2 * Another stupid program, this one parsing the headers of an
3 * email to figure out authorship and subject
10 static FILE *cmitmsg, *patchfile, *fin, *fout;
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;
20 TE_DONTCARE, TE_QP, TE_BASE64
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;
29 #define MAX_HDR_PARSED 10
30 #define MAX_BOUNDARIES 5
32 static void cleanup_space(struct strbuf *sb);
35 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
37 struct strbuf *src = name;
38 if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
39 strchr(name->buf, '<') || strchr(name->buf, '>'))
44 strbuf_addbuf(out, src);
47 static void parse_bogus_from(const struct strbuf *line)
49 /* John Doe <johndoe> */
52 /* This is fallback, so do not bother if we already have an
58 bra = strchr(line->buf, '<');
61 ket = strchr(bra, '>');
66 strbuf_add(&email, bra + 1, ket - bra - 1);
69 strbuf_add(&name, line->buf, bra - line->buf);
71 get_sane_name(&name, &name, &email);
74 static void handle_from(const struct strbuf *from)
80 strbuf_init(&f, from->len);
81 strbuf_addbuf(&f, from);
83 at = strchr(f.buf, '@');
85 parse_bogus_from(from);
90 * If we already have one email, don't take any confusing lines
92 if (email.len && strchr(at + 1, '@')) {
97 /* Pick up the string around '@', possibly delimited with <>
98 * pair; that is the email part.
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));
115 /* The remainder is name. It could be
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)
121 * but we have removed the email part, so
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').
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);
134 get_sane_name(&name, &f, &email);
138 static void handle_header(struct strbuf **out, const struct strbuf *line)
141 *out = xmalloc(sizeof(struct strbuf));
142 strbuf_init(*out, line->len);
146 strbuf_addbuf(*out, line);
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.
155 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
157 const char *ends, *ap = strcasestr(line, name);
160 strbuf_setlen(attr, 0);
170 sz = strcspn(ap, ends);
171 strbuf_add(attr, ap, sz);
175 static struct strbuf *content[MAX_BOUNDARIES];
177 static struct strbuf **content_top = content;
179 static void handle_content_type(struct strbuf *line)
181 struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
182 strbuf_init(boundary, line->len);
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");
190 *content_top = boundary;
193 slurp_attr(line->buf, "charset=", &charset);
196 strbuf_release(boundary);
201 static void handle_content_transfer_encoding(const struct strbuf *line)
203 if (strcasestr(line->buf, "base64"))
204 transfer_encoding = TE_BASE64;
205 else if (strcasestr(line->buf, "quoted-printable"))
206 transfer_encoding = TE_QP;
208 transfer_encoding = TE_DONTCARE;
211 static int is_multipart_boundary(const struct strbuf *line)
213 return (((*content_top)->len <= line->len) &&
214 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
217 static void cleanup_subject(struct strbuf *subject)
221 while (at < subject->len) {
225 switch (subject->buf[at]) {
227 if (subject->len <= at + 3)
229 if ((subject->buf[at + 1] == 'e' ||
230 subject->buf[at + 1] == 'E') &&
231 subject->buf[at + 2] == ':') {
232 strbuf_remove(subject, at, 3);
237 case ' ': case '\t': case ':':
238 strbuf_remove(subject, at, 1);
241 pos = strchr(subject->buf + at, ']');
244 remove = pos - subject->buf + at + 1;
245 if (!keep_non_patch_brackets_in_subject ||
247 memmem(subject->buf + at, remove, "PATCH", 5)))
248 strbuf_remove(subject, at, remove);
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
257 if (isspace(subject->buf[at]))
264 strbuf_trim(subject);
267 static void cleanup_space(struct strbuf *sb)
270 for (pos = 0; pos < sb->len; pos++) {
271 if (isspace(sb->buf[pos])) {
273 for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
274 strbuf_remove(sb, pos + 1, cnt);
279 static void decode_header(struct strbuf *line);
280 static const char *header[MAX_HDR_PARSED] = {
281 "From","Subject","Date",
284 static inline int cmp_header(const struct strbuf *line, const char *hdr)
286 int len = strlen(hdr);
287 return !strncasecmp(line->buf, hdr, len) && line->len > len &&
288 line->buf[len] == ':' && isspace(line->buf[len + 1]);
291 static int is_format_patch_separator(const char *line, int len)
293 static const char SAMPLE[] =
294 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
297 if (len != strlen(SAMPLE))
299 if (!skip_prefix(line, "From ", &cp))
301 if (strspn(cp, "0123456789abcdef") != 40)
304 return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
307 static int check_header(const struct strbuf *line,
308 struct strbuf *hdr_data[], int overwrite)
311 struct strbuf sb = STRBUF_INIT;
312 /* search for the interesting parts */
313 for (i = 0; header[i]; i++) {
314 int len = strlen(header[i]);
315 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
316 /* Unwrap inline B and Q encoding, and optionally
317 * normalize the meta information to utf8.
319 strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
321 handle_header(&hdr_data[i], &sb);
323 goto check_header_out;
328 if (cmp_header(line, "Content-Type")) {
329 len = strlen("Content-Type: ");
330 strbuf_add(&sb, line->buf + len, line->len - len);
332 strbuf_insert(&sb, 0, "Content-Type: ", len);
333 handle_content_type(&sb);
335 goto check_header_out;
337 if (cmp_header(line, "Content-Transfer-Encoding")) {
338 len = strlen("Content-Transfer-Encoding: ");
339 strbuf_add(&sb, line->buf + len, line->len - len);
341 handle_content_transfer_encoding(&sb);
343 goto check_header_out;
346 /* for inbody stuff */
347 if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
348 ret = is_format_patch_separator(line->buf + 1, line->len - 1);
349 goto check_header_out;
351 if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
352 for (i = 0; header[i]; i++) {
353 if (!strcmp("Subject", header[i])) {
354 handle_header(&hdr_data[i], line);
356 goto check_header_out;
366 static int is_rfc2822_header(const struct strbuf *line)
369 * The section that defines the loosest possible
370 * field name is "3.6.8 Optional fields".
372 * optional-field = field-name ":" unstructured CRLF
373 * field-name = 1*ftext
374 * ftext = %d33-57 / %59-126
377 char *cp = line->buf;
379 /* Count mbox From headers as headers */
380 if (starts_with(cp, "From ") || starts_with(cp, ">From "))
383 while ((ch = *cp++)) {
386 if ((33 <= ch && ch <= 57) ||
387 (59 <= ch && ch <= 126))
394 static int read_one_header_line(struct strbuf *line, FILE *in)
396 /* Get the first part of the line. */
397 if (strbuf_getline(line, in, '\n'))
401 * Is it an empty line or not a valid rfc2822 header?
402 * If so, stop here, and return false ("not a header")
405 if (!line->len || !is_rfc2822_header(line)) {
406 /* Re-add the newline */
407 strbuf_addch(line, '\n');
412 * Now we need to eat all the continuation lines..
413 * Yuck, 2822 header "folding"
417 struct strbuf continuation = STRBUF_INIT;
419 peek = fgetc(in); ungetc(peek, in);
420 if (peek != ' ' && peek != '\t')
422 if (strbuf_getline(&continuation, in, '\n'))
424 continuation.buf[0] = ' ';
425 strbuf_rtrim(&continuation);
426 strbuf_addbuf(line, &continuation);
432 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
434 const char *in = q_seg->buf;
436 struct strbuf *out = xmalloc(sizeof(struct strbuf));
437 strbuf_init(out, q_seg->len);
439 while ((c = *in++) != 0) {
443 break; /* drop trailing newline */
444 strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
447 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
449 strbuf_addch(out, c);
454 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
456 /* Decode in..ep, possibly in-place to ot */
457 int c, pos = 0, acc = 0;
458 const char *in = b_seg->buf;
459 struct strbuf *out = xmalloc(sizeof(struct strbuf));
460 strbuf_init(out, b_seg->len);
462 while ((c = *in++) != 0) {
467 else if ('A' <= c && c <= 'Z')
469 else if ('a' <= c && c <= 'z')
471 else if ('0' <= c && c <= '9')
474 continue; /* garbage */
480 strbuf_addch(out, (acc | (c >> 4)));
484 strbuf_addch(out, (acc | (c >> 2)));
488 strbuf_addch(out, (acc | c));
496 static void convert_to_utf8(struct strbuf *line, const char *charset)
500 if (!charset || !*charset)
503 if (same_encoding(metainfo_charset, charset))
505 out = reencode_string(line->buf, metainfo_charset, charset);
507 die("cannot convert from %s to %s",
508 charset, metainfo_charset);
509 strbuf_attach(line, out, strlen(out), strlen(out));
512 static int decode_header_bq(struct strbuf *it)
515 struct strbuf outbuf = STRBUF_INIT, *dec;
516 struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
520 while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
522 strbuf_reset(&charset_q);
523 strbuf_reset(&piecebuf);
528 * We are about to process an encoded-word
529 * that begins at ep, but there is something
530 * before the encoded word.
533 for (scan = in; scan < ep; scan++)
537 if (scan != ep || in == it->buf) {
539 * We should not lose that "something",
540 * unless we have just processed an
541 * encoded-word, and there is only LWS
542 * before the one we are about to process.
544 strbuf_add(&outbuf, in, ep - in);
548 * ep : "=?iso-2022-jp?B?GyR...?= foo"
549 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
553 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
554 goto decode_header_bq_out;
556 if (cp + 3 - it->buf > it->len)
557 goto decode_header_bq_out;
558 strbuf_add(&charset_q, ep, cp - ep);
561 if (!encoding || cp[2] != '?')
562 goto decode_header_bq_out;
563 ep = strstr(cp + 3, "?=");
565 goto decode_header_bq_out;
566 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
567 switch (tolower(encoding)) {
569 goto decode_header_bq_out;
571 dec = decode_b_segment(&piecebuf);
574 dec = decode_q_segment(&piecebuf, 1);
577 if (metainfo_charset)
578 convert_to_utf8(dec, charset_q.buf);
580 strbuf_addbuf(&outbuf, dec);
585 strbuf_addstr(&outbuf, in);
587 strbuf_addbuf(it, &outbuf);
588 decode_header_bq_out:
589 strbuf_release(&outbuf);
590 strbuf_release(&charset_q);
591 strbuf_release(&piecebuf);
595 static void decode_header(struct strbuf *it)
597 if (decode_header_bq(it))
599 /* otherwise "it" is a straight copy of the input.
600 * This can be binary guck but there is no charset specified.
602 if (metainfo_charset)
603 convert_to_utf8(it, "");
606 static void decode_transfer_encoding(struct strbuf *line)
610 switch (transfer_encoding) {
612 ret = decode_q_segment(line, 0);
615 ret = decode_b_segment(line);
622 strbuf_addbuf(line, ret);
627 static void handle_filter(struct strbuf *line);
629 static int find_boundary(void)
631 while (!strbuf_getline(&line, fin, '\n')) {
632 if (*content_top && is_multipart_boundary(&line))
638 static int handle_boundary(void)
640 struct strbuf newline = STRBUF_INIT;
642 strbuf_addch(&newline, '\n');
644 if (line.len >= (*content_top)->len + 2 &&
645 !memcmp(line.buf + (*content_top)->len, "--", 2)) {
646 /* we hit an end boundary */
647 /* pop the current boundary off the stack */
648 strbuf_release(*content_top);
652 /* technically won't happen as is_multipart_boundary()
653 will fail first. But just in case..
655 if (--content_top < content) {
656 fprintf(stderr, "Detected mismatched boundaries, "
660 handle_filter(&newline);
661 strbuf_release(&newline);
663 /* skip to the next boundary */
664 if (!find_boundary())
669 /* set some defaults */
670 transfer_encoding = TE_DONTCARE;
671 strbuf_reset(&charset);
673 /* slurp in this section's info */
674 while (read_one_header_line(&line, fin))
675 check_header(&line, p_hdr_data, 0);
677 strbuf_release(&newline);
679 if (strbuf_getline(&line, fin, '\n'))
681 strbuf_addch(&line, '\n');
685 static inline int patchbreak(const struct strbuf *line)
689 /* Beginning of a "diff -" header? */
690 if (starts_with(line->buf, "diff -"))
693 /* CVS "Index: " line? */
694 if (starts_with(line->buf, "Index: "))
698 * "--- <filename>" starts patches without headers
699 * "---<sp>*" is a manual separator
704 if (starts_with(line->buf, "---")) {
705 /* space followed by a filename? */
706 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
708 /* Just whitespace? */
709 for (i = 3; i < line->len; i++) {
710 unsigned char c = line->buf[i];
721 static int is_scissors_line(const struct strbuf *line)
723 size_t i, len = line->len;
724 int scissors = 0, gap = 0;
725 int first_nonblank = -1;
726 int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
727 const char *buf = line->buf;
729 for (i = 0; i < len; i++) {
730 if (isspace(buf[i])) {
731 if (in_perforation) {
738 if (first_nonblank < 0)
746 (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
747 !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
758 * The mark must be at least 8 bytes long (e.g. "-- >8 --").
759 * Even though there can be arbitrary cruft on the same line
760 * (e.g. "cut here"), in order to avoid misidentification, the
761 * perforation must occupy more than a third of the visible
762 * width of the line, and dashes and scissors must occupy more
763 * than half of the perforation.
766 visible = last_nonblank - first_nonblank + 1;
767 return (scissors && 8 <= visible &&
768 visible < perforation * 3 &&
769 gap * 2 < perforation);
772 static int handle_commit_msg(struct strbuf *line)
774 static int still_looking = 1;
780 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
784 if (use_inbody_headers && still_looking) {
785 still_looking = check_header(line, s_hdr_data, 0);
789 /* Only trim the first (blank) line of the commit message
790 * when ignoring in-body headers.
794 /* normalize the log message to UTF-8. */
795 if (metainfo_charset)
796 convert_to_utf8(line, charset.buf);
798 if (use_scissors && is_scissors_line(line)) {
800 if (fseek(cmitmsg, 0L, SEEK_SET))
801 die_errno("Could not rewind output message file");
802 if (ftruncate(fileno(cmitmsg), 0))
803 die_errno("Could not truncate output message file at scissors");
807 * We may have already read "secondary headers"; purge
808 * them to give ourselves a clean restart.
810 for (i = 0; header[i]; i++) {
812 strbuf_release(s_hdr_data[i]);
813 s_hdr_data[i] = NULL;
818 if (patchbreak(line)) {
824 fputs(line->buf, cmitmsg);
828 static void handle_patch(const struct strbuf *line)
830 fwrite(line->buf, 1, line->len, patchfile);
834 static void handle_filter(struct strbuf *line)
836 static int filter = 0;
838 /* filter tells us which part we left off on */
841 if (!handle_commit_msg(line))
850 static void handle_body(void)
852 struct strbuf prev = STRBUF_INIT;
854 /* Skip up to the first boundary */
856 if (!find_boundary())
857 goto handle_body_out;
861 /* process any boundary lines */
862 if (*content_top && is_multipart_boundary(&line)) {
863 /* flush any leftover */
865 handle_filter(&prev);
868 if (!handle_boundary())
869 goto handle_body_out;
872 /* Unwrap transfer encoding */
873 decode_transfer_encoding(&line);
875 switch (transfer_encoding) {
879 struct strbuf **lines, **it, *sb;
881 /* Prepend any previous partial lines */
882 strbuf_insert(&line, 0, prev.buf, prev.len);
886 * This is a decoded line that may contain
887 * multiple new lines. Pass only one chunk
888 * at a time to handle_filter()
890 lines = strbuf_split(&line, '\n');
891 for (it = lines; (sb = *it); it++) {
892 if (*(it + 1) == NULL) /* The last line */
893 if (sb->buf[sb->len - 1] != '\n') {
894 /* Partial line, save it for later. */
895 strbuf_addbuf(&prev, sb);
901 * The partial chunk is saved in "prev" and will be
902 * appended by the next iteration of read_line_with_nul().
904 strbuf_list_free(lines);
908 handle_filter(&line);
911 } while (!strbuf_getwholeline(&line, fin, '\n'));
914 strbuf_release(&prev);
917 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
919 const char *sp = data->buf;
921 char *ep = strchr(sp, '\n');
927 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
934 static void handle_info(void)
939 for (i = 0; header[i]; i++) {
940 /* only print inbody headers if we output a patch file */
941 if (patch_lines && s_hdr_data[i])
943 else if (p_hdr_data[i])
948 if (!strcmp(header[i], "Subject")) {
950 cleanup_subject(hdr);
953 output_header_lines(fout, "Subject", hdr);
954 } else if (!strcmp(header[i], "From")) {
957 fprintf(fout, "Author: %s\n", name.buf);
958 fprintf(fout, "Email: %s\n", email.buf);
961 fprintf(fout, "%s: %s\n", header[i], hdr->buf);
967 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
973 cmitmsg = fopen(msg, "w");
978 patchfile = fopen(patch, "w");
985 p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
986 s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
990 } while (isspace(peek));
993 /* process the email header */
994 while (read_one_header_line(&line, fin))
995 check_header(&line, p_hdr_data, 1);
1003 static int git_mailinfo_config(const char *var, const char *value, void *unused)
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);
1011 /* perhaps others here */
1015 static const char mailinfo_usage[] =
1016 "git mailinfo [-k|-b] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1018 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1020 const char *def_charset;
1022 /* NEEDSWORK: might want to do the optional .git/ directory
1025 git_config(git_mailinfo_config, NULL);
1027 def_charset = get_commit_output_encoding();
1028 metainfo_charset = def_charset;
1030 while (1 < argc && argv[1][0] == '-') {
1031 if (!strcmp(argv[1], "-k"))
1033 else if (!strcmp(argv[1], "-b"))
1034 keep_non_patch_brackets_in_subject = 1;
1035 else if (!strcmp(argv[1], "-u"))
1036 metainfo_charset = def_charset;
1037 else if (!strcmp(argv[1], "-n"))
1038 metainfo_charset = NULL;
1039 else if (starts_with(argv[1], "--encoding="))
1040 metainfo_charset = argv[1] + 11;
1041 else if (!strcmp(argv[1], "--scissors"))
1043 else if (!strcmp(argv[1], "--no-scissors"))
1045 else if (!strcmp(argv[1], "--no-inbody-headers"))
1046 use_inbody_headers = 0;
1048 usage(mailinfo_usage);
1053 usage(mailinfo_usage);
1055 return !!mailinfo(stdin, stdout, argv[1], argv[2]);