6d8c93360fe084a4209277312a99fbb40d1c04bc
1 /*
2 * Another stupid program, this one parsing the headers of an
3 * email to figure out authorship and subject
4 */
5 #define _GNU_SOURCE
6 #include <stdio.h>
7 #include <stdlib.h>
8 #include <string.h>
9 #include <ctype.h>
10 #include <iconv.h>
12 #ifdef NO_STRCASESTR
13 extern char *gitstrcasestr(const char *haystack, const char *needle);
14 #endif
16 static FILE *cmitmsg, *patchfile;
18 static int keep_subject = 0;
19 static char *metainfo_charset = NULL;
20 static char line[1000];
21 static char date[1000];
22 static char name[1000];
23 static char email[1000];
24 static char subject[1000];
26 static enum {
27 TE_DONTCARE, TE_QP, TE_BASE64,
28 } transfer_encoding;
29 static char charset[256];
31 static char multipart_boundary[1000];
32 static int multipart_boundary_len;
33 static int patch_lines = 0;
35 static char *sanity_check(char *name, char *email)
36 {
37 int len = strlen(name);
38 if (len < 3 || len > 60)
39 return email;
40 if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
41 return email;
42 return name;
43 }
45 static int handle_from(char *line)
46 {
47 char *at = strchr(line, '@');
48 char *dst;
50 if (!at)
51 return 0;
53 /*
54 * If we already have one email, don't take any confusing lines
55 */
56 if (*email && strchr(at+1, '@'))
57 return 0;
59 /* Pick up the string around '@', possibly delimited with <>
60 * pair; that is the email part. White them out while copying.
61 */
62 while (at > line) {
63 char c = at[-1];
64 if (isspace(c))
65 break;
66 if (c == '<') {
67 at[-1] = ' ';
68 break;
69 }
70 at--;
71 }
72 dst = email;
73 for (;;) {
74 unsigned char c = *at;
75 if (!c || c == '>' || isspace(c)) {
76 if (c == '>')
77 *at = ' ';
78 break;
79 }
80 *at++ = ' ';
81 *dst++ = c;
82 }
83 *dst++ = 0;
85 /* The remainder is name. It could be "John Doe <john.doe@xz>"
86 * or "john.doe@xz (John Doe)", but we have whited out the
87 * email part, so trim from both ends, possibly removing
88 * the () pair at the end.
89 */
90 at = line + strlen(line);
91 while (at > line) {
92 unsigned char c = *--at;
93 if (!isspace(c)) {
94 at[(c == ')') ? 0 : 1] = 0;
95 break;
96 }
97 }
99 at = line;
100 for (;;) {
101 unsigned char c = *at;
102 if (!c || !isspace(c)) {
103 if (c == '(')
104 at++;
105 break;
106 }
107 at++;
108 }
109 at = sanity_check(at, email);
110 strcpy(name, at);
111 return 1;
112 }
114 static int handle_date(char *line)
115 {
116 strcpy(date, line);
117 return 0;
118 }
120 static int handle_subject(char *line)
121 {
122 strcpy(subject, line);
123 return 0;
124 }
126 /* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
127 * to have enough heuristics to grok MIME encoded patches often found
128 * on our mailing lists. For example, we do not even treat header lines
129 * case insensitively.
130 */
132 static int slurp_attr(const char *line, const char *name, char *attr)
133 {
134 char *ends, *ap = strcasestr(line, name);
135 size_t sz;
137 if (!ap) {
138 *attr = 0;
139 return 0;
140 }
141 ap += strlen(name);
142 if (*ap == '"') {
143 ap++;
144 ends = "\"";
145 }
146 else
147 ends = "; \t";
148 sz = strcspn(ap, ends);
149 memcpy(attr, ap, sz);
150 attr[sz] = 0;
151 return 1;
152 }
154 static int handle_subcontent_type(char *line)
155 {
156 /* We do not want to mess with boundary. Note that we do not
157 * handle nested multipart.
158 */
159 if (strcasestr(line, "boundary=")) {
160 fprintf(stderr, "Not handling nested multipart message.\n");
161 exit(1);
162 }
163 slurp_attr(line, "charset=", charset);
164 if (*charset) {
165 int i, c;
166 for (i = 0; (c = charset[i]) != 0; i++)
167 charset[i] = tolower(c);
168 }
169 return 0;
170 }
172 static int handle_content_type(char *line)
173 {
174 *multipart_boundary = 0;
175 if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
176 memcpy(multipart_boundary, "--", 2);
177 multipart_boundary_len = strlen(multipart_boundary);
178 }
179 slurp_attr(line, "charset=", charset);
180 return 0;
181 }
183 static int handle_content_transfer_encoding(char *line)
184 {
185 if (strcasestr(line, "base64"))
186 transfer_encoding = TE_BASE64;
187 else if (strcasestr(line, "quoted-printable"))
188 transfer_encoding = TE_QP;
189 else
190 transfer_encoding = TE_DONTCARE;
191 return 0;
192 }
194 static int is_multipart_boundary(const char *line)
195 {
196 return (!memcmp(line, multipart_boundary, multipart_boundary_len));
197 }
199 static int eatspace(char *line)
200 {
201 int len = strlen(line);
202 while (len > 0 && isspace(line[len-1]))
203 line[--len] = 0;
204 return len;
205 }
207 #define SEEN_FROM 01
208 #define SEEN_DATE 02
209 #define SEEN_SUBJECT 04
211 /* First lines of body can have From:, Date:, and Subject: */
212 static int handle_inbody_header(int *seen, char *line)
213 {
214 if (!memcmp("From:", line, 5) && isspace(line[5])) {
215 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
216 *seen |= SEEN_FROM;
217 return 1;
218 }
219 }
220 if (!memcmp("Date:", line, 5) && isspace(line[5])) {
221 if (!(*seen & SEEN_DATE)) {
222 handle_date(line+6);
223 *seen |= SEEN_DATE;
224 return 1;
225 }
226 }
227 if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
228 if (!(*seen & SEEN_SUBJECT)) {
229 handle_subject(line+9);
230 *seen |= SEEN_SUBJECT;
231 return 1;
232 }
233 }
234 if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
235 if (!(*seen & SEEN_SUBJECT)) {
236 handle_subject(line);
237 *seen |= SEEN_SUBJECT;
238 return 1;
239 }
240 }
241 return 0;
242 }
244 static char *cleanup_subject(char *subject)
245 {
246 if (keep_subject)
247 return subject;
248 for (;;) {
249 char *p;
250 int len, remove;
251 switch (*subject) {
252 case 'r': case 'R':
253 if (!memcmp("e:", subject+1, 2)) {
254 subject +=3;
255 continue;
256 }
257 break;
258 case ' ': case '\t': case ':':
259 subject++;
260 continue;
262 case '[':
263 p = strchr(subject, ']');
264 if (!p) {
265 subject++;
266 continue;
267 }
268 len = strlen(p);
269 remove = p - subject;
270 if (remove <= len *2) {
271 subject = p+1;
272 continue;
273 }
274 break;
275 }
276 return subject;
277 }
278 }
280 static void cleanup_space(char *buf)
281 {
282 unsigned char c;
283 while ((c = *buf) != 0) {
284 buf++;
285 if (isspace(c)) {
286 buf[-1] = ' ';
287 c = *buf;
288 while (isspace(c)) {
289 int len = strlen(buf);
290 memmove(buf, buf+1, len);
291 c = *buf;
292 }
293 }
294 }
295 }
297 typedef int (*header_fn_t)(char *);
298 struct header_def {
299 const char *name;
300 header_fn_t func;
301 int namelen;
302 };
304 static void check_header(char *line, int len, struct header_def *header)
305 {
306 int i;
308 if (header[0].namelen <= 0) {
309 for (i = 0; header[i].name; i++)
310 header[i].namelen = strlen(header[i].name);
311 }
312 for (i = 0; header[i].name; i++) {
313 int len = header[i].namelen;
314 if (!strncasecmp(line, header[i].name, len) &&
315 line[len] == ':' && isspace(line[len + 1])) {
316 header[i].func(line + len + 2);
317 break;
318 }
319 }
320 }
322 static void check_subheader_line(char *line, int len)
323 {
324 static struct header_def header[] = {
325 { "Content-Type", handle_subcontent_type },
326 { "Content-Transfer-Encoding",
327 handle_content_transfer_encoding },
328 { NULL },
329 };
330 check_header(line, len, header);
331 }
332 static void check_header_line(char *line, int len)
333 {
334 static struct header_def header[] = {
335 { "From", handle_from },
336 { "Date", handle_date },
337 { "Subject", handle_subject },
338 { "Content-Type", handle_content_type },
339 { "Content-Transfer-Encoding",
340 handle_content_transfer_encoding },
341 { NULL },
342 };
343 check_header(line, len, header);
344 }
346 static int read_one_header_line(char *line, int sz, FILE *in)
347 {
348 int ofs = 0;
349 while (ofs < sz) {
350 int peek, len;
351 if (fgets(line + ofs, sz - ofs, in) == NULL)
352 return ofs;
353 len = eatspace(line + ofs);
354 if (len == 0)
355 return ofs;
356 peek = fgetc(in); ungetc(peek, in);
357 if (peek == ' ' || peek == '\t') {
358 /* Yuck, 2822 header "folding" */
359 ofs += len;
360 continue;
361 }
362 return ofs + len;
363 }
364 return ofs;
365 }
367 static unsigned hexval(int c)
368 {
369 if (c >= '0' && c <= '9')
370 return c - '0';
371 if (c >= 'a' && c <= 'f')
372 return c - 'a' + 10;
373 if (c >= 'A' && c <= 'F')
374 return c - 'A' + 10;
375 return ~0;
376 }
378 static int decode_q_segment(char *in, char *ot, char *ep)
379 {
380 int c;
381 while ((c = *in++) != 0 && (in <= ep)) {
382 if (c == '=') {
383 int d = *in++;
384 if (d == '\n' || !d)
385 break; /* drop trailing newline */
386 *ot++ = ((hexval(d) << 4) | hexval(*in++));
387 }
388 else
389 *ot++ = c;
390 }
391 *ot = 0;
392 return 0;
393 }
395 static int decode_b_segment(char *in, char *ot, char *ep)
396 {
397 /* Decode in..ep, possibly in-place to ot */
398 int c, pos = 0, acc = 0;
400 while ((c = *in++) != 0 && (in <= ep)) {
401 if (c == '+')
402 c = 62;
403 else if (c == '/')
404 c = 63;
405 else if ('A' <= c && c <= 'Z')
406 c -= 'A';
407 else if ('a' <= c && c <= 'z')
408 c -= 'a' - 26;
409 else if ('0' <= c && c <= '9')
410 c -= '0' - 52;
411 else if (c == '=') {
412 /* padding is almost like (c == 0), except we do
413 * not output NUL resulting only from it;
414 * for now we just trust the data.
415 */
416 c = 0;
417 }
418 else
419 continue; /* garbage */
420 switch (pos++) {
421 case 0:
422 acc = (c << 2);
423 break;
424 case 1:
425 *ot++ = (acc | (c >> 4));
426 acc = (c & 15) << 4;
427 break;
428 case 2:
429 *ot++ = (acc | (c >> 2));
430 acc = (c & 3) << 6;
431 break;
432 case 3:
433 *ot++ = (acc | c);
434 acc = pos = 0;
435 break;
436 }
437 }
438 *ot = 0;
439 return 0;
440 }
442 static void convert_to_utf8(char *line, char *charset)
443 {
444 char *in, *out;
445 size_t insize, outsize, nrc;
446 char outbuf[4096]; /* cheat */
447 static char latin_one[] = "latin-1";
448 char *input_charset = *charset ? charset : latin_one;
449 iconv_t conv = iconv_open(metainfo_charset, input_charset);
451 if (conv == (iconv_t) -1) {
452 static int warned_latin1_once = 0;
453 if (input_charset != latin_one) {
454 fprintf(stderr, "cannot convert from %s to %s\n",
455 input_charset, metainfo_charset);
456 *charset = 0;
457 }
458 else if (!warned_latin1_once) {
459 warned_latin1_once = 1;
460 fprintf(stderr, "tried to convert from %s to %s, "
461 "but your iconv does not work with it.\n",
462 input_charset, metainfo_charset);
463 }
464 return;
465 }
466 in = line;
467 insize = strlen(in);
468 out = outbuf;
469 outsize = sizeof(outbuf);
470 nrc = iconv(conv, &in, &insize, &out, &outsize);
471 iconv_close(conv);
472 if (nrc == (size_t) -1)
473 return;
474 *out = 0;
475 strcpy(line, outbuf);
476 }
478 static void decode_header_bq(char *it)
479 {
480 char *in, *out, *ep, *cp, *sp;
481 char outbuf[1000];
483 in = it;
484 out = outbuf;
485 while ((ep = strstr(in, "=?")) != NULL) {
486 int sz, encoding;
487 char charset_q[256], piecebuf[256];
488 if (in != ep) {
489 sz = ep - in;
490 memcpy(out, in, sz);
491 out += sz;
492 in += sz;
493 }
494 /* E.g.
495 * ep : "=?iso-2022-jp?B?GyR...?= foo"
496 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
497 */
498 ep += 2;
499 cp = strchr(ep, '?');
500 if (!cp)
501 return; /* no munging */
502 for (sp = ep; sp < cp; sp++)
503 charset_q[sp - ep] = tolower(*sp);
504 charset_q[cp - ep] = 0;
505 encoding = cp[1];
506 if (!encoding || cp[2] != '?')
507 return; /* no munging */
508 ep = strstr(cp + 3, "?=");
509 if (!ep)
510 return; /* no munging */
511 switch (tolower(encoding)) {
512 default:
513 return; /* no munging */
514 case 'b':
515 sz = decode_b_segment(cp + 3, piecebuf, ep);
516 break;
517 case 'q':
518 sz = decode_q_segment(cp + 3, piecebuf, ep);
519 break;
520 }
521 if (sz < 0)
522 return;
523 if (metainfo_charset)
524 convert_to_utf8(piecebuf, charset_q);
525 strcpy(out, piecebuf);
526 out += strlen(out);
527 in = ep + 2;
528 }
529 strcpy(out, in);
530 strcpy(it, outbuf);
531 }
533 static void decode_transfer_encoding(char *line)
534 {
535 char *ep;
537 switch (transfer_encoding) {
538 case TE_QP:
539 ep = line + strlen(line);
540 decode_q_segment(line, line, ep);
541 break;
542 case TE_BASE64:
543 ep = line + strlen(line);
544 decode_b_segment(line, line, ep);
545 break;
546 case TE_DONTCARE:
547 break;
548 }
549 }
551 static void handle_info(void)
552 {
553 char *sub;
554 static int done_info = 0;
556 if (done_info)
557 return;
559 done_info = 1;
560 sub = cleanup_subject(subject);
561 cleanup_space(name);
562 cleanup_space(date);
563 cleanup_space(email);
564 cleanup_space(sub);
566 /* Unwrap inline B and Q encoding, and optionally
567 * normalize the meta information to utf8.
568 */
569 decode_header_bq(name);
570 decode_header_bq(date);
571 decode_header_bq(email);
572 decode_header_bq(sub);
573 printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
574 name, email, sub, date);
575 }
577 /* We are inside message body and have read line[] already.
578 * Spit out the commit log.
579 */
580 static int handle_commit_msg(void)
581 {
582 if (!cmitmsg)
583 return 0;
584 do {
585 if (!memcmp("diff -", line, 6) ||
586 !memcmp("---", line, 3) ||
587 !memcmp("Index: ", line, 7))
588 break;
589 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
590 /* We come here when the first part had only
591 * the commit message without any patch. We
592 * pretend we have not seen this line yet, and
593 * go back to the loop.
594 */
595 return 1;
596 }
598 /* Unwrap transfer encoding and optionally
599 * normalize the log message to UTF-8.
600 */
601 decode_transfer_encoding(line);
602 if (metainfo_charset)
603 convert_to_utf8(line, charset);
604 fputs(line, cmitmsg);
605 } while (fgets(line, sizeof(line), stdin) != NULL);
606 fclose(cmitmsg);
607 cmitmsg = NULL;
608 return 0;
609 }
611 /* We have done the commit message and have the first
612 * line of the patch in line[].
613 */
614 static void handle_patch(void)
615 {
616 do {
617 if (multipart_boundary[0] && is_multipart_boundary(line))
618 break;
619 /* Only unwrap transfer encoding but otherwise do not
620 * do anything. We do *NOT* want UTF-8 conversion
621 * here; we are dealing with the user payload.
622 */
623 decode_transfer_encoding(line);
624 fputs(line, patchfile);
625 patch_lines++;
626 } while (fgets(line, sizeof(line), stdin) != NULL);
627 }
629 /* multipart boundary and transfer encoding are set up for us, and we
630 * are at the end of the sub header. do equivalent of handle_body up
631 * to the next boundary without closing patchfile --- we will expect
632 * that the first part to contain commit message and a patch, and
633 * handle other parts as pure patches.
634 */
635 static int handle_multipart_one_part(void)
636 {
637 int seen = 0;
638 int n = 0;
639 int len;
641 while (fgets(line, sizeof(line), stdin) != NULL) {
642 again:
643 len = eatspace(line);
644 n++;
645 if (!len)
646 continue;
647 if (is_multipart_boundary(line))
648 break;
649 if (0 <= seen && handle_inbody_header(&seen, line))
650 continue;
651 seen = -1; /* no more inbody headers */
652 line[len] = '\n';
653 handle_info();
654 if (handle_commit_msg())
655 goto again;
656 handle_patch();
657 break;
658 }
659 if (n == 0)
660 return -1;
661 return 0;
662 }
664 static void handle_multipart_body(void)
665 {
666 int part_num = 0;
668 /* Skip up to the first boundary */
669 while (fgets(line, sizeof(line), stdin) != NULL)
670 if (is_multipart_boundary(line)) {
671 part_num = 1;
672 break;
673 }
674 if (!part_num)
675 return;
676 /* We are on boundary line. Start slurping the subhead. */
677 while (1) {
678 int len = read_one_header_line(line, sizeof(line), stdin);
679 if (!len) {
680 if (handle_multipart_one_part() < 0)
681 return;
682 }
683 else
684 check_subheader_line(line, len);
685 }
686 fclose(patchfile);
687 if (!patch_lines) {
688 fprintf(stderr, "No patch found\n");
689 exit(1);
690 }
691 }
693 /* Non multipart message */
694 static void handle_body(void)
695 {
696 int seen = 0;
698 while (fgets(line, sizeof(line), stdin) != NULL) {
699 int len = eatspace(line);
700 if (!len)
701 continue;
702 if (0 <= seen && handle_inbody_header(&seen, line))
703 continue;
704 seen = -1; /* no more inbody headers */
705 line[len] = '\n';
706 handle_info();
707 handle_commit_msg();
708 handle_patch();
709 break;
710 }
711 fclose(patchfile);
712 if (!patch_lines) {
713 fprintf(stderr, "No patch found\n");
714 exit(1);
715 }
716 }
718 static const char mailinfo_usage[] =
719 "git-mailinfo [-k] [-u] msg patch <mail >info";
721 static void usage(void) {
722 fprintf(stderr, "%s\n", mailinfo_usage);
723 exit(1);
724 }
726 int main(int argc, char **argv)
727 {
728 while (1 < argc && argv[1][0] == '-') {
729 if (!strcmp(argv[1], "-k"))
730 keep_subject = 1;
731 else if (!strcmp(argv[1], "-u"))
732 metainfo_charset = "utf-8";
733 else if (!strncmp(argv[1], "-u=", 3))
734 metainfo_charset = argv[1] + 3;
735 else
736 usage();
737 argc--; argv++;
738 }
740 if (argc != 3)
741 usage();
742 cmitmsg = fopen(argv[1], "w");
743 if (!cmitmsg) {
744 perror(argv[1]);
745 exit(1);
746 }
747 patchfile = fopen(argv[2], "w");
748 if (!patchfile) {
749 perror(argv[2]);
750 exit(1);
751 }
752 while (1) {
753 int len = read_one_header_line(line, sizeof(line), stdin);
754 if (!len) {
755 if (multipart_boundary[0])
756 handle_multipart_body();
757 else
758 handle_body();
759 break;
760 }
761 check_header_line(line, len);
762 }
763 return 0;
764 }