show_ident_date: fix tz range check
[git/debian.git] / pretty.c
blob4d4c1e95625e5758efc8945a67a8ca0776b219fc
1 #include "cache.h"
2 #include "commit.h"
3 #include "utf8.h"
4 #include "diff.h"
5 #include "revision.h"
6 #include "string-list.h"
7 #include "mailmap.h"
8 #include "log-tree.h"
9 #include "notes.h"
10 #include "color.h"
11 #include "reflog-walk.h"
12 #include "gpg-interface.h"
14 static char *user_format;
15 static struct cmt_fmt_map {
16 const char *name;
17 enum cmit_fmt format;
18 int is_tformat;
19 int is_alias;
20 const char *user_format;
21 } *commit_formats;
22 static size_t builtin_formats_len;
23 static size_t commit_formats_len;
24 static size_t commit_formats_alloc;
25 static struct cmt_fmt_map *find_commit_format(const char *sought);
27 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
29 free(user_format);
30 user_format = xstrdup(cp);
31 if (is_tformat)
32 rev->use_terminator = 1;
33 rev->commit_format = CMIT_FMT_USERFORMAT;
36 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
38 struct cmt_fmt_map *commit_format = NULL;
39 const char *name;
40 const char *fmt;
41 int i;
43 if (prefixcmp(var, "pretty."))
44 return 0;
46 name = var + strlen("pretty.");
47 for (i = 0; i < builtin_formats_len; i++) {
48 if (!strcmp(commit_formats[i].name, name))
49 return 0;
52 for (i = builtin_formats_len; i < commit_formats_len; i++) {
53 if (!strcmp(commit_formats[i].name, name)) {
54 commit_format = &commit_formats[i];
55 break;
59 if (!commit_format) {
60 ALLOC_GROW(commit_formats, commit_formats_len+1,
61 commit_formats_alloc);
62 commit_format = &commit_formats[commit_formats_len];
63 memset(commit_format, 0, sizeof(*commit_format));
64 commit_formats_len++;
67 commit_format->name = xstrdup(name);
68 commit_format->format = CMIT_FMT_USERFORMAT;
69 git_config_string(&fmt, var, value);
70 if (!prefixcmp(fmt, "format:") || !prefixcmp(fmt, "tformat:")) {
71 commit_format->is_tformat = fmt[0] == 't';
72 fmt = strchr(fmt, ':') + 1;
73 } else if (strchr(fmt, '%'))
74 commit_format->is_tformat = 1;
75 else
76 commit_format->is_alias = 1;
77 commit_format->user_format = fmt;
79 return 0;
82 static void setup_commit_formats(void)
84 struct cmt_fmt_map builtin_formats[] = {
85 { "raw", CMIT_FMT_RAW, 0 },
86 { "medium", CMIT_FMT_MEDIUM, 0 },
87 { "short", CMIT_FMT_SHORT, 0 },
88 { "email", CMIT_FMT_EMAIL, 0 },
89 { "fuller", CMIT_FMT_FULLER, 0 },
90 { "full", CMIT_FMT_FULL, 0 },
91 { "oneline", CMIT_FMT_ONELINE, 1 }
93 commit_formats_len = ARRAY_SIZE(builtin_formats);
94 builtin_formats_len = commit_formats_len;
95 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
96 memcpy(commit_formats, builtin_formats,
97 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
99 git_config(git_pretty_formats_config, NULL);
102 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
103 const char *original,
104 int num_redirections)
106 struct cmt_fmt_map *found = NULL;
107 size_t found_match_len = 0;
108 int i;
110 if (num_redirections >= commit_formats_len)
111 die("invalid --pretty format: "
112 "'%s' references an alias which points to itself",
113 original);
115 for (i = 0; i < commit_formats_len; i++) {
116 size_t match_len;
118 if (prefixcmp(commit_formats[i].name, sought))
119 continue;
121 match_len = strlen(commit_formats[i].name);
122 if (found == NULL || found_match_len > match_len) {
123 found = &commit_formats[i];
124 found_match_len = match_len;
128 if (found && found->is_alias) {
129 found = find_commit_format_recursive(found->user_format,
130 original,
131 num_redirections+1);
134 return found;
137 static struct cmt_fmt_map *find_commit_format(const char *sought)
139 if (!commit_formats)
140 setup_commit_formats();
142 return find_commit_format_recursive(sought, sought, 0);
145 void get_commit_format(const char *arg, struct rev_info *rev)
147 struct cmt_fmt_map *commit_format;
149 rev->use_terminator = 0;
150 if (!arg || !*arg) {
151 rev->commit_format = CMIT_FMT_DEFAULT;
152 return;
154 if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
155 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
156 return;
159 if (strchr(arg, '%')) {
160 save_user_format(rev, arg, 1);
161 return;
164 commit_format = find_commit_format(arg);
165 if (!commit_format)
166 die("invalid --pretty format: %s", arg);
168 rev->commit_format = commit_format->format;
169 rev->use_terminator = commit_format->is_tformat;
170 if (commit_format->format == CMIT_FMT_USERFORMAT) {
171 save_user_format(rev, commit_format->user_format,
172 commit_format->is_tformat);
177 * Generic support for pretty-printing the header
179 static int get_one_line(const char *msg)
181 int ret = 0;
183 for (;;) {
184 char c = *msg++;
185 if (!c)
186 break;
187 ret++;
188 if (c == '\n')
189 break;
191 return ret;
194 /* High bit set, or ISO-2022-INT */
195 static int non_ascii(int ch)
197 return !isascii(ch) || ch == '\033';
200 int has_non_ascii(const char *s)
202 int ch;
203 if (!s)
204 return 0;
205 while ((ch = *s++) != '\0') {
206 if (non_ascii(ch))
207 return 1;
209 return 0;
212 static int is_rfc822_special(char ch)
214 switch (ch) {
215 case '(':
216 case ')':
217 case '<':
218 case '>':
219 case '[':
220 case ']':
221 case ':':
222 case ';':
223 case '@':
224 case ',':
225 case '.':
226 case '"':
227 case '\\':
228 return 1;
229 default:
230 return 0;
234 static int needs_rfc822_quoting(const char *s, int len)
236 int i;
237 for (i = 0; i < len; i++)
238 if (is_rfc822_special(s[i]))
239 return 1;
240 return 0;
243 static int last_line_length(struct strbuf *sb)
245 int i;
247 /* How many bytes are already used on the last line? */
248 for (i = sb->len - 1; i >= 0; i--)
249 if (sb->buf[i] == '\n')
250 break;
251 return sb->len - (i + 1);
254 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
256 int i;
258 /* just a guess, we may have to also backslash-quote */
259 strbuf_grow(out, len + 2);
261 strbuf_addch(out, '"');
262 for (i = 0; i < len; i++) {
263 switch (s[i]) {
264 case '"':
265 case '\\':
266 strbuf_addch(out, '\\');
267 /* fall through */
268 default:
269 strbuf_addch(out, s[i]);
272 strbuf_addch(out, '"');
275 enum rfc2047_type {
276 RFC2047_SUBJECT,
277 RFC2047_ADDRESS,
280 static int is_rfc2047_special(char ch, enum rfc2047_type type)
283 * rfc2047, section 4.2:
285 * 8-bit values which correspond to printable ASCII characters other
286 * than "=", "?", and "_" (underscore), MAY be represented as those
287 * characters. (But see section 5 for restrictions.) In
288 * particular, SPACE and TAB MUST NOT be represented as themselves
289 * within encoded words.
293 * rule out non-ASCII characters and non-printable characters (the
294 * non-ASCII check should be redundant as isprint() is not localized
295 * and only knows about ASCII, but be defensive about that)
297 if (non_ascii(ch) || !isprint(ch))
298 return 1;
301 * rule out special printable characters (' ' should be the only
302 * whitespace character considered printable, but be defensive and use
303 * isspace())
305 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
306 return 1;
309 * rfc2047, section 5.3:
311 * As a replacement for a 'word' entity within a 'phrase', for example,
312 * one that precedes an address in a From, To, or Cc header. The ABNF
313 * definition for 'phrase' from RFC 822 thus becomes:
315 * phrase = 1*( encoded-word / word )
317 * In this case the set of characters that may be used in a "Q"-encoded
318 * 'encoded-word' is restricted to: <upper and lower case ASCII
319 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
320 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
321 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
322 * 'special' by 'linear-white-space'.
325 if (type != RFC2047_ADDRESS)
326 return 0;
328 /* '=' and '_' are special cases and have been checked above */
329 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
332 static int needs_rfc2047_encoding(const char *line, int len,
333 enum rfc2047_type type)
335 int i;
337 for (i = 0; i < len; i++) {
338 int ch = line[i];
339 if (non_ascii(ch) || ch == '\n')
340 return 1;
341 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
342 return 1;
345 return 0;
348 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
349 const char *encoding, enum rfc2047_type type)
351 static const int max_encoded_length = 76; /* per rfc2047 */
352 int i;
353 int line_len = last_line_length(sb);
355 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
356 strbuf_addf(sb, "=?%s?q?", encoding);
357 line_len += strlen(encoding) + 5; /* 5 for =??q? */
359 while (len) {
361 * RFC 2047, section 5 (3):
363 * Each 'encoded-word' MUST represent an integral number of
364 * characters. A multi-octet character may not be split across
365 * adjacent 'encoded- word's.
367 const unsigned char *p = (const unsigned char *)line;
368 int chrlen = mbs_chrlen(&line, &len, encoding);
369 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
371 /* "=%02X" * chrlen, or the byte itself */
372 const char *encoded_fmt = is_special ? "=%02X" : "%c";
373 int encoded_len = is_special ? 3 * chrlen : 1;
376 * According to RFC 2047, we could encode the special character
377 * ' ' (space) with '_' (underscore) for readability. But many
378 * programs do not understand this and just leave the
379 * underscore in place. Thus, we do nothing special here, which
380 * causes ' ' to be encoded as '=20', avoiding this problem.
383 if (line_len + encoded_len + 2 > max_encoded_length) {
384 /* It won't fit with trailing "?=" --- break the line */
385 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
386 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
389 for (i = 0; i < chrlen; i++)
390 strbuf_addf(sb, encoded_fmt, p[i]);
391 line_len += encoded_len;
393 strbuf_addstr(sb, "?=");
396 static const char *show_ident_date(const struct ident_split *ident,
397 enum date_mode mode)
399 unsigned long date = 0;
400 long tz = 0;
402 if (ident->date_begin && ident->date_end)
403 date = strtoul(ident->date_begin, NULL, 10);
404 if (date_overflows(date))
405 date = 0;
406 else {
407 if (ident->tz_begin && ident->tz_end)
408 tz = strtol(ident->tz_begin, NULL, 10);
409 if (tz >= INT_MAX || tz <= INT_MIN)
410 tz = 0;
412 return show_date(date, tz, mode);
415 void pp_user_info(const struct pretty_print_context *pp,
416 const char *what, struct strbuf *sb,
417 const char *line, const char *encoding)
419 struct strbuf name;
420 struct strbuf mail;
421 struct ident_split ident;
422 int linelen;
423 char *line_end;
424 const char *mailbuf, *namebuf;
425 size_t namelen, maillen;
426 int max_length = 78; /* per rfc2822 */
428 if (pp->fmt == CMIT_FMT_ONELINE)
429 return;
431 line_end = strchr(line, '\n');
432 if (!line_end) {
433 line_end = strchr(line, '\0');
434 if (!line_end)
435 return;
438 linelen = ++line_end - line;
439 if (split_ident_line(&ident, line, linelen))
440 return;
443 mailbuf = ident.mail_begin;
444 maillen = ident.mail_end - ident.mail_begin;
445 namebuf = ident.name_begin;
446 namelen = ident.name_end - ident.name_begin;
448 if (pp->mailmap)
449 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
451 strbuf_init(&mail, 0);
452 strbuf_init(&name, 0);
454 strbuf_add(&mail, mailbuf, maillen);
455 strbuf_add(&name, namebuf, namelen);
457 namelen = name.len + mail.len + 3; /* ' ' + '<' + '>' */
459 if (pp->fmt == CMIT_FMT_EMAIL) {
460 strbuf_addstr(sb, "From: ");
461 if (needs_rfc2047_encoding(name.buf, name.len, RFC2047_ADDRESS)) {
462 add_rfc2047(sb, name.buf, name.len,
463 encoding, RFC2047_ADDRESS);
464 max_length = 76; /* per rfc2047 */
465 } else if (needs_rfc822_quoting(name.buf, name.len)) {
466 struct strbuf quoted = STRBUF_INIT;
467 add_rfc822_quoted(&quoted, name.buf, name.len);
468 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
469 -6, 1, max_length);
470 strbuf_release(&quoted);
471 } else {
472 strbuf_add_wrapped_bytes(sb, name.buf, name.len,
473 -6, 1, max_length);
475 if (namelen - name.len + last_line_length(sb) > max_length)
476 strbuf_addch(sb, '\n');
478 strbuf_addf(sb, " <%s>\n", mail.buf);
479 } else {
480 strbuf_addf(sb, "%s: %.*s%s <%s>\n", what,
481 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0,
482 " ", name.buf, mail.buf);
485 strbuf_release(&mail);
486 strbuf_release(&name);
488 switch (pp->fmt) {
489 case CMIT_FMT_MEDIUM:
490 strbuf_addf(sb, "Date: %s\n",
491 show_ident_date(&ident, pp->date_mode));
492 break;
493 case CMIT_FMT_EMAIL:
494 strbuf_addf(sb, "Date: %s\n",
495 show_ident_date(&ident, DATE_RFC2822));
496 break;
497 case CMIT_FMT_FULLER:
498 strbuf_addf(sb, "%sDate: %s\n", what,
499 show_ident_date(&ident, pp->date_mode));
500 break;
501 default:
502 /* notin' */
503 break;
507 static int is_empty_line(const char *line, int *len_p)
509 int len = *len_p;
510 while (len && isspace(line[len-1]))
511 len--;
512 *len_p = len;
513 return !len;
516 static const char *skip_empty_lines(const char *msg)
518 for (;;) {
519 int linelen = get_one_line(msg);
520 int ll = linelen;
521 if (!linelen)
522 break;
523 if (!is_empty_line(msg, &ll))
524 break;
525 msg += linelen;
527 return msg;
530 static void add_merge_info(const struct pretty_print_context *pp,
531 struct strbuf *sb, const struct commit *commit)
533 struct commit_list *parent = commit->parents;
535 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
536 !parent || !parent->next)
537 return;
539 strbuf_addstr(sb, "Merge:");
541 while (parent) {
542 struct commit *p = parent->item;
543 const char *hex = NULL;
544 if (pp->abbrev)
545 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
546 if (!hex)
547 hex = sha1_to_hex(p->object.sha1);
548 parent = parent->next;
550 strbuf_addf(sb, " %s", hex);
552 strbuf_addch(sb, '\n');
555 static char *get_header(const struct commit *commit, const char *msg,
556 const char *key)
558 int key_len = strlen(key);
559 const char *line = msg;
561 while (line) {
562 const char *eol = strchr(line, '\n'), *next;
564 if (line == eol)
565 return NULL;
566 if (!eol) {
567 warning("malformed commit (header is missing newline): %s",
568 sha1_to_hex(commit->object.sha1));
569 eol = line + strlen(line);
570 next = NULL;
571 } else
572 next = eol + 1;
573 if (eol - line > key_len &&
574 !strncmp(line, key, key_len) &&
575 line[key_len] == ' ') {
576 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
578 line = next;
580 return NULL;
583 static char *replace_encoding_header(char *buf, const char *encoding)
585 struct strbuf tmp = STRBUF_INIT;
586 size_t start, len;
587 char *cp = buf;
589 /* guess if there is an encoding header before a \n\n */
590 while (strncmp(cp, "encoding ", strlen("encoding "))) {
591 cp = strchr(cp, '\n');
592 if (!cp || *++cp == '\n')
593 return buf;
595 start = cp - buf;
596 cp = strchr(cp, '\n');
597 if (!cp)
598 return buf; /* should not happen but be defensive */
599 len = cp + 1 - (buf + start);
601 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
602 if (is_encoding_utf8(encoding)) {
603 /* we have re-coded to UTF-8; drop the header */
604 strbuf_remove(&tmp, start, len);
605 } else {
606 /* just replaces XXXX in 'encoding XXXX\n' */
607 strbuf_splice(&tmp, start + strlen("encoding "),
608 len - strlen("encoding \n"),
609 encoding, strlen(encoding));
611 return strbuf_detach(&tmp, NULL);
614 char *logmsg_reencode(const struct commit *commit,
615 const char *output_encoding)
617 static const char *utf8 = "UTF-8";
618 const char *use_encoding;
619 char *encoding;
620 char *msg = commit->buffer;
621 char *out;
623 if (!msg) {
624 enum object_type type;
625 unsigned long size;
627 msg = read_sha1_file(commit->object.sha1, &type, &size);
628 if (!msg)
629 die("Cannot read commit object %s",
630 sha1_to_hex(commit->object.sha1));
631 if (type != OBJ_COMMIT)
632 die("Expected commit for '%s', got %s",
633 sha1_to_hex(commit->object.sha1), typename(type));
636 if (!output_encoding || !*output_encoding)
637 return msg;
638 encoding = get_header(commit, msg, "encoding");
639 use_encoding = encoding ? encoding : utf8;
640 if (same_encoding(use_encoding, output_encoding)) {
642 * No encoding work to be done. If we have no encoding header
643 * at all, then there's nothing to do, and we can return the
644 * message verbatim (whether newly allocated or not).
646 if (!encoding)
647 return msg;
650 * Otherwise, we still want to munge the encoding header in the
651 * result, which will be done by modifying the buffer. If we
652 * are using a fresh copy, we can reuse it. But if we are using
653 * the cached copy from commit->buffer, we need to duplicate it
654 * to avoid munging commit->buffer.
656 out = msg;
657 if (out == commit->buffer)
658 out = xstrdup(out);
660 else {
662 * There's actual encoding work to do. Do the reencoding, which
663 * still leaves the header to be replaced in the next step. At
664 * this point, we are done with msg. If we allocated a fresh
665 * copy, we can free it.
667 out = reencode_string(msg, output_encoding, use_encoding);
668 if (out && msg != commit->buffer)
669 free(msg);
673 * This replacement actually consumes the buffer we hand it, so we do
674 * not have to worry about freeing the old "out" here.
676 if (out)
677 out = replace_encoding_header(out, output_encoding);
679 free(encoding);
681 * If the re-encoding failed, out might be NULL here; in that
682 * case we just return the commit message verbatim.
684 return out ? out : msg;
687 void logmsg_free(char *msg, const struct commit *commit)
689 if (msg != commit->buffer)
690 free(msg);
693 static int mailmap_name(const char **email, size_t *email_len,
694 const char **name, size_t *name_len)
696 static struct string_list *mail_map;
697 if (!mail_map) {
698 mail_map = xcalloc(1, sizeof(*mail_map));
699 read_mailmap(mail_map, NULL);
701 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
704 static size_t format_person_part(struct strbuf *sb, char part,
705 const char *msg, int len, enum date_mode dmode)
707 /* currently all placeholders have same length */
708 const int placeholder_len = 2;
709 struct ident_split s;
710 const char *name, *mail;
711 size_t maillen, namelen;
713 if (split_ident_line(&s, msg, len) < 0)
714 goto skip;
716 name = s.name_begin;
717 namelen = s.name_end - s.name_begin;
718 mail = s.mail_begin;
719 maillen = s.mail_end - s.mail_begin;
721 if (part == 'N' || part == 'E') /* mailmap lookup */
722 mailmap_name(&mail, &maillen, &name, &namelen);
723 if (part == 'n' || part == 'N') { /* name */
724 strbuf_add(sb, name, namelen);
725 return placeholder_len;
727 if (part == 'e' || part == 'E') { /* email */
728 strbuf_add(sb, mail, maillen);
729 return placeholder_len;
732 if (!s.date_begin)
733 goto skip;
735 if (part == 't') { /* date, UNIX timestamp */
736 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
737 return placeholder_len;
740 switch (part) {
741 case 'd': /* date */
742 strbuf_addstr(sb, show_ident_date(&s, dmode));
743 return placeholder_len;
744 case 'D': /* date, RFC2822 style */
745 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
746 return placeholder_len;
747 case 'r': /* date, relative */
748 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
749 return placeholder_len;
750 case 'i': /* date, ISO 8601 */
751 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
752 return placeholder_len;
755 skip:
757 * reading from either a bogus commit, or a reflog entry with
758 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
759 * to compute a valid return value.
761 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
762 || part == 'D' || part == 'r' || part == 'i')
763 return placeholder_len;
765 return 0; /* unknown placeholder */
768 struct chunk {
769 size_t off;
770 size_t len;
773 struct format_commit_context {
774 const struct commit *commit;
775 const struct pretty_print_context *pretty_ctx;
776 unsigned commit_header_parsed:1;
777 unsigned commit_message_parsed:1;
778 unsigned commit_signature_parsed:1;
779 struct {
780 char *gpg_output;
781 char *gpg_status;
782 char good_bad;
783 char *signer;
784 char *key;
785 } signature;
786 char *message;
787 size_t width, indent1, indent2;
789 /* These offsets are relative to the start of the commit message. */
790 struct chunk author;
791 struct chunk committer;
792 struct chunk encoding;
793 size_t message_off;
794 size_t subject_off;
795 size_t body_off;
797 /* The following ones are relative to the result struct strbuf. */
798 struct chunk abbrev_commit_hash;
799 struct chunk abbrev_tree_hash;
800 struct chunk abbrev_parent_hashes;
801 size_t wrap_start;
804 static int add_again(struct strbuf *sb, struct chunk *chunk)
806 if (chunk->len) {
807 strbuf_adddup(sb, chunk->off, chunk->len);
808 return 1;
812 * We haven't seen this chunk before. Our caller is surely
813 * going to add it the hard way now. Remember the most likely
814 * start of the to-be-added chunk: the current end of the
815 * struct strbuf.
817 chunk->off = sb->len;
818 return 0;
821 static void parse_commit_header(struct format_commit_context *context)
823 const char *msg = context->message;
824 int i;
826 for (i = 0; msg[i]; i++) {
827 int eol;
828 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
829 ; /* do nothing */
831 if (i == eol) {
832 break;
833 } else if (!prefixcmp(msg + i, "author ")) {
834 context->author.off = i + 7;
835 context->author.len = eol - i - 7;
836 } else if (!prefixcmp(msg + i, "committer ")) {
837 context->committer.off = i + 10;
838 context->committer.len = eol - i - 10;
839 } else if (!prefixcmp(msg + i, "encoding ")) {
840 context->encoding.off = i + 9;
841 context->encoding.len = eol - i - 9;
843 i = eol;
845 context->message_off = i;
846 context->commit_header_parsed = 1;
849 static int istitlechar(char c)
851 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
852 (c >= '0' && c <= '9') || c == '.' || c == '_';
855 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
857 size_t trimlen;
858 size_t start_len = sb->len;
859 int space = 2;
861 for (; *msg && *msg != '\n'; msg++) {
862 if (istitlechar(*msg)) {
863 if (space == 1)
864 strbuf_addch(sb, '-');
865 space = 0;
866 strbuf_addch(sb, *msg);
867 if (*msg == '.')
868 while (*(msg+1) == '.')
869 msg++;
870 } else
871 space |= 1;
874 /* trim any trailing '.' or '-' characters */
875 trimlen = 0;
876 while (sb->len - trimlen > start_len &&
877 (sb->buf[sb->len - 1 - trimlen] == '.'
878 || sb->buf[sb->len - 1 - trimlen] == '-'))
879 trimlen++;
880 strbuf_remove(sb, sb->len - trimlen, trimlen);
883 const char *format_subject(struct strbuf *sb, const char *msg,
884 const char *line_separator)
886 int first = 1;
888 for (;;) {
889 const char *line = msg;
890 int linelen = get_one_line(line);
892 msg += linelen;
893 if (!linelen || is_empty_line(line, &linelen))
894 break;
896 if (!sb)
897 continue;
898 strbuf_grow(sb, linelen + 2);
899 if (!first)
900 strbuf_addstr(sb, line_separator);
901 strbuf_add(sb, line, linelen);
902 first = 0;
904 return msg;
907 static void parse_commit_message(struct format_commit_context *c)
909 const char *msg = c->message + c->message_off;
910 const char *start = c->message;
912 msg = skip_empty_lines(msg);
913 c->subject_off = msg - start;
915 msg = format_subject(NULL, msg, NULL);
916 msg = skip_empty_lines(msg);
917 c->body_off = msg - start;
919 c->commit_message_parsed = 1;
922 static void format_decoration(struct strbuf *sb, const struct commit *commit)
924 struct name_decoration *d;
925 const char *prefix = " (";
927 load_ref_decorations(DECORATE_SHORT_REFS);
928 d = lookup_decoration(&name_decoration, &commit->object);
929 while (d) {
930 strbuf_addstr(sb, prefix);
931 prefix = ", ";
932 strbuf_addstr(sb, d->name);
933 d = d->next;
935 if (prefix[0] == ',')
936 strbuf_addch(sb, ')');
939 static void strbuf_wrap(struct strbuf *sb, size_t pos,
940 size_t width, size_t indent1, size_t indent2)
942 struct strbuf tmp = STRBUF_INIT;
944 if (pos)
945 strbuf_add(&tmp, sb->buf, pos);
946 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
947 (int) indent1, (int) indent2, (int) width);
948 strbuf_swap(&tmp, sb);
949 strbuf_release(&tmp);
952 static void rewrap_message_tail(struct strbuf *sb,
953 struct format_commit_context *c,
954 size_t new_width, size_t new_indent1,
955 size_t new_indent2)
957 if (c->width == new_width && c->indent1 == new_indent1 &&
958 c->indent2 == new_indent2)
959 return;
960 if (c->wrap_start < sb->len)
961 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
962 c->wrap_start = sb->len;
963 c->width = new_width;
964 c->indent1 = new_indent1;
965 c->indent2 = new_indent2;
968 static struct {
969 char result;
970 const char *check;
971 } signature_check[] = {
972 { 'G', "\n[GNUPG:] GOODSIG " },
973 { 'B', "\n[GNUPG:] BADSIG " },
976 static void parse_signature_lines(struct format_commit_context *ctx)
978 const char *buf = ctx->signature.gpg_status;
979 int i;
981 for (i = 0; i < ARRAY_SIZE(signature_check); i++) {
982 const char *found = strstr(buf, signature_check[i].check);
983 const char *next;
984 if (!found)
985 continue;
986 ctx->signature.good_bad = signature_check[i].result;
987 found += strlen(signature_check[i].check);
988 ctx->signature.key = xmemdupz(found, 16);
989 found += 17;
990 next = strchrnul(found, '\n');
991 ctx->signature.signer = xmemdupz(found, next - found);
992 break;
996 static void parse_commit_signature(struct format_commit_context *ctx)
998 struct strbuf payload = STRBUF_INIT;
999 struct strbuf signature = STRBUF_INIT;
1000 struct strbuf gpg_output = STRBUF_INIT;
1001 struct strbuf gpg_status = STRBUF_INIT;
1002 int status;
1004 ctx->commit_signature_parsed = 1;
1006 if (parse_signed_commit(ctx->commit->object.sha1,
1007 &payload, &signature) <= 0)
1008 goto out;
1009 status = verify_signed_buffer(payload.buf, payload.len,
1010 signature.buf, signature.len,
1011 &gpg_output, &gpg_status);
1012 if (status && !gpg_output.len)
1013 goto out;
1014 ctx->signature.gpg_output = strbuf_detach(&gpg_output, NULL);
1015 ctx->signature.gpg_status = strbuf_detach(&gpg_status, NULL);
1016 parse_signature_lines(ctx);
1018 out:
1019 strbuf_release(&gpg_status);
1020 strbuf_release(&gpg_output);
1021 strbuf_release(&payload);
1022 strbuf_release(&signature);
1026 static int format_reflog_person(struct strbuf *sb,
1027 char part,
1028 struct reflog_walk_info *log,
1029 enum date_mode dmode)
1031 const char *ident;
1033 if (!log)
1034 return 2;
1036 ident = get_reflog_ident(log);
1037 if (!ident)
1038 return 2;
1040 return format_person_part(sb, part, ident, strlen(ident), dmode);
1043 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
1044 void *context)
1046 struct format_commit_context *c = context;
1047 const struct commit *commit = c->commit;
1048 const char *msg = c->message;
1049 struct commit_list *p;
1050 int h1, h2;
1052 /* these are independent of the commit */
1053 switch (placeholder[0]) {
1054 case 'C':
1055 if (placeholder[1] == '(') {
1056 const char *begin = placeholder + 2;
1057 const char *end = strchr(begin, ')');
1058 char color[COLOR_MAXLEN];
1060 if (!end)
1061 return 0;
1062 if (!prefixcmp(begin, "auto,")) {
1063 if (!want_color(c->pretty_ctx->color))
1064 return end - placeholder + 1;
1065 begin += 5;
1067 color_parse_mem(begin,
1068 end - begin,
1069 "--pretty format", color);
1070 strbuf_addstr(sb, color);
1071 return end - placeholder + 1;
1073 if (!prefixcmp(placeholder + 1, "red")) {
1074 strbuf_addstr(sb, GIT_COLOR_RED);
1075 return 4;
1076 } else if (!prefixcmp(placeholder + 1, "green")) {
1077 strbuf_addstr(sb, GIT_COLOR_GREEN);
1078 return 6;
1079 } else if (!prefixcmp(placeholder + 1, "blue")) {
1080 strbuf_addstr(sb, GIT_COLOR_BLUE);
1081 return 5;
1082 } else if (!prefixcmp(placeholder + 1, "reset")) {
1083 strbuf_addstr(sb, GIT_COLOR_RESET);
1084 return 6;
1085 } else
1086 return 0;
1087 case 'n': /* newline */
1088 strbuf_addch(sb, '\n');
1089 return 1;
1090 case 'x':
1091 /* %x00 == NUL, %x0a == LF, etc. */
1092 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1093 h1 <= 16 &&
1094 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1095 h2 <= 16) {
1096 strbuf_addch(sb, (h1<<4)|h2);
1097 return 3;
1098 } else
1099 return 0;
1100 case 'w':
1101 if (placeholder[1] == '(') {
1102 unsigned long width = 0, indent1 = 0, indent2 = 0;
1103 char *next;
1104 const char *start = placeholder + 2;
1105 const char *end = strchr(start, ')');
1106 if (!end)
1107 return 0;
1108 if (end > start) {
1109 width = strtoul(start, &next, 10);
1110 if (*next == ',') {
1111 indent1 = strtoul(next + 1, &next, 10);
1112 if (*next == ',') {
1113 indent2 = strtoul(next + 1,
1114 &next, 10);
1117 if (*next != ')')
1118 return 0;
1120 rewrap_message_tail(sb, c, width, indent1, indent2);
1121 return end - placeholder + 1;
1122 } else
1123 return 0;
1126 /* these depend on the commit */
1127 if (!commit->object.parsed)
1128 parse_object(commit->object.sha1);
1130 switch (placeholder[0]) {
1131 case 'H': /* commit hash */
1132 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1133 return 1;
1134 case 'h': /* abbreviated commit hash */
1135 if (add_again(sb, &c->abbrev_commit_hash))
1136 return 1;
1137 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1138 c->pretty_ctx->abbrev));
1139 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1140 return 1;
1141 case 'T': /* tree hash */
1142 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1143 return 1;
1144 case 't': /* abbreviated tree hash */
1145 if (add_again(sb, &c->abbrev_tree_hash))
1146 return 1;
1147 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1148 c->pretty_ctx->abbrev));
1149 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1150 return 1;
1151 case 'P': /* parent hashes */
1152 for (p = commit->parents; p; p = p->next) {
1153 if (p != commit->parents)
1154 strbuf_addch(sb, ' ');
1155 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1157 return 1;
1158 case 'p': /* abbreviated parent hashes */
1159 if (add_again(sb, &c->abbrev_parent_hashes))
1160 return 1;
1161 for (p = commit->parents; p; p = p->next) {
1162 if (p != commit->parents)
1163 strbuf_addch(sb, ' ');
1164 strbuf_addstr(sb, find_unique_abbrev(
1165 p->item->object.sha1,
1166 c->pretty_ctx->abbrev));
1168 c->abbrev_parent_hashes.len = sb->len -
1169 c->abbrev_parent_hashes.off;
1170 return 1;
1171 case 'm': /* left/right/bottom */
1172 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1173 return 1;
1174 case 'd':
1175 format_decoration(sb, commit);
1176 return 1;
1177 case 'g': /* reflog info */
1178 switch(placeholder[1]) {
1179 case 'd': /* reflog selector */
1180 case 'D':
1181 if (c->pretty_ctx->reflog_info)
1182 get_reflog_selector(sb,
1183 c->pretty_ctx->reflog_info,
1184 c->pretty_ctx->date_mode,
1185 c->pretty_ctx->date_mode_explicit,
1186 (placeholder[1] == 'd'));
1187 return 2;
1188 case 's': /* reflog message */
1189 if (c->pretty_ctx->reflog_info)
1190 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1191 return 2;
1192 case 'n':
1193 case 'N':
1194 case 'e':
1195 case 'E':
1196 return format_reflog_person(sb,
1197 placeholder[1],
1198 c->pretty_ctx->reflog_info,
1199 c->pretty_ctx->date_mode);
1201 return 0; /* unknown %g placeholder */
1202 case 'N':
1203 if (c->pretty_ctx->notes_message) {
1204 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1205 return 1;
1207 return 0;
1210 if (placeholder[0] == 'G') {
1211 if (!c->commit_signature_parsed)
1212 parse_commit_signature(c);
1213 switch (placeholder[1]) {
1214 case 'G':
1215 if (c->signature.gpg_output)
1216 strbuf_addstr(sb, c->signature.gpg_output);
1217 break;
1218 case '?':
1219 switch (c->signature.good_bad) {
1220 case 'G':
1221 case 'B':
1222 strbuf_addch(sb, c->signature.good_bad);
1224 break;
1225 case 'S':
1226 if (c->signature.signer)
1227 strbuf_addstr(sb, c->signature.signer);
1228 break;
1229 case 'K':
1230 if (c->signature.key)
1231 strbuf_addstr(sb, c->signature.key);
1232 break;
1234 return 2;
1238 /* For the rest we have to parse the commit header. */
1239 if (!c->commit_header_parsed)
1240 parse_commit_header(c);
1242 switch (placeholder[0]) {
1243 case 'a': /* author ... */
1244 return format_person_part(sb, placeholder[1],
1245 msg + c->author.off, c->author.len,
1246 c->pretty_ctx->date_mode);
1247 case 'c': /* committer ... */
1248 return format_person_part(sb, placeholder[1],
1249 msg + c->committer.off, c->committer.len,
1250 c->pretty_ctx->date_mode);
1251 case 'e': /* encoding */
1252 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1253 return 1;
1254 case 'B': /* raw body */
1255 /* message_off is always left at the initial newline */
1256 strbuf_addstr(sb, msg + c->message_off + 1);
1257 return 1;
1260 /* Now we need to parse the commit message. */
1261 if (!c->commit_message_parsed)
1262 parse_commit_message(c);
1264 switch (placeholder[0]) {
1265 case 's': /* subject */
1266 format_subject(sb, msg + c->subject_off, " ");
1267 return 1;
1268 case 'f': /* sanitized subject */
1269 format_sanitized_subject(sb, msg + c->subject_off);
1270 return 1;
1271 case 'b': /* body */
1272 strbuf_addstr(sb, msg + c->body_off);
1273 return 1;
1275 return 0; /* unknown placeholder */
1278 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1279 void *context)
1281 int consumed;
1282 size_t orig_len;
1283 enum {
1284 NO_MAGIC,
1285 ADD_LF_BEFORE_NON_EMPTY,
1286 DEL_LF_BEFORE_EMPTY,
1287 ADD_SP_BEFORE_NON_EMPTY
1288 } magic = NO_MAGIC;
1290 switch (placeholder[0]) {
1291 case '-':
1292 magic = DEL_LF_BEFORE_EMPTY;
1293 break;
1294 case '+':
1295 magic = ADD_LF_BEFORE_NON_EMPTY;
1296 break;
1297 case ' ':
1298 magic = ADD_SP_BEFORE_NON_EMPTY;
1299 break;
1300 default:
1301 break;
1303 if (magic != NO_MAGIC)
1304 placeholder++;
1306 orig_len = sb->len;
1307 consumed = format_commit_one(sb, placeholder, context);
1308 if (magic == NO_MAGIC)
1309 return consumed;
1311 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1312 while (sb->len && sb->buf[sb->len - 1] == '\n')
1313 strbuf_setlen(sb, sb->len - 1);
1314 } else if (orig_len != sb->len) {
1315 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1316 strbuf_insert(sb, orig_len, "\n", 1);
1317 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1318 strbuf_insert(sb, orig_len, " ", 1);
1320 return consumed + 1;
1323 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1324 void *context)
1326 struct userformat_want *w = context;
1328 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1329 placeholder++;
1331 switch (*placeholder) {
1332 case 'N':
1333 w->notes = 1;
1334 break;
1336 return 0;
1339 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1341 struct strbuf dummy = STRBUF_INIT;
1343 if (!fmt) {
1344 if (!user_format)
1345 return;
1346 fmt = user_format;
1348 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1349 strbuf_release(&dummy);
1352 void format_commit_message(const struct commit *commit,
1353 const char *format, struct strbuf *sb,
1354 const struct pretty_print_context *pretty_ctx)
1356 struct format_commit_context context;
1357 const char *output_enc = pretty_ctx->output_encoding;
1359 memset(&context, 0, sizeof(context));
1360 context.commit = commit;
1361 context.pretty_ctx = pretty_ctx;
1362 context.wrap_start = sb->len;
1363 context.message = logmsg_reencode(commit, output_enc);
1365 strbuf_expand(sb, format, format_commit_item, &context);
1366 rewrap_message_tail(sb, &context, 0, 0, 0);
1368 logmsg_free(context.message, commit);
1369 free(context.signature.gpg_output);
1370 free(context.signature.signer);
1373 static void pp_header(const struct pretty_print_context *pp,
1374 const char *encoding,
1375 const struct commit *commit,
1376 const char **msg_p,
1377 struct strbuf *sb)
1379 int parents_shown = 0;
1381 for (;;) {
1382 const char *line = *msg_p;
1383 int linelen = get_one_line(*msg_p);
1385 if (!linelen)
1386 return;
1387 *msg_p += linelen;
1389 if (linelen == 1)
1390 /* End of header */
1391 return;
1393 if (pp->fmt == CMIT_FMT_RAW) {
1394 strbuf_add(sb, line, linelen);
1395 continue;
1398 if (!prefixcmp(line, "parent ")) {
1399 if (linelen != 48)
1400 die("bad parent line in commit");
1401 continue;
1404 if (!parents_shown) {
1405 struct commit_list *parent;
1406 int num;
1407 for (parent = commit->parents, num = 0;
1408 parent;
1409 parent = parent->next, num++)
1411 /* with enough slop */
1412 strbuf_grow(sb, num * 50 + 20);
1413 add_merge_info(pp, sb, commit);
1414 parents_shown = 1;
1418 * MEDIUM == DEFAULT shows only author with dates.
1419 * FULL shows both authors but not dates.
1420 * FULLER shows both authors and dates.
1422 if (!prefixcmp(line, "author ")) {
1423 strbuf_grow(sb, linelen + 80);
1424 pp_user_info(pp, "Author", sb, line + 7, encoding);
1426 if (!prefixcmp(line, "committer ") &&
1427 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1428 strbuf_grow(sb, linelen + 80);
1429 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1434 void pp_title_line(const struct pretty_print_context *pp,
1435 const char **msg_p,
1436 struct strbuf *sb,
1437 const char *encoding,
1438 int need_8bit_cte)
1440 static const int max_length = 78; /* per rfc2047 */
1441 struct strbuf title;
1443 strbuf_init(&title, 80);
1444 *msg_p = format_subject(&title, *msg_p,
1445 pp->preserve_subject ? "\n" : " ");
1447 strbuf_grow(sb, title.len + 1024);
1448 if (pp->subject) {
1449 strbuf_addstr(sb, pp->subject);
1450 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1451 add_rfc2047(sb, title.buf, title.len,
1452 encoding, RFC2047_SUBJECT);
1453 else
1454 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1455 -last_line_length(sb), 1, max_length);
1456 } else {
1457 strbuf_addbuf(sb, &title);
1459 strbuf_addch(sb, '\n');
1461 if (need_8bit_cte > 0) {
1462 const char *header_fmt =
1463 "MIME-Version: 1.0\n"
1464 "Content-Type: text/plain; charset=%s\n"
1465 "Content-Transfer-Encoding: 8bit\n";
1466 strbuf_addf(sb, header_fmt, encoding);
1468 if (pp->after_subject) {
1469 strbuf_addstr(sb, pp->after_subject);
1471 if (pp->fmt == CMIT_FMT_EMAIL) {
1472 strbuf_addch(sb, '\n');
1474 strbuf_release(&title);
1477 void pp_remainder(const struct pretty_print_context *pp,
1478 const char **msg_p,
1479 struct strbuf *sb,
1480 int indent)
1482 int first = 1;
1483 for (;;) {
1484 const char *line = *msg_p;
1485 int linelen = get_one_line(line);
1486 *msg_p += linelen;
1488 if (!linelen)
1489 break;
1491 if (is_empty_line(line, &linelen)) {
1492 if (first)
1493 continue;
1494 if (pp->fmt == CMIT_FMT_SHORT)
1495 break;
1497 first = 0;
1499 strbuf_grow(sb, linelen + indent + 20);
1500 if (indent) {
1501 memset(sb->buf + sb->len, ' ', indent);
1502 strbuf_setlen(sb, sb->len + indent);
1504 strbuf_add(sb, line, linelen);
1505 strbuf_addch(sb, '\n');
1509 void pretty_print_commit(const struct pretty_print_context *pp,
1510 const struct commit *commit,
1511 struct strbuf *sb)
1513 unsigned long beginning_of_body;
1514 int indent = 4;
1515 const char *msg;
1516 char *reencoded;
1517 const char *encoding;
1518 int need_8bit_cte = pp->need_8bit_cte;
1520 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1521 format_commit_message(commit, user_format, sb, pp);
1522 return;
1525 encoding = get_log_output_encoding();
1526 msg = reencoded = logmsg_reencode(commit, encoding);
1528 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1529 indent = 0;
1532 * We need to check and emit Content-type: to mark it
1533 * as 8-bit if we haven't done so.
1535 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1536 int i, ch, in_body;
1538 for (in_body = i = 0; (ch = msg[i]); i++) {
1539 if (!in_body) {
1540 /* author could be non 7-bit ASCII but
1541 * the log may be so; skip over the
1542 * header part first.
1544 if (ch == '\n' && msg[i+1] == '\n')
1545 in_body = 1;
1547 else if (non_ascii(ch)) {
1548 need_8bit_cte = 1;
1549 break;
1554 pp_header(pp, encoding, commit, &msg, sb);
1555 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1556 strbuf_addch(sb, '\n');
1559 /* Skip excess blank lines at the beginning of body, if any... */
1560 msg = skip_empty_lines(msg);
1562 /* These formats treat the title line specially. */
1563 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1564 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1566 beginning_of_body = sb->len;
1567 if (pp->fmt != CMIT_FMT_ONELINE)
1568 pp_remainder(pp, &msg, sb, indent);
1569 strbuf_rtrim(sb);
1571 /* Make sure there is an EOLN for the non-oneline case */
1572 if (pp->fmt != CMIT_FMT_ONELINE)
1573 strbuf_addch(sb, '\n');
1576 * The caller may append additional body text in e-mail
1577 * format. Make sure we did not strip the blank line
1578 * between the header and the body.
1580 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1581 strbuf_addch(sb, '\n');
1583 logmsg_free(reencoded, commit);
1586 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1587 struct strbuf *sb)
1589 struct pretty_print_context pp = {0};
1590 pp.fmt = fmt;
1591 pretty_print_commit(&pp, commit, sb);