Fix "git diff --stat" with long filenames
[git/gitweb.git] / mailinfo.c
blobb27651935db85cf0ab651d523c502943d137b7cd
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 #ifndef NO_ICONV
11 #include <iconv.h>
12 #endif
13 #include "git-compat-util.h"
14 #include "cache.h"
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)
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;
45 static int bogus_from(char *line)
47 /* John Doe <johndoe> */
48 char *bra, *ket, *dst, *cp;
50 /* This is fallback, so do not bother if we already have an
51 * e-mail address.
52 */
53 if (*email)
54 return 0;
56 bra = strchr(line, '<');
57 if (!bra)
58 return 0;
59 ket = strchr(bra, '>');
60 if (!ket)
61 return 0;
63 for (dst = email, cp = bra+1; cp < ket; )
64 *dst++ = *cp++;
65 *dst = 0;
66 for (cp = line; isspace(*cp); cp++)
68 for (bra--; isspace(*bra); bra--)
69 *bra = 0;
70 cp = sanity_check(cp, email);
71 strcpy(name, cp);
72 return 1;
75 static int handle_from(char *line)
77 char *at = strchr(line, '@');
78 char *dst;
80 if (!at)
81 return bogus_from(line);
84 * If we already have one email, don't take any confusing lines
86 if (*email && strchr(at+1, '@'))
87 return 0;
89 /* Pick up the string around '@', possibly delimited with <>
90 * pair; that is the email part. White them out while copying.
92 while (at > line) {
93 char c = at[-1];
94 if (isspace(c))
95 break;
96 if (c == '<') {
97 at[-1] = ' ';
98 break;
100 at--;
102 dst = email;
103 for (;;) {
104 unsigned char c = *at;
105 if (!c || c == '>' || isspace(c)) {
106 if (c == '>')
107 *at = ' ';
108 break;
110 *at++ = ' ';
111 *dst++ = c;
113 *dst++ = 0;
115 /* The remainder is name. It could be "John Doe <john.doe@xz>"
116 * or "john.doe@xz (John Doe)", but we have whited out the
117 * email part, so trim from both ends, possibly removing
118 * the () pair at the end.
120 at = line + strlen(line);
121 while (at > line) {
122 unsigned char c = *--at;
123 if (!isspace(c)) {
124 at[(c == ')') ? 0 : 1] = 0;
125 break;
129 at = line;
130 for (;;) {
131 unsigned char c = *at;
132 if (!c || !isspace(c)) {
133 if (c == '(')
134 at++;
135 break;
137 at++;
139 at = sanity_check(at, email);
140 strcpy(name, at);
141 return 1;
144 static int handle_date(char *line)
146 strcpy(date, line);
147 return 0;
150 static int handle_subject(char *line)
152 strcpy(subject, line);
153 return 0;
156 /* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
157 * to have enough heuristics to grok MIME encoded patches often found
158 * on our mailing lists. For example, we do not even treat header lines
159 * case insensitively.
162 static int slurp_attr(const char *line, const char *name, char *attr)
164 char *ends, *ap = strcasestr(line, name);
165 size_t sz;
167 if (!ap) {
168 *attr = 0;
169 return 0;
171 ap += strlen(name);
172 if (*ap == '"') {
173 ap++;
174 ends = "\"";
176 else
177 ends = "; \t";
178 sz = strcspn(ap, ends);
179 memcpy(attr, ap, sz);
180 attr[sz] = 0;
181 return 1;
184 static int handle_subcontent_type(char *line)
186 /* We do not want to mess with boundary. Note that we do not
187 * handle nested multipart.
189 if (strcasestr(line, "boundary=")) {
190 fprintf(stderr, "Not handling nested multipart message.\n");
191 exit(1);
193 slurp_attr(line, "charset=", charset);
194 if (*charset) {
195 int i, c;
196 for (i = 0; (c = charset[i]) != 0; i++)
197 charset[i] = tolower(c);
199 return 0;
202 static int handle_content_type(char *line)
204 *multipart_boundary = 0;
205 if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
206 memcpy(multipart_boundary, "--", 2);
207 multipart_boundary_len = strlen(multipart_boundary);
209 slurp_attr(line, "charset=", charset);
210 return 0;
213 static int handle_content_transfer_encoding(char *line)
215 if (strcasestr(line, "base64"))
216 transfer_encoding = TE_BASE64;
217 else if (strcasestr(line, "quoted-printable"))
218 transfer_encoding = TE_QP;
219 else
220 transfer_encoding = TE_DONTCARE;
221 return 0;
224 static int is_multipart_boundary(const char *line)
226 return (!memcmp(line, multipart_boundary, multipart_boundary_len));
229 static int eatspace(char *line)
231 int len = strlen(line);
232 while (len > 0 && isspace(line[len-1]))
233 line[--len] = 0;
234 return len;
237 #define SEEN_FROM 01
238 #define SEEN_DATE 02
239 #define SEEN_SUBJECT 04
241 /* First lines of body can have From:, Date:, and Subject: */
242 static int handle_inbody_header(int *seen, char *line)
244 if (!memcmp("From:", line, 5) && isspace(line[5])) {
245 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
246 *seen |= SEEN_FROM;
247 return 1;
250 if (!memcmp("Date:", line, 5) && isspace(line[5])) {
251 if (!(*seen & SEEN_DATE)) {
252 handle_date(line+6);
253 *seen |= SEEN_DATE;
254 return 1;
257 if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
258 if (!(*seen & SEEN_SUBJECT)) {
259 handle_subject(line+9);
260 *seen |= SEEN_SUBJECT;
261 return 1;
264 if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
265 if (!(*seen & SEEN_SUBJECT)) {
266 handle_subject(line);
267 *seen |= SEEN_SUBJECT;
268 return 1;
271 return 0;
274 static char *cleanup_subject(char *subject)
276 if (keep_subject)
277 return subject;
278 for (;;) {
279 char *p;
280 int len, remove;
281 switch (*subject) {
282 case 'r': case 'R':
283 if (!memcmp("e:", subject+1, 2)) {
284 subject +=3;
285 continue;
287 break;
288 case ' ': case '\t': case ':':
289 subject++;
290 continue;
292 case '[':
293 p = strchr(subject, ']');
294 if (!p) {
295 subject++;
296 continue;
298 len = strlen(p);
299 remove = p - subject;
300 if (remove <= len *2) {
301 subject = p+1;
302 continue;
304 break;
306 return subject;
310 static void cleanup_space(char *buf)
312 unsigned char c;
313 while ((c = *buf) != 0) {
314 buf++;
315 if (isspace(c)) {
316 buf[-1] = ' ';
317 c = *buf;
318 while (isspace(c)) {
319 int len = strlen(buf);
320 memmove(buf, buf+1, len);
321 c = *buf;
327 typedef int (*header_fn_t)(char *);
328 struct header_def {
329 const char *name;
330 header_fn_t func;
331 int namelen;
334 static void check_header(char *line, int len, struct header_def *header)
336 int i;
338 if (header[0].namelen <= 0) {
339 for (i = 0; header[i].name; i++)
340 header[i].namelen = strlen(header[i].name);
342 for (i = 0; header[i].name; i++) {
343 int len = header[i].namelen;
344 if (!strncasecmp(line, header[i].name, len) &&
345 line[len] == ':' && isspace(line[len + 1])) {
346 header[i].func(line + len + 2);
347 break;
352 static void check_subheader_line(char *line, int len)
354 static struct header_def header[] = {
355 { "Content-Type", handle_subcontent_type },
356 { "Content-Transfer-Encoding",
357 handle_content_transfer_encoding },
358 { NULL },
360 check_header(line, len, header);
362 static void check_header_line(char *line, int len)
364 static struct header_def header[] = {
365 { "From", handle_from },
366 { "Date", handle_date },
367 { "Subject", handle_subject },
368 { "Content-Type", handle_content_type },
369 { "Content-Transfer-Encoding",
370 handle_content_transfer_encoding },
371 { NULL },
373 check_header(line, len, header);
376 static int read_one_header_line(char *line, int sz, FILE *in)
378 int ofs = 0;
379 while (ofs < sz) {
380 int peek, len;
381 if (fgets(line + ofs, sz - ofs, in) == NULL)
382 return ofs;
383 len = eatspace(line + ofs);
384 if (len == 0)
385 return ofs;
386 peek = fgetc(in); ungetc(peek, in);
387 if (peek == ' ' || peek == '\t') {
388 /* Yuck, 2822 header "folding" */
389 ofs += len;
390 continue;
392 return ofs + len;
394 return ofs;
397 static unsigned hexval(int c)
399 if (c >= '0' && c <= '9')
400 return c - '0';
401 if (c >= 'a' && c <= 'f')
402 return c - 'a' + 10;
403 if (c >= 'A' && c <= 'F')
404 return c - 'A' + 10;
405 return ~0;
408 static int decode_q_segment(char *in, char *ot, char *ep, int rfc2047)
410 int c;
411 while ((c = *in++) != 0 && (in <= ep)) {
412 if (c == '=') {
413 int d = *in++;
414 if (d == '\n' || !d)
415 break; /* drop trailing newline */
416 *ot++ = ((hexval(d) << 4) | hexval(*in++));
417 continue;
419 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
420 c = 0x20;
421 *ot++ = c;
423 *ot = 0;
424 return 0;
427 static int decode_b_segment(char *in, char *ot, char *ep)
429 /* Decode in..ep, possibly in-place to ot */
430 int c, pos = 0, acc = 0;
432 while ((c = *in++) != 0 && (in <= ep)) {
433 if (c == '+')
434 c = 62;
435 else if (c == '/')
436 c = 63;
437 else if ('A' <= c && c <= 'Z')
438 c -= 'A';
439 else if ('a' <= c && c <= 'z')
440 c -= 'a' - 26;
441 else if ('0' <= c && c <= '9')
442 c -= '0' - 52;
443 else if (c == '=') {
444 /* padding is almost like (c == 0), except we do
445 * not output NUL resulting only from it;
446 * for now we just trust the data.
448 c = 0;
450 else
451 continue; /* garbage */
452 switch (pos++) {
453 case 0:
454 acc = (c << 2);
455 break;
456 case 1:
457 *ot++ = (acc | (c >> 4));
458 acc = (c & 15) << 4;
459 break;
460 case 2:
461 *ot++ = (acc | (c >> 2));
462 acc = (c & 3) << 6;
463 break;
464 case 3:
465 *ot++ = (acc | c);
466 acc = pos = 0;
467 break;
470 *ot = 0;
471 return 0;
474 static void convert_to_utf8(char *line, char *charset)
476 #ifndef NO_ICONV
477 char *in, *out;
478 size_t insize, outsize, nrc;
479 char outbuf[4096]; /* cheat */
480 static char latin_one[] = "latin1";
481 char *input_charset = *charset ? charset : latin_one;
482 iconv_t conv = iconv_open(metainfo_charset, input_charset);
484 if (conv == (iconv_t) -1) {
485 static int warned_latin1_once = 0;
486 if (input_charset != latin_one) {
487 fprintf(stderr, "cannot convert from %s to %s\n",
488 input_charset, metainfo_charset);
489 *charset = 0;
491 else if (!warned_latin1_once) {
492 warned_latin1_once = 1;
493 fprintf(stderr, "tried to convert from %s to %s, "
494 "but your iconv does not work with it.\n",
495 input_charset, metainfo_charset);
497 return;
499 in = line;
500 insize = strlen(in);
501 out = outbuf;
502 outsize = sizeof(outbuf);
503 nrc = iconv(conv, &in, &insize, &out, &outsize);
504 iconv_close(conv);
505 if (nrc == (size_t) -1)
506 return;
507 *out = 0;
508 strcpy(line, outbuf);
509 #endif
512 static void decode_header_bq(char *it)
514 char *in, *out, *ep, *cp, *sp;
515 char outbuf[1000];
517 in = it;
518 out = outbuf;
519 while ((ep = strstr(in, "=?")) != NULL) {
520 int sz, encoding;
521 char charset_q[256], piecebuf[256];
522 if (in != ep) {
523 sz = ep - in;
524 memcpy(out, in, sz);
525 out += sz;
526 in += sz;
528 /* E.g.
529 * ep : "=?iso-2022-jp?B?GyR...?= foo"
530 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
532 ep += 2;
533 cp = strchr(ep, '?');
534 if (!cp)
535 return; /* no munging */
536 for (sp = ep; sp < cp; sp++)
537 charset_q[sp - ep] = tolower(*sp);
538 charset_q[cp - ep] = 0;
539 encoding = cp[1];
540 if (!encoding || cp[2] != '?')
541 return; /* no munging */
542 ep = strstr(cp + 3, "?=");
543 if (!ep)
544 return; /* no munging */
545 switch (tolower(encoding)) {
546 default:
547 return; /* no munging */
548 case 'b':
549 sz = decode_b_segment(cp + 3, piecebuf, ep);
550 break;
551 case 'q':
552 sz = decode_q_segment(cp + 3, piecebuf, ep, 1);
553 break;
555 if (sz < 0)
556 return;
557 if (metainfo_charset)
558 convert_to_utf8(piecebuf, charset_q);
559 strcpy(out, piecebuf);
560 out += strlen(out);
561 in = ep + 2;
563 strcpy(out, in);
564 strcpy(it, outbuf);
567 static void decode_transfer_encoding(char *line)
569 char *ep;
571 switch (transfer_encoding) {
572 case TE_QP:
573 ep = line + strlen(line);
574 decode_q_segment(line, line, ep, 0);
575 break;
576 case TE_BASE64:
577 ep = line + strlen(line);
578 decode_b_segment(line, line, ep);
579 break;
580 case TE_DONTCARE:
581 break;
585 static void handle_info(void)
587 char *sub;
588 static int done_info = 0;
590 if (done_info)
591 return;
593 done_info = 1;
594 sub = cleanup_subject(subject);
595 cleanup_space(name);
596 cleanup_space(date);
597 cleanup_space(email);
598 cleanup_space(sub);
600 /* Unwrap inline B and Q encoding, and optionally
601 * normalize the meta information to utf8.
603 decode_header_bq(name);
604 decode_header_bq(date);
605 decode_header_bq(email);
606 decode_header_bq(sub);
607 printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
608 name, email, sub, date);
611 /* We are inside message body and have read line[] already.
612 * Spit out the commit log.
614 static int handle_commit_msg(void)
616 if (!cmitmsg)
617 return 0;
618 do {
619 if (!memcmp("diff -", line, 6) ||
620 !memcmp("---", line, 3) ||
621 !memcmp("Index: ", line, 7))
622 break;
623 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
624 /* We come here when the first part had only
625 * the commit message without any patch. We
626 * pretend we have not seen this line yet, and
627 * go back to the loop.
629 return 1;
632 /* Unwrap transfer encoding and optionally
633 * normalize the log message to UTF-8.
635 decode_transfer_encoding(line);
636 if (metainfo_charset)
637 convert_to_utf8(line, charset);
638 fputs(line, cmitmsg);
639 } while (fgets(line, sizeof(line), stdin) != NULL);
640 fclose(cmitmsg);
641 cmitmsg = NULL;
642 return 0;
645 /* We have done the commit message and have the first
646 * line of the patch in line[].
648 static void handle_patch(void)
650 do {
651 if (multipart_boundary[0] && is_multipart_boundary(line))
652 break;
653 /* Only unwrap transfer encoding but otherwise do not
654 * do anything. We do *NOT* want UTF-8 conversion
655 * here; we are dealing with the user payload.
657 decode_transfer_encoding(line);
658 fputs(line, patchfile);
659 patch_lines++;
660 } while (fgets(line, sizeof(line), stdin) != NULL);
663 /* multipart boundary and transfer encoding are set up for us, and we
664 * are at the end of the sub header. do equivalent of handle_body up
665 * to the next boundary without closing patchfile --- we will expect
666 * that the first part to contain commit message and a patch, and
667 * handle other parts as pure patches.
669 static int handle_multipart_one_part(void)
671 int seen = 0;
672 int n = 0;
673 int len;
675 while (fgets(line, sizeof(line), stdin) != NULL) {
676 again:
677 len = eatspace(line);
678 n++;
679 if (!len)
680 continue;
681 if (is_multipart_boundary(line))
682 break;
683 if (0 <= seen && handle_inbody_header(&seen, line))
684 continue;
685 seen = -1; /* no more inbody headers */
686 line[len] = '\n';
687 handle_info();
688 if (handle_commit_msg())
689 goto again;
690 handle_patch();
691 break;
693 if (n == 0)
694 return -1;
695 return 0;
698 static void handle_multipart_body(void)
700 int part_num = 0;
702 /* Skip up to the first boundary */
703 while (fgets(line, sizeof(line), stdin) != NULL)
704 if (is_multipart_boundary(line)) {
705 part_num = 1;
706 break;
708 if (!part_num)
709 return;
710 /* We are on boundary line. Start slurping the subhead. */
711 while (1) {
712 int len = read_one_header_line(line, sizeof(line), stdin);
713 if (!len) {
714 if (handle_multipart_one_part() < 0)
715 return;
716 /* Reset per part headers */
717 transfer_encoding = TE_DONTCARE;
718 charset[0] = 0;
720 else
721 check_subheader_line(line, len);
723 fclose(patchfile);
724 if (!patch_lines) {
725 fprintf(stderr, "No patch found\n");
726 exit(1);
730 /* Non multipart message */
731 static void handle_body(void)
733 int seen = 0;
735 while (fgets(line, sizeof(line), stdin) != NULL) {
736 int len = eatspace(line);
737 if (!len)
738 continue;
739 if (0 <= seen && handle_inbody_header(&seen, line))
740 continue;
741 seen = -1; /* no more inbody headers */
742 line[len] = '\n';
743 handle_info();
744 handle_commit_msg();
745 handle_patch();
746 break;
748 fclose(patchfile);
749 if (!patch_lines) {
750 fprintf(stderr, "No patch found\n");
751 exit(1);
755 static const char mailinfo_usage[] =
756 "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
758 int main(int argc, char **argv)
760 /* NEEDSWORK: might want to do the optional .git/ directory
761 * discovery
763 git_config(git_default_config);
765 while (1 < argc && argv[1][0] == '-') {
766 if (!strcmp(argv[1], "-k"))
767 keep_subject = 1;
768 else if (!strcmp(argv[1], "-u"))
769 metainfo_charset = git_commit_encoding;
770 else if (!strncmp(argv[1], "--encoding=", 11))
771 metainfo_charset = argv[1] + 11;
772 else
773 usage(mailinfo_usage);
774 argc--; argv++;
777 if (argc != 3)
778 usage(mailinfo_usage);
779 cmitmsg = fopen(argv[1], "w");
780 if (!cmitmsg) {
781 perror(argv[1]);
782 exit(1);
784 patchfile = fopen(argv[2], "w");
785 if (!patchfile) {
786 perror(argv[2]);
787 exit(1);
789 while (1) {
790 int len = read_one_header_line(line, sizeof(line), stdin);
791 if (!len) {
792 if (multipart_boundary[0])
793 handle_multipart_body();
794 else
795 handle_body();
796 break;
798 check_header_line(line, len);
800 return 0;