Merge branch 'pw/maint-p4-rcs-expansion-newline' into maint
[git] / builtin / mailinfo.c
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
10 static FILE *cmitmsg, *patchfile, *fin, *fout;
11
12 static int keep_subject;
13 static int keep_non_patch_brackets_in_subject;
14 static const char *metainfo_charset;
15 static struct strbuf line = STRBUF_INIT;
16 static struct strbuf name = STRBUF_INIT;
17 static struct strbuf email = STRBUF_INIT;
18
19 static enum  {
20         TE_DONTCARE, TE_QP, TE_BASE64
21 } transfer_encoding;
22
23 static struct strbuf charset = STRBUF_INIT;
24 static int patch_lines;
25 static struct strbuf **p_hdr_data, **s_hdr_data;
26 static int use_scissors;
27 static int use_inbody_headers = 1;
28
29 #define MAX_HDR_PARSED 10
30 #define MAX_BOUNDARIES 5
31
32 static void cleanup_space(struct strbuf *sb);
33
34
35 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
36 {
37         struct strbuf *src = name;
38         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
39                 strchr(name->buf, '<') || strchr(name->buf, '>'))
40                 src = email;
41         else if (name == out)
42                 return;
43         strbuf_reset(out);
44         strbuf_addbuf(out, src);
45 }
46
47 static void parse_bogus_from(const struct strbuf *line)
48 {
49         /* John Doe <johndoe> */
50
51         char *bra, *ket;
52         /* This is fallback, so do not bother if we already have an
53          * e-mail address.
54          */
55         if (email.len)
56                 return;
57
58         bra = strchr(line->buf, '<');
59         if (!bra)
60                 return;
61         ket = strchr(bra, '>');
62         if (!ket)
63                 return;
64
65         strbuf_reset(&email);
66         strbuf_add(&email, bra + 1, ket - bra - 1);
67
68         strbuf_reset(&name);
69         strbuf_add(&name, line->buf, bra - line->buf);
70         strbuf_trim(&name);
71         get_sane_name(&name, &name, &email);
72 }
73
74 static void handle_from(const struct strbuf *from)
75 {
76         char *at;
77         size_t el;
78         struct strbuf f;
79
80         strbuf_init(&f, from->len);
81         strbuf_addbuf(&f, from);
82
83         at = strchr(f.buf, '@');
84         if (!at) {
85                 parse_bogus_from(from);
86                 return;
87         }
88
89         /*
90          * If we already have one email, don't take any confusing lines
91          */
92         if (email.len && strchr(at + 1, '@')) {
93                 strbuf_release(&f);
94                 return;
95         }
96
97         /* Pick up the string around '@', possibly delimited with <>
98          * pair; that is the email part.
99          */
100         while (at > f.buf) {
101                 char c = at[-1];
102                 if (isspace(c))
103                         break;
104                 if (c == '<') {
105                         at[-1] = ' ';
106                         break;
107                 }
108                 at--;
109         }
110         el = strcspn(at, " \n\t\r\v\f>");
111         strbuf_reset(&email);
112         strbuf_add(&email, at, el);
113         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
114
115         /* The remainder is name.  It could be
116          *
117          * - "John Doe <john.doe@xz>"                   (a), or
118          * - "john.doe@xz (John Doe)"                   (b), or
119          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
120          *
121          * but we have removed the email part, so
122          *
123          * - remove extra spaces which could stay after email (case 'c'), and
124          * - trim from both ends, possibly removing the () pair at the end
125          *   (cases 'a' and 'b').
126          */
127         cleanup_space(&f);
128         strbuf_trim(&f);
129         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
130                 strbuf_remove(&f, 0, 1);
131                 strbuf_setlen(&f, f.len - 1);
132         }
133
134         get_sane_name(&name, &f, &email);
135         strbuf_release(&f);
136 }
137
138 static void handle_header(struct strbuf **out, const struct strbuf *line)
139 {
140         if (!*out) {
141                 *out = xmalloc(sizeof(struct strbuf));
142                 strbuf_init(*out, line->len);
143         } else
144                 strbuf_reset(*out);
145
146         strbuf_addbuf(*out, line);
147 }
148
149 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
150  * to have enough heuristics to grok MIME encoded patches often found
151  * on our mailing lists.  For example, we do not even treat header lines
152  * case insensitively.
153  */
154
155 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
156 {
157         const char *ends, *ap = strcasestr(line, name);
158         size_t sz;
159
160         strbuf_setlen(attr, 0);
161         if (!ap)
162                 return 0;
163         ap += strlen(name);
164         if (*ap == '"') {
165                 ap++;
166                 ends = "\"";
167         }
168         else
169                 ends = "; \t";
170         sz = strcspn(ap, ends);
171         strbuf_add(attr, ap, sz);
172         return 1;
173 }
174
175 static struct strbuf *content[MAX_BOUNDARIES];
176
177 static struct strbuf **content_top = content;
178
179 static void handle_content_type(struct strbuf *line)
180 {
181         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
182         strbuf_init(boundary, line->len);
183
184         if (slurp_attr(line->buf, "boundary=", boundary)) {
185                 strbuf_insert(boundary, 0, "--", 2);
186                 if (++content_top > &content[MAX_BOUNDARIES]) {
187                         fprintf(stderr, "Too many boundaries to handle\n");
188                         exit(1);
189                 }
190                 *content_top = boundary;
191                 boundary = NULL;
192         }
193         slurp_attr(line->buf, "charset=", &charset);
194
195         if (boundary) {
196                 strbuf_release(boundary);
197                 free(boundary);
198         }
199 }
200
201 static void handle_content_transfer_encoding(const struct strbuf *line)
202 {
203         if (strcasestr(line->buf, "base64"))
204                 transfer_encoding = TE_BASE64;
205         else if (strcasestr(line->buf, "quoted-printable"))
206                 transfer_encoding = TE_QP;
207         else
208                 transfer_encoding = TE_DONTCARE;
209 }
210
211 static int is_multipart_boundary(const struct strbuf *line)
212 {
213         return (((*content_top)->len <= line->len) &&
214                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
215 }
216
217 static void cleanup_subject(struct strbuf *subject)
218 {
219         size_t at = 0;
220
221         while (at < subject->len) {
222                 char *pos;
223                 size_t remove;
224
225                 switch (subject->buf[at]) {
226                 case 'r': case 'R':
227                         if (subject->len <= at + 3)
228                                 break;
229                         if ((subject->buf[at + 1] == 'e' ||
230                              subject->buf[at + 1] == 'E') &&
231                             subject->buf[at + 2] == ':') {
232                                 strbuf_remove(subject, at, 3);
233                                 continue;
234                         }
235                         at++;
236                         break;
237                 case ' ': case '\t': case ':':
238                         strbuf_remove(subject, at, 1);
239                         continue;
240                 case '[':
241                         pos = strchr(subject->buf + at, ']');
242                         if (!pos)
243                                 break;
244                         remove = pos - subject->buf + at + 1;
245                         if (!keep_non_patch_brackets_in_subject ||
246                             (7 <= remove &&
247                              memmem(subject->buf + at, remove, "PATCH", 5)))
248                                 strbuf_remove(subject, at, remove);
249                         else {
250                                 at += remove;
251                                 /*
252                                  * If the input had a space after the ], keep
253                                  * it.  We don't bother with finding the end of
254                                  * the space, since we later normalize it
255                                  * anyway.
256                                  */
257                                 if (isspace(subject->buf[at]))
258                                         at += 1;
259                         }
260                         continue;
261                 }
262                 break;
263         }
264         strbuf_trim(subject);
265 }
266
267 static void cleanup_space(struct strbuf *sb)
268 {
269         size_t pos, cnt;
270         for (pos = 0; pos < sb->len; pos++) {
271                 if (isspace(sb->buf[pos])) {
272                         sb->buf[pos] = ' ';
273                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
274                         strbuf_remove(sb, pos + 1, cnt);
275                 }
276         }
277 }
278
279 static void decode_header(struct strbuf *line);
280 static const char *header[MAX_HDR_PARSED] = {
281         "From","Subject","Date",
282 };
283
284 static inline int cmp_header(const struct strbuf *line, const char *hdr)
285 {
286         int len = strlen(hdr);
287         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
288                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
289 }
290
291 static int check_header(const struct strbuf *line,
292                                 struct strbuf *hdr_data[], int overwrite)
293 {
294         int i, ret = 0, len;
295         struct strbuf sb = STRBUF_INIT;
296         /* search for the interesting parts */
297         for (i = 0; header[i]; i++) {
298                 int len = strlen(header[i]);
299                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
300                         /* Unwrap inline B and Q encoding, and optionally
301                          * normalize the meta information to utf8.
302                          */
303                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
304                         decode_header(&sb);
305                         handle_header(&hdr_data[i], &sb);
306                         ret = 1;
307                         goto check_header_out;
308                 }
309         }
310
311         /* Content stuff */
312         if (cmp_header(line, "Content-Type")) {
313                 len = strlen("Content-Type: ");
314                 strbuf_add(&sb, line->buf + len, line->len - len);
315                 decode_header(&sb);
316                 strbuf_insert(&sb, 0, "Content-Type: ", len);
317                 handle_content_type(&sb);
318                 ret = 1;
319                 goto check_header_out;
320         }
321         if (cmp_header(line, "Content-Transfer-Encoding")) {
322                 len = strlen("Content-Transfer-Encoding: ");
323                 strbuf_add(&sb, line->buf + len, line->len - len);
324                 decode_header(&sb);
325                 handle_content_transfer_encoding(&sb);
326                 ret = 1;
327                 goto check_header_out;
328         }
329
330         /* for inbody stuff */
331         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
332                 ret = 1; /* Should this return 0? */
333                 goto check_header_out;
334         }
335         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
336                 for (i = 0; header[i]; i++) {
337                         if (!memcmp("Subject", header[i], 7)) {
338                                 handle_header(&hdr_data[i], line);
339                                 ret = 1;
340                                 goto check_header_out;
341                         }
342                 }
343         }
344
345 check_header_out:
346         strbuf_release(&sb);
347         return ret;
348 }
349
350 static int is_rfc2822_header(const struct strbuf *line)
351 {
352         /*
353          * The section that defines the loosest possible
354          * field name is "3.6.8 Optional fields".
355          *
356          * optional-field = field-name ":" unstructured CRLF
357          * field-name = 1*ftext
358          * ftext = %d33-57 / %59-126
359          */
360         int ch;
361         char *cp = line->buf;
362
363         /* Count mbox From headers as headers */
364         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
365                 return 1;
366
367         while ((ch = *cp++)) {
368                 if (ch == ':')
369                         return 1;
370                 if ((33 <= ch && ch <= 57) ||
371                     (59 <= ch && ch <= 126))
372                         continue;
373                 break;
374         }
375         return 0;
376 }
377
378 static int read_one_header_line(struct strbuf *line, FILE *in)
379 {
380         /* Get the first part of the line. */
381         if (strbuf_getline(line, in, '\n'))
382                 return 0;
383
384         /*
385          * Is it an empty line or not a valid rfc2822 header?
386          * If so, stop here, and return false ("not a header")
387          */
388         strbuf_rtrim(line);
389         if (!line->len || !is_rfc2822_header(line)) {
390                 /* Re-add the newline */
391                 strbuf_addch(line, '\n');
392                 return 0;
393         }
394
395         /*
396          * Now we need to eat all the continuation lines..
397          * Yuck, 2822 header "folding"
398          */
399         for (;;) {
400                 int peek;
401                 struct strbuf continuation = STRBUF_INIT;
402
403                 peek = fgetc(in); ungetc(peek, in);
404                 if (peek != ' ' && peek != '\t')
405                         break;
406                 if (strbuf_getline(&continuation, in, '\n'))
407                         break;
408                 continuation.buf[0] = ' ';
409                 strbuf_rtrim(&continuation);
410                 strbuf_addbuf(line, &continuation);
411         }
412
413         return 1;
414 }
415
416 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
417 {
418         const char *in = q_seg->buf;
419         int c;
420         struct strbuf *out = xmalloc(sizeof(struct strbuf));
421         strbuf_init(out, q_seg->len);
422
423         while ((c = *in++) != 0) {
424                 if (c == '=') {
425                         int d = *in++;
426                         if (d == '\n' || !d)
427                                 break; /* drop trailing newline */
428                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
429                         continue;
430                 }
431                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
432                         c = 0x20;
433                 strbuf_addch(out, c);
434         }
435         return out;
436 }
437
438 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
439 {
440         /* Decode in..ep, possibly in-place to ot */
441         int c, pos = 0, acc = 0;
442         const char *in = b_seg->buf;
443         struct strbuf *out = xmalloc(sizeof(struct strbuf));
444         strbuf_init(out, b_seg->len);
445
446         while ((c = *in++) != 0) {
447                 if (c == '+')
448                         c = 62;
449                 else if (c == '/')
450                         c = 63;
451                 else if ('A' <= c && c <= 'Z')
452                         c -= 'A';
453                 else if ('a' <= c && c <= 'z')
454                         c -= 'a' - 26;
455                 else if ('0' <= c && c <= '9')
456                         c -= '0' - 52;
457                 else
458                         continue; /* garbage */
459                 switch (pos++) {
460                 case 0:
461                         acc = (c << 2);
462                         break;
463                 case 1:
464                         strbuf_addch(out, (acc | (c >> 4)));
465                         acc = (c & 15) << 4;
466                         break;
467                 case 2:
468                         strbuf_addch(out, (acc | (c >> 2)));
469                         acc = (c & 3) << 6;
470                         break;
471                 case 3:
472                         strbuf_addch(out, (acc | c));
473                         acc = pos = 0;
474                         break;
475                 }
476         }
477         return out;
478 }
479
480 static void convert_to_utf8(struct strbuf *line, const char *charset)
481 {
482         char *out;
483
484         if (!charset || !*charset)
485                 return;
486         if (!strcasecmp(metainfo_charset, charset))
487                 return;
488         out = reencode_string(line->buf, metainfo_charset, charset);
489         if (!out)
490                 die("cannot convert from %s to %s",
491                     charset, metainfo_charset);
492         strbuf_attach(line, out, strlen(out), strlen(out));
493 }
494
495 static int decode_header_bq(struct strbuf *it)
496 {
497         char *in, *ep, *cp;
498         struct strbuf outbuf = STRBUF_INIT, *dec;
499         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
500         int rfc2047 = 0;
501
502         in = it->buf;
503         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
504                 int encoding;
505                 strbuf_reset(&charset_q);
506                 strbuf_reset(&piecebuf);
507                 rfc2047 = 1;
508
509                 if (in != ep) {
510                         /*
511                          * We are about to process an encoded-word
512                          * that begins at ep, but there is something
513                          * before the encoded word.
514                          */
515                         char *scan;
516                         for (scan = in; scan < ep; scan++)
517                                 if (!isspace(*scan))
518                                         break;
519
520                         if (scan != ep || in == it->buf) {
521                                 /*
522                                  * We should not lose that "something",
523                                  * unless we have just processed an
524                                  * encoded-word, and there is only LWS
525                                  * before the one we are about to process.
526                                  */
527                                 strbuf_add(&outbuf, in, ep - in);
528                         }
529                 }
530                 /* E.g.
531                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
532                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
533                  */
534                 ep += 2;
535
536                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
537                         goto decode_header_bq_out;
538
539                 if (cp + 3 - it->buf > it->len)
540                         goto decode_header_bq_out;
541                 strbuf_add(&charset_q, ep, cp - ep);
542
543                 encoding = cp[1];
544                 if (!encoding || cp[2] != '?')
545                         goto decode_header_bq_out;
546                 ep = strstr(cp + 3, "?=");
547                 if (!ep)
548                         goto decode_header_bq_out;
549                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
550                 switch (tolower(encoding)) {
551                 default:
552                         goto decode_header_bq_out;
553                 case 'b':
554                         dec = decode_b_segment(&piecebuf);
555                         break;
556                 case 'q':
557                         dec = decode_q_segment(&piecebuf, 1);
558                         break;
559                 }
560                 if (metainfo_charset)
561                         convert_to_utf8(dec, charset_q.buf);
562
563                 strbuf_addbuf(&outbuf, dec);
564                 strbuf_release(dec);
565                 free(dec);
566                 in = ep + 2;
567         }
568         strbuf_addstr(&outbuf, in);
569         strbuf_reset(it);
570         strbuf_addbuf(it, &outbuf);
571 decode_header_bq_out:
572         strbuf_release(&outbuf);
573         strbuf_release(&charset_q);
574         strbuf_release(&piecebuf);
575         return rfc2047;
576 }
577
578 static void decode_header(struct strbuf *it)
579 {
580         if (decode_header_bq(it))
581                 return;
582         /* otherwise "it" is a straight copy of the input.
583          * This can be binary guck but there is no charset specified.
584          */
585         if (metainfo_charset)
586                 convert_to_utf8(it, "");
587 }
588
589 static void decode_transfer_encoding(struct strbuf *line)
590 {
591         struct strbuf *ret;
592
593         switch (transfer_encoding) {
594         case TE_QP:
595                 ret = decode_q_segment(line, 0);
596                 break;
597         case TE_BASE64:
598                 ret = decode_b_segment(line);
599                 break;
600         case TE_DONTCARE:
601         default:
602                 return;
603         }
604         strbuf_reset(line);
605         strbuf_addbuf(line, ret);
606         strbuf_release(ret);
607         free(ret);
608 }
609
610 static void handle_filter(struct strbuf *line);
611
612 static int find_boundary(void)
613 {
614         while (!strbuf_getline(&line, fin, '\n')) {
615                 if (*content_top && is_multipart_boundary(&line))
616                         return 1;
617         }
618         return 0;
619 }
620
621 static int handle_boundary(void)
622 {
623         struct strbuf newline = STRBUF_INIT;
624
625         strbuf_addch(&newline, '\n');
626 again:
627         if (line.len >= (*content_top)->len + 2 &&
628             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
629                 /* we hit an end boundary */
630                 /* pop the current boundary off the stack */
631                 strbuf_release(*content_top);
632                 free(*content_top);
633                 *content_top = NULL;
634
635                 /* technically won't happen as is_multipart_boundary()
636                    will fail first.  But just in case..
637                  */
638                 if (--content_top < content) {
639                         fprintf(stderr, "Detected mismatched boundaries, "
640                                         "can't recover\n");
641                         exit(1);
642                 }
643                 handle_filter(&newline);
644                 strbuf_release(&newline);
645
646                 /* skip to the next boundary */
647                 if (!find_boundary())
648                         return 0;
649                 goto again;
650         }
651
652         /* set some defaults */
653         transfer_encoding = TE_DONTCARE;
654         strbuf_reset(&charset);
655
656         /* slurp in this section's info */
657         while (read_one_header_line(&line, fin))
658                 check_header(&line, p_hdr_data, 0);
659
660         strbuf_release(&newline);
661         /* replenish line */
662         if (strbuf_getline(&line, fin, '\n'))
663                 return 0;
664         strbuf_addch(&line, '\n');
665         return 1;
666 }
667
668 static inline int patchbreak(const struct strbuf *line)
669 {
670         size_t i;
671
672         /* Beginning of a "diff -" header? */
673         if (!prefixcmp(line->buf, "diff -"))
674                 return 1;
675
676         /* CVS "Index: " line? */
677         if (!prefixcmp(line->buf, "Index: "))
678                 return 1;
679
680         /*
681          * "--- <filename>" starts patches without headers
682          * "---<sp>*" is a manual separator
683          */
684         if (line->len < 4)
685                 return 0;
686
687         if (!prefixcmp(line->buf, "---")) {
688                 /* space followed by a filename? */
689                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
690                         return 1;
691                 /* Just whitespace? */
692                 for (i = 3; i < line->len; i++) {
693                         unsigned char c = line->buf[i];
694                         if (c == '\n')
695                                 return 1;
696                         if (!isspace(c))
697                                 break;
698                 }
699                 return 0;
700         }
701         return 0;
702 }
703
704 static int is_scissors_line(const struct strbuf *line)
705 {
706         size_t i, len = line->len;
707         int scissors = 0, gap = 0;
708         int first_nonblank = -1;
709         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
710         const char *buf = line->buf;
711
712         for (i = 0; i < len; i++) {
713                 if (isspace(buf[i])) {
714                         if (in_perforation) {
715                                 perforation++;
716                                 gap++;
717                         }
718                         continue;
719                 }
720                 last_nonblank = i;
721                 if (first_nonblank < 0)
722                         first_nonblank = i;
723                 if (buf[i] == '-') {
724                         in_perforation = 1;
725                         perforation++;
726                         continue;
727                 }
728                 if (i + 1 < len &&
729                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
730                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
731                         in_perforation = 1;
732                         perforation += 2;
733                         scissors += 2;
734                         i++;
735                         continue;
736                 }
737                 in_perforation = 0;
738         }
739
740         /*
741          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
742          * Even though there can be arbitrary cruft on the same line
743          * (e.g. "cut here"), in order to avoid misidentification, the
744          * perforation must occupy more than a third of the visible
745          * width of the line, and dashes and scissors must occupy more
746          * than half of the perforation.
747          */
748
749         visible = last_nonblank - first_nonblank + 1;
750         return (scissors && 8 <= visible &&
751                 visible < perforation * 3 &&
752                 gap * 2 < perforation);
753 }
754
755 static int handle_commit_msg(struct strbuf *line)
756 {
757         static int still_looking = 1;
758
759         if (!cmitmsg)
760                 return 0;
761
762         if (still_looking) {
763                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
764                         return 0;
765         }
766
767         if (use_inbody_headers && still_looking) {
768                 still_looking = check_header(line, s_hdr_data, 0);
769                 if (still_looking)
770                         return 0;
771         } else
772                 /* Only trim the first (blank) line of the commit message
773                  * when ignoring in-body headers.
774                  */
775                 still_looking = 0;
776
777         /* normalize the log message to UTF-8. */
778         if (metainfo_charset)
779                 convert_to_utf8(line, charset.buf);
780
781         if (use_scissors && is_scissors_line(line)) {
782                 int i;
783                 if (fseek(cmitmsg, 0L, SEEK_SET))
784                         die_errno("Could not rewind output message file");
785                 if (ftruncate(fileno(cmitmsg), 0))
786                         die_errno("Could not truncate output message file at scissors");
787                 still_looking = 1;
788
789                 /*
790                  * We may have already read "secondary headers"; purge
791                  * them to give ourselves a clean restart.
792                  */
793                 for (i = 0; header[i]; i++) {
794                         if (s_hdr_data[i])
795                                 strbuf_release(s_hdr_data[i]);
796                         s_hdr_data[i] = NULL;
797                 }
798                 return 0;
799         }
800
801         if (patchbreak(line)) {
802                 fclose(cmitmsg);
803                 cmitmsg = NULL;
804                 return 1;
805         }
806
807         fputs(line->buf, cmitmsg);
808         return 0;
809 }
810
811 static void handle_patch(const struct strbuf *line)
812 {
813         fwrite(line->buf, 1, line->len, patchfile);
814         patch_lines++;
815 }
816
817 static void handle_filter(struct strbuf *line)
818 {
819         static int filter = 0;
820
821         /* filter tells us which part we left off on */
822         switch (filter) {
823         case 0:
824                 if (!handle_commit_msg(line))
825                         break;
826                 filter++;
827         case 1:
828                 handle_patch(line);
829                 break;
830         }
831 }
832
833 static void handle_body(void)
834 {
835         struct strbuf prev = STRBUF_INIT;
836
837         /* Skip up to the first boundary */
838         if (*content_top) {
839                 if (!find_boundary())
840                         goto handle_body_out;
841         }
842
843         do {
844                 /* process any boundary lines */
845                 if (*content_top && is_multipart_boundary(&line)) {
846                         /* flush any leftover */
847                         if (prev.len) {
848                                 handle_filter(&prev);
849                                 strbuf_reset(&prev);
850                         }
851                         if (!handle_boundary())
852                                 goto handle_body_out;
853                 }
854
855                 /* Unwrap transfer encoding */
856                 decode_transfer_encoding(&line);
857
858                 switch (transfer_encoding) {
859                 case TE_BASE64:
860                 case TE_QP:
861                 {
862                         struct strbuf **lines, **it, *sb;
863
864                         /* Prepend any previous partial lines */
865                         strbuf_insert(&line, 0, prev.buf, prev.len);
866                         strbuf_reset(&prev);
867
868                         /*
869                          * This is a decoded line that may contain
870                          * multiple new lines.  Pass only one chunk
871                          * at a time to handle_filter()
872                          */
873                         lines = strbuf_split(&line, '\n');
874                         for (it = lines; (sb = *it); it++) {
875                                 if (*(it + 1) == NULL) /* The last line */
876                                         if (sb->buf[sb->len - 1] != '\n') {
877                                                 /* Partial line, save it for later. */
878                                                 strbuf_addbuf(&prev, sb);
879                                                 break;
880                                         }
881                                 handle_filter(sb);
882                         }
883                         /*
884                          * The partial chunk is saved in "prev" and will be
885                          * appended by the next iteration of read_line_with_nul().
886                          */
887                         strbuf_list_free(lines);
888                         break;
889                 }
890                 default:
891                         handle_filter(&line);
892                 }
893
894         } while (!strbuf_getwholeline(&line, fin, '\n'));
895
896 handle_body_out:
897         strbuf_release(&prev);
898 }
899
900 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
901 {
902         const char *sp = data->buf;
903         while (1) {
904                 char *ep = strchr(sp, '\n');
905                 int len;
906                 if (!ep)
907                         len = strlen(sp);
908                 else
909                         len = ep - sp;
910                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
911                 if (!ep)
912                         break;
913                 sp = ep + 1;
914         }
915 }
916
917 static void handle_info(void)
918 {
919         struct strbuf *hdr;
920         int i;
921
922         for (i = 0; header[i]; i++) {
923                 /* only print inbody headers if we output a patch file */
924                 if (patch_lines && s_hdr_data[i])
925                         hdr = s_hdr_data[i];
926                 else if (p_hdr_data[i])
927                         hdr = p_hdr_data[i];
928                 else
929                         continue;
930
931                 if (!memcmp(header[i], "Subject", 7)) {
932                         if (!keep_subject) {
933                                 cleanup_subject(hdr);
934                                 cleanup_space(hdr);
935                         }
936                         output_header_lines(fout, "Subject", hdr);
937                 } else if (!memcmp(header[i], "From", 4)) {
938                         cleanup_space(hdr);
939                         handle_from(hdr);
940                         fprintf(fout, "Author: %s\n", name.buf);
941                         fprintf(fout, "Email: %s\n", email.buf);
942                 } else {
943                         cleanup_space(hdr);
944                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
945                 }
946         }
947         fprintf(fout, "\n");
948 }
949
950 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
951 {
952         int peek;
953         fin = in;
954         fout = out;
955
956         cmitmsg = fopen(msg, "w");
957         if (!cmitmsg) {
958                 perror(msg);
959                 return -1;
960         }
961         patchfile = fopen(patch, "w");
962         if (!patchfile) {
963                 perror(patch);
964                 fclose(cmitmsg);
965                 return -1;
966         }
967
968         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
969         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
970
971         do {
972                 peek = fgetc(in);
973         } while (isspace(peek));
974         ungetc(peek, in);
975
976         /* process the email header */
977         while (read_one_header_line(&line, fin))
978                 check_header(&line, p_hdr_data, 1);
979
980         handle_body();
981         handle_info();
982
983         return 0;
984 }
985
986 static int git_mailinfo_config(const char *var, const char *value, void *unused)
987 {
988         if (prefixcmp(var, "mailinfo."))
989                 return git_default_config(var, value, unused);
990         if (!strcmp(var, "mailinfo.scissors")) {
991                 use_scissors = git_config_bool(var, value);
992                 return 0;
993         }
994         /* perhaps others here */
995         return 0;
996 }
997
998 static const char mailinfo_usage[] =
999         "git mailinfo [-k|-b] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1000
1001 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1002 {
1003         const char *def_charset;
1004
1005         /* NEEDSWORK: might want to do the optional .git/ directory
1006          * discovery
1007          */
1008         git_config(git_mailinfo_config, NULL);
1009
1010         def_charset = get_commit_output_encoding();
1011         metainfo_charset = def_charset;
1012
1013         while (1 < argc && argv[1][0] == '-') {
1014                 if (!strcmp(argv[1], "-k"))
1015                         keep_subject = 1;
1016                 else if (!strcmp(argv[1], "-b"))
1017                         keep_non_patch_brackets_in_subject = 1;
1018                 else if (!strcmp(argv[1], "-u"))
1019                         metainfo_charset = def_charset;
1020                 else if (!strcmp(argv[1], "-n"))
1021                         metainfo_charset = NULL;
1022                 else if (!prefixcmp(argv[1], "--encoding="))
1023                         metainfo_charset = argv[1] + 11;
1024                 else if (!strcmp(argv[1], "--scissors"))
1025                         use_scissors = 1;
1026                 else if (!strcmp(argv[1], "--no-scissors"))
1027                         use_scissors = 0;
1028                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1029                         use_inbody_headers = 0;
1030                 else
1031                         usage(mailinfo_usage);
1032                 argc--; argv++;
1033         }
1034
1035         if (argc != 3)
1036                 usage(mailinfo_usage);
1037
1038         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1039 }