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"
8#include "strbuf.h"
9
10static FILE *cmitmsg, *patchfile;
11
12static const char *metainfo_charset;
13
14struct mailinfo {
15 FILE *input;
16 FILE *output;
17
18 struct strbuf name;
19 struct strbuf email;
20 int keep_subject;
21 int keep_non_patch_brackets_in_subject;
22
23 int filter_stage; /* still reading log or are we copying patch? */
24 int header_stage; /* still checking in-body headers? */
25};
26static char *message_id;
27
28static enum {
29 TE_DONTCARE, TE_QP, TE_BASE64
30} transfer_encoding;
31
32static struct strbuf charset = STRBUF_INIT;
33static int patch_lines;
34static struct strbuf **p_hdr_data, **s_hdr_data;
35static int use_scissors;
36static int add_message_id;
37static int use_inbody_headers = 1;
38
39#define MAX_BOUNDARIES 5
40
41static void cleanup_space(struct strbuf *sb)
42{
43 size_t pos, cnt;
44 for (pos = 0; pos < sb->len; pos++) {
45 if (isspace(sb->buf[pos])) {
46 sb->buf[pos] = ' ';
47 for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
48 strbuf_remove(sb, pos + 1, cnt);
49 }
50 }
51}
52
53static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
54{
55 struct strbuf *src = name;
56 if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
57 strchr(name->buf, '<') || strchr(name->buf, '>'))
58 src = email;
59 else if (name == out)
60 return;
61 strbuf_reset(out);
62 strbuf_addbuf(out, src);
63}
64
65static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
66{
67 /* John Doe <johndoe> */
68
69 char *bra, *ket;
70 /* This is fallback, so do not bother if we already have an
71 * e-mail address.
72 */
73 if (mi->email.len)
74 return;
75
76 bra = strchr(line->buf, '<');
77 if (!bra)
78 return;
79 ket = strchr(bra, '>');
80 if (!ket)
81 return;
82
83 strbuf_reset(&mi->email);
84 strbuf_add(&mi->email, bra + 1, ket - bra - 1);
85
86 strbuf_reset(&mi->name);
87 strbuf_add(&mi->name, line->buf, bra - line->buf);
88 strbuf_trim(&mi->name);
89 get_sane_name(&mi->name, &mi->name, &mi->email);
90}
91
92static void handle_from(struct mailinfo *mi, const struct strbuf *from)
93{
94 char *at;
95 size_t el;
96 struct strbuf f;
97
98 strbuf_init(&f, from->len);
99 strbuf_addbuf(&f, from);
100
101 at = strchr(f.buf, '@');
102 if (!at) {
103 parse_bogus_from(mi, from);
104 return;
105 }
106
107 /*
108 * If we already have one email, don't take any confusing lines
109 */
110 if (mi->email.len && strchr(at + 1, '@')) {
111 strbuf_release(&f);
112 return;
113 }
114
115 /* Pick up the string around '@', possibly delimited with <>
116 * pair; that is the email part.
117 */
118 while (at > f.buf) {
119 char c = at[-1];
120 if (isspace(c))
121 break;
122 if (c == '<') {
123 at[-1] = ' ';
124 break;
125 }
126 at--;
127 }
128 el = strcspn(at, " \n\t\r\v\f>");
129 strbuf_reset(&mi->email);
130 strbuf_add(&mi->email, at, el);
131 strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
132
133 /* The remainder is name. It could be
134 *
135 * - "John Doe <john.doe@xz>" (a), or
136 * - "john.doe@xz (John Doe)" (b), or
137 * - "John (zzz) Doe <john.doe@xz> (Comment)" (c)
138 *
139 * but we have removed the email part, so
140 *
141 * - remove extra spaces which could stay after email (case 'c'), and
142 * - trim from both ends, possibly removing the () pair at the end
143 * (cases 'a' and 'b').
144 */
145 cleanup_space(&f);
146 strbuf_trim(&f);
147 if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
148 strbuf_remove(&f, 0, 1);
149 strbuf_setlen(&f, f.len - 1);
150 }
151
152 get_sane_name(&mi->name, &f, &mi->email);
153 strbuf_release(&f);
154}
155
156static void handle_header(struct strbuf **out, const struct strbuf *line)
157{
158 if (!*out) {
159 *out = xmalloc(sizeof(struct strbuf));
160 strbuf_init(*out, line->len);
161 } else
162 strbuf_reset(*out);
163
164 strbuf_addbuf(*out, line);
165}
166
167/* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
168 * to have enough heuristics to grok MIME encoded patches often found
169 * on our mailing lists. For example, we do not even treat header lines
170 * case insensitively.
171 */
172
173static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
174{
175 const char *ends, *ap = strcasestr(line, name);
176 size_t sz;
177
178 strbuf_setlen(attr, 0);
179 if (!ap)
180 return 0;
181 ap += strlen(name);
182 if (*ap == '"') {
183 ap++;
184 ends = "\"";
185 }
186 else
187 ends = "; \t";
188 sz = strcspn(ap, ends);
189 strbuf_add(attr, ap, sz);
190 return 1;
191}
192
193static struct strbuf *content[MAX_BOUNDARIES];
194
195static struct strbuf **content_top = content;
196
197static void handle_content_type(struct strbuf *line)
198{
199 struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
200 strbuf_init(boundary, line->len);
201
202 if (slurp_attr(line->buf, "boundary=", boundary)) {
203 strbuf_insert(boundary, 0, "--", 2);
204 if (++content_top >= &content[MAX_BOUNDARIES]) {
205 fprintf(stderr, "Too many boundaries to handle\n");
206 exit(1);
207 }
208 *content_top = boundary;
209 boundary = NULL;
210 }
211 slurp_attr(line->buf, "charset=", &charset);
212
213 if (boundary) {
214 strbuf_release(boundary);
215 free(boundary);
216 }
217}
218
219static void handle_message_id(const struct strbuf *line)
220{
221 if (add_message_id)
222 message_id = strdup(line->buf);
223}
224
225static void handle_content_transfer_encoding(const struct strbuf *line)
226{
227 if (strcasestr(line->buf, "base64"))
228 transfer_encoding = TE_BASE64;
229 else if (strcasestr(line->buf, "quoted-printable"))
230 transfer_encoding = TE_QP;
231 else
232 transfer_encoding = TE_DONTCARE;
233}
234
235static int is_multipart_boundary(const struct strbuf *line)
236{
237 return (((*content_top)->len <= line->len) &&
238 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
239}
240
241static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
242{
243 size_t at = 0;
244
245 while (at < subject->len) {
246 char *pos;
247 size_t remove;
248
249 switch (subject->buf[at]) {
250 case 'r': case 'R':
251 if (subject->len <= at + 3)
252 break;
253 if ((subject->buf[at + 1] == 'e' ||
254 subject->buf[at + 1] == 'E') &&
255 subject->buf[at + 2] == ':') {
256 strbuf_remove(subject, at, 3);
257 continue;
258 }
259 at++;
260 break;
261 case ' ': case '\t': case ':':
262 strbuf_remove(subject, at, 1);
263 continue;
264 case '[':
265 pos = strchr(subject->buf + at, ']');
266 if (!pos)
267 break;
268 remove = pos - subject->buf + at + 1;
269 if (!mi->keep_non_patch_brackets_in_subject ||
270 (7 <= remove &&
271 memmem(subject->buf + at, remove, "PATCH", 5)))
272 strbuf_remove(subject, at, remove);
273 else {
274 at += remove;
275 /*
276 * If the input had a space after the ], keep
277 * it. We don't bother with finding the end of
278 * the space, since we later normalize it
279 * anyway.
280 */
281 if (isspace(subject->buf[at]))
282 at += 1;
283 }
284 continue;
285 }
286 break;
287 }
288 strbuf_trim(subject);
289}
290
291#define MAX_HDR_PARSED 10
292static const char *header[MAX_HDR_PARSED] = {
293 "From","Subject","Date",
294};
295
296static inline int cmp_header(const struct strbuf *line, const char *hdr)
297{
298 int len = strlen(hdr);
299 return !strncasecmp(line->buf, hdr, len) && line->len > len &&
300 line->buf[len] == ':' && isspace(line->buf[len + 1]);
301}
302
303static int is_format_patch_separator(const char *line, int len)
304{
305 static const char SAMPLE[] =
306 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
307 const char *cp;
308
309 if (len != strlen(SAMPLE))
310 return 0;
311 if (!skip_prefix(line, "From ", &cp))
312 return 0;
313 if (strspn(cp, "0123456789abcdef") != 40)
314 return 0;
315 cp += 40;
316 return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
317}
318
319static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
320{
321 const char *in = q_seg->buf;
322 int c;
323 struct strbuf *out = xmalloc(sizeof(struct strbuf));
324 strbuf_init(out, q_seg->len);
325
326 while ((c = *in++) != 0) {
327 if (c == '=') {
328 int d = *in++;
329 if (d == '\n' || !d)
330 break; /* drop trailing newline */
331 strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
332 continue;
333 }
334 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
335 c = 0x20;
336 strbuf_addch(out, c);
337 }
338 return out;
339}
340
341static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
342{
343 /* Decode in..ep, possibly in-place to ot */
344 int c, pos = 0, acc = 0;
345 const char *in = b_seg->buf;
346 struct strbuf *out = xmalloc(sizeof(struct strbuf));
347 strbuf_init(out, b_seg->len);
348
349 while ((c = *in++) != 0) {
350 if (c == '+')
351 c = 62;
352 else if (c == '/')
353 c = 63;
354 else if ('A' <= c && c <= 'Z')
355 c -= 'A';
356 else if ('a' <= c && c <= 'z')
357 c -= 'a' - 26;
358 else if ('0' <= c && c <= '9')
359 c -= '0' - 52;
360 else
361 continue; /* garbage */
362 switch (pos++) {
363 case 0:
364 acc = (c << 2);
365 break;
366 case 1:
367 strbuf_addch(out, (acc | (c >> 4)));
368 acc = (c & 15) << 4;
369 break;
370 case 2:
371 strbuf_addch(out, (acc | (c >> 2)));
372 acc = (c & 3) << 6;
373 break;
374 case 3:
375 strbuf_addch(out, (acc | c));
376 acc = pos = 0;
377 break;
378 }
379 }
380 return out;
381}
382
383static void convert_to_utf8(struct strbuf *line, const char *charset)
384{
385 char *out;
386
387 if (!charset || !*charset)
388 return;
389
390 if (same_encoding(metainfo_charset, charset))
391 return;
392 out = reencode_string(line->buf, metainfo_charset, charset);
393 if (!out)
394 die("cannot convert from %s to %s",
395 charset, metainfo_charset);
396 strbuf_attach(line, out, strlen(out), strlen(out));
397}
398
399static void decode_header(struct strbuf *it)
400{
401 char *in, *ep, *cp;
402 struct strbuf outbuf = STRBUF_INIT, *dec;
403 struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
404
405 in = it->buf;
406 while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
407 int encoding;
408 strbuf_reset(&charset_q);
409 strbuf_reset(&piecebuf);
410
411 if (in != ep) {
412 /*
413 * We are about to process an encoded-word
414 * that begins at ep, but there is something
415 * before the encoded word.
416 */
417 char *scan;
418 for (scan = in; scan < ep; scan++)
419 if (!isspace(*scan))
420 break;
421
422 if (scan != ep || in == it->buf) {
423 /*
424 * We should not lose that "something",
425 * unless we have just processed an
426 * encoded-word, and there is only LWS
427 * before the one we are about to process.
428 */
429 strbuf_add(&outbuf, in, ep - in);
430 }
431 }
432 /* E.g.
433 * ep : "=?iso-2022-jp?B?GyR...?= foo"
434 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
435 */
436 ep += 2;
437
438 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
439 goto release_return;
440
441 if (cp + 3 - it->buf > it->len)
442 goto release_return;
443 strbuf_add(&charset_q, ep, cp - ep);
444
445 encoding = cp[1];
446 if (!encoding || cp[2] != '?')
447 goto release_return;
448 ep = strstr(cp + 3, "?=");
449 if (!ep)
450 goto release_return;
451 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
452 switch (tolower(encoding)) {
453 default:
454 goto release_return;
455 case 'b':
456 dec = decode_b_segment(&piecebuf);
457 break;
458 case 'q':
459 dec = decode_q_segment(&piecebuf, 1);
460 break;
461 }
462 if (metainfo_charset)
463 convert_to_utf8(dec, charset_q.buf);
464
465 strbuf_addbuf(&outbuf, dec);
466 strbuf_release(dec);
467 free(dec);
468 in = ep + 2;
469 }
470 strbuf_addstr(&outbuf, in);
471 strbuf_reset(it);
472 strbuf_addbuf(it, &outbuf);
473release_return:
474 strbuf_release(&outbuf);
475 strbuf_release(&charset_q);
476 strbuf_release(&piecebuf);
477}
478
479static int check_header(const struct strbuf *line,
480 struct strbuf *hdr_data[], int overwrite)
481{
482 int i, ret = 0, len;
483 struct strbuf sb = STRBUF_INIT;
484 /* search for the interesting parts */
485 for (i = 0; header[i]; i++) {
486 int len = strlen(header[i]);
487 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
488 /* Unwrap inline B and Q encoding, and optionally
489 * normalize the meta information to utf8.
490 */
491 strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
492 decode_header(&sb);
493 handle_header(&hdr_data[i], &sb);
494 ret = 1;
495 goto check_header_out;
496 }
497 }
498
499 /* Content stuff */
500 if (cmp_header(line, "Content-Type")) {
501 len = strlen("Content-Type: ");
502 strbuf_add(&sb, line->buf + len, line->len - len);
503 decode_header(&sb);
504 strbuf_insert(&sb, 0, "Content-Type: ", len);
505 handle_content_type(&sb);
506 ret = 1;
507 goto check_header_out;
508 }
509 if (cmp_header(line, "Content-Transfer-Encoding")) {
510 len = strlen("Content-Transfer-Encoding: ");
511 strbuf_add(&sb, line->buf + len, line->len - len);
512 decode_header(&sb);
513 handle_content_transfer_encoding(&sb);
514 ret = 1;
515 goto check_header_out;
516 }
517 if (cmp_header(line, "Message-Id")) {
518 len = strlen("Message-Id: ");
519 strbuf_add(&sb, line->buf + len, line->len - len);
520 decode_header(&sb);
521 handle_message_id(&sb);
522 ret = 1;
523 goto check_header_out;
524 }
525
526 /* for inbody stuff */
527 if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
528 ret = is_format_patch_separator(line->buf + 1, line->len - 1);
529 goto check_header_out;
530 }
531 if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
532 for (i = 0; header[i]; i++) {
533 if (!strcmp("Subject", header[i])) {
534 handle_header(&hdr_data[i], line);
535 ret = 1;
536 goto check_header_out;
537 }
538 }
539 }
540
541check_header_out:
542 strbuf_release(&sb);
543 return ret;
544}
545
546static void decode_transfer_encoding(struct strbuf *line)
547{
548 struct strbuf *ret;
549
550 switch (transfer_encoding) {
551 case TE_QP:
552 ret = decode_q_segment(line, 0);
553 break;
554 case TE_BASE64:
555 ret = decode_b_segment(line);
556 break;
557 case TE_DONTCARE:
558 default:
559 return;
560 }
561 strbuf_reset(line);
562 strbuf_addbuf(line, ret);
563 strbuf_release(ret);
564 free(ret);
565}
566
567static inline int patchbreak(const struct strbuf *line)
568{
569 size_t i;
570
571 /* Beginning of a "diff -" header? */
572 if (starts_with(line->buf, "diff -"))
573 return 1;
574
575 /* CVS "Index: " line? */
576 if (starts_with(line->buf, "Index: "))
577 return 1;
578
579 /*
580 * "--- <filename>" starts patches without headers
581 * "---<sp>*" is a manual separator
582 */
583 if (line->len < 4)
584 return 0;
585
586 if (starts_with(line->buf, "---")) {
587 /* space followed by a filename? */
588 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
589 return 1;
590 /* Just whitespace? */
591 for (i = 3; i < line->len; i++) {
592 unsigned char c = line->buf[i];
593 if (c == '\n')
594 return 1;
595 if (!isspace(c))
596 break;
597 }
598 return 0;
599 }
600 return 0;
601}
602
603static int is_scissors_line(const struct strbuf *line)
604{
605 size_t i, len = line->len;
606 int scissors = 0, gap = 0;
607 int first_nonblank = -1;
608 int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
609 const char *buf = line->buf;
610
611 for (i = 0; i < len; i++) {
612 if (isspace(buf[i])) {
613 if (in_perforation) {
614 perforation++;
615 gap++;
616 }
617 continue;
618 }
619 last_nonblank = i;
620 if (first_nonblank < 0)
621 first_nonblank = i;
622 if (buf[i] == '-') {
623 in_perforation = 1;
624 perforation++;
625 continue;
626 }
627 if (i + 1 < len &&
628 (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
629 !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
630 in_perforation = 1;
631 perforation += 2;
632 scissors += 2;
633 i++;
634 continue;
635 }
636 in_perforation = 0;
637 }
638
639 /*
640 * The mark must be at least 8 bytes long (e.g. "-- >8 --").
641 * Even though there can be arbitrary cruft on the same line
642 * (e.g. "cut here"), in order to avoid misidentification, the
643 * perforation must occupy more than a third of the visible
644 * width of the line, and dashes and scissors must occupy more
645 * than half of the perforation.
646 */
647
648 visible = last_nonblank - first_nonblank + 1;
649 return (scissors && 8 <= visible &&
650 visible < perforation * 3 &&
651 gap * 2 < perforation);
652}
653
654static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
655{
656 if (!cmitmsg)
657 return 0;
658
659 if (mi->header_stage) {
660 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
661 return 0;
662 }
663
664 if (use_inbody_headers && mi->header_stage) {
665 mi->header_stage = check_header(line, s_hdr_data, 0);
666 if (mi->header_stage)
667 return 0;
668 } else
669 /* Only trim the first (blank) line of the commit message
670 * when ignoring in-body headers.
671 */
672 mi->header_stage = 0;
673
674 /* normalize the log message to UTF-8. */
675 if (metainfo_charset)
676 convert_to_utf8(line, charset.buf);
677
678 if (use_scissors && is_scissors_line(line)) {
679 int i;
680 if (fseek(cmitmsg, 0L, SEEK_SET))
681 die_errno("Could not rewind output message file");
682 if (ftruncate(fileno(cmitmsg), 0))
683 die_errno("Could not truncate output message file at scissors");
684 mi->header_stage = 1;
685
686 /*
687 * We may have already read "secondary headers"; purge
688 * them to give ourselves a clean restart.
689 */
690 for (i = 0; header[i]; i++) {
691 if (s_hdr_data[i])
692 strbuf_release(s_hdr_data[i]);
693 s_hdr_data[i] = NULL;
694 }
695 return 0;
696 }
697
698 if (patchbreak(line)) {
699 if (message_id)
700 fprintf(cmitmsg, "Message-Id: %s\n", message_id);
701 fclose(cmitmsg);
702 cmitmsg = NULL;
703 return 1;
704 }
705
706 fputs(line->buf, cmitmsg);
707 return 0;
708}
709
710static void handle_patch(const struct strbuf *line)
711{
712 fwrite(line->buf, 1, line->len, patchfile);
713 patch_lines++;
714}
715
716static void handle_filter(struct mailinfo *mi, struct strbuf *line)
717{
718 switch (mi->filter_stage) {
719 case 0:
720 if (!handle_commit_msg(mi, line))
721 break;
722 mi->filter_stage++;
723 case 1:
724 handle_patch(line);
725 break;
726 }
727}
728
729static int is_rfc2822_header(const struct strbuf *line)
730{
731 /*
732 * The section that defines the loosest possible
733 * field name is "3.6.8 Optional fields".
734 *
735 * optional-field = field-name ":" unstructured CRLF
736 * field-name = 1*ftext
737 * ftext = %d33-57 / %59-126
738 */
739 int ch;
740 char *cp = line->buf;
741
742 /* Count mbox From headers as headers */
743 if (starts_with(cp, "From ") || starts_with(cp, ">From "))
744 return 1;
745
746 while ((ch = *cp++)) {
747 if (ch == ':')
748 return 1;
749 if ((33 <= ch && ch <= 57) ||
750 (59 <= ch && ch <= 126))
751 continue;
752 break;
753 }
754 return 0;
755}
756
757static int read_one_header_line(struct strbuf *line, FILE *in)
758{
759 struct strbuf continuation = STRBUF_INIT;
760
761 /* Get the first part of the line. */
762 if (strbuf_getline(line, in, '\n'))
763 return 0;
764
765 /*
766 * Is it an empty line or not a valid rfc2822 header?
767 * If so, stop here, and return false ("not a header")
768 */
769 strbuf_rtrim(line);
770 if (!line->len || !is_rfc2822_header(line)) {
771 /* Re-add the newline */
772 strbuf_addch(line, '\n');
773 return 0;
774 }
775
776 /*
777 * Now we need to eat all the continuation lines..
778 * Yuck, 2822 header "folding"
779 */
780 for (;;) {
781 int peek;
782
783 peek = fgetc(in); ungetc(peek, in);
784 if (peek != ' ' && peek != '\t')
785 break;
786 if (strbuf_getline(&continuation, in, '\n'))
787 break;
788 continuation.buf[0] = ' ';
789 strbuf_rtrim(&continuation);
790 strbuf_addbuf(line, &continuation);
791 }
792 strbuf_release(&continuation);
793
794 return 1;
795}
796
797static int find_boundary(struct mailinfo *mi, struct strbuf *line)
798{
799 while (!strbuf_getline(line, mi->input, '\n')) {
800 if (*content_top && is_multipart_boundary(line))
801 return 1;
802 }
803 return 0;
804}
805
806static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
807{
808 struct strbuf newline = STRBUF_INIT;
809
810 strbuf_addch(&newline, '\n');
811again:
812 if (line->len >= (*content_top)->len + 2 &&
813 !memcmp(line->buf + (*content_top)->len, "--", 2)) {
814 /* we hit an end boundary */
815 /* pop the current boundary off the stack */
816 strbuf_release(*content_top);
817 free(*content_top);
818 *content_top = NULL;
819
820 /* technically won't happen as is_multipart_boundary()
821 will fail first. But just in case..
822 */
823 if (--content_top < content) {
824 fprintf(stderr, "Detected mismatched boundaries, "
825 "can't recover\n");
826 exit(1);
827 }
828 handle_filter(mi, &newline);
829 strbuf_release(&newline);
830
831 /* skip to the next boundary */
832 if (!find_boundary(mi, line))
833 return 0;
834 goto again;
835 }
836
837 /* set some defaults */
838 transfer_encoding = TE_DONTCARE;
839 strbuf_reset(&charset);
840
841 /* slurp in this section's info */
842 while (read_one_header_line(line, mi->input))
843 check_header(line, p_hdr_data, 0);
844
845 strbuf_release(&newline);
846 /* replenish line */
847 if (strbuf_getline(line, mi->input, '\n'))
848 return 0;
849 strbuf_addch(line, '\n');
850 return 1;
851}
852
853static void handle_body(struct mailinfo *mi, struct strbuf *line)
854{
855 struct strbuf prev = STRBUF_INIT;
856
857 /* Skip up to the first boundary */
858 if (*content_top) {
859 if (!find_boundary(mi, line))
860 goto handle_body_out;
861 }
862
863 do {
864 /* process any boundary lines */
865 if (*content_top && is_multipart_boundary(line)) {
866 /* flush any leftover */
867 if (prev.len) {
868 handle_filter(mi, &prev);
869 strbuf_reset(&prev);
870 }
871 if (!handle_boundary(mi, line))
872 goto handle_body_out;
873 }
874
875 /* Unwrap transfer encoding */
876 decode_transfer_encoding(line);
877
878 switch (transfer_encoding) {
879 case TE_BASE64:
880 case TE_QP:
881 {
882 struct strbuf **lines, **it, *sb;
883
884 /* Prepend any previous partial lines */
885 strbuf_insert(line, 0, prev.buf, prev.len);
886 strbuf_reset(&prev);
887
888 /*
889 * This is a decoded line that may contain
890 * multiple new lines. Pass only one chunk
891 * at a time to handle_filter()
892 */
893 lines = strbuf_split(line, '\n');
894 for (it = lines; (sb = *it); it++) {
895 if (*(it + 1) == NULL) /* The last line */
896 if (sb->buf[sb->len - 1] != '\n') {
897 /* Partial line, save it for later. */
898 strbuf_addbuf(&prev, sb);
899 break;
900 }
901 handle_filter(mi, sb);
902 }
903 /*
904 * The partial chunk is saved in "prev" and will be
905 * appended by the next iteration of read_line_with_nul().
906 */
907 strbuf_list_free(lines);
908 break;
909 }
910 default:
911 handle_filter(mi, line);
912 }
913
914 } while (!strbuf_getwholeline(line, mi->input, '\n'));
915
916handle_body_out:
917 strbuf_release(&prev);
918}
919
920static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
921{
922 const char *sp = data->buf;
923 while (1) {
924 char *ep = strchr(sp, '\n');
925 int len;
926 if (!ep)
927 len = strlen(sp);
928 else
929 len = ep - sp;
930 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
931 if (!ep)
932 break;
933 sp = ep + 1;
934 }
935}
936
937static void handle_info(struct mailinfo *mi)
938{
939 struct strbuf *hdr;
940 int i;
941
942 for (i = 0; header[i]; i++) {
943 /* only print inbody headers if we output a patch file */
944 if (patch_lines && s_hdr_data[i])
945 hdr = s_hdr_data[i];
946 else if (p_hdr_data[i])
947 hdr = p_hdr_data[i];
948 else
949 continue;
950
951 if (!strcmp(header[i], "Subject")) {
952 if (!mi->keep_subject) {
953 cleanup_subject(mi, hdr);
954 cleanup_space(hdr);
955 }
956 output_header_lines(mi->output, "Subject", hdr);
957 } else if (!strcmp(header[i], "From")) {
958 cleanup_space(hdr);
959 handle_from(mi, hdr);
960 fprintf(mi->output, "Author: %s\n", mi->name.buf);
961 fprintf(mi->output, "Email: %s\n", mi->email.buf);
962 } else {
963 cleanup_space(hdr);
964 fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
965 }
966 }
967 fprintf(mi->output, "\n");
968}
969
970static int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
971{
972 int peek;
973 struct strbuf line = STRBUF_INIT;
974
975 cmitmsg = fopen(msg, "w");
976 if (!cmitmsg) {
977 perror(msg);
978 return -1;
979 }
980 patchfile = fopen(patch, "w");
981 if (!patchfile) {
982 perror(patch);
983 fclose(cmitmsg);
984 return -1;
985 }
986
987 p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
988 s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
989
990 do {
991 peek = fgetc(mi->input);
992 } while (isspace(peek));
993 ungetc(peek, mi->input);
994
995 /* process the email header */
996 while (read_one_header_line(&line, mi->input))
997 check_header(&line, p_hdr_data, 1);
998
999 handle_body(mi, &line);
1000 fclose(patchfile);
1001
1002 handle_info(mi);
1003 strbuf_release(&line);
1004 return 0;
1005}
1006
1007static int git_mailinfo_config(const char *var, const char *value, void *unused)
1008{
1009 if (!starts_with(var, "mailinfo."))
1010 return git_default_config(var, value, unused);
1011 if (!strcmp(var, "mailinfo.scissors")) {
1012 use_scissors = git_config_bool(var, value);
1013 return 0;
1014 }
1015 /* perhaps others here */
1016 return 0;
1017}
1018
1019static void setup_mailinfo(struct mailinfo *mi)
1020{
1021 memset(mi, 0, sizeof(*mi));
1022 strbuf_init(&mi->name, 0);
1023 strbuf_init(&mi->email, 0);
1024 mi->header_stage = 1;
1025 git_config(git_mailinfo_config, &mi);
1026}
1027
1028static void clear_mailinfo(struct mailinfo *mi)
1029{
1030 strbuf_release(&mi->name);
1031 strbuf_release(&mi->email);
1032}
1033
1034static const char mailinfo_usage[] =
1035 "git mailinfo [-k | -b] [-m | --message-id] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] <msg> <patch> < mail >info";
1036
1037int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1038{
1039 const char *def_charset;
1040 struct mailinfo mi;
1041 int status;
1042
1043 /* NEEDSWORK: might want to do the optional .git/ directory
1044 * discovery
1045 */
1046 setup_mailinfo(&mi);
1047
1048 def_charset = get_commit_output_encoding();
1049 metainfo_charset = def_charset;
1050
1051 while (1 < argc && argv[1][0] == '-') {
1052 if (!strcmp(argv[1], "-k"))
1053 mi.keep_subject = 1;
1054 else if (!strcmp(argv[1], "-b"))
1055 mi.keep_non_patch_brackets_in_subject = 1;
1056 else if (!strcmp(argv[1], "-m") || !strcmp(argv[1], "--message-id"))
1057 add_message_id = 1;
1058 else if (!strcmp(argv[1], "-u"))
1059 metainfo_charset = def_charset;
1060 else if (!strcmp(argv[1], "-n"))
1061 metainfo_charset = NULL;
1062 else if (starts_with(argv[1], "--encoding="))
1063 metainfo_charset = argv[1] + 11;
1064 else if (!strcmp(argv[1], "--scissors"))
1065 use_scissors = 1;
1066 else if (!strcmp(argv[1], "--no-scissors"))
1067 use_scissors = 0;
1068 else if (!strcmp(argv[1], "--no-inbody-headers"))
1069 use_inbody_headers = 0;
1070 else
1071 usage(mailinfo_usage);
1072 argc--; argv++;
1073 }
1074
1075 if (argc != 3)
1076 usage(mailinfo_usage);
1077
1078 mi.input = stdin;
1079 mi.output = stdout;
1080 status = !!mailinfo(&mi, argv[1], argv[2]);
1081 clear_mailinfo(&mi);
1082
1083 return status;
1084}