a0416610c95b5c5653eb394f5337379c122cf619
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, *fin, *fout;
11
12static int keep_subject;
13static int keep_non_patch_brackets_in_subject;
14static const char *metainfo_charset;
15static struct strbuf name = STRBUF_INIT;
16static struct strbuf email = STRBUF_INIT;
17static char *message_id;
18
19static enum {
20 TE_DONTCARE, TE_QP, TE_BASE64
21} transfer_encoding;
22
23static struct strbuf charset = STRBUF_INIT;
24static int patch_lines;
25static struct strbuf **p_hdr_data, **s_hdr_data;
26static int use_scissors;
27static int add_message_id;
28static int use_inbody_headers = 1;
29
30#define MAX_BOUNDARIES 5
31
32static void cleanup_space(struct strbuf *sb)
33{
34 size_t pos, cnt;
35 for (pos = 0; pos < sb->len; pos++) {
36 if (isspace(sb->buf[pos])) {
37 sb->buf[pos] = ' ';
38 for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
39 strbuf_remove(sb, pos + 1, cnt);
40 }
41 }
42}
43
44static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
45{
46 struct strbuf *src = name;
47 if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
48 strchr(name->buf, '<') || strchr(name->buf, '>'))
49 src = email;
50 else if (name == out)
51 return;
52 strbuf_reset(out);
53 strbuf_addbuf(out, src);
54}
55
56static void parse_bogus_from(const struct strbuf *line)
57{
58 /* John Doe <johndoe> */
59
60 char *bra, *ket;
61 /* This is fallback, so do not bother if we already have an
62 * e-mail address.
63 */
64 if (email.len)
65 return;
66
67 bra = strchr(line->buf, '<');
68 if (!bra)
69 return;
70 ket = strchr(bra, '>');
71 if (!ket)
72 return;
73
74 strbuf_reset(&email);
75 strbuf_add(&email, bra + 1, ket - bra - 1);
76
77 strbuf_reset(&name);
78 strbuf_add(&name, line->buf, bra - line->buf);
79 strbuf_trim(&name);
80 get_sane_name(&name, &name, &email);
81}
82
83static void handle_from(const struct strbuf *from)
84{
85 char *at;
86 size_t el;
87 struct strbuf f;
88
89 strbuf_init(&f, from->len);
90 strbuf_addbuf(&f, from);
91
92 at = strchr(f.buf, '@');
93 if (!at) {
94 parse_bogus_from(from);
95 return;
96 }
97
98 /*
99 * If we already have one email, don't take any confusing lines
100 */
101 if (email.len && strchr(at + 1, '@')) {
102 strbuf_release(&f);
103 return;
104 }
105
106 /* Pick up the string around '@', possibly delimited with <>
107 * pair; that is the email part.
108 */
109 while (at > f.buf) {
110 char c = at[-1];
111 if (isspace(c))
112 break;
113 if (c == '<') {
114 at[-1] = ' ';
115 break;
116 }
117 at--;
118 }
119 el = strcspn(at, " \n\t\r\v\f>");
120 strbuf_reset(&email);
121 strbuf_add(&email, at, el);
122 strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
123
124 /* The remainder is name. It could be
125 *
126 * - "John Doe <john.doe@xz>" (a), or
127 * - "john.doe@xz (John Doe)" (b), or
128 * - "John (zzz) Doe <john.doe@xz> (Comment)" (c)
129 *
130 * but we have removed the email part, so
131 *
132 * - remove extra spaces which could stay after email (case 'c'), and
133 * - trim from both ends, possibly removing the () pair at the end
134 * (cases 'a' and 'b').
135 */
136 cleanup_space(&f);
137 strbuf_trim(&f);
138 if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
139 strbuf_remove(&f, 0, 1);
140 strbuf_setlen(&f, f.len - 1);
141 }
142
143 get_sane_name(&name, &f, &email);
144 strbuf_release(&f);
145}
146
147static void handle_header(struct strbuf **out, const struct strbuf *line)
148{
149 if (!*out) {
150 *out = xmalloc(sizeof(struct strbuf));
151 strbuf_init(*out, line->len);
152 } else
153 strbuf_reset(*out);
154
155 strbuf_addbuf(*out, line);
156}
157
158/* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
159 * to have enough heuristics to grok MIME encoded patches often found
160 * on our mailing lists. For example, we do not even treat header lines
161 * case insensitively.
162 */
163
164static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
165{
166 const char *ends, *ap = strcasestr(line, name);
167 size_t sz;
168
169 strbuf_setlen(attr, 0);
170 if (!ap)
171 return 0;
172 ap += strlen(name);
173 if (*ap == '"') {
174 ap++;
175 ends = "\"";
176 }
177 else
178 ends = "; \t";
179 sz = strcspn(ap, ends);
180 strbuf_add(attr, ap, sz);
181 return 1;
182}
183
184static struct strbuf *content[MAX_BOUNDARIES];
185
186static struct strbuf **content_top = content;
187
188static void handle_content_type(struct strbuf *line)
189{
190 struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
191 strbuf_init(boundary, line->len);
192
193 if (slurp_attr(line->buf, "boundary=", boundary)) {
194 strbuf_insert(boundary, 0, "--", 2);
195 if (++content_top >= &content[MAX_BOUNDARIES]) {
196 fprintf(stderr, "Too many boundaries to handle\n");
197 exit(1);
198 }
199 *content_top = boundary;
200 boundary = NULL;
201 }
202 slurp_attr(line->buf, "charset=", &charset);
203
204 if (boundary) {
205 strbuf_release(boundary);
206 free(boundary);
207 }
208}
209
210static void handle_message_id(const struct strbuf *line)
211{
212 if (add_message_id)
213 message_id = strdup(line->buf);
214}
215
216static void handle_content_transfer_encoding(const struct strbuf *line)
217{
218 if (strcasestr(line->buf, "base64"))
219 transfer_encoding = TE_BASE64;
220 else if (strcasestr(line->buf, "quoted-printable"))
221 transfer_encoding = TE_QP;
222 else
223 transfer_encoding = TE_DONTCARE;
224}
225
226static int is_multipart_boundary(const struct strbuf *line)
227{
228 return (((*content_top)->len <= line->len) &&
229 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
230}
231
232static void cleanup_subject(struct strbuf *subject)
233{
234 size_t at = 0;
235
236 while (at < subject->len) {
237 char *pos;
238 size_t remove;
239
240 switch (subject->buf[at]) {
241 case 'r': case 'R':
242 if (subject->len <= at + 3)
243 break;
244 if ((subject->buf[at + 1] == 'e' ||
245 subject->buf[at + 1] == 'E') &&
246 subject->buf[at + 2] == ':') {
247 strbuf_remove(subject, at, 3);
248 continue;
249 }
250 at++;
251 break;
252 case ' ': case '\t': case ':':
253 strbuf_remove(subject, at, 1);
254 continue;
255 case '[':
256 pos = strchr(subject->buf + at, ']');
257 if (!pos)
258 break;
259 remove = pos - subject->buf + at + 1;
260 if (!keep_non_patch_brackets_in_subject ||
261 (7 <= remove &&
262 memmem(subject->buf + at, remove, "PATCH", 5)))
263 strbuf_remove(subject, at, remove);
264 else {
265 at += remove;
266 /*
267 * If the input had a space after the ], keep
268 * it. We don't bother with finding the end of
269 * the space, since we later normalize it
270 * anyway.
271 */
272 if (isspace(subject->buf[at]))
273 at += 1;
274 }
275 continue;
276 }
277 break;
278 }
279 strbuf_trim(subject);
280}
281
282#define MAX_HDR_PARSED 10
283static const char *header[MAX_HDR_PARSED] = {
284 "From","Subject","Date",
285};
286
287static inline int cmp_header(const struct strbuf *line, const char *hdr)
288{
289 int len = strlen(hdr);
290 return !strncasecmp(line->buf, hdr, len) && line->len > len &&
291 line->buf[len] == ':' && isspace(line->buf[len + 1]);
292}
293
294static int is_format_patch_separator(const char *line, int len)
295{
296 static const char SAMPLE[] =
297 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
298 const char *cp;
299
300 if (len != strlen(SAMPLE))
301 return 0;
302 if (!skip_prefix(line, "From ", &cp))
303 return 0;
304 if (strspn(cp, "0123456789abcdef") != 40)
305 return 0;
306 cp += 40;
307 return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
308}
309
310static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
311{
312 const char *in = q_seg->buf;
313 int c;
314 struct strbuf *out = xmalloc(sizeof(struct strbuf));
315 strbuf_init(out, q_seg->len);
316
317 while ((c = *in++) != 0) {
318 if (c == '=') {
319 int d = *in++;
320 if (d == '\n' || !d)
321 break; /* drop trailing newline */
322 strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
323 continue;
324 }
325 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
326 c = 0x20;
327 strbuf_addch(out, c);
328 }
329 return out;
330}
331
332static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
333{
334 /* Decode in..ep, possibly in-place to ot */
335 int c, pos = 0, acc = 0;
336 const char *in = b_seg->buf;
337 struct strbuf *out = xmalloc(sizeof(struct strbuf));
338 strbuf_init(out, b_seg->len);
339
340 while ((c = *in++) != 0) {
341 if (c == '+')
342 c = 62;
343 else if (c == '/')
344 c = 63;
345 else if ('A' <= c && c <= 'Z')
346 c -= 'A';
347 else if ('a' <= c && c <= 'z')
348 c -= 'a' - 26;
349 else if ('0' <= c && c <= '9')
350 c -= '0' - 52;
351 else
352 continue; /* garbage */
353 switch (pos++) {
354 case 0:
355 acc = (c << 2);
356 break;
357 case 1:
358 strbuf_addch(out, (acc | (c >> 4)));
359 acc = (c & 15) << 4;
360 break;
361 case 2:
362 strbuf_addch(out, (acc | (c >> 2)));
363 acc = (c & 3) << 6;
364 break;
365 case 3:
366 strbuf_addch(out, (acc | c));
367 acc = pos = 0;
368 break;
369 }
370 }
371 return out;
372}
373
374static void convert_to_utf8(struct strbuf *line, const char *charset)
375{
376 char *out;
377
378 if (!charset || !*charset)
379 return;
380
381 if (same_encoding(metainfo_charset, charset))
382 return;
383 out = reencode_string(line->buf, metainfo_charset, charset);
384 if (!out)
385 die("cannot convert from %s to %s",
386 charset, metainfo_charset);
387 strbuf_attach(line, out, strlen(out), strlen(out));
388}
389
390static void decode_header(struct strbuf *it)
391{
392 char *in, *ep, *cp;
393 struct strbuf outbuf = STRBUF_INIT, *dec;
394 struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
395
396 in = it->buf;
397 while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
398 int encoding;
399 strbuf_reset(&charset_q);
400 strbuf_reset(&piecebuf);
401
402 if (in != ep) {
403 /*
404 * We are about to process an encoded-word
405 * that begins at ep, but there is something
406 * before the encoded word.
407 */
408 char *scan;
409 for (scan = in; scan < ep; scan++)
410 if (!isspace(*scan))
411 break;
412
413 if (scan != ep || in == it->buf) {
414 /*
415 * We should not lose that "something",
416 * unless we have just processed an
417 * encoded-word, and there is only LWS
418 * before the one we are about to process.
419 */
420 strbuf_add(&outbuf, in, ep - in);
421 }
422 }
423 /* E.g.
424 * ep : "=?iso-2022-jp?B?GyR...?= foo"
425 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
426 */
427 ep += 2;
428
429 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
430 goto release_return;
431
432 if (cp + 3 - it->buf > it->len)
433 goto release_return;
434 strbuf_add(&charset_q, ep, cp - ep);
435
436 encoding = cp[1];
437 if (!encoding || cp[2] != '?')
438 goto release_return;
439 ep = strstr(cp + 3, "?=");
440 if (!ep)
441 goto release_return;
442 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
443 switch (tolower(encoding)) {
444 default:
445 goto release_return;
446 case 'b':
447 dec = decode_b_segment(&piecebuf);
448 break;
449 case 'q':
450 dec = decode_q_segment(&piecebuf, 1);
451 break;
452 }
453 if (metainfo_charset)
454 convert_to_utf8(dec, charset_q.buf);
455
456 strbuf_addbuf(&outbuf, dec);
457 strbuf_release(dec);
458 free(dec);
459 in = ep + 2;
460 }
461 strbuf_addstr(&outbuf, in);
462 strbuf_reset(it);
463 strbuf_addbuf(it, &outbuf);
464release_return:
465 strbuf_release(&outbuf);
466 strbuf_release(&charset_q);
467 strbuf_release(&piecebuf);
468}
469
470static int check_header(const struct strbuf *line,
471 struct strbuf *hdr_data[], int overwrite)
472{
473 int i, ret = 0, len;
474 struct strbuf sb = STRBUF_INIT;
475 /* search for the interesting parts */
476 for (i = 0; header[i]; i++) {
477 int len = strlen(header[i]);
478 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
479 /* Unwrap inline B and Q encoding, and optionally
480 * normalize the meta information to utf8.
481 */
482 strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
483 decode_header(&sb);
484 handle_header(&hdr_data[i], &sb);
485 ret = 1;
486 goto check_header_out;
487 }
488 }
489
490 /* Content stuff */
491 if (cmp_header(line, "Content-Type")) {
492 len = strlen("Content-Type: ");
493 strbuf_add(&sb, line->buf + len, line->len - len);
494 decode_header(&sb);
495 strbuf_insert(&sb, 0, "Content-Type: ", len);
496 handle_content_type(&sb);
497 ret = 1;
498 goto check_header_out;
499 }
500 if (cmp_header(line, "Content-Transfer-Encoding")) {
501 len = strlen("Content-Transfer-Encoding: ");
502 strbuf_add(&sb, line->buf + len, line->len - len);
503 decode_header(&sb);
504 handle_content_transfer_encoding(&sb);
505 ret = 1;
506 goto check_header_out;
507 }
508 if (cmp_header(line, "Message-Id")) {
509 len = strlen("Message-Id: ");
510 strbuf_add(&sb, line->buf + len, line->len - len);
511 decode_header(&sb);
512 handle_message_id(&sb);
513 ret = 1;
514 goto check_header_out;
515 }
516
517 /* for inbody stuff */
518 if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
519 ret = is_format_patch_separator(line->buf + 1, line->len - 1);
520 goto check_header_out;
521 }
522 if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
523 for (i = 0; header[i]; i++) {
524 if (!strcmp("Subject", header[i])) {
525 handle_header(&hdr_data[i], line);
526 ret = 1;
527 goto check_header_out;
528 }
529 }
530 }
531
532check_header_out:
533 strbuf_release(&sb);
534 return ret;
535}
536
537static void decode_transfer_encoding(struct strbuf *line)
538{
539 struct strbuf *ret;
540
541 switch (transfer_encoding) {
542 case TE_QP:
543 ret = decode_q_segment(line, 0);
544 break;
545 case TE_BASE64:
546 ret = decode_b_segment(line);
547 break;
548 case TE_DONTCARE:
549 default:
550 return;
551 }
552 strbuf_reset(line);
553 strbuf_addbuf(line, ret);
554 strbuf_release(ret);
555 free(ret);
556}
557
558static inline int patchbreak(const struct strbuf *line)
559{
560 size_t i;
561
562 /* Beginning of a "diff -" header? */
563 if (starts_with(line->buf, "diff -"))
564 return 1;
565
566 /* CVS "Index: " line? */
567 if (starts_with(line->buf, "Index: "))
568 return 1;
569
570 /*
571 * "--- <filename>" starts patches without headers
572 * "---<sp>*" is a manual separator
573 */
574 if (line->len < 4)
575 return 0;
576
577 if (starts_with(line->buf, "---")) {
578 /* space followed by a filename? */
579 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
580 return 1;
581 /* Just whitespace? */
582 for (i = 3; i < line->len; i++) {
583 unsigned char c = line->buf[i];
584 if (c == '\n')
585 return 1;
586 if (!isspace(c))
587 break;
588 }
589 return 0;
590 }
591 return 0;
592}
593
594static int is_scissors_line(const struct strbuf *line)
595{
596 size_t i, len = line->len;
597 int scissors = 0, gap = 0;
598 int first_nonblank = -1;
599 int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
600 const char *buf = line->buf;
601
602 for (i = 0; i < len; i++) {
603 if (isspace(buf[i])) {
604 if (in_perforation) {
605 perforation++;
606 gap++;
607 }
608 continue;
609 }
610 last_nonblank = i;
611 if (first_nonblank < 0)
612 first_nonblank = i;
613 if (buf[i] == '-') {
614 in_perforation = 1;
615 perforation++;
616 continue;
617 }
618 if (i + 1 < len &&
619 (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
620 !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
621 in_perforation = 1;
622 perforation += 2;
623 scissors += 2;
624 i++;
625 continue;
626 }
627 in_perforation = 0;
628 }
629
630 /*
631 * The mark must be at least 8 bytes long (e.g. "-- >8 --").
632 * Even though there can be arbitrary cruft on the same line
633 * (e.g. "cut here"), in order to avoid misidentification, the
634 * perforation must occupy more than a third of the visible
635 * width of the line, and dashes and scissors must occupy more
636 * than half of the perforation.
637 */
638
639 visible = last_nonblank - first_nonblank + 1;
640 return (scissors && 8 <= visible &&
641 visible < perforation * 3 &&
642 gap * 2 < perforation);
643}
644
645static int handle_commit_msg(struct strbuf *line, int *still_looking)
646{
647 if (!cmitmsg)
648 return 0;
649
650 if (*still_looking) {
651 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
652 return 0;
653 }
654
655 if (use_inbody_headers && *still_looking) {
656 *still_looking = check_header(line, s_hdr_data, 0);
657 if (*still_looking)
658 return 0;
659 } else
660 /* Only trim the first (blank) line of the commit message
661 * when ignoring in-body headers.
662 */
663 *still_looking = 0;
664
665 /* normalize the log message to UTF-8. */
666 if (metainfo_charset)
667 convert_to_utf8(line, charset.buf);
668
669 if (use_scissors && is_scissors_line(line)) {
670 int i;
671 if (fseek(cmitmsg, 0L, SEEK_SET))
672 die_errno("Could not rewind output message file");
673 if (ftruncate(fileno(cmitmsg), 0))
674 die_errno("Could not truncate output message file at scissors");
675 *still_looking = 1;
676
677 /*
678 * We may have already read "secondary headers"; purge
679 * them to give ourselves a clean restart.
680 */
681 for (i = 0; header[i]; i++) {
682 if (s_hdr_data[i])
683 strbuf_release(s_hdr_data[i]);
684 s_hdr_data[i] = NULL;
685 }
686 return 0;
687 }
688
689 if (patchbreak(line)) {
690 if (message_id)
691 fprintf(cmitmsg, "Message-Id: %s\n", message_id);
692 fclose(cmitmsg);
693 cmitmsg = NULL;
694 return 1;
695 }
696
697 fputs(line->buf, cmitmsg);
698 return 0;
699}
700
701static void handle_patch(const struct strbuf *line)
702{
703 fwrite(line->buf, 1, line->len, patchfile);
704 patch_lines++;
705}
706
707static void handle_filter(struct strbuf *line, int *filter_stage, int *header_stage)
708{
709 switch (*filter_stage) {
710 case 0:
711 if (!handle_commit_msg(line, header_stage))
712 break;
713 (*filter_stage)++;
714 case 1:
715 handle_patch(line);
716 break;
717 }
718}
719
720static int is_rfc2822_header(const struct strbuf *line)
721{
722 /*
723 * The section that defines the loosest possible
724 * field name is "3.6.8 Optional fields".
725 *
726 * optional-field = field-name ":" unstructured CRLF
727 * field-name = 1*ftext
728 * ftext = %d33-57 / %59-126
729 */
730 int ch;
731 char *cp = line->buf;
732
733 /* Count mbox From headers as headers */
734 if (starts_with(cp, "From ") || starts_with(cp, ">From "))
735 return 1;
736
737 while ((ch = *cp++)) {
738 if (ch == ':')
739 return 1;
740 if ((33 <= ch && ch <= 57) ||
741 (59 <= ch && ch <= 126))
742 continue;
743 break;
744 }
745 return 0;
746}
747
748static int read_one_header_line(struct strbuf *line, FILE *in)
749{
750 struct strbuf continuation = STRBUF_INIT;
751
752 /* Get the first part of the line. */
753 if (strbuf_getline(line, in, '\n'))
754 return 0;
755
756 /*
757 * Is it an empty line or not a valid rfc2822 header?
758 * If so, stop here, and return false ("not a header")
759 */
760 strbuf_rtrim(line);
761 if (!line->len || !is_rfc2822_header(line)) {
762 /* Re-add the newline */
763 strbuf_addch(line, '\n');
764 return 0;
765 }
766
767 /*
768 * Now we need to eat all the continuation lines..
769 * Yuck, 2822 header "folding"
770 */
771 for (;;) {
772 int peek;
773
774 peek = fgetc(in); ungetc(peek, in);
775 if (peek != ' ' && peek != '\t')
776 break;
777 if (strbuf_getline(&continuation, in, '\n'))
778 break;
779 continuation.buf[0] = ' ';
780 strbuf_rtrim(&continuation);
781 strbuf_addbuf(line, &continuation);
782 }
783 strbuf_release(&continuation);
784
785 return 1;
786}
787
788static int find_boundary(struct strbuf *line)
789{
790 while (!strbuf_getline(line, fin, '\n')) {
791 if (*content_top && is_multipart_boundary(line))
792 return 1;
793 }
794 return 0;
795}
796
797static int handle_boundary(struct strbuf *line, int *filter_stage, int *header_stage)
798{
799 struct strbuf newline = STRBUF_INIT;
800
801 strbuf_addch(&newline, '\n');
802again:
803 if (line->len >= (*content_top)->len + 2 &&
804 !memcmp(line->buf + (*content_top)->len, "--", 2)) {
805 /* we hit an end boundary */
806 /* pop the current boundary off the stack */
807 strbuf_release(*content_top);
808 free(*content_top);
809 *content_top = NULL;
810
811 /* technically won't happen as is_multipart_boundary()
812 will fail first. But just in case..
813 */
814 if (--content_top < content) {
815 fprintf(stderr, "Detected mismatched boundaries, "
816 "can't recover\n");
817 exit(1);
818 }
819 handle_filter(&newline, filter_stage, header_stage);
820 strbuf_release(&newline);
821
822 /* skip to the next boundary */
823 if (!find_boundary(line))
824 return 0;
825 goto again;
826 }
827
828 /* set some defaults */
829 transfer_encoding = TE_DONTCARE;
830 strbuf_reset(&charset);
831
832 /* slurp in this section's info */
833 while (read_one_header_line(line, fin))
834 check_header(line, p_hdr_data, 0);
835
836 strbuf_release(&newline);
837 /* replenish line */
838 if (strbuf_getline(line, fin, '\n'))
839 return 0;
840 strbuf_addch(line, '\n');
841 return 1;
842}
843
844static void handle_body(struct strbuf *line)
845{
846 struct strbuf prev = STRBUF_INIT;
847 int filter_stage = 0;
848 int header_stage = 1;
849
850 /* Skip up to the first boundary */
851 if (*content_top) {
852 if (!find_boundary(line))
853 goto handle_body_out;
854 }
855
856 do {
857 /* process any boundary lines */
858 if (*content_top && is_multipart_boundary(line)) {
859 /* flush any leftover */
860 if (prev.len) {
861 handle_filter(&prev, &filter_stage, &header_stage);
862 strbuf_reset(&prev);
863 }
864 if (!handle_boundary(line, &filter_stage, &header_stage))
865 goto handle_body_out;
866 }
867
868 /* Unwrap transfer encoding */
869 decode_transfer_encoding(line);
870
871 switch (transfer_encoding) {
872 case TE_BASE64:
873 case TE_QP:
874 {
875 struct strbuf **lines, **it, *sb;
876
877 /* Prepend any previous partial lines */
878 strbuf_insert(line, 0, prev.buf, prev.len);
879 strbuf_reset(&prev);
880
881 /*
882 * This is a decoded line that may contain
883 * multiple new lines. Pass only one chunk
884 * at a time to handle_filter()
885 */
886 lines = strbuf_split(line, '\n');
887 for (it = lines; (sb = *it); it++) {
888 if (*(it + 1) == NULL) /* The last line */
889 if (sb->buf[sb->len - 1] != '\n') {
890 /* Partial line, save it for later. */
891 strbuf_addbuf(&prev, sb);
892 break;
893 }
894 handle_filter(sb, &filter_stage, &header_stage);
895 }
896 /*
897 * The partial chunk is saved in "prev" and will be
898 * appended by the next iteration of read_line_with_nul().
899 */
900 strbuf_list_free(lines);
901 break;
902 }
903 default:
904 handle_filter(line, &filter_stage, &header_stage);
905 }
906
907 } while (!strbuf_getwholeline(line, fin, '\n'));
908
909handle_body_out:
910 strbuf_release(&prev);
911}
912
913static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
914{
915 const char *sp = data->buf;
916 while (1) {
917 char *ep = strchr(sp, '\n');
918 int len;
919 if (!ep)
920 len = strlen(sp);
921 else
922 len = ep - sp;
923 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
924 if (!ep)
925 break;
926 sp = ep + 1;
927 }
928}
929
930static void handle_info(void)
931{
932 struct strbuf *hdr;
933 int i;
934
935 for (i = 0; header[i]; i++) {
936 /* only print inbody headers if we output a patch file */
937 if (patch_lines && s_hdr_data[i])
938 hdr = s_hdr_data[i];
939 else if (p_hdr_data[i])
940 hdr = p_hdr_data[i];
941 else
942 continue;
943
944 if (!strcmp(header[i], "Subject")) {
945 if (!keep_subject) {
946 cleanup_subject(hdr);
947 cleanup_space(hdr);
948 }
949 output_header_lines(fout, "Subject", hdr);
950 } else if (!strcmp(header[i], "From")) {
951 cleanup_space(hdr);
952 handle_from(hdr);
953 fprintf(fout, "Author: %s\n", name.buf);
954 fprintf(fout, "Email: %s\n", email.buf);
955 } else {
956 cleanup_space(hdr);
957 fprintf(fout, "%s: %s\n", header[i], hdr->buf);
958 }
959 }
960 fprintf(fout, "\n");
961}
962
963static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
964{
965 int peek;
966 struct strbuf line = STRBUF_INIT;
967
968 fin = in;
969 fout = out;
970
971 cmitmsg = fopen(msg, "w");
972 if (!cmitmsg) {
973 perror(msg);
974 return -1;
975 }
976 patchfile = fopen(patch, "w");
977 if (!patchfile) {
978 perror(patch);
979 fclose(cmitmsg);
980 return -1;
981 }
982
983 p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
984 s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
985
986 do {
987 peek = fgetc(in);
988 } while (isspace(peek));
989 ungetc(peek, in);
990
991 /* process the email header */
992 while (read_one_header_line(&line, fin))
993 check_header(&line, p_hdr_data, 1);
994
995 handle_body(&line);
996 fclose(patchfile);
997
998 handle_info();
999 strbuf_release(&line);
1000 return 0;
1001}
1002
1003static int git_mailinfo_config(const char *var, const char *value, void *unused)
1004{
1005 if (!starts_with(var, "mailinfo."))
1006 return git_default_config(var, value, unused);
1007 if (!strcmp(var, "mailinfo.scissors")) {
1008 use_scissors = git_config_bool(var, value);
1009 return 0;
1010 }
1011 /* perhaps others here */
1012 return 0;
1013}
1014
1015static const char mailinfo_usage[] =
1016 "git mailinfo [-k | -b] [-m | --message-id] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] <msg> <patch> < mail >info";
1017
1018int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1019{
1020 const char *def_charset;
1021
1022 /* NEEDSWORK: might want to do the optional .git/ directory
1023 * discovery
1024 */
1025 git_config(git_mailinfo_config, NULL);
1026
1027 def_charset = get_commit_output_encoding();
1028 metainfo_charset = def_charset;
1029
1030 while (1 < argc && argv[1][0] == '-') {
1031 if (!strcmp(argv[1], "-k"))
1032 keep_subject = 1;
1033 else if (!strcmp(argv[1], "-b"))
1034 keep_non_patch_brackets_in_subject = 1;
1035 else if (!strcmp(argv[1], "-m") || !strcmp(argv[1], "--message-id"))
1036 add_message_id = 1;
1037 else if (!strcmp(argv[1], "-u"))
1038 metainfo_charset = def_charset;
1039 else if (!strcmp(argv[1], "-n"))
1040 metainfo_charset = NULL;
1041 else if (starts_with(argv[1], "--encoding="))
1042 metainfo_charset = argv[1] + 11;
1043 else if (!strcmp(argv[1], "--scissors"))
1044 use_scissors = 1;
1045 else if (!strcmp(argv[1], "--no-scissors"))
1046 use_scissors = 0;
1047 else if (!strcmp(argv[1], "--no-inbody-headers"))
1048 use_inbody_headers = 0;
1049 else
1050 usage(mailinfo_usage);
1051 argc--; argv++;
1052 }
1053
1054 if (argc != 3)
1055 usage(mailinfo_usage);
1056
1057 return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1058}