2 * Another stupid program, this one parsing the headers of an
3 * email to figure out authorship and subject
13 extern char *gitstrcasestr(const char *haystack, const char *needle);
16 static FILE *cmitmsg, *patchfile;
18 static int keep_subject = 0;
19 static char *metainfo_charset = NULL;
20 static char line[1000];
21 static char date[1000];
22 static char name[1000];
23 static char email[1000];
24 static char subject[1000];
27 TE_DONTCARE, TE_QP, TE_BASE64,
29 static char charset[256];
31 static char multipart_boundary[1000];
32 static int multipart_boundary_len;
33 static int patch_lines = 0;
35 static char *sanity_check(char *name, char *email)
37 int len = strlen(name);
38 if (len < 3 || len > 60)
40 if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
45 static int handle_from(char *line)
47 char *at = strchr(line, '@');
54 * If we already have one email, don't take any confusing lines
56 if (*email && strchr(at+1, '@'))
59 /* Pick up the string around '@', possibly delimited with <>
60 * pair; that is the email part. White them out while copying.
74 unsigned char c = *at;
75 if (!c || c == '>' || isspace(c)) {
85 /* The remainder is name. It could be "John Doe <john.doe@xz>"
86 * or "john.doe@xz (John Doe)", but we have whited out the
87 * email part, so trim from both ends, possibly removing
88 * the () pair at the end.
90 at = line + strlen(line);
92 unsigned char c = *--at;
94 at[(c == ')') ? 0 : 1] = 0;
101 unsigned char c = *at;
102 if (!c || !isspace(c)) {
109 at = sanity_check(at, email);
114 static int handle_date(char *line)
120 static int handle_subject(char *line)
122 strcpy(subject, line);
126 /* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
127 * to have enough heuristics to grok MIME encoded patches often found
128 * on our mailing lists. For example, we do not even treat header lines
129 * case insensitively.
132 static int slurp_attr(const char *line, const char *name, char *attr)
134 char *ends, *ap = strcasestr(line, name);
148 sz = strcspn(ap, ends);
149 memcpy(attr, ap, sz);
154 static int handle_subcontent_type(char *line)
156 /* We do not want to mess with boundary. Note that we do not
157 * handle nested multipart.
159 if (strcasestr(line, "boundary=")) {
160 fprintf(stderr, "Not handling nested multipart message.\n");
163 slurp_attr(line, "charset=", charset);
166 for (i = 0; (c = charset[i]) != 0; i++)
167 charset[i] = tolower(c);
172 static int handle_content_type(char *line)
174 *multipart_boundary = 0;
175 if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
176 memcpy(multipart_boundary, "--", 2);
177 multipart_boundary_len = strlen(multipart_boundary);
179 slurp_attr(line, "charset=", charset);
183 static int handle_content_transfer_encoding(char *line)
185 if (strcasestr(line, "base64"))
186 transfer_encoding = TE_BASE64;
187 else if (strcasestr(line, "quoted-printable"))
188 transfer_encoding = TE_QP;
190 transfer_encoding = TE_DONTCARE;
194 static int is_multipart_boundary(const char *line)
196 return (!memcmp(line, multipart_boundary, multipart_boundary_len));
199 static int eatspace(char *line)
201 int len = strlen(line);
202 while (len > 0 && isspace(line[len-1]))
209 #define SEEN_SUBJECT 04
211 /* First lines of body can have From:, Date:, and Subject: */
212 static int handle_inbody_header(int *seen, char *line)
214 if (!memcmp("From:", line, 5) && isspace(line[5])) {
215 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
220 if (!memcmp("Date:", line, 5) && isspace(line[5])) {
221 if (!(*seen & SEEN_DATE)) {
227 if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
228 if (!(*seen & SEEN_SUBJECT)) {
229 handle_subject(line+9);
230 *seen |= SEEN_SUBJECT;
234 if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
235 if (!(*seen & SEEN_SUBJECT)) {
236 handle_subject(line);
237 *seen |= SEEN_SUBJECT;
244 static char *cleanup_subject(char *subject)
253 if (!memcmp("e:", subject+1, 2)) {
258 case ' ': case '\t': case ':':
263 p = strchr(subject, ']');
269 remove = p - subject;
270 if (remove <= len *2) {
280 static void cleanup_space(char *buf)
283 while ((c = *buf) != 0) {
289 int len = strlen(buf);
290 memmove(buf, buf+1, len);
297 typedef int (*header_fn_t)(char *);
304 static void check_header(char *line, int len, struct header_def *header)
308 if (header[0].namelen <= 0) {
309 for (i = 0; header[i].name; i++)
310 header[i].namelen = strlen(header[i].name);
312 for (i = 0; header[i].name; i++) {
313 int len = header[i].namelen;
314 if (!strncasecmp(line, header[i].name, len) &&
315 line[len] == ':' && isspace(line[len + 1])) {
316 header[i].func(line + len + 2);
322 static void check_subheader_line(char *line, int len)
324 static struct header_def header[] = {
325 { "Content-Type", handle_subcontent_type },
326 { "Content-Transfer-Encoding",
327 handle_content_transfer_encoding },
330 check_header(line, len, header);
332 static void check_header_line(char *line, int len)
334 static struct header_def header[] = {
335 { "From", handle_from },
336 { "Date", handle_date },
337 { "Subject", handle_subject },
338 { "Content-Type", handle_content_type },
339 { "Content-Transfer-Encoding",
340 handle_content_transfer_encoding },
343 check_header(line, len, header);
346 static int read_one_header_line(char *line, int sz, FILE *in)
351 if (fgets(line + ofs, sz - ofs, in) == NULL)
353 len = eatspace(line + ofs);
356 peek = fgetc(in); ungetc(peek, in);
357 if (peek == ' ' || peek == '\t') {
358 /* Yuck, 2822 header "folding" */
367 static unsigned hexval(int c)
369 if (c >= '0' && c <= '9')
371 if (c >= 'a' && c <= 'f')
373 if (c >= 'A' && c <= 'F')
378 static int decode_q_segment(char *in, char *ot, char *ep)
381 while ((c = *in++) != 0 && (in <= ep)) {
385 break; /* drop trailing newline */
386 *ot++ = ((hexval(d) << 4) | hexval(*in++));
395 static int decode_b_segment(char *in, char *ot, char *ep)
397 /* Decode in..ep, possibly in-place to ot */
398 int c, pos = 0, acc = 0;
400 while ((c = *in++) != 0 && (in <= ep)) {
405 else if ('A' <= c && c <= 'Z')
407 else if ('a' <= c && c <= 'z')
409 else if ('0' <= c && c <= '9')
412 /* padding is almost like (c == 0), except we do
413 * not output NUL resulting only from it;
414 * for now we just trust the data.
419 continue; /* garbage */
425 *ot++ = (acc | (c >> 4));
429 *ot++ = (acc | (c >> 2));
442 static void convert_to_utf8(char *line, char *charset)
445 size_t insize, outsize, nrc;
446 char outbuf[4096]; /* cheat */
447 static char latin_one[] = "latin-1";
448 char *input_charset = *charset ? charset : latin_one;
449 iconv_t conv = iconv_open(metainfo_charset, input_charset);
451 if (conv == (iconv_t) -1) {
452 static int warned_latin1_once = 0;
453 if (input_charset != latin_one) {
454 fprintf(stderr, "cannot convert from %s to %s\n",
455 input_charset, metainfo_charset);
458 else if (!warned_latin1_once) {
459 warned_latin1_once = 1;
460 fprintf(stderr, "tried to convert from %s to %s, "
461 "but your iconv does not work with it.\n",
462 input_charset, metainfo_charset);
469 outsize = sizeof(outbuf);
470 nrc = iconv(conv, &in, &insize, &out, &outsize);
472 if (nrc == (size_t) -1)
475 strcpy(line, outbuf);
478 static void decode_header_bq(char *it)
480 char *in, *out, *ep, *cp, *sp;
485 while ((ep = strstr(in, "=?")) != NULL) {
487 char charset_q[256], piecebuf[256];
495 * ep : "=?iso-2022-jp?B?GyR...?= foo"
496 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
499 cp = strchr(ep, '?');
501 return; /* no munging */
502 for (sp = ep; sp < cp; sp++)
503 charset_q[sp - ep] = tolower(*sp);
504 charset_q[cp - ep] = 0;
506 if (!encoding || cp[2] != '?')
507 return; /* no munging */
508 ep = strstr(cp + 3, "?=");
510 return; /* no munging */
511 switch (tolower(encoding)) {
513 return; /* no munging */
515 sz = decode_b_segment(cp + 3, piecebuf, ep);
518 sz = decode_q_segment(cp + 3, piecebuf, ep);
523 if (metainfo_charset)
524 convert_to_utf8(piecebuf, charset_q);
525 strcpy(out, piecebuf);
533 static void decode_transfer_encoding(char *line)
537 switch (transfer_encoding) {
539 ep = line + strlen(line);
540 decode_q_segment(line, line, ep);
543 ep = line + strlen(line);
544 decode_b_segment(line, line, ep);
551 static void handle_info(void)
554 static int done_info = 0;
560 sub = cleanup_subject(subject);
563 cleanup_space(email);
566 /* Unwrap inline B and Q encoding, and optionally
567 * normalize the meta information to utf8.
569 decode_header_bq(name);
570 decode_header_bq(date);
571 decode_header_bq(email);
572 decode_header_bq(sub);
573 printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
574 name, email, sub, date);
577 /* We are inside message body and have read line[] already.
578 * Spit out the commit log.
580 static int handle_commit_msg(void)
585 if (!memcmp("diff -", line, 6) ||
586 !memcmp("---", line, 3) ||
587 !memcmp("Index: ", line, 7))
589 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
590 /* We come here when the first part had only
591 * the commit message without any patch. We
592 * pretend we have not seen this line yet, and
593 * go back to the loop.
598 /* Unwrap transfer encoding and optionally
599 * normalize the log message to UTF-8.
601 decode_transfer_encoding(line);
602 if (metainfo_charset)
603 convert_to_utf8(line, charset);
604 fputs(line, cmitmsg);
605 } while (fgets(line, sizeof(line), stdin) != NULL);
611 /* We have done the commit message and have the first
612 * line of the patch in line[].
614 static void handle_patch(void)
617 if (multipart_boundary[0] && is_multipart_boundary(line))
619 /* Only unwrap transfer encoding but otherwise do not
620 * do anything. We do *NOT* want UTF-8 conversion
621 * here; we are dealing with the user payload.
623 decode_transfer_encoding(line);
624 fputs(line, patchfile);
626 } while (fgets(line, sizeof(line), stdin) != NULL);
629 /* multipart boundary and transfer encoding are set up for us, and we
630 * are at the end of the sub header. do equivalent of handle_body up
631 * to the next boundary without closing patchfile --- we will expect
632 * that the first part to contain commit message and a patch, and
633 * handle other parts as pure patches.
635 static int handle_multipart_one_part(void)
641 while (fgets(line, sizeof(line), stdin) != NULL) {
643 len = eatspace(line);
647 if (is_multipart_boundary(line))
649 if (0 <= seen && handle_inbody_header(&seen, line))
651 seen = -1; /* no more inbody headers */
654 if (handle_commit_msg())
664 static void handle_multipart_body(void)
668 /* Skip up to the first boundary */
669 while (fgets(line, sizeof(line), stdin) != NULL)
670 if (is_multipart_boundary(line)) {
676 /* We are on boundary line. Start slurping the subhead. */
678 int len = read_one_header_line(line, sizeof(line), stdin);
680 if (handle_multipart_one_part() < 0)
684 check_subheader_line(line, len);
688 fprintf(stderr, "No patch found\n");
693 /* Non multipart message */
694 static void handle_body(void)
698 while (fgets(line, sizeof(line), stdin) != NULL) {
699 int len = eatspace(line);
702 if (0 <= seen && handle_inbody_header(&seen, line))
704 seen = -1; /* no more inbody headers */
713 fprintf(stderr, "No patch found\n");
718 static const char mailinfo_usage[] =
719 "git-mailinfo [-k] [-u] msg patch <mail >info";
721 static void usage(void) {
722 fprintf(stderr, "%s\n", mailinfo_usage);
726 int main(int argc, char **argv)
728 while (1 < argc && argv[1][0] == '-') {
729 if (!strcmp(argv[1], "-k"))
731 else if (!strcmp(argv[1], "-u"))
732 metainfo_charset = "utf-8";
733 else if (!strncmp(argv[1], "-u=", 3))
734 metainfo_charset = argv[1] + 3;
742 cmitmsg = fopen(argv[1], "w");
747 patchfile = fopen(argv[2], "w");
753 int len = read_one_header_line(line, sizeof(line), stdin);
755 if (multipart_boundary[0])
756 handle_multipart_body();
761 check_header_line(line, len);