NULL on error | |
619 | */ | |
620 | ||
621 | #define FAILED(s) { *errorptr = s; goto PARSE_FAILED; } | |
622 | ||
623 | uschar * | |
624 | parse_extract_address(uschar *mailbox, uschar **errorptr, int *start, int *end, | |
625 | int *domain, BOOL allow_null) | |
626 | { | |
627 | uschar *yield = store_get(Ustrlen(mailbox) + 1); | |
628 | uschar *startptr, *endptr; | |
629 | uschar *s = (uschar *)mailbox; | |
630 | uschar *t = (uschar *)yield; | |
631 | ||
632 | *domain = 0; | |
633 | ||
634 | /* At the start of the string we expect either an addr-spec or a phrase | |
635 | preceding a <route-addr>. If groups are allowed, we might also find a phrase | |
636 | preceding a colon and an address. If we find an initial word followed by | |
637 | a dot, strict interpretation of the RFC would cause it to be taken | |
638 | as the start of an addr-spec. However, many mailers break the rules | |
639 | and use addresses of the form "a.n.other <ano@somewhere>" and so we | |
640 | allow this case. */ | |
641 | ||
642 | RESTART: /* Come back here after passing a group name */ | |
643 | ||
644 | s = skip_comment(s); | |
645 | startptr = s; /* In case addr-spec */ | |
646 | s = read_local_part(s, t, errorptr, TRUE); /* Dot separated words */ | |
647 | if (*errorptr != NULL) goto PARSE_FAILED; | |
648 | ||
649 | /* If the terminator is neither < nor @ then the format of the address | |
650 | must either be a bare local-part (we are now at the end), or a phrase | |
651 | followed by a route-addr (more words must follow). */ | |
652 | ||
653 | if (*s != '@' && *s != '<') | |
654 | { | |
655 | if (*s == 0 || *s == ';') | |
656 | { | |
657 | if (*t == 0) FAILED(US"empty address"); | |
658 | endptr = last_comment_position; | |
659 | goto PARSE_SUCCEEDED; /* Bare local part */ | |
660 | } | |
661 | ||
662 | /* Expect phrase route-addr, or phrase : if groups permitted, but allow | |
663 | dots in the phrase; complete the loop only when '<' or ':' is encountered - | |
664 | end of string will produce a null local_part and therefore fail. We don't | |
665 | need to keep updating t, as the phrase isn't to be kept. */ | |
666 | ||
667 | while (*s != '<' && (!parse_allow_group || *s != ':')) | |
668 | { | |
669 | s = read_local_part(s, t, errorptr, FALSE); | |
670 | if (*errorptr != NULL) | |
671 | { | |
672 | *errorptr = string_sprintf("%s (expected word or \"<\")", *errorptr); | |
673 | goto PARSE_FAILED; | |
674 | } | |
675 | } | |
676 | ||
677 | if (*s == ':') | |
678 | { | |
679 | parse_found_group = TRUE; | |
680 | parse_allow_group = FALSE; | |
681 | s++; | |
682 | goto RESTART; | |
683 | } | |
684 | ||
685 | /* Assert *s == '<' */ | |
686 | } | |
687 | ||
688 | /* At this point the next character is either '@' or '<'. If it is '@', only a | |
689 | single local-part has previously been read. An angle bracket signifies the | |
690 | start of an <addr-spec>. Throw away anything we have saved so far before | |
691 | processing it. Note that this is "if" rather than "else if" because it's also | |
692 | used after reading a preceding phrase. | |
693 | ||
694 | There are a lot of broken sendmails out there that put additional pairs of <> | |
695 | round <route-addr>s. If strip_excess_angle_brackets is set, allow any number of | |
696 | them, as long as they match. */ | |
697 | ||
698 | if (*s == '<') | |
699 | { | |
700 | uschar *domainptr = yield; | |
701 | BOOL source_routed = FALSE; | |
702 | int bracket_count = 1; | |
703 | ||
704 | s++; | |
705 | if (strip_excess_angle_brackets) | |
706 | while (*s == '<') { bracket_count++; s++; } | |
707 | ||
708 | t = yield; | |
709 | startptr = s; | |
710 | s = skip_comment(s); | |
711 | ||
712 | /* Read an optional series of routes, each of which is a domain. They | |
713 | are separated by commas and terminated by a colon. However, we totally ignore | |
714 | such routes (RFC 1123 says we MAY, and the revision of RFC 821 says we | |
715 | SHOULD). */ | |
716 | ||
717 | if (*s == '@') | |
718 | { | |
719 | s = read_route(s, t, errorptr); | |
720 | if (*errorptr != NULL) goto PARSE_FAILED; | |
721 | *t = 0; /* Ensure route is ignored - probably overkill */ | |
722 | source_routed = TRUE; | |
723 | } | |
724 | ||
725 | /* Now an addr-spec, terminated by '>'. If there is no preceding route, | |
726 | we must allow an empty addr-spec if allow_null is TRUE, to permit the | |
727 | address "<>" in some circumstances. A source-routed address MUST have | |
728 | a domain in the final part. */ | |
729 | ||
730 | if (allow_null && !source_routed && *s == '>') | |
731 | { | |
732 | *t = 0; | |
733 | *errorptr = NULL; | |
734 | } | |
735 | else | |
736 | { | |
737 | s = read_addr_spec(s, t, '>', errorptr, &domainptr); | |
738 | if (*errorptr != NULL) goto PARSE_FAILED; | |
739 | *domain = domainptr - yield; | |
740 | if (source_routed && *domain == 0) | |
741 | FAILED(US"domain missing in source-routed address"); | |
742 | } | |
743 | ||
744 | endptr = s; | |
745 | if (*errorptr != NULL) goto PARSE_FAILED; | |
746 | while (bracket_count-- > 0) if (*s++ != '>') | |
747 | { | |
748 | *errorptr = (s[-1] == 0)? US"'>' missing at end of address" : | |
749 | string_sprintf("malformed address: %.32s may not follow %.*s", | |
750 | s-1, s - (uschar *)mailbox - 1, mailbox); | |
751 | goto PARSE_FAILED; | |
752 | } | |
753 | ||
754 | s = skip_comment(s); | |
755 | } | |
756 | ||
757 | /* Hitting '@' after the first local-part means we have definitely got an | |
758 | addr-spec, on a strict reading of the RFC, and the rest of the string | |
759 | should be the domain. However, for flexibility we allow for a route-address | |
760 | not enclosed in <> as well, which is indicated by an empty first local | |
761 | part preceding '@'. The source routing is, however, ignored. */ | |
762 | ||
763 | else if (*t == 0) | |
764 | { | |
765 | uschar *domainptr = yield; | |
766 | s = read_route(s, t, errorptr); | |
767 | if (*errorptr != NULL) goto PARSE_FAILED; | |
768 | *t = 0; /* Ensure route is ignored - probably overkill */ | |
769 | s = read_addr_spec(s, t, 0, errorptr, &domainptr); | |
770 | if (*errorptr != NULL) goto PARSE_FAILED; | |
771 | *domain = domainptr - yield; | |
772 | endptr = last_comment_position; | |
773 | if (*domain == 0) FAILED(US"domain missing in source-routed address"); | |
774 | } | |
775 | ||
776 | /* This is the strict case of local-part@domain. */ | |
777 | ||
778 | else | |
779 | { | |
780 | t += Ustrlen((const uschar *)t); | |
781 | *t++ = *s++; | |
782 | *domain = t - yield; | |
783 | s = read_domain(s, t, errorptr); | |
784 | if (*t == 0) goto PARSE_FAILED; | |
785 | endptr = last_comment_position; | |
786 | } | |
787 | ||
788 | /* Use goto to get here from the bare local part case. Arrive by falling | |
789 | through for other cases. Endptr may have been moved over whitespace, so | |
790 | move it back past white space if necessary. */ | |
791 | ||
792 | PARSE_SUCCEEDED: | |
793 | if (*s != 0) | |
794 | { | |
795 | if (parse_found_group && *s == ';') | |
796 | { | |
797 | parse_found_group = FALSE; | |
798 | parse_allow_group = TRUE; | |
799 | } | |
800 | else | |
801 | { | |
802 | *errorptr = string_sprintf("malformed address: %.32s may not follow %.*s", | |
803 | s, s - (uschar *)mailbox, mailbox); | |
804 | goto PARSE_FAILED; | |
805 | } | |
806 | } | |
807 | *start = startptr - (uschar *)mailbox; /* Return offsets */ | |
808 | while (isspace(endptr[-1])) endptr--; | |
809 | *end = endptr - (uschar *)mailbox; | |
810 | ||
811 | /* Although this code has no limitation on the length of address extracted, | |
812 | other parts of Exim may have limits, and in any case, RFC 2821 limits local | |
813 | parts to 64 and domains to 255, so we do a check here, giving an error if the | |
814 | address is ridiculously long. */ | |
815 | ||
816 | if (*end - *start > ADDRESS_MAXLENGTH) | |
817 | { | |
818 | *errorptr = string_sprintf("address is ridiculously long: %.64s...", yield); | |
819 | return NULL; | |
820 | } | |
821 | ||
822 | return (uschar *)yield; | |
823 | ||
824 | /* Use goto (via the macro FAILED) to get to here from a variety of places. | |
825 | We might have an empty address in a group - the caller can choose to ignore | |
826 | this. We must, however, keep the flags correct. */ | |
827 | ||
828 | PARSE_FAILED: | |
829 | if (parse_found_group && *s == ';') | |
830 | { | |
831 | parse_found_group = FALSE; | |
832 | parse_allow_group = TRUE; | |
833 | } | |
834 | return NULL; | |
835 | } | |
836 | ||
837 | #undef FAILED | |
838 | ||
839 | ||
840 | ||
841 | /************************************************* | |
842 | * Quote according to RFC 2047 * | |
843 | *************************************************/ | |
844 | ||
845 | /* This function is used for quoting text in headers according to RFC 2047. | |
846 | If the only characters that strictly need quoting are spaces, we return the | |
847 | original string, unmodified. If a quoted string is too long for the buffer, it | |
848 | is truncated. (This shouldn't happen: this is normally handling short strings.) | |
849 | ||
8e669ac1 PH |
850 | Hmmph. As always, things get perverted for other uses. This function was |
851 | originally for the "phrase" part of addresses. Now it is being used for much | |
852 | longer texts in ACLs and via the ${rfc2047: expansion item. This means we have | |
14702f5b PH |
853 | to check for overlong "encoded-word"s and split them. November 2004. |
854 | ||
059ec3d9 PH |
855 | Arguments: |
856 | string the string to quote - already checked to contain non-printing | |
857 | chars | |
858 | len the length of the string | |
859 | charset the name of the character set; NULL => iso-8859-1 | |
860 | buffer the buffer to put the answer in | |
861 | buffer_size the size of the buffer | |
46218253 PH |
862 | fold if TRUE, a newline is inserted before the separating space when |
863 | more than one encoded-word is generated | |
059ec3d9 PH |
864 | |
865 | Returns: pointer to the original string, if no quoting needed, or | |
866 | pointer to buffer containing the quoted string, or | |
867 | a pointer to "String too long" if the buffer can't even hold | |
868 | the introduction | |
869 | */ | |
870 | ||
871 | uschar * | |
872 | parse_quote_2047(uschar *string, int len, uschar *charset, uschar *buffer, | |
46218253 | 873 | int buffer_size, BOOL fold) |
059ec3d9 PH |
874 | { |
875 | uschar *s = string; | |
14702f5b PH |
876 | uschar *p, *t; |
877 | int hlen; | |
059ec3d9 PH |
878 | BOOL coded = FALSE; |
879 | ||
880 | if (charset == NULL) charset = US"iso-8859-1"; | |
881 | ||
882 | /* We don't expect this to fail! */ | |
883 | ||
884 | if (!string_format(buffer, buffer_size, "=?%s?Q?", charset)) | |
885 | return US"String too long"; | |
886 | ||
14702f5b PH |
887 | hlen = Ustrlen(buffer); |
888 | t = buffer + hlen; | |
889 | p = buffer; | |
890 | ||
059ec3d9 PH |
891 | for (; len > 0; len--) |
892 | { | |
893 | int ch = *s++; | |
14702f5b | 894 | if (t > buffer + buffer_size - hlen - 8) break; |
8e669ac1 | 895 | |
14702f5b PH |
896 | if (t - p > 70) |
897 | { | |
898 | *t++ = '?'; | |
899 | *t++ = '='; | |
46218253 | 900 | if (fold) *t++ = '\n'; |
14702f5b PH |
901 | *t++ = ' '; |
902 | p = t; | |
903 | Ustrncpy(p, buffer, hlen); | |
904 | t += hlen; | |
8e669ac1 PH |
905 | } |
906 | ||
059ec3d9 PH |
907 | if (ch < 33 || ch > 126 || |
908 | Ustrchr("?=()<>@,;:\\\".[]_", ch) != NULL) | |
909 | { | |
910 | if (ch == ' ') *t++ = '_'; else | |
911 | { | |
912 | sprintf(CS t, "=%02X", ch); | |
913 | while (*t != 0) t++; | |
914 | coded = TRUE; | |
915 | } | |
916 | } | |
917 | else *t++ = ch; | |
918 | } | |
8e669ac1 | 919 | |
14702f5b | 920 | *t++ = '?'; |
8e669ac1 | 921 | *t++ = '='; |
14702f5b | 922 | *t = 0; |
8e669ac1 | 923 | |
059ec3d9 PH |
924 | return coded? buffer : string; |
925 | } | |
926 | ||
927 | ||
928 | ||
929 | ||
930 | /************************************************* | |
931 | * Fix up an RFC 822 "phrase" * | |
932 | *************************************************/ | |
933 | ||
934 | /* This function is called to repair any syntactic defects in the "phrase" part | |
935 | of an RFC822 address. In particular, it is applied to the user's name as read | |
936 | from the passwd file when accepting a local message, and to the data from the | |
937 | -F option. | |
938 | ||
939 | If the string contains existing quoted strings or comments containing | |
940 | freestanding quotes, then we just quote those bits that need quoting - | |
941 | otherwise it would get awfully messy and probably not look good. If not, we | |
942 | quote the whole thing if necessary. Thus | |
943 | ||
944 | John Q. Smith => "John Q. Smith" | |
945 | John "Jack" Smith => John "Jack" Smith | |
946 | John "Jack" Q. Smith => John "Jack" "Q." Smith | |
947 | John (Jack) Q. Smith => "John (Jack) Q. Smith" | |
948 | John ("Jack") Q. Smith => John ("Jack") "Q." Smith | |
949 | but | |
950 | John (\"Jack\") Q. Smith => "John (\"Jack\") Q. Smith" | |
951 | ||
952 | Sheesh! This is tedious code. It is a great pity that the syntax of RFC822 is | |
953 | the way it is... | |
954 | ||
955 | August 2000: Additional code added: | |
956 | ||
957 | Previously, non-printing characters were turned into question marks, which do | |
958 | not need to be quoted. | |
959 | ||
960 | Now, a different tactic is used if there are any non-printing ASCII | |
961 | characters. The encoding method from RFC 2047 is used, assuming iso-8859-1 as | |
962 | the character set. | |
963 | ||
964 | We *could* use this for all cases, getting rid of the messy original code, | |
965 | but leave it for now. It would complicate simple cases like "John Q. Smith". | |
966 | ||
967 | The result is passed back in the buffer; it is usually going to be added to | |
968 | some other string. In order to be sure there is going to be no overflow, | |
969 | restrict the length of the input to 1/4 of the buffer size - this allows for | |
970 | every single character to be quoted or encoded without overflowing, and that | |
971 | wouldn't happen because of amalgamation. If the phrase is too long, return a | |
972 | fixed string. | |
973 | ||
974 | Arguments: | |
975 | phrase an RFC822 phrase | |
976 | len the length of the phrase | |
977 | buffer a buffer to put the result in | |
978 | buffer_size the size of the buffer | |
979 | ||
980 | Returns: the fixed RFC822 phrase | |
981 | */ | |
982 | ||
983 | uschar * | |
984 | parse_fix_phrase(uschar *phrase, int len, uschar *buffer, int buffer_size) | |
985 | { | |
986 | int ch, i; | |
987 | BOOL quoted = FALSE; | |
988 | uschar *s, *t, *end, *yield; | |
989 | ||
990 | while (len > 0 && isspace(*phrase)) { phrase++; len--; } | |
991 | if (len > buffer_size/4) return US"Name too long"; | |
992 | ||
993 | /* See if there are any non-printing characters, and if so, use the RFC 2047 | |
994 | encoding for the whole thing. */ | |
995 | ||
996 | for (i = 0, s = phrase; i < len; i++, s++) | |
997 | if ((*s < 32 && *s != '\t') || *s > 126) break; | |
998 | ||
999 | if (i < len) return parse_quote_2047(phrase, len, headers_charset, buffer, | |
46218253 | 1000 | buffer_size, FALSE); |
059ec3d9 PH |
1001 | |
1002 | /* No non-printers; use the RFC 822 quoting rules */ | |
1003 | ||
1004 | s = phrase; | |
1005 | end = s + len; | |
1006 | yield = t = buffer + 1; | |
1007 | ||
1008 | while (s < end) | |
1009 | { | |
1010 | ch = *s++; | |
1011 | ||
1012 | /* Copy over quoted strings, remembering we encountered one */ | |
1013 | ||
1014 | if (ch == '\"') | |
1015 | { | |
1016 | *t++ = '\"'; | |
1017 | while (s < end && (ch = *s++) != '\"') | |
1018 | { | |
1019 | *t++ = ch; | |
1020 | if (ch == '\\' && s < end) *t++ = *s++; | |
1021 | } | |
1022 | *t++ = '\"'; | |
1023 | if (s >= end) break; | |
1024 | quoted = TRUE; | |
1025 | } | |
1026 | ||
1027 | /* Copy over comments, noting if they contain freestanding quote | |
1028 | characters */ | |
1029 | ||
1030 | else if (ch == '(') | |
1031 | { | |
1032 | int level = 1; | |
1033 | *t++ = '('; | |
1034 | while (s < end) | |
1035 | { | |
1036 | ch = *s++; | |
1037 | *t++ = ch; | |
1038 | if (ch == '(') level++; | |
1039 | else if (ch == ')') { if (--level <= 0) break; } | |
1040 | else if (ch == '\\' && s < end) *t++ = *s++ & 127; | |
1041 | else if (ch == '\"') quoted = TRUE; | |
1042 | } | |
1043 | if (ch == 0) | |
1044 | { | |
1045 | while (level--) *t++ = ')'; | |
1046 | break; | |
1047 | } | |
1048 | } | |
1049 | ||
1050 | /* Handle special characters that need to be quoted */ | |
1051 | ||
1052 | else if (Ustrchr(")<>@,;:\\.[]", ch) != NULL) | |
1053 | { | |
1054 | /* If hit previous quotes just make one quoted "word" */ | |
1055 | ||
1056 | if (quoted) | |
1057 | { | |
1058 | uschar *tt = t++; | |
1059 | while (*(--tt) != ' ' && *tt != '\"' && *tt != ')') tt[1] = *tt; | |
1060 | tt[1] = '\"'; | |
1061 | *t++ = ch; | |
1062 | while (s < end) | |
1063 | { | |
1064 | ch = *s++; | |
1065 | if (ch == ' ' || ch == '\"') { s--; break; } else *t++ = ch; | |
1066 | } | |
1067 | *t++ = '\"'; | |
1068 | } | |
1069 | ||
1070 | /* Else quote the whole string so far, and the rest up to any following | |
1071 | quotes. We must treat anything following a backslash as a literal. */ | |
1072 | ||
1073 | else | |
1074 | { | |
1075 | BOOL escaped = (ch == '\\'); | |
1076 | *(--yield) = '\"'; | |
1077 | *t++ = ch; | |
1078 | ||
1079 | /* Now look for the end or a quote */ | |
1080 | ||
1081 | while (s < end) | |
1082 | { | |
1083 | ch = *s++; | |
1084 | ||
1085 | /* Handle escaped pairs */ | |
1086 | ||
1087 | if (escaped) | |
1088 | { | |
1089 | *t++ = ch; | |
1090 | escaped = FALSE; | |
1091 | } | |
1092 | ||
1093 | else if (ch == '\\') | |
1094 | { | |
1095 | *t++ = ch; | |
1096 | escaped = TRUE; | |
1097 | } | |
1098 | ||
1099 | /* If hit subsequent quotes, insert our quote before any trailing | |
1100 | spaces and back up to re-handle the quote in the outer loop. */ | |
1101 | ||
1102 | else if (ch == '\"') | |
1103 | { | |
1104 | int count = 0; | |
1105 | while (t[-1] == ' ') { t--; count++; } | |
1106 | *t++ = '\"'; | |
1107 | while (count-- > 0) *t++ = ' '; | |
1108 | s--; | |
1109 | break; | |
1110 | } | |
1111 | ||
1112 | /* If hit a subsequent comment, check it for unescaped quotes, | |
1113 | and if so, end our quote before it. */ | |
1114 | ||
1115 | else if (ch == '(') | |
1116 | { | |
1117 | uschar *ss = s; /* uschar after '(' */ | |
1118 | int level = 1; | |
1119 | while(ss < end) | |
1120 | { | |
1121 | ch = *ss++; | |
1122 | if (ch == '(') level++; | |
1123 | else if (ch == ')') { if (--level <= 0) break; } | |
1124 | else if (ch == '\\' && ss+1 < end) ss++; | |
1125 | else if (ch == '\"') { quoted = TRUE; break; } | |
1126 | } | |
1127 | ||
1128 | /* Comment contains unescaped quotes; end our quote before | |
1129 | the start of the comment. */ | |
1130 | ||
1131 | if (quoted) | |
1132 | { | |
1133 | int count = 0; | |
1134 | while (t[-1] == ' ') { t--; count++; } | |
1135 | *t++ = '\"'; | |
1136 | while (count-- > 0) *t++ = ' '; | |
1137 | break; | |
1138 | } | |
1139 | ||
1140 | /* Comment does not contain unescaped quotes; include it in | |
1141 | our quote. */ | |
1142 | ||
1143 | else | |
1144 | { | |
1145 | if (ss >= end) ss--; | |
1146 | *t++ = '('; | |
1147 | Ustrncpy(t, s, ss-s); | |
1148 | t += ss-s; | |
1149 | s = ss; | |
1150 | } | |
1151 | } | |
1152 | ||
1153 | /* Not a comment or quote; include this character in our quotes. */ | |
1154 | ||
1155 | else *t++ = ch; | |
1156 | } | |
1157 | } | |
1158 | ||
1159 | /* Add a final quote if we hit the end of the string. */ | |
1160 | ||
1161 | if (s >= end) *t++ = '\"'; | |
1162 | } | |
1163 | ||
1164 | /* Non-special character; just copy it over */ | |
1165 | ||
1166 | else *t++ = ch; | |
1167 | } | |
1168 | ||
1169 | *t = 0; | |
1170 | return yield; | |
1171 | } | |
1172 | ||
1173 | ||
1174 | /************************************************* | |
1175 | * Extract addresses from a list * | |
1176 | *************************************************/ | |
1177 | ||
1178 | /* This function is called by the redirect router to scan a string containing a | |
1179 | list of addresses separated by commas (with optional white space) or by | |
1180 | newlines, and to generate a chain of address items from them. In other words, | |
1181 | to unpick data from an alias or .forward file. | |
1182 | ||
1183 | The SunOS5 documentation for alias files is not very clear on the syntax; it | |
1184 | does not say that either a comma or a newline can be used for separation. | |
1185 | However, that is the way Smail does it, so we follow suit. | |
1186 | ||
1187 | If a # character is encountered in a white space position, then characters from | |
1188 | there to the next newline are skipped. | |
1189 | ||
1190 | If an unqualified address begins with '\', just skip that character. This gives | |
1191 | compatibility with Sendmail's use of \ to prevent looping. Exim has its own | |
1192 | loop prevention scheme which handles other cases too - see the code in | |
1193 | route_address(). | |
1194 | ||
1195 | An "address" can be a specification of a file or a pipe; the latter may often | |
1196 | need to be quoted because it may contain spaces, but we don't want to retain | |
1197 | the quotes. Quotes may appear in normal addresses too, and should be retained. | |
1198 | We can distinguish between these cases, because in addresses, quotes are used | |
1199 | only for parts of the address, not the whole thing. Therefore, we remove quotes | |
1200 | from items when they entirely enclose them, but not otherwise. | |
1201 | ||
1202 | An "address" can also be of the form :include:pathname to include a list of | |
1203 | addresses contained in the specified file. | |
1204 | ||
1205 | Any unqualified addresses are qualified with and rewritten if necessary, via | |
1206 | the rewrite_address() function. | |
1207 | ||
1208 | Arguments: | |
1209 | s the list of addresses (typically a complete | |
1210 | .forward file or a list of entries in an alias file) | |
1211 | options option bits for permitting or denying various special cases; | |
1212 | not all bits are relevant here - some are for filter | |
1213 | files; those we use here are: | |
1214 | RDO_DEFER | |
1215 | RDO_FREEZE | |
1216 | RDO_FAIL | |
1217 | RDO_BLACKHOLE | |
1218 | RDO_REWRITE | |
1219 | RDO_INCLUDE | |
1220 | anchor where to hang the chain of newly-created addresses. This | |
1221 | should be initialized to NULL. | |
1222 | error where to return an error text | |
1223 | incoming domain domain of the incoming address; used to qualify unqualified | |
1224 | local parts preceded by \ | |
1225 | directory if NULL, no checks are done on :include: files | |
1226 | otherwise, included file names must start with the given | |
1227 | directory | |
1228 | syntax_errors if not NULL, it carries on after syntax errors in addresses, | |
1229 | building up a list of errors as error blocks chained on | |
1230 | here. | |
1231 | ||
1232 | Returns: FF_DELIVERED addresses extracted | |
1233 | FF_NOTDELIVERED no addresses extracted, but no errors | |
1234 | FF_BLACKHOLE :blackhole: | |
1235 | FF_DEFER :defer: | |
1236 | FF_FAIL :fail: | |
1237 | FF_INCLUDEFAIL some problem with :include:; *error set | |
1238 | FF_ERROR other problems; *error is set | |
1239 | */ | |
1240 | ||
1241 | int | |
1242 | parse_forward_list(uschar *s, int options, address_item **anchor, | |
1243 | uschar **error, uschar *incoming_domain, uschar *directory, | |
1244 | error_block **syntax_errors) | |
1245 | { | |
1246 | int count = 0; | |
1247 | ||
1248 | DEBUG(D_route) debug_printf("parse_forward_list: %s\n", s); | |
1249 | ||
1250 | for (;;) | |
1251 | { | |
1252 | int len; | |
1253 | int special = 0; | |
1254 | int specopt = 0; | |
1255 | int specbit = 0; | |
1256 | uschar *ss, *nexts; | |
1257 | address_item *addr; | |
1258 | BOOL inquote = FALSE; | |
1259 | ||
1260 | for (;;) | |
1261 | { | |
1262 | while (isspace(*s) || *s == ',') s++; | |
1263 | if (*s == '#') { while (*s != 0 && *s != '\n') s++; } else break; | |
1264 | } | |
1265 | ||
1266 | /* When we reach the end of the list, we return FF_DELIVERED if any child | |
1267 | addresses have been generated. If nothing has been generated, there are two | |
1268 | possibilities: either the list is really empty, or there were syntax errors | |
1269 | that are being skipped. (If syntax errors are not being skipped, an FF_ERROR | |
1270 | return is generated on hitting a syntax error and we don't get here.) For a | |
1271 | truly empty list we return FF_NOTDELIVERED so that the router can decline. | |
1272 | However, if the list is empty only because syntax errors were skipped, we | |
1273 | return FF_DELIVERED. */ | |
1274 | ||
1275 | if (*s == 0) | |
1276 | { | |
1277 | return (count > 0 || (syntax_errors != NULL && *syntax_errors != NULL))? | |
1278 | FF_DELIVERED : FF_NOTDELIVERED; | |
1279 | ||
1280 | /* This previous code returns FF_ERROR if nothing is generated but a | |
1281 | syntax error has been skipped. I now think it is the wrong approach, but | |
1282 | have left this here just in case, and for the record. */ | |
1283 | ||
1284 | #ifdef NEVER | |
1285 | if (count > 0) return FF_DELIVERED; /* Something was generated */ | |
1286 | ||
1287 | if (syntax_errors == NULL || /* Not skipping syntax errors, or */ | |
1288 | *syntax_errors == NULL) /* we didn't actually skip any */ | |
1289 | return FF_NOTDELIVERED; | |
1290 | ||
1291 | *error = string_sprintf("no addresses generated: syntax error in %s: %s", | |
1292 | (*syntax_errors)->text2, (*syntax_errors)->text1); | |
1293 | return FF_ERROR; | |
1294 | #endif | |
1295 | ||
1296 | } | |
1297 | ||
1298 | /* Find the end of the next address. Quoted strings in addresses may contain | |
1299 | escaped characters; I haven't found a proper specification of .forward or | |
1300 | alias files that mentions the quoting properties, but it seems right to do | |
1301 | the escaping thing in all cases, so use the function that finds the end of an | |
1302 | address. However, don't let a quoted string extend over the end of a line. */ | |
1303 | ||
1304 | ss = parse_find_address_end(s, TRUE); | |
1305 | ||
1306 | /* Remember where we finished, for starting the next one. */ | |
1307 | ||
1308 | nexts = ss; | |
1309 | ||
1310 | /* Remove any trailing spaces; we know there's at least one non-space. */ | |
1311 | ||
1312 | while (isspace((ss[-1]))) ss--; | |
1313 | ||
1314 | /* We now have s->start and ss->end of the next address. Remove quotes | |
1315 | if they completely enclose, remembering the address started with a quote | |
1316 | for handling pipes and files. Another round of removal of leading and | |
1317 | trailing spaces is then required. */ | |
1318 | ||
1319 | if (*s == '\"' && ss[-1] == '\"') | |
1320 | { | |
1321 | s++; | |
1322 | ss--; | |
1323 | inquote = TRUE; | |
1324 | while (s < ss && isspace(*s)) s++; | |
1325 | while (ss > s && isspace((ss[-1]))) ss--; | |
1326 | } | |
1327 | ||
1328 | /* Set up the length of the address. */ | |
1329 | ||
1330 | len = ss - s; | |
1331 | ||
1332 | DEBUG(D_route) | |
1333 | { | |
1334 | int save = s[len]; | |
1335 | s[len] = 0; | |
1336 | debug_printf("extract item: %s\n", s); | |
1337 | s[len] = save; | |
1338 | } | |
1339 | ||
1340 | /* Handle special addresses if permitted. If the address is :unknown: | |
1341 | ignore it - this is for backward compatibility with old alias files. You | |
1342 | don't need to use it nowadays - just generate an empty string. For :defer:, | |
1343 | :blackhole:, or :fail: we have to set up the error message and give up right | |
1344 | away. */ | |
1345 | ||
1346 | if (Ustrncmp(s, ":unknown:", len) == 0) | |
1347 | { | |
1348 | s = nexts; | |
1349 | continue; | |
1350 | } | |
1351 | ||
1352 | if (Ustrncmp(s, ":defer:", 7) == 0) | |
1353 | { special = FF_DEFER; specopt = RDO_DEFER; } /* specbit is 0 */ | |
1354 | else if (Ustrncmp(s, ":blackhole:", 11) == 0) | |
1355 | { special = FF_BLACKHOLE; specopt = specbit = RDO_BLACKHOLE; } | |
1356 | else if (Ustrncmp(s, ":fail:", 6) == 0) | |
1357 | { special = FF_FAIL; specopt = RDO_FAIL; } /* specbit is 0 */ | |
1358 | ||
1359 | if (special != 0) | |
1360 | { | |
1361 | uschar *ss = Ustrchr(s+1, ':') + 1; | |
1362 | if ((options & specopt) == specbit) | |
1363 | { | |
1364 | *error = string_sprintf("\"%.*s\" is not permitted", len, s); | |
1365 | return FF_ERROR; | |
1366 | } | |
1367 | while (*ss != 0 && isspace(*ss)) ss++; | |
1368 | while (s[len] != 0 && s[len] != '\n') len++; | |
1369 | s[len] = 0; | |
1370 | *error = string_copy(ss); | |
1371 | return special; | |
1372 | } | |
1373 | ||
1374 | /* If the address is of the form :include:pathname, read the file, and call | |
1375 | this function recursively to extract the addresses from it. If directory is | |
1376 | NULL, do no checks. Otherwise, insist that the file name starts with the | |
1377 | given directory and is a regular file. */ | |
1378 | ||
1379 | if (Ustrncmp(s, ":include:", 9) == 0) | |
1380 | { | |
1381 | uschar *filebuf; | |
1382 | uschar filename[256]; | |
1383 | uschar *t = s+9; | |
1384 | int flen = len - 9; | |
1385 | int frc; | |
1386 | struct stat statbuf; | |
1387 | address_item *last; | |
1388 | FILE *f; | |
1389 | ||
1390 | while (flen > 0 && isspace(*t)) { t++; flen--; } | |
1391 | ||
1392 | if (flen <= 0) | |
1393 | { | |
1394 | *error = string_sprintf("file name missing after :include:"); | |
1395 | return FF_ERROR; | |
1396 | } | |
1397 | ||
1398 | if (flen > 255) | |
1399 | { | |
1400 | *error = string_sprintf("included file name \"%s\" is too long", t); | |
1401 | return FF_ERROR; | |
1402 | } | |
1403 | ||
1404 | Ustrncpy(filename, t, flen); | |
1405 | filename[flen] = 0; | |
1406 | ||
1407 | /* Insist on absolute path */ | |
1408 | ||
1409 | if (filename[0]!= '/') | |
1410 | { | |
1411 | *error = string_sprintf("included file \"%s\" is not an absolute path", | |
1412 | filename); | |
1413 | return FF_ERROR; | |
1414 | } | |
1415 | ||
1416 | /* Check if include is permitted */ | |
1417 | ||
1418 | if ((options & RDO_INCLUDE) != 0) | |
1419 | { | |
1420 | *error = US"included files not permitted"; | |
1421 | return FF_ERROR; | |
1422 | } | |
1423 | ||
1424 | /* Check file name if required */ | |
1425 | ||
1426 | if (directory != NULL) | |
1427 | { | |
1428 | int len = Ustrlen(directory); | |
1429 | uschar *p = filename + len; | |
1430 | ||
1431 | if (Ustrncmp(filename, directory, len) != 0 || *p != '/') | |
1432 | { | |
1433 | *error = string_sprintf("included file %s is not in directory %s", | |
1434 | filename, directory); | |
1435 | return FF_ERROR; | |
1436 | } | |
1437 | ||
1438 | /* It is necessary to check that every component inside the directory | |
1439 | is NOT a symbolic link, in order to keep the file inside the directory. | |
1440 | This is mighty tedious. It is also not totally foolproof in that it | |
1441 | leaves the possibility of a race attack, but I don't know how to do | |
1442 | any better. */ | |
1443 | ||
1444 | while (*p != 0) | |
1445 | { | |
1446 | int temp; | |
1447 | while (*(++p) != 0 && *p != '/'); | |
1448 | temp = *p; | |
1449 | *p = 0; | |
1450 | if (Ulstat(filename, &statbuf) != 0) | |
1451 | { | |
1452 | *error = string_sprintf("failed to stat %s (component of included " | |
1453 | "file)", filename); | |
1454 | *p = temp; | |
1455 | return FF_ERROR; | |
1456 | } | |
1457 | ||
1458 | *p = temp; | |
1459 | ||
1460 | if ((statbuf.st_mode & S_IFMT) == S_IFLNK) | |
1461 | { | |
1462 | *error = string_sprintf("included file %s in the %s directory " | |
1463 | "involves a symbolic link", filename, directory); | |
1464 | return FF_ERROR; | |
1465 | } | |
1466 | } | |
1467 | } | |
1468 | ||
1469 | /* Open and stat the file */ | |
1470 | ||
1471 | if ((f = Ufopen(filename, "rb")) == NULL) | |
1472 | { | |
1473 | *error = string_open_failed(errno, "included file %s", filename); | |
1474 | return FF_INCLUDEFAIL; | |
1475 | } | |
1476 | ||
1477 | if (fstat(fileno(f), &statbuf) != 0) | |
1478 | { | |
1479 | *error = string_sprintf("failed to stat included file %s: %s", | |
1480 | filename, strerror(errno)); | |
f1e894f3 | 1481 | (void)fclose(f); |
059ec3d9 PH |
1482 | return FF_INCLUDEFAIL; |
1483 | } | |
1484 | ||
1485 | /* If directory was checked, double check that we opened a regular file */ | |
1486 | ||
1487 | if (directory != NULL && (statbuf.st_mode & S_IFMT) != S_IFREG) | |
1488 | { | |
1489 | *error = string_sprintf("included file %s is not a regular file in " | |
1490 | "the %s directory", filename, directory); | |
1491 | return FF_ERROR; | |
1492 | } | |
1493 | ||
1494 | /* Get a buffer and read the contents */ | |
1495 | ||
1496 | if (statbuf.st_size > MAX_INCLUDE_SIZE) | |
1497 | { | |
1498 | *error = string_sprintf("included file %s is too big (max %d)", | |
1499 | filename, MAX_INCLUDE_SIZE); | |
1500 | return FF_ERROR; | |
1501 | } | |
1502 | ||
1503 | filebuf = store_get(statbuf.st_size + 1); | |
1504 | if (fread(filebuf, 1, statbuf.st_size, f) != statbuf.st_size) | |
1505 | { | |
1506 | *error = string_sprintf("error while reading included file %s: %s", | |
1507 | filename, strerror(errno)); | |
f1e894f3 | 1508 | (void)fclose(f); |
059ec3d9 PH |
1509 | return FF_ERROR; |
1510 | } | |
1511 | filebuf[statbuf.st_size] = 0; | |
f1e894f3 | 1512 | (void)fclose(f); |
059ec3d9 PH |
1513 | |
1514 | addr = NULL; | |
1515 | frc = parse_forward_list(filebuf, options, &addr, | |
1516 | error, incoming_domain, directory, syntax_errors); | |
1517 | if (frc != FF_DELIVERED && frc != FF_NOTDELIVERED) return frc; | |
1518 | ||
1519 | if (addr != NULL) | |
1520 | { | |
1521 | last = addr; | |
1522 | while (last->next != NULL) { count++; last = last->next; } | |
1523 | last->next = *anchor; | |
1524 | *anchor = addr; | |
1525 | count++; | |
1526 | } | |
1527 | } | |
1528 | ||
1529 | /* Else (not :include:) ensure address is syntactically correct and fully | |
1530 | qualified if not a pipe or a file, removing a leading \ if present on an | |
1531 | unqualified address. For pipes and files we must handle quoting. It's | |
1532 | not quite clear exactly what to do for partially quoted things, but the | |
1533 | common case of having the whole thing in quotes is straightforward. If this | |
1534 | was the case, inquote will have been set TRUE above and the quotes removed. | |
1535 | ||
1536 | There is a possible ambiguity over addresses whose local parts start with | |
1537 | a vertical bar or a slash, and the latter do in fact occur, thanks to X.400. | |
1538 | Consider a .forward file that contains the line | |
1539 | ||
1540 | /X=xxx/Y=xxx/OU=xxx/@some.gate.way | |
1541 | ||
1542 | Is this a file or an X.400 address? Does it make any difference if it is in | |
1543 | quotes? On the grounds that file names of this type are rare, Exim treats | |
1544 | something that parses as an RFC 822 address and has a domain as an address | |
1545 | rather than a file or a pipe. This is also how an address such as the above | |
1546 | would be treated if it came in from outside. */ | |
1547 | ||
1548 | else | |
1549 | { | |
1550 | int start, end, domain; | |
1551 | uschar *recipient = NULL; | |
1552 | int save = s[len]; | |
1553 | s[len] = 0; | |
1554 | ||
1555 | /* If it starts with \ and the rest of it parses as a valid mail address | |
1556 | without a domain, carry on with that address, but qualify it with the | |
1557 | incoming domain. Otherwise arrange for the address to fall through, | |
1558 | causing an error message on the re-parse. */ | |
1559 | ||
1560 | if (*s == '\\') | |
1561 | { | |
1562 | recipient = | |
1563 | parse_extract_address(s+1, error, &start, &end, &domain, FALSE); | |
1564 | if (recipient != NULL) | |
1565 | recipient = (domain != 0)? NULL : | |
1566 | string_sprintf("%s@%s", recipient, incoming_domain); | |
1567 | } | |
1568 | ||
1569 | /* Try parsing the item as an address. */ | |
1570 | ||
1571 | if (recipient == NULL) recipient = | |
1572 | parse_extract_address(s, error, &start, &end, &domain, FALSE); | |
1573 | ||
1574 | /* If item starts with / or | and is not a valid address, or there | |
1575 | is no domain, treat it as a file or pipe. If it was a quoted item, | |
1576 | remove the quoting occurrences of \ within it. */ | |
1577 | ||
1578 | if ((*s == '|' || *s == '/') && (recipient == NULL || domain == 0)) | |
1579 | { | |
1580 | uschar *t = store_get(Ustrlen(s) + 1); | |
1581 | uschar *p = t; | |
1582 | uschar *q = s; | |
1583 | while (*q != 0) | |
1584 | { | |
1585 | if (inquote) | |
1586 | { | |
1587 | *p++ = (*q == '\\')? *(++q) : *q; | |
1588 | q++; | |
1589 | } | |
1590 | else *p++ = *q++; | |
1591 | } | |
1592 | *p = 0; | |
1593 | addr = deliver_make_addr(t, TRUE); | |
1594 | setflag(addr, af_pfr); /* indicates pipe/file/reply */ | |
1595 | if (*s != '|') setflag(addr, af_file); /* indicates file */ | |
1596 | } | |
1597 | ||
1598 | /* Item must be an address. Complain if not, else qualify, rewrite and set | |
1599 | up the control block. It appears that people are in the habit of using | |
1600 | empty addresses but with comments as a way of putting comments into | |
1601 | alias and forward files. Therefore, ignore the error "empty address". | |
1602 | Mailing lists might want to tolerate syntax errors; there is therefore | |
1603 | an option to do so. */ | |
1604 | ||
1605 | else | |
1606 | { | |
1607 | if (recipient == NULL) | |
1608 | { | |
1609 | if (Ustrcmp(*error, "empty address") == 0) | |
1610 | { | |
1611 | *error = NULL; | |
1612 | s[len] = save; | |
1613 | s = nexts; | |
1614 | continue; | |
1615 | } | |
1616 | ||
1617 | if (syntax_errors != NULL) | |
1618 | { | |
1619 | error_block *e = store_get(sizeof(error_block)); | |
1620 | error_block *last = *syntax_errors; | |
1621 | if (last == NULL) *syntax_errors = e; else | |
1622 | { | |
1623 | while (last->next != NULL) last = last->next; | |
1624 | last->next = e; | |
1625 | } | |
1626 | e->next = NULL; | |
1627 | e->text1 = *error; | |
1628 | e->text2 = string_copy(s); | |
1629 | s[len] = save; | |
1630 | s = nexts; | |
1631 | continue; | |
1632 | } | |
1633 | else | |
1634 | { | |
1635 | *error = string_sprintf("%s in \"%s\"", *error, s); | |
1636 | s[len] = save; /* _after_ using it for *error */ | |
1637 | return FF_ERROR; | |
1638 | } | |
1639 | } | |
1640 | ||
1641 | /* Address was successfully parsed. Rewrite, and then make an address | |
1642 | block. */ | |
1643 | ||
1644 | recipient = ((options & RDO_REWRITE) != 0)? | |
1645 | rewrite_address(recipient, TRUE, FALSE, global_rewrite_rules, | |
1646 | rewrite_existflags) : | |
1647 | rewrite_address_qualify(recipient, TRUE); | |
1648 | addr = deliver_make_addr(recipient, TRUE); /* TRUE => copy recipient */ | |
1649 | } | |
1650 | ||
1651 | /* Restore the final character in the original data, and add to the | |
1652 | output chain. */ | |
1653 | ||
1654 | s[len] = save; | |
1655 | addr->next = *anchor; | |
1656 | *anchor = addr; | |
1657 | count++; | |
1658 | } | |
1659 | ||
1660 | /* Advance pointer for the next address */ | |
1661 | ||
1662 | s = nexts; | |
1663 | } | |
1664 | } | |
1665 | ||
1666 | ||
30dba1e6 PH |
1667 | |
1668 | /************************************************* | |
1669 | * Extract a Message-ID * | |
1670 | *************************************************/ | |
1671 | ||
1672 | /* This function is used to extract message ids from In-Reply-To: and | |
1673 | References: header lines. | |
1674 | ||
1675 | Arguments: | |
1676 | str pointer to the start of the message-id | |
1677 | yield put pointer to the message id (in dynamic memory) here | |
1678 | error put error message here on failure | |
1679 | ||
1680 | Returns: points after the processed message-id or NULL on error | |
1681 | */ | |
1682 | ||
1683 | uschar * | |
1684 | parse_message_id(uschar *str, uschar **yield, uschar **error) | |
1685 | { | |
1686 | uschar *domain = NULL; | |
1687 | uschar *id; | |
1688 | ||
1689 | str = skip_comment(str); | |
1690 | if (*str != '<') | |
1691 | { | |
1692 | *error = US"Missing '<' before message-id"; | |
1693 | return NULL; | |
1694 | } | |
1695 | ||
1696 | /* Getting a block the size of the input string will definitely be sufficient | |
1697 | for the answer, but it may also be very long if we are processing a header | |
1698 | line. Therefore, take care to release unwanted store afterwards. */ | |
1699 | ||
1700 | id = *yield = store_get(Ustrlen(str) + 1); | |
1701 | *id++ = *str++; | |
1702 | ||
1703 | str = read_addr_spec(str, id, '>', error, &domain); | |
1704 | ||
1705 | if (*error == NULL) | |
1706 | { | |
1707 | if (*str != '>') *error = US"Missing '>' after message-id"; | |
1708 | else if (domain == NULL) *error = US"domain missing in message-id"; | |
1709 | } | |
1710 | ||
1711 | if (*error != NULL) | |
1712 | { | |
1713 | store_reset(*yield); | |
1714 | return NULL; | |
1715 | } | |
1716 | ||
1717 | while (*id != 0) id++; | |
1718 | *id++ = *str++; | |
1719 | *id++ = 0; | |
1720 | store_reset(id); | |
1721 | ||
1722 | str = skip_comment(str); | |
1723 | return str; | |
1724 | } | |
1725 | ||
1726 | ||
1727 | ||
1728 | ||
059ec3d9 PH |
1729 | /************************************************* |
1730 | ************************************************** | |
1731 | * Stand-alone test program * | |
1732 | ************************************************** | |
1733 | *************************************************/ | |
1734 | ||
1735 | #if defined STAND_ALONE | |
1736 | int main(void) | |
1737 | { | |
1738 | int start, end, domain; | |
1739 | uschar buffer[1024]; | |
1740 | uschar outbuff[1024]; | |
1741 | ||
1742 | big_buffer = store_malloc(big_buffer_size); | |
1743 | ||
1744 | /* strip_trailing_dot = TRUE; */ | |
1745 | allow_domain_literals = TRUE; | |
1746 | ||
1747 | printf("Testing parse_fix_phrase\n"); | |
1748 | ||
1749 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
1750 | { | |
1751 | buffer[Ustrlen(buffer)-1] = 0; | |
1752 | if (buffer[0] == 0) break; | |
1753 | printf("%s\n", CS parse_fix_phrase(buffer, Ustrlen(buffer), outbuff, | |
1754 | sizeof(outbuff))); | |
1755 | } | |
1756 | ||
1757 | printf("Testing parse_extract_address without group syntax and without UTF-8\n"); | |
1758 | ||
1759 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
1760 | { | |
1761 | uschar *out; | |
1762 | uschar *errmess; | |
1763 | buffer[Ustrlen(buffer) - 1] = 0; | |
1764 | if (buffer[0] == 0) break; | |
1765 | out = parse_extract_address(buffer, &errmess, &start, &end, &domain, FALSE); | |
1766 | if (out == NULL) printf("*** bad address: %s\n", errmess); else | |
1767 | { | |
1768 | uschar extract[1024]; | |
1769 | Ustrncpy(extract, buffer+start, end-start); | |
1770 | extract[end-start] = 0; | |
1771 | printf("%s %d %d %d \"%s\"\n", out, start, end, domain, extract); | |
1772 | } | |
1773 | } | |
1774 | ||
1775 | printf("Testing parse_extract_address without group syntax but with UTF-8\n"); | |
1776 | ||
1777 | allow_utf8_domains = TRUE; | |
1778 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
1779 | { | |
1780 | uschar *out; | |
1781 | uschar *errmess; | |
1782 | buffer[Ustrlen(buffer) - 1] = 0; | |
1783 | if (buffer[0] == 0) break; | |
1784 | out = parse_extract_address(buffer, &errmess, &start, &end, &domain, FALSE); | |
1785 | if (out == NULL) printf("*** bad address: %s\n", errmess); else | |
1786 | { | |
1787 | uschar extract[1024]; | |
1788 | Ustrncpy(extract, buffer+start, end-start); | |
1789 | extract[end-start] = 0; | |
1790 | printf("%s %d %d %d \"%s\"\n", out, start, end, domain, extract); | |
1791 | } | |
1792 | } | |
1793 | allow_utf8_domains = FALSE; | |
1794 | ||
1795 | printf("Testing parse_extract_address with group syntax\n"); | |
1796 | ||
1797 | parse_allow_group = TRUE; | |
1798 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
1799 | { | |
1800 | uschar *out; | |
1801 | uschar *errmess; | |
1802 | uschar *s; | |
1803 | buffer[Ustrlen(buffer) - 1] = 0; | |
1804 | if (buffer[0] == 0) break; | |
1805 | s = buffer; | |
1806 | while (*s != 0) | |
1807 | { | |
1808 | uschar *ss = parse_find_address_end(s, FALSE); | |
1809 | int terminator = *ss; | |
1810 | *ss = 0; | |
1811 | out = parse_extract_address(buffer, &errmess, &start, &end, &domain, FALSE); | |
1812 | *ss = terminator; | |
1813 | ||
1814 | if (out == NULL) printf("*** bad address: %s\n", errmess); else | |
1815 | { | |
1816 | uschar extract[1024]; | |
1817 | Ustrncpy(extract, buffer+start, end-start); | |
1818 | extract[end-start] = 0; | |
1819 | printf("%s %d %d %d \"%s\"\n", out, start, end, domain, extract); | |
1820 | } | |
1821 | ||
1822 | s = ss + (terminator? 1:0); | |
1823 | while (isspace(*s)) s++; | |
1824 | } | |
1825 | } | |
1826 | ||
1827 | printf("Testing parse_find_at\n"); | |
1828 | ||
1829 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
1830 | { | |
1831 | uschar *s; | |
1832 | buffer[Ustrlen(buffer)-1] = 0; | |
1833 | if (buffer[0] == 0) break; | |
1834 | s = parse_find_at(buffer); | |
1835 | if (s == NULL) printf("no @ found\n"); | |
1836 | else printf("offset = %d\n", s - buffer); | |
1837 | } | |
1838 | ||
1839 | printf("Testing parse_extract_addresses\n"); | |
1840 | ||
1841 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
1842 | { | |
1843 | uschar *errmess; | |
1844 | int extracted; | |
1845 | address_item *anchor = NULL; | |
1846 | buffer[Ustrlen(buffer) - 1] = 0; | |
1847 | if (buffer[0] == 0) break; | |
1848 | if ((extracted = parse_forward_list(buffer, -1, &anchor, | |
1849 | &errmess, US"incoming.domain", NULL, NULL)) == FF_DELIVERED) | |
1850 | { | |
1851 | while (anchor != NULL) | |
1852 | { | |
1853 | address_item *addr = anchor; | |
1854 | anchor = anchor->next; | |
1855 | printf("%d %s\n", testflag(addr, af_pfr), addr->address); | |
1856 | } | |
1857 | } | |
1858 | else printf("Failed: %d %s\n", extracted, errmess); | |
1859 | } | |
1860 | ||
30dba1e6 PH |
1861 | printf("Testing parse_message_id\n"); |
1862 | ||
1863 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
1864 | { | |
1865 | uschar *s, *t, *errmess; | |
1866 | buffer[Ustrlen(buffer) - 1] = 0; | |
1867 | if (buffer[0] == 0) break; | |
1868 | s = buffer; | |
1869 | while (*s != 0) | |
1870 | { | |
1871 | s = parse_message_id(s, &t, &errmess); | |
1872 | if (errmess != NULL) | |
1873 | { | |
1874 | printf("Failed: %s\n", errmess); | |
1875 | break; | |
1876 | } | |
1877 | printf("%s\n", t); | |
1878 | } | |
1879 | } | |
1880 | ||
059ec3d9 PH |
1881 | return 0; |
1882 | } | |
1883 | ||
1884 | #endif | |
1885 | ||
1886 | /* End of parse.c */ |