Documentation: update git-clone.txt for clone's new default behavior
[alt-git.git] / builtin-mailinfo.c
bloba67f3eb90b6f715714c6fa7bb931044630c74111
1 /*
2 * Another stupid program, this one parsing the headers of an
3 * email to figure out authorship and subject
4 */
5 #include "cache.h"
6 #include "builtin.h"
7 #include "utf8.h"
9 static FILE *cmitmsg, *patchfile, *fin, *fout;
11 static int keep_subject;
12 static const char *metainfo_charset;
13 static char line[1000];
14 static char date[1000];
15 static char name[1000];
16 static char email[1000];
17 static char subject[1000];
19 static enum {
20 TE_DONTCARE, TE_QP, TE_BASE64,
21 } transfer_encoding;
22 static char charset[256];
24 static char multipart_boundary[1000];
25 static int multipart_boundary_len;
26 static int patch_lines;
28 static char *sanity_check(char *name, char *email)
30 int len = strlen(name);
31 if (len < 3 || len > 60)
32 return email;
33 if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
34 return email;
35 return name;
38 static int bogus_from(char *line)
40 /* John Doe <johndoe> */
41 char *bra, *ket, *dst, *cp;
43 /* This is fallback, so do not bother if we already have an
44 * e-mail address.
46 if (*email)
47 return 0;
49 bra = strchr(line, '<');
50 if (!bra)
51 return 0;
52 ket = strchr(bra, '>');
53 if (!ket)
54 return 0;
56 for (dst = email, cp = bra+1; cp < ket; )
57 *dst++ = *cp++;
58 *dst = 0;
59 for (cp = line; isspace(*cp); cp++)
61 for (bra--; isspace(*bra); bra--)
62 *bra = 0;
63 cp = sanity_check(cp, email);
64 strcpy(name, cp);
65 return 1;
68 static int handle_from(char *in_line)
70 char line[1000];
71 char *at;
72 char *dst;
74 strcpy(line, in_line);
75 at = strchr(line, '@');
76 if (!at)
77 return bogus_from(line);
80 * If we already have one email, don't take any confusing lines
82 if (*email && strchr(at+1, '@'))
83 return 0;
85 /* Pick up the string around '@', possibly delimited with <>
86 * pair; that is the email part. White them out while copying.
88 while (at > line) {
89 char c = at[-1];
90 if (isspace(c))
91 break;
92 if (c == '<') {
93 at[-1] = ' ';
94 break;
96 at--;
98 dst = email;
99 for (;;) {
100 unsigned char c = *at;
101 if (!c || c == '>' || isspace(c)) {
102 if (c == '>')
103 *at = ' ';
104 break;
106 *at++ = ' ';
107 *dst++ = c;
109 *dst++ = 0;
111 /* The remainder is name. It could be "John Doe <john.doe@xz>"
112 * or "john.doe@xz (John Doe)", but we have whited out the
113 * email part, so trim from both ends, possibly removing
114 * the () pair at the end.
116 at = line + strlen(line);
117 while (at > line) {
118 unsigned char c = *--at;
119 if (!isspace(c)) {
120 at[(c == ')') ? 0 : 1] = 0;
121 break;
125 at = line;
126 for (;;) {
127 unsigned char c = *at;
128 if (!c || !isspace(c)) {
129 if (c == '(')
130 at++;
131 break;
133 at++;
135 at = sanity_check(at, email);
136 strcpy(name, at);
137 return 1;
140 static int handle_date(char *line)
142 strcpy(date, line);
143 return 0;
146 static int handle_subject(char *line)
148 strcpy(subject, line);
149 return 0;
152 /* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
153 * to have enough heuristics to grok MIME encoded patches often found
154 * on our mailing lists. For example, we do not even treat header lines
155 * case insensitively.
158 static int slurp_attr(const char *line, const char *name, char *attr)
160 const char *ends, *ap = strcasestr(line, name);
161 size_t sz;
163 if (!ap) {
164 *attr = 0;
165 return 0;
167 ap += strlen(name);
168 if (*ap == '"') {
169 ap++;
170 ends = "\"";
172 else
173 ends = "; \t";
174 sz = strcspn(ap, ends);
175 memcpy(attr, ap, sz);
176 attr[sz] = 0;
177 return 1;
180 static int handle_subcontent_type(char *line)
182 /* We do not want to mess with boundary. Note that we do not
183 * handle nested multipart.
185 if (strcasestr(line, "boundary=")) {
186 fprintf(stderr, "Not handling nested multipart message.\n");
187 exit(1);
189 slurp_attr(line, "charset=", charset);
190 if (*charset) {
191 int i, c;
192 for (i = 0; (c = charset[i]) != 0; i++)
193 charset[i] = tolower(c);
195 return 0;
198 static int handle_content_type(char *line)
200 *multipart_boundary = 0;
201 if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
202 memcpy(multipart_boundary, "--", 2);
203 multipart_boundary_len = strlen(multipart_boundary);
205 slurp_attr(line, "charset=", charset);
206 return 0;
209 static int handle_content_transfer_encoding(char *line)
211 if (strcasestr(line, "base64"))
212 transfer_encoding = TE_BASE64;
213 else if (strcasestr(line, "quoted-printable"))
214 transfer_encoding = TE_QP;
215 else
216 transfer_encoding = TE_DONTCARE;
217 return 0;
220 static int is_multipart_boundary(const char *line)
222 return (!memcmp(line, multipart_boundary, multipart_boundary_len));
225 static int eatspace(char *line)
227 int len = strlen(line);
228 while (len > 0 && isspace(line[len-1]))
229 line[--len] = 0;
230 return len;
233 #define SEEN_FROM 01
234 #define SEEN_DATE 02
235 #define SEEN_SUBJECT 04
236 #define SEEN_BOGUS_UNIX_FROM 010
237 #define SEEN_PREFIX 020
239 /* First lines of body can have From:, Date:, and Subject: or empty */
240 static void handle_inbody_header(int *seen, char *line)
242 if (*seen & SEEN_PREFIX)
243 return;
244 if (isspace(*line)) {
245 char *cp;
246 for (cp = line + 1; *cp; cp++) {
247 if (!isspace(*cp))
248 break;
250 if (!*cp)
251 return;
253 if (!memcmp(">From", line, 5) && isspace(line[5])) {
254 if (!(*seen & SEEN_BOGUS_UNIX_FROM)) {
255 *seen |= SEEN_BOGUS_UNIX_FROM;
256 return;
259 if (!memcmp("From:", line, 5) && isspace(line[5])) {
260 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
261 *seen |= SEEN_FROM;
262 return;
265 if (!memcmp("Date:", line, 5) && isspace(line[5])) {
266 if (!(*seen & SEEN_DATE)) {
267 handle_date(line+6);
268 *seen |= SEEN_DATE;
269 return;
272 if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
273 if (!(*seen & SEEN_SUBJECT)) {
274 handle_subject(line+9);
275 *seen |= SEEN_SUBJECT;
276 return;
279 if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
280 if (!(*seen & SEEN_SUBJECT)) {
281 handle_subject(line);
282 *seen |= SEEN_SUBJECT;
283 return;
286 *seen |= SEEN_PREFIX;
289 static char *cleanup_subject(char *subject)
291 if (keep_subject)
292 return subject;
293 for (;;) {
294 char *p;
295 int len, remove;
296 switch (*subject) {
297 case 'r': case 'R':
298 if (!memcmp("e:", subject+1, 2)) {
299 subject +=3;
300 continue;
302 break;
303 case ' ': case '\t': case ':':
304 subject++;
305 continue;
307 case '[':
308 p = strchr(subject, ']');
309 if (!p) {
310 subject++;
311 continue;
313 len = strlen(p);
314 remove = p - subject;
315 if (remove <= len *2) {
316 subject = p+1;
317 continue;
319 break;
321 eatspace(subject);
322 return subject;
326 static void cleanup_space(char *buf)
328 unsigned char c;
329 while ((c = *buf) != 0) {
330 buf++;
331 if (isspace(c)) {
332 buf[-1] = ' ';
333 c = *buf;
334 while (isspace(c)) {
335 int len = strlen(buf);
336 memmove(buf, buf+1, len);
337 c = *buf;
343 static void decode_header(char *it);
344 typedef int (*header_fn_t)(char *);
345 struct header_def {
346 const char *name;
347 header_fn_t func;
348 int namelen;
351 static void check_header(char *line, struct header_def *header)
353 int i;
355 if (header[0].namelen <= 0) {
356 for (i = 0; header[i].name; i++)
357 header[i].namelen = strlen(header[i].name);
359 for (i = 0; header[i].name; i++) {
360 int len = header[i].namelen;
361 if (!strncasecmp(line, header[i].name, len) &&
362 line[len] == ':' && isspace(line[len + 1])) {
363 /* Unwrap inline B and Q encoding, and optionally
364 * normalize the meta information to utf8.
366 decode_header(line + len + 2);
367 header[i].func(line + len + 2);
368 break;
373 static void check_subheader_line(char *line)
375 static struct header_def header[] = {
376 { "Content-Type", handle_subcontent_type },
377 { "Content-Transfer-Encoding",
378 handle_content_transfer_encoding },
379 { NULL },
381 check_header(line, header);
383 static void check_header_line(char *line)
385 static struct header_def header[] = {
386 { "From", handle_from },
387 { "Date", handle_date },
388 { "Subject", handle_subject },
389 { "Content-Type", handle_content_type },
390 { "Content-Transfer-Encoding",
391 handle_content_transfer_encoding },
392 { NULL },
394 check_header(line, header);
397 static int is_rfc2822_header(char *line)
400 * The section that defines the loosest possible
401 * field name is "3.6.8 Optional fields".
403 * optional-field = field-name ":" unstructured CRLF
404 * field-name = 1*ftext
405 * ftext = %d33-57 / %59-126
407 int ch;
408 char *cp = line;
409 while ((ch = *cp++)) {
410 if (ch == ':')
411 return cp != line;
412 if ((33 <= ch && ch <= 57) ||
413 (59 <= ch && ch <= 126))
414 continue;
415 break;
417 return 0;
420 static int read_one_header_line(char *line, int sz, FILE *in)
422 int ofs = 0;
423 while (ofs < sz) {
424 int peek, len;
425 if (fgets(line + ofs, sz - ofs, in) == NULL)
426 break;
427 len = eatspace(line + ofs);
428 if ((len == 0) || !is_rfc2822_header(line)) {
429 /* Re-add the newline */
430 line[ofs + len] = '\n';
431 line[ofs + len + 1] = '\0';
432 break;
434 ofs += len;
435 /* Yuck, 2822 header "folding" */
436 peek = fgetc(in); ungetc(peek, in);
437 if (peek != ' ' && peek != '\t')
438 break;
440 /* Count mbox From headers as headers */
441 if (!ofs && (!memcmp(line, "From ", 5) || !memcmp(line, ">From ", 6)))
442 ofs = 1;
443 return ofs;
446 static int decode_q_segment(char *in, char *ot, char *ep, int rfc2047)
448 int c;
449 while ((c = *in++) != 0 && (in <= ep)) {
450 if (c == '=') {
451 int d = *in++;
452 if (d == '\n' || !d)
453 break; /* drop trailing newline */
454 *ot++ = ((hexval(d) << 4) | hexval(*in++));
455 continue;
457 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
458 c = 0x20;
459 *ot++ = c;
461 *ot = 0;
462 return 0;
465 static int decode_b_segment(char *in, char *ot, char *ep)
467 /* Decode in..ep, possibly in-place to ot */
468 int c, pos = 0, acc = 0;
470 while ((c = *in++) != 0 && (in <= ep)) {
471 if (c == '+')
472 c = 62;
473 else if (c == '/')
474 c = 63;
475 else if ('A' <= c && c <= 'Z')
476 c -= 'A';
477 else if ('a' <= c && c <= 'z')
478 c -= 'a' - 26;
479 else if ('0' <= c && c <= '9')
480 c -= '0' - 52;
481 else if (c == '=') {
482 /* padding is almost like (c == 0), except we do
483 * not output NUL resulting only from it;
484 * for now we just trust the data.
486 c = 0;
488 else
489 continue; /* garbage */
490 switch (pos++) {
491 case 0:
492 acc = (c << 2);
493 break;
494 case 1:
495 *ot++ = (acc | (c >> 4));
496 acc = (c & 15) << 4;
497 break;
498 case 2:
499 *ot++ = (acc | (c >> 2));
500 acc = (c & 3) << 6;
501 break;
502 case 3:
503 *ot++ = (acc | c);
504 acc = pos = 0;
505 break;
508 *ot = 0;
509 return 0;
512 static void convert_to_utf8(char *line, char *charset)
514 static char latin_one[] = "latin1";
515 char *input_charset = *charset ? charset : latin_one;
516 char *out = reencode_string(line, metainfo_charset, input_charset);
518 if (!out) {
519 fprintf(stderr, "cannot convert from %s to %s\n",
520 input_charset, metainfo_charset);
521 *charset = 0;
522 return;
524 strcpy(line, out);
525 free(out);
528 static int decode_header_bq(char *it)
530 char *in, *out, *ep, *cp, *sp;
531 char outbuf[1000];
532 int rfc2047 = 0;
534 in = it;
535 out = outbuf;
536 while ((ep = strstr(in, "=?")) != NULL) {
537 int sz, encoding;
538 char charset_q[256], piecebuf[256];
539 rfc2047 = 1;
541 if (in != ep) {
542 sz = ep - in;
543 memcpy(out, in, sz);
544 out += sz;
545 in += sz;
547 /* E.g.
548 * ep : "=?iso-2022-jp?B?GyR...?= foo"
549 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
551 ep += 2;
552 cp = strchr(ep, '?');
553 if (!cp)
554 return rfc2047; /* no munging */
555 for (sp = ep; sp < cp; sp++)
556 charset_q[sp - ep] = tolower(*sp);
557 charset_q[cp - ep] = 0;
558 encoding = cp[1];
559 if (!encoding || cp[2] != '?')
560 return rfc2047; /* no munging */
561 ep = strstr(cp + 3, "?=");
562 if (!ep)
563 return rfc2047; /* no munging */
564 switch (tolower(encoding)) {
565 default:
566 return rfc2047; /* no munging */
567 case 'b':
568 sz = decode_b_segment(cp + 3, piecebuf, ep);
569 break;
570 case 'q':
571 sz = decode_q_segment(cp + 3, piecebuf, ep, 1);
572 break;
574 if (sz < 0)
575 return rfc2047;
576 if (metainfo_charset)
577 convert_to_utf8(piecebuf, charset_q);
578 strcpy(out, piecebuf);
579 out += strlen(out);
580 in = ep + 2;
582 strcpy(out, in);
583 strcpy(it, outbuf);
584 return rfc2047;
587 static void decode_header(char *it)
590 if (decode_header_bq(it))
591 return;
592 /* otherwise "it" is a straight copy of the input.
593 * This can be binary guck but there is no charset specified.
595 if (metainfo_charset)
596 convert_to_utf8(it, "");
599 static void decode_transfer_encoding(char *line)
601 char *ep;
603 switch (transfer_encoding) {
604 case TE_QP:
605 ep = line + strlen(line);
606 decode_q_segment(line, line, ep, 0);
607 break;
608 case TE_BASE64:
609 ep = line + strlen(line);
610 decode_b_segment(line, line, ep);
611 break;
612 case TE_DONTCARE:
613 break;
617 static void handle_info(void)
619 char *sub;
621 sub = cleanup_subject(subject);
622 cleanup_space(name);
623 cleanup_space(date);
624 cleanup_space(email);
625 cleanup_space(sub);
627 fprintf(fout, "Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
628 name, email, sub, date);
631 /* We are inside message body and have read line[] already.
632 * Spit out the commit log.
634 static int handle_commit_msg(int *seen)
636 if (!cmitmsg)
637 return 0;
638 do {
639 if (!memcmp("diff -", line, 6) ||
640 !memcmp("---", line, 3) ||
641 !memcmp("Index: ", line, 7))
642 break;
643 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
644 /* We come here when the first part had only
645 * the commit message without any patch. We
646 * pretend we have not seen this line yet, and
647 * go back to the loop.
649 return 1;
652 /* Unwrap transfer encoding and optionally
653 * normalize the log message to UTF-8.
655 decode_transfer_encoding(line);
656 if (metainfo_charset)
657 convert_to_utf8(line, charset);
659 handle_inbody_header(seen, line);
660 if (!(*seen & SEEN_PREFIX))
661 continue;
663 fputs(line, cmitmsg);
664 } while (fgets(line, sizeof(line), fin) != NULL);
665 fclose(cmitmsg);
666 cmitmsg = NULL;
667 return 0;
670 /* We have done the commit message and have the first
671 * line of the patch in line[].
673 static void handle_patch(void)
675 do {
676 if (multipart_boundary[0] && is_multipart_boundary(line))
677 break;
678 /* Only unwrap transfer encoding but otherwise do not
679 * do anything. We do *NOT* want UTF-8 conversion
680 * here; we are dealing with the user payload.
682 decode_transfer_encoding(line);
683 fputs(line, patchfile);
684 patch_lines++;
685 } while (fgets(line, sizeof(line), fin) != NULL);
688 /* multipart boundary and transfer encoding are set up for us, and we
689 * are at the end of the sub header. do equivalent of handle_body up
690 * to the next boundary without closing patchfile --- we will expect
691 * that the first part to contain commit message and a patch, and
692 * handle other parts as pure patches.
694 static int handle_multipart_one_part(int *seen)
696 int n = 0;
698 while (fgets(line, sizeof(line), fin) != NULL) {
699 again:
700 n++;
701 if (is_multipart_boundary(line))
702 break;
703 if (handle_commit_msg(seen))
704 goto again;
705 handle_patch();
706 break;
708 if (n == 0)
709 return -1;
710 return 0;
713 static void handle_multipart_body(void)
715 int seen = 0;
716 int part_num = 0;
718 /* Skip up to the first boundary */
719 while (fgets(line, sizeof(line), fin) != NULL)
720 if (is_multipart_boundary(line)) {
721 part_num = 1;
722 break;
724 if (!part_num)
725 return;
726 /* We are on boundary line. Start slurping the subhead. */
727 while (1) {
728 int hdr = read_one_header_line(line, sizeof(line), fin);
729 if (!hdr) {
730 if (handle_multipart_one_part(&seen) < 0)
731 return;
732 /* Reset per part headers */
733 transfer_encoding = TE_DONTCARE;
734 charset[0] = 0;
736 else
737 check_subheader_line(line);
739 fclose(patchfile);
740 if (!patch_lines) {
741 fprintf(stderr, "No patch found\n");
742 exit(1);
746 /* Non multipart message */
747 static void handle_body(void)
749 int seen = 0;
751 handle_commit_msg(&seen);
752 handle_patch();
753 fclose(patchfile);
754 if (!patch_lines) {
755 fprintf(stderr, "No patch found\n");
756 exit(1);
760 int mailinfo(FILE *in, FILE *out, int ks, const char *encoding,
761 const char *msg, const char *patch)
763 keep_subject = ks;
764 metainfo_charset = encoding;
765 fin = in;
766 fout = out;
768 cmitmsg = fopen(msg, "w");
769 if (!cmitmsg) {
770 perror(msg);
771 return -1;
773 patchfile = fopen(patch, "w");
774 if (!patchfile) {
775 perror(patch);
776 fclose(cmitmsg);
777 return -1;
779 while (1) {
780 int hdr = read_one_header_line(line, sizeof(line), fin);
781 if (!hdr) {
782 if (multipart_boundary[0])
783 handle_multipart_body();
784 else
785 handle_body();
786 handle_info();
787 break;
789 check_header_line(line);
792 return 0;
795 static const char mailinfo_usage[] =
796 "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
798 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
800 /* NEEDSWORK: might want to do the optional .git/ directory
801 * discovery
803 git_config(git_default_config);
805 while (1 < argc && argv[1][0] == '-') {
806 if (!strcmp(argv[1], "-k"))
807 keep_subject = 1;
808 else if (!strcmp(argv[1], "-u"))
809 metainfo_charset = (git_commit_encoding
810 ? git_commit_encoding : "utf-8");
811 else if (!strncmp(argv[1], "--encoding=", 11))
812 metainfo_charset = argv[1] + 11;
813 else
814 usage(mailinfo_usage);
815 argc--; argv++;
818 if (argc != 3)
819 usage(mailinfo_usage);
821 return !!mailinfo(stdin, stdout, keep_subject, metainfo_charset, argv[1], argv[2]);