2 * Another stupid program, this one parsing the headers of an
3 * email to figure out authorship and subject
11 #include "git-compat-util.h"
14 static FILE *cmitmsg, *patchfile;
16 static int keep_subject = 0;
17 static char *metainfo_charset = NULL;
18 static char line[1000];
19 static char date[1000];
20 static char name[1000];
21 static char email[1000];
22 static char subject[1000];
25 TE_DONTCARE, TE_QP, TE_BASE64,
27 static char charset[256];
29 static char multipart_boundary[1000];
30 static int multipart_boundary_len;
31 static int patch_lines = 0;
33 static char *sanity_check(char *name, char *email)
35 int len = strlen(name);
36 if (len < 3 || len > 60)
38 if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
43 static int handle_from(char *line)
45 char *at = strchr(line, '@');
52 * If we already have one email, don't take any confusing lines
54 if (*email && strchr(at+1, '@'))
57 /* Pick up the string around '@', possibly delimited with <>
58 * pair; that is the email part. White them out while copying.
72 unsigned char c = *at;
73 if (!c || c == '>' || isspace(c)) {
83 /* The remainder is name. It could be "John Doe <john.doe@xz>"
84 * or "john.doe@xz (John Doe)", but we have whited out the
85 * email part, so trim from both ends, possibly removing
86 * the () pair at the end.
88 at = line + strlen(line);
90 unsigned char c = *--at;
92 at[(c == ')') ? 0 : 1] = 0;
99 unsigned char c = *at;
100 if (!c || !isspace(c)) {
107 at = sanity_check(at, email);
112 static int handle_date(char *line)
118 static int handle_subject(char *line)
120 strcpy(subject, line);
124 /* NOTE NOTE NOTE. We do not claim we do full MIME. We just attempt
125 * to have enough heuristics to grok MIME encoded patches often found
126 * on our mailing lists. For example, we do not even treat header lines
127 * case insensitively.
130 static int slurp_attr(const char *line, const char *name, char *attr)
132 char *ends, *ap = strcasestr(line, name);
146 sz = strcspn(ap, ends);
147 memcpy(attr, ap, sz);
152 static int handle_subcontent_type(char *line)
154 /* We do not want to mess with boundary. Note that we do not
155 * handle nested multipart.
157 if (strcasestr(line, "boundary=")) {
158 fprintf(stderr, "Not handling nested multipart message.\n");
161 slurp_attr(line, "charset=", charset);
164 for (i = 0; (c = charset[i]) != 0; i++)
165 charset[i] = tolower(c);
170 static int handle_content_type(char *line)
172 *multipart_boundary = 0;
173 if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
174 memcpy(multipart_boundary, "--", 2);
175 multipart_boundary_len = strlen(multipart_boundary);
177 slurp_attr(line, "charset=", charset);
181 static int handle_content_transfer_encoding(char *line)
183 if (strcasestr(line, "base64"))
184 transfer_encoding = TE_BASE64;
185 else if (strcasestr(line, "quoted-printable"))
186 transfer_encoding = TE_QP;
188 transfer_encoding = TE_DONTCARE;
192 static int is_multipart_boundary(const char *line)
194 return (!memcmp(line, multipart_boundary, multipart_boundary_len));
197 static int eatspace(char *line)
199 int len = strlen(line);
200 while (len > 0 && isspace(line[len-1]))
207 #define SEEN_SUBJECT 04
209 /* First lines of body can have From:, Date:, and Subject: */
210 static int handle_inbody_header(int *seen, char *line)
212 if (!memcmp("From:", line, 5) && isspace(line[5])) {
213 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
218 if (!memcmp("Date:", line, 5) && isspace(line[5])) {
219 if (!(*seen & SEEN_DATE)) {
225 if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
226 if (!(*seen & SEEN_SUBJECT)) {
227 handle_subject(line+9);
228 *seen |= SEEN_SUBJECT;
232 if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
233 if (!(*seen & SEEN_SUBJECT)) {
234 handle_subject(line);
235 *seen |= SEEN_SUBJECT;
242 static char *cleanup_subject(char *subject)
251 if (!memcmp("e:", subject+1, 2)) {
256 case ' ': case '\t': case ':':
261 p = strchr(subject, ']');
267 remove = p - subject;
268 if (remove <= len *2) {
278 static void cleanup_space(char *buf)
281 while ((c = *buf) != 0) {
287 int len = strlen(buf);
288 memmove(buf, buf+1, len);
295 typedef int (*header_fn_t)(char *);
302 static void check_header(char *line, int len, struct header_def *header)
306 if (header[0].namelen <= 0) {
307 for (i = 0; header[i].name; i++)
308 header[i].namelen = strlen(header[i].name);
310 for (i = 0; header[i].name; i++) {
311 int len = header[i].namelen;
312 if (!strncasecmp(line, header[i].name, len) &&
313 line[len] == ':' && isspace(line[len + 1])) {
314 header[i].func(line + len + 2);
320 static void check_subheader_line(char *line, int len)
322 static struct header_def header[] = {
323 { "Content-Type", handle_subcontent_type },
324 { "Content-Transfer-Encoding",
325 handle_content_transfer_encoding },
328 check_header(line, len, header);
330 static void check_header_line(char *line, int len)
332 static struct header_def header[] = {
333 { "From", handle_from },
334 { "Date", handle_date },
335 { "Subject", handle_subject },
336 { "Content-Type", handle_content_type },
337 { "Content-Transfer-Encoding",
338 handle_content_transfer_encoding },
341 check_header(line, len, header);
344 static int read_one_header_line(char *line, int sz, FILE *in)
349 if (fgets(line + ofs, sz - ofs, in) == NULL)
351 len = eatspace(line + ofs);
354 peek = fgetc(in); ungetc(peek, in);
355 if (peek == ' ' || peek == '\t') {
356 /* Yuck, 2822 header "folding" */
365 static unsigned hexval(int c)
367 if (c >= '0' && c <= '9')
369 if (c >= 'a' && c <= 'f')
371 if (c >= 'A' && c <= 'F')
376 static int decode_q_segment(char *in, char *ot, char *ep)
379 while ((c = *in++) != 0 && (in <= ep)) {
383 break; /* drop trailing newline */
384 *ot++ = ((hexval(d) << 4) | hexval(*in++));
393 static int decode_b_segment(char *in, char *ot, char *ep)
395 /* Decode in..ep, possibly in-place to ot */
396 int c, pos = 0, acc = 0;
398 while ((c = *in++) != 0 && (in <= ep)) {
403 else if ('A' <= c && c <= 'Z')
405 else if ('a' <= c && c <= 'z')
407 else if ('0' <= c && c <= '9')
410 /* padding is almost like (c == 0), except we do
411 * not output NUL resulting only from it;
412 * for now we just trust the data.
417 continue; /* garbage */
423 *ot++ = (acc | (c >> 4));
427 *ot++ = (acc | (c >> 2));
440 static void convert_to_utf8(char *line, char *charset)
443 size_t insize, outsize, nrc;
444 char outbuf[4096]; /* cheat */
445 static char latin_one[] = "latin-1";
446 char *input_charset = *charset ? charset : latin_one;
447 iconv_t conv = iconv_open(metainfo_charset, input_charset);
449 if (conv == (iconv_t) -1) {
450 static int warned_latin1_once = 0;
451 if (input_charset != latin_one) {
452 fprintf(stderr, "cannot convert from %s to %s\n",
453 input_charset, metainfo_charset);
456 else if (!warned_latin1_once) {
457 warned_latin1_once = 1;
458 fprintf(stderr, "tried to convert from %s to %s, "
459 "but your iconv does not work with it.\n",
460 input_charset, metainfo_charset);
467 outsize = sizeof(outbuf);
468 nrc = iconv(conv, &in, &insize, &out, &outsize);
470 if (nrc == (size_t) -1)
473 strcpy(line, outbuf);
476 static void decode_header_bq(char *it)
478 char *in, *out, *ep, *cp, *sp;
483 while ((ep = strstr(in, "=?")) != NULL) {
485 char charset_q[256], piecebuf[256];
493 * ep : "=?iso-2022-jp?B?GyR...?= foo"
494 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
497 cp = strchr(ep, '?');
499 return; /* no munging */
500 for (sp = ep; sp < cp; sp++)
501 charset_q[sp - ep] = tolower(*sp);
502 charset_q[cp - ep] = 0;
504 if (!encoding || cp[2] != '?')
505 return; /* no munging */
506 ep = strstr(cp + 3, "?=");
508 return; /* no munging */
509 switch (tolower(encoding)) {
511 return; /* no munging */
513 sz = decode_b_segment(cp + 3, piecebuf, ep);
516 sz = decode_q_segment(cp + 3, piecebuf, ep);
521 if (metainfo_charset)
522 convert_to_utf8(piecebuf, charset_q);
523 strcpy(out, piecebuf);
531 static void decode_transfer_encoding(char *line)
535 switch (transfer_encoding) {
537 ep = line + strlen(line);
538 decode_q_segment(line, line, ep);
541 ep = line + strlen(line);
542 decode_b_segment(line, line, ep);
549 static void handle_info(void)
552 static int done_info = 0;
558 sub = cleanup_subject(subject);
561 cleanup_space(email);
564 /* Unwrap inline B and Q encoding, and optionally
565 * normalize the meta information to utf8.
567 decode_header_bq(name);
568 decode_header_bq(date);
569 decode_header_bq(email);
570 decode_header_bq(sub);
571 printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
572 name, email, sub, date);
575 /* We are inside message body and have read line[] already.
576 * Spit out the commit log.
578 static int handle_commit_msg(void)
583 if (!memcmp("diff -", line, 6) ||
584 !memcmp("---", line, 3) ||
585 !memcmp("Index: ", line, 7))
587 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
588 /* We come here when the first part had only
589 * the commit message without any patch. We
590 * pretend we have not seen this line yet, and
591 * go back to the loop.
596 /* Unwrap transfer encoding and optionally
597 * normalize the log message to UTF-8.
599 decode_transfer_encoding(line);
600 if (metainfo_charset)
601 convert_to_utf8(line, charset);
602 fputs(line, cmitmsg);
603 } while (fgets(line, sizeof(line), stdin) != NULL);
609 /* We have done the commit message and have the first
610 * line of the patch in line[].
612 static void handle_patch(void)
615 if (multipart_boundary[0] && is_multipart_boundary(line))
617 /* Only unwrap transfer encoding but otherwise do not
618 * do anything. We do *NOT* want UTF-8 conversion
619 * here; we are dealing with the user payload.
621 decode_transfer_encoding(line);
622 fputs(line, patchfile);
624 } while (fgets(line, sizeof(line), stdin) != NULL);
627 /* multipart boundary and transfer encoding are set up for us, and we
628 * are at the end of the sub header. do equivalent of handle_body up
629 * to the next boundary without closing patchfile --- we will expect
630 * that the first part to contain commit message and a patch, and
631 * handle other parts as pure patches.
633 static int handle_multipart_one_part(void)
639 while (fgets(line, sizeof(line), stdin) != NULL) {
641 len = eatspace(line);
645 if (is_multipart_boundary(line))
647 if (0 <= seen && handle_inbody_header(&seen, line))
649 seen = -1; /* no more inbody headers */
652 if (handle_commit_msg())
662 static void handle_multipart_body(void)
666 /* Skip up to the first boundary */
667 while (fgets(line, sizeof(line), stdin) != NULL)
668 if (is_multipart_boundary(line)) {
674 /* We are on boundary line. Start slurping the subhead. */
676 int len = read_one_header_line(line, sizeof(line), stdin);
678 if (handle_multipart_one_part() < 0)
682 check_subheader_line(line, len);
686 fprintf(stderr, "No patch found\n");
691 /* Non multipart message */
692 static void handle_body(void)
696 while (fgets(line, sizeof(line), stdin) != NULL) {
697 int len = eatspace(line);
700 if (0 <= seen && handle_inbody_header(&seen, line))
702 seen = -1; /* no more inbody headers */
711 fprintf(stderr, "No patch found\n");
716 static const char mailinfo_usage[] =
717 "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
719 int main(int argc, char **argv)
721 /* NEEDSWORK: might want to do the optional .git/ directory
724 git_config(git_default_config);
726 while (1 < argc && argv[1][0] == '-') {
727 if (!strcmp(argv[1], "-k"))
729 else if (!strcmp(argv[1], "-u"))
730 metainfo_charset = git_commit_encoding;
731 else if (!strncmp(argv[1], "--encoding=", 11))
732 metainfo_charset = argv[1] + 11;
734 usage(mailinfo_usage);
739 usage(mailinfo_usage);
740 cmitmsg = fopen(argv[1], "w");
745 patchfile = fopen(argv[2], "w");
751 int len = read_one_header_line(line, sizeof(line), stdin);
753 if (multipart_boundary[0])
754 handle_multipart_body();
759 check_header_line(line, len);