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