2 * Another stupid program, this one parsing the headers of an
3 * email to figure out authorship and subject
13 static FILE *cmitmsg, *patchfile;
15 static int keep_subject = 0;
16 static char *metainfo_charset = NULL;
17 static char line[1000];
18 static char date[1000];
19 static char name[1000];
20 static char email[1000];
21 static char subject[1000];
24 TE_DONTCARE, TE_QP, TE_BASE64,
26 static char charset[256];
28 static char multipart_boundary[1000];
29 static int multipart_boundary_len;
30 static int patch_lines = 0;
32 static char *sanity_check(char *name, char *email)
34 int len = strlen(name);
35 if (len < 3 || len > 60)
37 if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
42 static int handle_from(char *line)
44 char *at = strchr(line, '@');
51 * If we already have one email, don't take any confusing lines
53 if (*email && strchr(at+1, '@'))
56 /* Pick up the string around '@', possibly delimited with <>
57 * pair; that is the email part. White them out while copying.
71 unsigned char c = *at;
72 if (!c || c == '>' || isspace(c)) {
82 /* The remainder is name. It could be "John Doe <john.doe@xz>"
83 * or "john.doe@xz (John Doe)", but we have whited out the
84 * email part, so trim from both ends, possibly removing
85 * the () pair at the end.
87 at = line + strlen(line);
89 unsigned char c = *--at;
91 at[(c == ')') ? 0 : 1] = 0;
98 unsigned char c = *at;
99 if (!c || !isspace(c)) {
106 at = sanity_check(at, email);
111 static int handle_date(char *line)
117 static int handle_subject(char *line)
119 strcpy(subject, line);
123 /* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
124 * to have enough heuristics to grok MIME encoded patches often found
125 * on our mailing lists. For example, we do not even treat header lines
126 * case insensitively.
129 static int slurp_attr(const char *line, const char *name, char *attr)
131 char *ends, *ap = strcasestr(line, name);
145 sz = strcspn(ap, ends);
146 memcpy(attr, ap, sz);
151 static int handle_subcontent_type(char *line)
153 /* We do not want to mess with boundary. Note that we do not
154 * handle nested multipart.
156 if (strcasestr(line, "boundary=")) {
157 fprintf(stderr, "Not handling nested multipart message.\n");
160 slurp_attr(line, "charset=", charset);
163 for (i = 0; (c = charset[i]) != 0; i++)
164 charset[i] = tolower(c);
169 static int handle_content_type(char *line)
171 *multipart_boundary = 0;
172 if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
173 memcpy(multipart_boundary, "--", 2);
174 multipart_boundary_len = strlen(multipart_boundary);
176 slurp_attr(line, "charset=", charset);
180 static int handle_content_transfer_encoding(char *line)
182 if (strcasestr(line, "base64"))
183 transfer_encoding = TE_BASE64;
184 else if (strcasestr(line, "quoted-printable"))
185 transfer_encoding = TE_QP;
187 transfer_encoding = TE_DONTCARE;
191 static int is_multipart_boundary(const char *line)
193 return (!memcmp(line, multipart_boundary, multipart_boundary_len));
196 static int eatspace(char *line)
198 int len = strlen(line);
199 while (len > 0 && isspace(line[len-1]))
206 #define SEEN_SUBJECT 04
208 /* First lines of body can have From:, Date:, and Subject: */
209 static int handle_inbody_header(int *seen, char *line)
211 if (!memcmp("From:", line, 5) && isspace(line[5])) {
212 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
217 if (!memcmp("Date:", line, 5) && isspace(line[5])) {
218 if (!(*seen & SEEN_DATE)) {
224 if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
225 if (!(*seen & SEEN_SUBJECT)) {
226 handle_subject(line+9);
227 *seen |= SEEN_SUBJECT;
231 if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
232 if (!(*seen & SEEN_SUBJECT)) {
233 handle_subject(line);
234 *seen |= SEEN_SUBJECT;
241 static char *cleanup_subject(char *subject)
250 if (!memcmp("e:", subject+1, 2)) {
255 case ' ': case '\t': case ':':
260 p = strchr(subject, ']');
266 remove = p - subject;
267 if (remove <= len *2) {
277 static void cleanup_space(char *buf)
280 while ((c = *buf) != 0) {
286 int len = strlen(buf);
287 memmove(buf, buf+1, len);
294 typedef int (*header_fn_t)(char *);
301 static void check_header(char *line, int len, struct header_def *header)
305 if (header[0].namelen <= 0) {
306 for (i = 0; header[i].name; i++)
307 header[i].namelen = strlen(header[i].name);
309 for (i = 0; header[i].name; i++) {
310 int len = header[i].namelen;
311 if (!strncasecmp(line, header[i].name, len) &&
312 line[len] == ':' && isspace(line[len + 1])) {
313 header[i].func(line + len + 2);
319 static void check_subheader_line(char *line, int len)
321 static struct header_def header[] = {
322 { "Content-Type", handle_subcontent_type },
323 { "Content-Transfer-Encoding",
324 handle_content_transfer_encoding },
327 check_header(line, len, header);
329 static void check_header_line(char *line, int len)
331 static struct header_def header[] = {
332 { "From", handle_from },
333 { "Date", handle_date },
334 { "Subject", handle_subject },
335 { "Content-Type", handle_content_type },
336 { "Content-Transfer-Encoding",
337 handle_content_transfer_encoding },
340 check_header(line, len, header);
343 static int read_one_header_line(char *line, int sz, FILE *in)
348 if (fgets(line + ofs, sz - ofs, in) == NULL)
350 len = eatspace(line + ofs);
353 peek = fgetc(in); ungetc(peek, in);
354 if (peek == ' ' || peek == '\t') {
355 /* Yuck, 2822 header "folding" */
364 static unsigned hexval(int c)
366 if (c >= '0' && c <= '9')
368 if (c >= 'a' && c <= 'f')
370 if (c >= 'A' && c <= 'F')
375 static int decode_q_segment(char *in, char *ot, char *ep)
378 while ((c = *in++) != 0 && (in <= ep)) {
382 break; /* drop trailing newline */
383 *ot++ = ((hexval(d) << 4) | hexval(*in++));
392 static int decode_b_segment(char *in, char *ot, char *ep)
394 /* Decode in..ep, possibly in-place to ot */
395 int c, pos = 0, acc = 0;
397 while ((c = *in++) != 0 && (in <= ep)) {
402 else if ('A' <= c && c <= 'Z')
404 else if ('a' <= c && c <= 'z')
406 else if ('0' <= c && c <= '9')
409 /* padding is almost like (c == 0), except we do
410 * not output NUL resulting only from it;
411 * for now we just trust the data.
416 continue; /* garbage */
422 *ot++ = (acc | (c >> 4));
426 *ot++ = (acc | (c >> 2));
439 static void convert_to_utf8(char *line, char *charset)
442 size_t insize, outsize, nrc;
443 char outbuf[4096]; /* cheat */
444 static char latin_one[] = "latin-1";
445 char *input_charset = *charset ? charset : latin_one;
446 iconv_t conv = iconv_open(metainfo_charset, input_charset);
448 if (conv == (iconv_t) -1) {
449 static int warned_latin1_once = 0;
450 if (input_charset != latin_one) {
451 fprintf(stderr, "cannot convert from %s to %s\n",
452 input_charset, metainfo_charset);
455 else if (!warned_latin1_once) {
456 warned_latin1_once = 1;
457 fprintf(stderr, "tried to convert from %s to %s, "
458 "but your iconv does not work with it.\n",
459 input_charset, metainfo_charset);
466 outsize = sizeof(outbuf);
467 nrc = iconv(conv, &in, &insize, &out, &outsize);
469 if (nrc == (size_t) -1)
472 strcpy(line, outbuf);
475 static void decode_header_bq(char *it)
477 char *in, *out, *ep, *cp, *sp;
482 while ((ep = strstr(in, "=?")) != NULL) {
484 char charset_q[256], piecebuf[256];
492 * ep : "=?iso-2022-jp?B?GyR...?= foo"
493 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
496 cp = strchr(ep, '?');
498 return; /* no munging */
499 for (sp = ep; sp < cp; sp++)
500 charset_q[sp - ep] = tolower(*sp);
501 charset_q[cp - ep] = 0;
503 if (!encoding || cp[2] != '?')
504 return; /* no munging */
505 ep = strstr(cp + 3, "?=");
507 return; /* no munging */
508 switch (tolower(encoding)) {
510 return; /* no munging */
512 sz = decode_b_segment(cp + 3, piecebuf, ep);
515 sz = decode_q_segment(cp + 3, piecebuf, ep);
520 if (metainfo_charset)
521 convert_to_utf8(piecebuf, charset_q);
522 strcpy(out, piecebuf);
530 static void decode_transfer_encoding(char *line)
534 switch (transfer_encoding) {
536 ep = line + strlen(line);
537 decode_q_segment(line, line, ep);
540 ep = line + strlen(line);
541 decode_b_segment(line, line, ep);
548 static void handle_info(void)
551 static int done_info = 0;
557 sub = cleanup_subject(subject);
560 cleanup_space(email);
563 /* Unwrap inline B and Q encoding, and optionally
564 * normalize the meta information to utf8.
566 decode_header_bq(name);
567 decode_header_bq(date);
568 decode_header_bq(email);
569 decode_header_bq(sub);
570 printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
571 name, email, sub, date);
574 /* We are inside message body and have read line[] already.
575 * Spit out the commit log.
577 static int handle_commit_msg(void)
582 if (!memcmp("diff -", line, 6) ||
583 !memcmp("---", line, 3) ||
584 !memcmp("Index: ", line, 7))
586 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
587 /* We come here when the first part had only
588 * the commit message without any patch. We
589 * pretend we have not seen this line yet, and
590 * go back to the loop.
595 /* Unwrap transfer encoding and optionally
596 * normalize the log message to UTF-8.
598 decode_transfer_encoding(line);
599 if (metainfo_charset)
600 convert_to_utf8(line, charset);
601 fputs(line, cmitmsg);
602 } while (fgets(line, sizeof(line), stdin) != NULL);
608 /* We have done the commit message and have the first
609 * line of the patch in line[].
611 static void handle_patch(void)
614 if (multipart_boundary[0] && is_multipart_boundary(line))
616 /* Only unwrap transfer encoding but otherwise do not
617 * do anything. We do *NOT* want UTF-8 conversion
618 * here; we are dealing with the user payload.
620 decode_transfer_encoding(line);
621 fputs(line, patchfile);
623 } while (fgets(line, sizeof(line), stdin) != NULL);
626 /* multipart boundary and transfer encoding are set up for us, and we
627 * are at the end of the sub header. do equivalent of handle_body up
628 * to the next boundary without closing patchfile --- we will expect
629 * that the first part to contain commit message and a patch, and
630 * handle other parts as pure patches.
632 static int handle_multipart_one_part(void)
638 while (fgets(line, sizeof(line), stdin) != NULL) {
640 len = eatspace(line);
644 if (is_multipart_boundary(line))
646 if (0 <= seen && handle_inbody_header(&seen, line))
648 seen = -1; /* no more inbody headers */
651 if (handle_commit_msg())
661 static void handle_multipart_body(void)
665 /* Skip up to the first boundary */
666 while (fgets(line, sizeof(line), stdin) != NULL)
667 if (is_multipart_boundary(line)) {
673 /* We are on boundary line. Start slurping the subhead. */
675 int len = read_one_header_line(line, sizeof(line), stdin);
677 if (handle_multipart_one_part() < 0)
681 check_subheader_line(line, len);
685 fprintf(stderr, "No patch found\n");
690 /* Non multipart message */
691 static void handle_body(void)
695 while (fgets(line, sizeof(line), stdin) != NULL) {
696 int len = eatspace(line);
699 if (0 <= seen && handle_inbody_header(&seen, line))
701 seen = -1; /* no more inbody headers */
710 fprintf(stderr, "No patch found\n");
715 static const char mailinfo_usage[] =
716 "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
718 int main(int argc, char **argv)
720 /* NEEDSWORK: might want to do the optional .git/ directory
723 git_config(git_default_config);
725 while (1 < argc && argv[1][0] == '-') {
726 if (!strcmp(argv[1], "-k"))
728 else if (!strcmp(argv[1], "-u"))
729 metainfo_charset = git_commit_encoding;
730 else if (!strncmp(argv[1], "--encoding=", 11))
731 metainfo_charset = argv[1] + 11;
733 usage(mailinfo_usage);
738 usage(mailinfo_usage);
739 cmitmsg = fopen(argv[1], "w");
744 patchfile = fopen(argv[2], "w");
750 int len = read_one_header_line(line, sizeof(line), stdin);
752 if (multipart_boundary[0])
753 handle_multipart_body();
758 check_header_line(line, len);