?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
diff-highlight: do not split multibyte characters
[git/gitweb.git] / pretty.c
blob6182ca9aed79503a4c24ed267b8cf69673cac16b
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"
13
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);
26
27 int commit_format_is_empty(enum cmit_fmt fmt)
28 {
29 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
30 }
31
32 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
33 {
34 free(user_format);
35 user_format = xstrdup(cp);
36 if (is_tformat)
37 rev->use_terminator = 1;
38 rev->commit_format = CMIT_FMT_USERFORMAT;
39 }
40
41 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
42 {
43 struct cmt_fmt_map *commit_format = NULL;
44 const char *name;
45 const char *fmt;
46 int i;
47
48 if (!skip_prefix(var, "pretty.", &name))
49 return 0;
50
51 for (i = 0; i < builtin_formats_len; i++) {
52 if (!strcmp(commit_formats[i].name, name))
53 return 0;
54 }
55
56 for (i = builtin_formats_len; i < commit_formats_len; i++) {
57 if (!strcmp(commit_formats[i].name, name)) {
58 commit_format = &commit_formats[i];
59 break;
60 }
61 }
62
63 if (!commit_format) {
64 ALLOC_GROW(commit_formats, commit_formats_len+1,
65 commit_formats_alloc);
66 commit_format = &commit_formats[commit_formats_len];
67 memset(commit_format, 0, sizeof(*commit_format));
68 commit_formats_len++;
69 }
70
71 commit_format->name = xstrdup(name);
72 commit_format->format = CMIT_FMT_USERFORMAT;
73 git_config_string(&fmt, var, value);
74 if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
75 commit_format->is_tformat = fmt[0] == 't';
76 fmt = strchr(fmt, ':') + 1;
77 } else if (strchr(fmt, '%'))
78 commit_format->is_tformat = 1;
79 else
80 commit_format->is_alias = 1;
81 commit_format->user_format = fmt;
82
83 return 0;
84 }
85
86 static void setup_commit_formats(void)
87 {
88 struct cmt_fmt_map builtin_formats[] = {
89 { "raw", CMIT_FMT_RAW, 0 },
90 { "medium", CMIT_FMT_MEDIUM, 0 },
91 { "short", CMIT_FMT_SHORT, 0 },
92 { "email", CMIT_FMT_EMAIL, 0 },
93 { "fuller", CMIT_FMT_FULLER, 0 },
94 { "full", CMIT_FMT_FULL, 0 },
95 { "oneline", CMIT_FMT_ONELINE, 1 }
96 };
97 commit_formats_len = ARRAY_SIZE(builtin_formats);
98 builtin_formats_len = commit_formats_len;
99 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
100 memcpy(commit_formats, builtin_formats,
101 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
102
103 git_config(git_pretty_formats_config, NULL);
104 }
105
106 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
107 const char *original,
108 int num_redirections)
109 {
110 struct cmt_fmt_map *found = NULL;
111 size_t found_match_len = 0;
112 int i;
113
114 if (num_redirections >= commit_formats_len)
115 die("invalid --pretty format: "
116 "'%s' references an alias which points to itself",
117 original);
118
119 for (i = 0; i < commit_formats_len; i++) {
120 size_t match_len;
121
122 if (!starts_with(commit_formats[i].name, sought))
123 continue;
124
125 match_len = strlen(commit_formats[i].name);
126 if (found == NULL || found_match_len > match_len) {
127 found = &commit_formats[i];
128 found_match_len = match_len;
129 }
130 }
131
132 if (found && found->is_alias) {
133 found = find_commit_format_recursive(found->user_format,
134 original,
135 num_redirections+1);
136 }
137
138 return found;
139 }
140
141 static struct cmt_fmt_map *find_commit_format(const char *sought)
142 {
143 if (!commit_formats)
144 setup_commit_formats();
145
146 return find_commit_format_recursive(sought, sought, 0);
147 }
148
149 void get_commit_format(const char *arg, struct rev_info *rev)
150 {
151 struct cmt_fmt_map *commit_format;
152
153 rev->use_terminator = 0;
154 if (!arg) {
155 rev->commit_format = CMIT_FMT_DEFAULT;
156 return;
157 }
158 if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
159 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
160 return;
161 }
162
163 if (!*arg || strchr(arg, '%')) {
164 save_user_format(rev, arg, 1);
165 return;
166 }
167
168 commit_format = find_commit_format(arg);
169 if (!commit_format)
170 die("invalid --pretty format: %s", arg);
171
172 rev->commit_format = commit_format->format;
173 rev->use_terminator = commit_format->is_tformat;
174 if (commit_format->format == CMIT_FMT_USERFORMAT) {
175 save_user_format(rev, commit_format->user_format,
176 commit_format->is_tformat);
177 }
178 }
179
180 /*
181 * Generic support for pretty-printing the header
182 */
183 static int get_one_line(const char *msg)
184 {
185 int ret = 0;
186
187 for (;;) {
188 char c = *msg++;
189 if (!c)
190 break;
191 ret++;
192 if (c == '\n')
193 break;
194 }
195 return ret;
196 }
197
198 /* High bit set, or ISO-2022-INT */
199 static int non_ascii(int ch)
200 {
201 return !isascii(ch) || ch == '\033';
202 }
203
204 int has_non_ascii(const char *s)
205 {
206 int ch;
207 if (!s)
208 return 0;
209 while ((ch = *s++) != '\0') {
210 if (non_ascii(ch))
211 return 1;
212 }
213 return 0;
214 }
215
216 static int is_rfc822_special(char ch)
217 {
218 switch (ch) {
219 case '(':
220 case ')':
221 case '<':
222 case '>':
223 case '[':
224 case ']':
225 case ':':
226 case ';':
227 case '@':
228 case ',':
229 case '.':
230 case '"':
231 case '\\':
232 return 1;
233 default:
234 return 0;
235 }
236 }
237
238 static int needs_rfc822_quoting(const char *s, int len)
239 {
240 int i;
241 for (i = 0; i < len; i++)
242 if (is_rfc822_special(s[i]))
243 return 1;
244 return 0;
245 }
246
247 static int last_line_length(struct strbuf *sb)
248 {
249 int i;
250
251 /* How many bytes are already used on the last line? */
252 for (i = sb->len - 1; i >= 0; i--)
253 if (sb->buf[i] == '\n')
254 break;
255 return sb->len - (i + 1);
256 }
257
258 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
259 {
260 int i;
261
262 /* just a guess, we may have to also backslash-quote */
263 strbuf_grow(out, len + 2);
264
265 strbuf_addch(out, '"');
266 for (i = 0; i < len; i++) {
267 switch (s[i]) {
268 case '"':
269 case '\\':
270 strbuf_addch(out, '\\');
271 /* fall through */
272 default:
273 strbuf_addch(out, s[i]);
274 }
275 }
276 strbuf_addch(out, '"');
277 }
278
279 enum rfc2047_type {
280 RFC2047_SUBJECT,
281 RFC2047_ADDRESS
282 };
283
284 static int is_rfc2047_special(char ch, enum rfc2047_type type)
285 {
286 /*
287 * rfc2047, section 4.2:
288 *
289 * 8-bit values which correspond to printable ASCII characters other
290 * than "=", "?", and "_" (underscore), MAY be represented as those
291 * characters. (But see section 5 for restrictions.) In
292 * particular, SPACE and TAB MUST NOT be represented as themselves
293 * within encoded words.
294 */
295
296 /*
297 * rule out non-ASCII characters and non-printable characters (the
298 * non-ASCII check should be redundant as isprint() is not localized
299 * and only knows about ASCII, but be defensive about that)
300 */
301 if (non_ascii(ch) || !isprint(ch))
302 return 1;
303
304 /*
305 * rule out special printable characters (' ' should be the only
306 * whitespace character considered printable, but be defensive and use
307 * isspace())
308 */
309 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
310 return 1;
311
312 /*
313 * rfc2047, section 5.3:
314 *
315 * As a replacement for a 'word' entity within a 'phrase', for example,
316 * one that precedes an address in a From, To, or Cc header. The ABNF
317 * definition for 'phrase' from RFC 822 thus becomes:
318 *
319 * phrase = 1*( encoded-word / word )
320 *
321 * In this case the set of characters that may be used in a "Q"-encoded
322 * 'encoded-word' is restricted to: <upper and lower case ASCII
323 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
324 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
325 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
326 * 'special' by 'linear-white-space'.
327 */
328
329 if (type != RFC2047_ADDRESS)
330 return 0;
331
332 /* '=' and '_' are special cases and have been checked above */
333 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
334 }
335
336 static int needs_rfc2047_encoding(const char *line, int len,
337 enum rfc2047_type type)
338 {
339 int i;
340
341 for (i = 0; i < len; i++) {
342 int ch = line[i];
343 if (non_ascii(ch) || ch == '\n')
344 return 1;
345 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
346 return 1;
347 }
348
349 return 0;
350 }
351
352 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
353 const char *encoding, enum rfc2047_type type)
354 {
355 static const int max_encoded_length = 76; /* per rfc2047 */
356 int i;
357 int line_len = last_line_length(sb);
358
359 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
360 strbuf_addf(sb, "=?%s?q?", encoding);
361 line_len += strlen(encoding) + 5; /* 5 for =??q? */
362
363 while (len) {
364 /*
365 * RFC 2047, section 5 (3):
366 *
367 * Each 'encoded-word' MUST represent an integral number of
368 * characters. A multi-octet character may not be split across
369 * adjacent 'encoded- word's.
370 */
371 const unsigned char *p = (const unsigned char *)line;
372 int chrlen = mbs_chrlen(&line, &len, encoding);
373 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
374
375 /* "=%02X" * chrlen, or the byte itself */
376 const char *encoded_fmt = is_special ? "=%02X" : "%c";
377 int encoded_len = is_special ? 3 * chrlen : 1;
378
379 /*
380 * According to RFC 2047, we could encode the special character
381 * ' ' (space) with '_' (underscore) for readability. But many
382 * programs do not understand this and just leave the
383 * underscore in place. Thus, we do nothing special here, which
384 * causes ' ' to be encoded as '=20', avoiding this problem.
385 */
386
387 if (line_len + encoded_len + 2 > max_encoded_length) {
388 /* It won't fit with trailing "?=" --- break the line */
389 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
390 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
391 }
392
393 for (i = 0; i < chrlen; i++)
394 strbuf_addf(sb, encoded_fmt, p[i]);
395 line_len += encoded_len;
396 }
397 strbuf_addstr(sb, "?=");
398 }
399
400 const char *show_ident_date(const struct ident_split *ident,
401 enum date_mode mode)
402 {
403 unsigned long date = 0;
404 long tz = 0;
405
406 if (ident->date_begin && ident->date_end)
407 date = strtoul(ident->date_begin, NULL, 10);
408 if (date_overflows(date))
409 date = 0;
410 else {
411 if (ident->tz_begin && ident->tz_end)
412 tz = strtol(ident->tz_begin, NULL, 10);
413 if (tz >= INT_MAX || tz <= INT_MIN)
414 tz = 0;
415 }
416 return show_date(date, tz, mode);
417 }
418
419 void pp_user_info(struct pretty_print_context *pp,
420 const char *what, struct strbuf *sb,
421 const char *line, const char *encoding)
422 {
423 struct ident_split ident;
424 char *line_end;
425 const char *mailbuf, *namebuf;
426 size_t namelen, maillen;
427 int max_length = 78; /* per rfc2822 */
428
429 if (pp->fmt == CMIT_FMT_ONELINE)
430 return;
431
432 line_end = strchrnul(line, '\n');
433 if (split_ident_line(&ident, line, line_end - line))
434 return;
435
436 mailbuf = ident.mail_begin;
437 maillen = ident.mail_end - ident.mail_begin;
438 namebuf = ident.name_begin;
439 namelen = ident.name_end - ident.name_begin;
440
441 if (pp->mailmap)
442 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
443
444 if (pp->fmt == CMIT_FMT_EMAIL) {
445 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
446 struct strbuf buf = STRBUF_INIT;
447
448 strbuf_addstr(&buf, "From: ");
449 strbuf_add(&buf, namebuf, namelen);
450 strbuf_addstr(&buf, " <");
451 strbuf_add(&buf, mailbuf, maillen);
452 strbuf_addstr(&buf, ">\n");
453 string_list_append(&pp->in_body_headers,
454 strbuf_detach(&buf, NULL));
455
456 mailbuf = pp->from_ident->mail_begin;
457 maillen = pp->from_ident->mail_end - mailbuf;
458 namebuf = pp->from_ident->name_begin;
459 namelen = pp->from_ident->name_end - namebuf;
460 }
461
462 strbuf_addstr(sb, "From: ");
463 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
464 add_rfc2047(sb, namebuf, namelen,
465 encoding, RFC2047_ADDRESS);
466 max_length = 76; /* per rfc2047 */
467 } else if (needs_rfc822_quoting(namebuf, namelen)) {
468 struct strbuf quoted = STRBUF_INIT;
469 add_rfc822_quoted(&quoted, namebuf, namelen);
470 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
471 -6, 1, max_length);
472 strbuf_release(&quoted);
473 } else {
474 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
475 -6, 1, max_length);
476 }
477
478 if (max_length <
479 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
480 strbuf_addch(sb, '\n');
481 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
482 } else {
483 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
484 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
485 (int)namelen, namebuf, (int)maillen, mailbuf);
486 }
487
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;
504 }
505 }
506
507 static int is_empty_line(const char *line, int *len_p)
508 {
509 int len = *len_p;
510 while (len && isspace(line[len - 1]))
511 len--;
512 *len_p = len;
513 return !len;
514 }
515
516 static const char *skip_empty_lines(const char *msg)
517 {
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;
526 }
527 return msg;
528 }
529
530 static void add_merge_info(const struct pretty_print_context *pp,
531 struct strbuf *sb, const struct commit *commit)
532 {
533 struct commit_list *parent = commit->parents;
534
535 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
536 !parent || !parent->next)
537 return;
538
539 strbuf_addstr(sb, "Merge:");
540
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;
549
550 strbuf_addf(sb, " %s", hex);
551 }
552 strbuf_addch(sb, '\n');
553 }
554
555 static char *get_header(const struct commit *commit, const char *msg,
556 const char *key)
557 {
558 int key_len = strlen(key);
559 const char *line = msg;
560
561 while (line) {
562 const char *eol = strchrnul(line, '\n'), *next;
563
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 next = NULL;
570 } else
571 next = eol + 1;
572 if (eol - line > key_len &&
573 !strncmp(line, key, key_len) &&
574 line[key_len] == ' ') {
575 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
576 }
577 line = next;
578 }
579 return NULL;
580 }
581
582 static char *replace_encoding_header(char *buf, const char *encoding)
583 {
584 struct strbuf tmp = STRBUF_INIT;
585 size_t start, len;
586 char *cp = buf;
587
588 /* guess if there is an encoding header before a \n\n */
589 while (strncmp(cp, "encoding ", strlen("encoding "))) {
590 cp = strchr(cp, '\n');
591 if (!cp || *++cp == '\n')
592 return buf;
593 }
594 start = cp - buf;
595 cp = strchr(cp, '\n');
596 if (!cp)
597 return buf; /* should not happen but be defensive */
598 len = cp + 1 - (buf + start);
599
600 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
601 if (is_encoding_utf8(encoding)) {
602 /* we have re-coded to UTF-8; drop the header */
603 strbuf_remove(&tmp, start, len);
604 } else {
605 /* just replaces XXXX in 'encoding XXXX\n' */
606 strbuf_splice(&tmp, start + strlen("encoding "),
607 len - strlen("encoding \n"),
608 encoding, strlen(encoding));
609 }
610 return strbuf_detach(&tmp, NULL);
611 }
612
613 const char *logmsg_reencode(const struct commit *commit,
614 char **commit_encoding,
615 const char *output_encoding)
616 {
617 static const char *utf8 = "UTF-8";
618 const char *use_encoding;
619 char *encoding;
620 const char *msg = get_commit_buffer(commit, NULL);
621 char *out;
622
623 if (!output_encoding || !*output_encoding) {
624 if (commit_encoding)
625 *commit_encoding =
626 get_header(commit, msg, "encoding");
627 return msg;
628 }
629 encoding = get_header(commit, msg, "encoding");
630 if (commit_encoding)
631 *commit_encoding = encoding;
632 use_encoding = encoding ? encoding : utf8;
633 if (same_encoding(use_encoding, output_encoding)) {
634 /*
635 * No encoding work to be done. If we have no encoding header
636 * at all, then there's nothing to do, and we can return the
637 * message verbatim (whether newly allocated or not).
638 */
639 if (!encoding)
640 return msg;
641
642 /*
643 * Otherwise, we still want to munge the encoding header in the
644 * result, which will be done by modifying the buffer. If we
645 * are using a fresh copy, we can reuse it. But if we are using
646 * the cached copy from get_commit_buffer, we need to duplicate it
647 * to avoid munging the cached copy.
648 */
649 if (msg == get_cached_commit_buffer(commit, NULL))
650 out = xstrdup(msg);
651 else
652 out = (char *)msg;
653 }
654 else {
655 /*
656 * There's actual encoding work to do. Do the reencoding, which
657 * still leaves the header to be replaced in the next step. At
658 * this point, we are done with msg. If we allocated a fresh
659 * copy, we can free it.
660 */
661 out = reencode_string(msg, output_encoding, use_encoding);
662 if (out)
663 unuse_commit_buffer(commit, msg);
664 }
665
666 /*
667 * This replacement actually consumes the buffer we hand it, so we do
668 * not have to worry about freeing the old "out" here.
669 */
670 if (out)
671 out = replace_encoding_header(out, output_encoding);
672
673 if (!commit_encoding)
674 free(encoding);
675 /*
676 * If the re-encoding failed, out might be NULL here; in that
677 * case we just return the commit message verbatim.
678 */
679 return out ? out : msg;
680 }
681
682 static int mailmap_name(const char **email, size_t *email_len,
683 const char **name, size_t *name_len)
684 {
685 static struct string_list *mail_map;
686 if (!mail_map) {
687 mail_map = xcalloc(1, sizeof(*mail_map));
688 read_mailmap(mail_map, NULL);
689 }
690 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
691 }
692
693 static size_t format_person_part(struct strbuf *sb, char part,
694 const char *msg, int len, enum date_mode dmode)
695 {
696 /* currently all placeholders have same length */
697 const int placeholder_len = 2;
698 struct ident_split s;
699 const char *name, *mail;
700 size_t maillen, namelen;
701
702 if (split_ident_line(&s, msg, len) < 0)
703 goto skip;
704
705 name = s.name_begin;
706 namelen = s.name_end - s.name_begin;
707 mail = s.mail_begin;
708 maillen = s.mail_end - s.mail_begin;
709
710 if (part == 'N' || part == 'E') /* mailmap lookup */
711 mailmap_name(&mail, &maillen, &name, &namelen);
712 if (part == 'n' || part == 'N') { /* name */
713 strbuf_add(sb, name, namelen);
714 return placeholder_len;
715 }
716 if (part == 'e' || part == 'E') { /* email */
717 strbuf_add(sb, mail, maillen);
718 return placeholder_len;
719 }
720
721 if (!s.date_begin)
722 goto skip;
723
724 if (part == 't') { /* date, UNIX timestamp */
725 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
726 return placeholder_len;
727 }
728
729 switch (part) {
730 case 'd': /* date */
731 strbuf_addstr(sb, show_ident_date(&s, dmode));
732 return placeholder_len;
733 case 'D': /* date, RFC2822 style */
734 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
735 return placeholder_len;
736 case 'r': /* date, relative */
737 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
738 return placeholder_len;
739 case 'i': /* date, ISO 8601 */
740 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
741 return placeholder_len;
742 }
743
744 skip:
745 /*
746 * reading from either a bogus commit, or a reflog entry with
747 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
748 * to compute a valid return value.
749 */
750 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
751 || part == 'D' || part == 'r' || part == 'i')
752 return placeholder_len;
753
754 return 0; /* unknown placeholder */
755 }
756
757 struct chunk {
758 size_t off;
759 size_t len;
760 };
761
762 enum flush_type {
763 no_flush,
764 flush_right,
765 flush_left,
766 flush_left_and_steal,
767 flush_both
768 };
769
770 enum trunc_type {
771 trunc_none,
772 trunc_left,
773 trunc_middle,
774 trunc_right
775 };
776
777 struct format_commit_context {
778 const struct commit *commit;
779 const struct pretty_print_context *pretty_ctx;
780 unsigned commit_header_parsed:1;
781 unsigned commit_message_parsed:1;
782 struct signature_check signature_check;
783 enum flush_type flush_type;
784 enum trunc_type truncate;
785 const char *message;
786 char *commit_encoding;
787 size_t width, indent1, indent2;
788 int auto_color;
789 int padding;
790
791 /* These offsets are relative to the start of the commit message. */
792 struct chunk author;
793 struct chunk committer;
794 size_t message_off;
795 size_t subject_off;
796 size_t body_off;
797
798 /* The following ones are relative to the result struct strbuf. */
799 struct chunk abbrev_commit_hash;
800 struct chunk abbrev_tree_hash;
801 struct chunk abbrev_parent_hashes;
802 size_t wrap_start;
803 };
804
805 static int add_again(struct strbuf *sb, struct chunk *chunk)
806 {
807 if (chunk->len) {
808 strbuf_adddup(sb, chunk->off, chunk->len);
809 return 1;
810 }
811
812 /*
813 * We haven't seen this chunk before. Our caller is surely
814 * going to add it the hard way now. Remember the most likely
815 * start of the to-be-added chunk: the current end of the
816 * struct strbuf.
817 */
818 chunk->off = sb->len;
819 return 0;
820 }
821
822 static void parse_commit_header(struct format_commit_context *context)
823 {
824 const char *msg = context->message;
825 int i;
826
827 for (i = 0; msg[i]; i++) {
828 int eol;
829 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
830 ; /* do nothing */
831
832 if (i == eol) {
833 break;
834 } else if (starts_with(msg + i, "author ")) {
835 context->author.off = i + 7;
836 context->author.len = eol - i - 7;
837 } else if (starts_with(msg + i, "committer ")) {
838 context->committer.off = i + 10;
839 context->committer.len = eol - i - 10;
840 }
841 i = eol;
842 }
843 context->message_off = i;
844 context->commit_header_parsed = 1;
845 }
846
847 static int istitlechar(char c)
848 {
849 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
850 (c >= '0' && c <= '9') || c == '.' || c == '_';
851 }
852
853 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
854 {
855 size_t trimlen;
856 size_t start_len = sb->len;
857 int space = 2;
858
859 for (; *msg && *msg != '\n'; msg++) {
860 if (istitlechar(*msg)) {
861 if (space == 1)
862 strbuf_addch(sb, '-');
863 space = 0;
864 strbuf_addch(sb, *msg);
865 if (*msg == '.')
866 while (*(msg+1) == '.')
867 msg++;
868 } else
869 space |= 1;
870 }
871
872 /* trim any trailing '.' or '-' characters */
873 trimlen = 0;
874 while (sb->len - trimlen > start_len &&
875 (sb->buf[sb->len - 1 - trimlen] == '.'
876 || sb->buf[sb->len - 1 - trimlen] == '-'))
877 trimlen++;
878 strbuf_remove(sb, sb->len - trimlen, trimlen);
879 }
880
881 const char *format_subject(struct strbuf *sb, const char *msg,
882 const char *line_separator)
883 {
884 int first = 1;
885
886 for (;;) {
887 const char *line = msg;
888 int linelen = get_one_line(line);
889
890 msg += linelen;
891 if (!linelen || is_empty_line(line, &linelen))
892 break;
893
894 if (!sb)
895 continue;
896 strbuf_grow(sb, linelen + 2);
897 if (!first)
898 strbuf_addstr(sb, line_separator);
899 strbuf_add(sb, line, linelen);
900 first = 0;
901 }
902 return msg;
903 }
904
905 static void parse_commit_message(struct format_commit_context *c)
906 {
907 const char *msg = c->message + c->message_off;
908 const char *start = c->message;
909
910 msg = skip_empty_lines(msg);
911 c->subject_off = msg - start;
912
913 msg = format_subject(NULL, msg, NULL);
914 msg = skip_empty_lines(msg);
915 c->body_off = msg - start;
916
917 c->commit_message_parsed = 1;
918 }
919
920 static void strbuf_wrap(struct strbuf *sb, size_t pos,
921 size_t width, size_t indent1, size_t indent2)
922 {
923 struct strbuf tmp = STRBUF_INIT;
924
925 if (pos)
926 strbuf_add(&tmp, sb->buf, pos);
927 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
928 (int) indent1, (int) indent2, (int) width);
929 strbuf_swap(&tmp, sb);
930 strbuf_release(&tmp);
931 }
932
933 static void rewrap_message_tail(struct strbuf *sb,
934 struct format_commit_context *c,
935 size_t new_width, size_t new_indent1,
936 size_t new_indent2)
937 {
938 if (c->width == new_width && c->indent1 == new_indent1 &&
939 c->indent2 == new_indent2)
940 return;
941 if (c->wrap_start < sb->len)
942 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
943 c->wrap_start = sb->len;
944 c->width = new_width;
945 c->indent1 = new_indent1;
946 c->indent2 = new_indent2;
947 }
948
949 static int format_reflog_person(struct strbuf *sb,
950 char part,
951 struct reflog_walk_info *log,
952 enum date_mode dmode)
953 {
954 const char *ident;
955
956 if (!log)
957 return 2;
958
959 ident = get_reflog_ident(log);
960 if (!ident)
961 return 2;
962
963 return format_person_part(sb, part, ident, strlen(ident), dmode);
964 }
965
966 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
967 const char *placeholder,
968 struct format_commit_context *c)
969 {
970 if (placeholder[1] == '(') {
971 const char *begin = placeholder + 2;
972 const char *end = strchr(begin, ')');
973 char color[COLOR_MAXLEN];
974
975 if (!end)
976 return 0;
977 if (starts_with(begin, "auto,")) {
978 if (!want_color(c->pretty_ctx->color))
979 return end - placeholder + 1;
980 begin += 5;
981 }
982 if (color_parse_mem(begin, end - begin, color) < 0)
983 die(_("unable to parse --pretty format"));
984 strbuf_addstr(sb, color);
985 return end - placeholder + 1;
986 }
987 if (starts_with(placeholder + 1, "red")) {
988 strbuf_addstr(sb, GIT_COLOR_RED);
989 return 4;
990 } else if (starts_with(placeholder + 1, "green")) {
991 strbuf_addstr(sb, GIT_COLOR_GREEN);
992 return 6;
993 } else if (starts_with(placeholder + 1, "blue")) {
994 strbuf_addstr(sb, GIT_COLOR_BLUE);
995 return 5;
996 } else if (starts_with(placeholder + 1, "reset")) {
997 strbuf_addstr(sb, GIT_COLOR_RESET);
998 return 6;
999 } else
1000 return 0;
1001 }
1002
1003 static size_t parse_padding_placeholder(struct strbuf *sb,
1004 const char *placeholder,
1005 struct format_commit_context *c)
1006 {
1007 const char *ch = placeholder;
1008 enum flush_type flush_type;
1009 int to_column = 0;
1010
1011 switch (*ch++) {
1012 case '<':
1013 flush_type = flush_right;
1014 break;
1015 case '>':
1016 if (*ch == '<') {
1017 flush_type = flush_both;
1018 ch++;
1019 } else if (*ch == '>') {
1020 flush_type = flush_left_and_steal;
1021 ch++;
1022 } else
1023 flush_type = flush_left;
1024 break;
1025 default:
1026 return 0;
1027 }
1028
1029 /* the next value means "wide enough to that column" */
1030 if (*ch == '|') {
1031 to_column = 1;
1032 ch++;
1033 }
1034
1035 if (*ch == '(') {
1036 const char *start = ch + 1;
1037 const char *end = start + strcspn(start, ",)");
1038 char *next;
1039 int width;
1040 if (!end || end == start)
1041 return 0;
1042 width = strtoul(start, &next, 10);
1043 if (next == start || width == 0)
1044 return 0;
1045 c->padding = to_column ? -width : width;
1046 c->flush_type = flush_type;
1047
1048 if (*end == ',') {
1049 start = end + 1;
1050 end = strchr(start, ')');
1051 if (!end || end == start)
1052 return 0;
1053 if (starts_with(start, "trunc)"))
1054 c->truncate = trunc_right;
1055 else if (starts_with(start, "ltrunc)"))
1056 c->truncate = trunc_left;
1057 else if (starts_with(start, "mtrunc)"))
1058 c->truncate = trunc_middle;
1059 else
1060 return 0;
1061 } else
1062 c->truncate = trunc_none;
1063
1064 return end - placeholder + 1;
1065 }
1066 return 0;
1067 }
1068
1069 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1070 const char *placeholder,
1071 void *context)
1072 {
1073 struct format_commit_context *c = context;
1074 const struct commit *commit = c->commit;
1075 const char *msg = c->message;
1076 struct commit_list *p;
1077 int h1, h2;
1078
1079 /* these are independent of the commit */
1080 switch (placeholder[0]) {
1081 case 'C':
1082 if (starts_with(placeholder + 1, "(auto)")) {
1083 c->auto_color = 1;
1084 return 7; /* consumed 7 bytes, "C(auto)" */
1085 } else {
1086 int ret = parse_color(sb, placeholder, c);
1087 if (ret)
1088 c->auto_color = 0;
1089 /*
1090 * Otherwise, we decided to treat %C<unknown>
1091 * as a literal string, and the previous
1092 * %C(auto) is still valid.
1093 */
1094 return ret;
1095 }
1096 case 'n': /* newline */
1097 strbuf_addch(sb, '\n');
1098 return 1;
1099 case 'x':
1100 /* %x00 == NUL, %x0a == LF, etc. */
1101 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1102 h1 <= 16 &&
1103 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1104 h2 <= 16) {
1105 strbuf_addch(sb, (h1<<4)|h2);
1106 return 3;
1107 } else
1108 return 0;
1109 case 'w':
1110 if (placeholder[1] == '(') {
1111 unsigned long width = 0, indent1 = 0, indent2 = 0;
1112 char *next;
1113 const char *start = placeholder + 2;
1114 const char *end = strchr(start, ')');
1115 if (!end)
1116 return 0;
1117 if (end > start) {
1118 width = strtoul(start, &next, 10);
1119 if (*next == ',') {
1120 indent1 = strtoul(next + 1, &next, 10);
1121 if (*next == ',') {
1122 indent2 = strtoul(next + 1,
1123 &next, 10);
1124 }
1125 }
1126 if (*next != ')')
1127 return 0;
1128 }
1129 rewrap_message_tail(sb, c, width, indent1, indent2);
1130 return end - placeholder + 1;
1131 } else
1132 return 0;
1133
1134 case '<':
1135 case '>':
1136 return parse_padding_placeholder(sb, placeholder, c);
1137 }
1138
1139 /* these depend on the commit */
1140 if (!commit->object.parsed)
1141 parse_object(commit->object.sha1);
1142
1143 switch (placeholder[0]) {
1144 case 'H': /* commit hash */
1145 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1146 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1147 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1148 return 1;
1149 case 'h': /* abbreviated commit hash */
1150 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1151 if (add_again(sb, &c->abbrev_commit_hash)) {
1152 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1153 return 1;
1154 }
1155 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1156 c->pretty_ctx->abbrev));
1157 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1158 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1159 return 1;
1160 case 'T': /* tree hash */
1161 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1162 return 1;
1163 case 't': /* abbreviated tree hash */
1164 if (add_again(sb, &c->abbrev_tree_hash))
1165 return 1;
1166 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1167 c->pretty_ctx->abbrev));
1168 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1169 return 1;
1170 case 'P': /* parent hashes */
1171 for (p = commit->parents; p; p = p->next) {
1172 if (p != commit->parents)
1173 strbuf_addch(sb, ' ');
1174 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1175 }
1176 return 1;
1177 case 'p': /* abbreviated parent hashes */
1178 if (add_again(sb, &c->abbrev_parent_hashes))
1179 return 1;
1180 for (p = commit->parents; p; p = p->next) {
1181 if (p != commit->parents)
1182 strbuf_addch(sb, ' ');
1183 strbuf_addstr(sb, find_unique_abbrev(
1184 p->item->object.sha1,
1185 c->pretty_ctx->abbrev));
1186 }
1187 c->abbrev_parent_hashes.len = sb->len -
1188 c->abbrev_parent_hashes.off;
1189 return 1;
1190 case 'm': /* left/right/bottom */
1191 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1192 return 1;
1193 case 'd':
1194 load_ref_decorations(DECORATE_SHORT_REFS);
1195 format_decorations(sb, commit, c->auto_color);
1196 return 1;
1197 case 'g': /* reflog info */
1198 switch(placeholder[1]) {
1199 case 'd': /* reflog selector */
1200 case 'D':
1201 if (c->pretty_ctx->reflog_info)
1202 get_reflog_selector(sb,
1203 c->pretty_ctx->reflog_info,
1204 c->pretty_ctx->date_mode,
1205 c->pretty_ctx->date_mode_explicit,
1206 (placeholder[1] == 'd'));
1207 return 2;
1208 case 's': /* reflog message */
1209 if (c->pretty_ctx->reflog_info)
1210 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1211 return 2;
1212 case 'n':
1213 case 'N':
1214 case 'e':
1215 case 'E':
1216 return format_reflog_person(sb,
1217 placeholder[1],
1218 c->pretty_ctx->reflog_info,
1219 c->pretty_ctx->date_mode);
1220 }
1221 return 0; /* unknown %g placeholder */
1222 case 'N':
1223 if (c->pretty_ctx->notes_message) {
1224 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1225 return 1;
1226 }
1227 return 0;
1228 }
1229
1230 if (placeholder[0] == 'G') {
1231 if (!c->signature_check.result)
1232 check_commit_signature(c->commit, &(c->signature_check));
1233 switch (placeholder[1]) {
1234 case 'G':
1235 if (c->signature_check.gpg_output)
1236 strbuf_addstr(sb, c->signature_check.gpg_output);
1237 break;
1238 case '?':
1239 switch (c->signature_check.result) {
1240 case 'G':
1241 case 'B':
1242 case 'U':
1243 case 'N':
1244 strbuf_addch(sb, c->signature_check.result);
1245 }
1246 break;
1247 case 'S':
1248 if (c->signature_check.signer)
1249 strbuf_addstr(sb, c->signature_check.signer);
1250 break;
1251 case 'K':
1252 if (c->signature_check.key)
1253 strbuf_addstr(sb, c->signature_check.key);
1254 break;
1255 default:
1256 return 0;
1257 }
1258 return 2;
1259 }
1260
1261
1262 /* For the rest we have to parse the commit header. */
1263 if (!c->commit_header_parsed)
1264 parse_commit_header(c);
1265
1266 switch (placeholder[0]) {
1267 case 'a': /* author ... */
1268 return format_person_part(sb, placeholder[1],
1269 msg + c->author.off, c->author.len,
1270 c->pretty_ctx->date_mode);
1271 case 'c': /* committer ... */
1272 return format_person_part(sb, placeholder[1],
1273 msg + c->committer.off, c->committer.len,
1274 c->pretty_ctx->date_mode);
1275 case 'e': /* encoding */
1276 if (c->commit_encoding)
1277 strbuf_addstr(sb, c->commit_encoding);
1278 return 1;
1279 case 'B': /* raw body */
1280 /* message_off is always left at the initial newline */
1281 strbuf_addstr(sb, msg + c->message_off + 1);
1282 return 1;
1283 }
1284
1285 /* Now we need to parse the commit message. */
1286 if (!c->commit_message_parsed)
1287 parse_commit_message(c);
1288
1289 switch (placeholder[0]) {
1290 case 's': /* subject */
1291 format_subject(sb, msg + c->subject_off, " ");
1292 return 1;
1293 case 'f': /* sanitized subject */
1294 format_sanitized_subject(sb, msg + c->subject_off);
1295 return 1;
1296 case 'b': /* body */
1297 strbuf_addstr(sb, msg + c->body_off);
1298 return 1;
1299 }
1300 return 0; /* unknown placeholder */
1301 }
1302
1303 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1304 const char *placeholder,
1305 struct format_commit_context *c)
1306 {
1307 struct strbuf local_sb = STRBUF_INIT;
1308 int total_consumed = 0, len, padding = c->padding;
1309 if (padding < 0) {
1310 const char *start = strrchr(sb->buf, '\n');
1311 int occupied;
1312 if (!start)
1313 start = sb->buf;
1314 occupied = utf8_strnwidth(start, -1, 1);
1315 padding = (-padding) - occupied;
1316 }
1317 while (1) {
1318 int modifier = *placeholder == 'C';
1319 int consumed = format_commit_one(&local_sb, placeholder, c);
1320 total_consumed += consumed;
1321
1322 if (!modifier)
1323 break;
1324
1325 placeholder += consumed;
1326 if (*placeholder != '%')
1327 break;
1328 placeholder++;
1329 total_consumed++;
1330 }
1331 len = utf8_strnwidth(local_sb.buf, -1, 1);
1332
1333 if (c->flush_type == flush_left_and_steal) {
1334 const char *ch = sb->buf + sb->len - 1;
1335 while (len > padding && ch > sb->buf) {
1336 const char *p;
1337 if (*ch == ' ') {
1338 ch--;
1339 padding++;
1340 continue;
1341 }
1342 /* check for trailing ansi sequences */
1343 if (*ch != 'm')
1344 break;
1345 p = ch - 1;
1346 while (ch - p < 10 && *p != '\033')
1347 p--;
1348 if (*p != '\033' ||
1349 ch + 1 - p != display_mode_esc_sequence_len(p))
1350 break;
1351 /*
1352 * got a good ansi sequence, put it back to
1353 * local_sb as we're cutting sb
1354 */
1355 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1356 ch = p - 1;
1357 }
1358 strbuf_setlen(sb, ch + 1 - sb->buf);
1359 c->flush_type = flush_left;
1360 }
1361
1362 if (len > padding) {
1363 switch (c->truncate) {
1364 case trunc_left:
1365 strbuf_utf8_replace(&local_sb,
1366 0, len - (padding - 2),
1367 "..");
1368 break;
1369 case trunc_middle:
1370 strbuf_utf8_replace(&local_sb,
1371 padding / 2 - 1,
1372 len - (padding - 2),
1373 "..");
1374 break;
1375 case trunc_right:
1376 strbuf_utf8_replace(&local_sb,
1377 padding - 2, len - (padding - 2),
1378 "..");
1379 break;
1380 case trunc_none:
1381 break;
1382 }
1383 strbuf_addbuf(sb, &local_sb);
1384 } else {
1385 int sb_len = sb->len, offset = 0;
1386 if (c->flush_type == flush_left)
1387 offset = padding - len;
1388 else if (c->flush_type == flush_both)
1389 offset = (padding - len) / 2;
1390 /*
1391 * we calculate padding in columns, now
1392 * convert it back to chars
1393 */
1394 padding = padding - len + local_sb.len;
1395 strbuf_grow(sb, padding);
1396 strbuf_setlen(sb, sb_len + padding);
1397 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1398 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1399 local_sb.len);
1400 }
1401 strbuf_release(&local_sb);
1402 c->flush_type = no_flush;
1403 return total_consumed;
1404 }
1405
1406 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1407 const char *placeholder,
1408 void *context)
1409 {
1410 int consumed;
1411 size_t orig_len;
1412 enum {
1413 NO_MAGIC,
1414 ADD_LF_BEFORE_NON_EMPTY,
1415 DEL_LF_BEFORE_EMPTY,
1416 ADD_SP_BEFORE_NON_EMPTY
1417 } magic = NO_MAGIC;
1418
1419 switch (placeholder[0]) {
1420 case '-':
1421 magic = DEL_LF_BEFORE_EMPTY;
1422 break;
1423 case '+':
1424 magic = ADD_LF_BEFORE_NON_EMPTY;
1425 break;
1426 case ' ':
1427 magic = ADD_SP_BEFORE_NON_EMPTY;
1428 break;
1429 default:
1430 break;
1431 }
1432 if (magic != NO_MAGIC)
1433 placeholder++;
1434
1435 orig_len = sb->len;
1436 if (((struct format_commit_context *)context)->flush_type != no_flush)
1437 consumed = format_and_pad_commit(sb, placeholder, context);
1438 else
1439 consumed = format_commit_one(sb, placeholder, context);
1440 if (magic == NO_MAGIC)
1441 return consumed;
1442
1443 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1444 while (sb->len && sb->buf[sb->len - 1] == '\n')
1445 strbuf_setlen(sb, sb->len - 1);
1446 } else if (orig_len != sb->len) {
1447 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1448 strbuf_insert(sb, orig_len, "\n", 1);
1449 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1450 strbuf_insert(sb, orig_len, " ", 1);
1451 }
1452 return consumed + 1;
1453 }
1454
1455 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1456 void *context)
1457 {
1458 struct userformat_want *w = context;
1459
1460 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1461 placeholder++;
1462
1463 switch (*placeholder) {
1464 case 'N':
1465 w->notes = 1;
1466 break;
1467 }
1468 return 0;
1469 }
1470
1471 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1472 {
1473 struct strbuf dummy = STRBUF_INIT;
1474
1475 if (!fmt) {
1476 if (!user_format)
1477 return;
1478 fmt = user_format;
1479 }
1480 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1481 strbuf_release(&dummy);
1482 }
1483
1484 void format_commit_message(const struct commit *commit,
1485 const char *format, struct strbuf *sb,
1486 const struct pretty_print_context *pretty_ctx)
1487 {
1488 struct format_commit_context context;
1489 const char *output_enc = pretty_ctx->output_encoding;
1490 const char *utf8 = "UTF-8";
1491
1492 memset(&context, 0, sizeof(context));
1493 context.commit = commit;
1494 context.pretty_ctx = pretty_ctx;
1495 context.wrap_start = sb->len;
1496 /*
1497 * convert a commit message to UTF-8 first
1498 * as far as 'format_commit_item' assumes it in UTF-8
1499 */
1500 context.message = logmsg_reencode(commit,
1501 &context.commit_encoding,
1502 utf8);
1503
1504 strbuf_expand(sb, format, format_commit_item, &context);
1505 rewrap_message_tail(sb, &context, 0, 0, 0);
1506
1507 /* then convert a commit message to an actual output encoding */
1508 if (output_enc) {
1509 if (same_encoding(utf8, output_enc))
1510 output_enc = NULL;
1511 } else {
1512 if (context.commit_encoding &&
1513 !same_encoding(context.commit_encoding, utf8))
1514 output_enc = context.commit_encoding;
1515 }
1516
1517 if (output_enc) {
1518 int outsz;
1519 char *out = reencode_string_len(sb->buf, sb->len,
1520 output_enc, utf8, &outsz);
1521 if (out)
1522 strbuf_attach(sb, out, outsz, outsz + 1);
1523 }
1524
1525 free(context.commit_encoding);
1526 unuse_commit_buffer(commit, context.message);
1527 }
1528
1529 static void pp_header(struct pretty_print_context *pp,
1530 const char *encoding,
1531 const struct commit *commit,
1532 const char **msg_p,
1533 struct strbuf *sb)
1534 {
1535 int parents_shown = 0;
1536
1537 for (;;) {
1538 const char *line = *msg_p;
1539 int linelen = get_one_line(*msg_p);
1540
1541 if (!linelen)
1542 return;
1543 *msg_p += linelen;
1544
1545 if (linelen == 1)
1546 /* End of header */
1547 return;
1548
1549 if (pp->fmt == CMIT_FMT_RAW) {
1550 strbuf_add(sb, line, linelen);
1551 continue;
1552 }
1553
1554 if (starts_with(line, "parent ")) {
1555 if (linelen != 48)
1556 die("bad parent line in commit");
1557 continue;
1558 }
1559
1560 if (!parents_shown) {
1561 unsigned num = commit_list_count(commit->parents);
1562 /* with enough slop */
1563 strbuf_grow(sb, num * 50 + 20);
1564 add_merge_info(pp, sb, commit);
1565 parents_shown = 1;
1566 }
1567
1568 /*
1569 * MEDIUM == DEFAULT shows only author with dates.
1570 * FULL shows both authors but not dates.
1571 * FULLER shows both authors and dates.
1572 */
1573 if (starts_with(line, "author ")) {
1574 strbuf_grow(sb, linelen + 80);
1575 pp_user_info(pp, "Author", sb, line + 7, encoding);
1576 }
1577 if (starts_with(line, "committer ") &&
1578 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1579 strbuf_grow(sb, linelen + 80);
1580 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1581 }
1582 }
1583 }
1584
1585 void pp_title_line(struct pretty_print_context *pp,
1586 const char **msg_p,
1587 struct strbuf *sb,
1588 const char *encoding,
1589 int need_8bit_cte)
1590 {
1591 static const int max_length = 78; /* per rfc2047 */
1592 struct strbuf title;
1593
1594 strbuf_init(&title, 80);
1595 *msg_p = format_subject(&title, *msg_p,
1596 pp->preserve_subject ? "\n" : " ");
1597
1598 strbuf_grow(sb, title.len + 1024);
1599 if (pp->subject) {
1600 strbuf_addstr(sb, pp->subject);
1601 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1602 add_rfc2047(sb, title.buf, title.len,
1603 encoding, RFC2047_SUBJECT);
1604 else
1605 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1606 -last_line_length(sb), 1, max_length);
1607 } else {
1608 strbuf_addbuf(sb, &title);
1609 }
1610 strbuf_addch(sb, '\n');
1611
1612 if (need_8bit_cte == 0) {
1613 int i;
1614 for (i = 0; i < pp->in_body_headers.nr; i++) {
1615 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1616 need_8bit_cte = 1;
1617 break;
1618 }
1619 }
1620 }
1621
1622 if (need_8bit_cte > 0) {
1623 const char *header_fmt =
1624 "MIME-Version: 1.0\n"
1625 "Content-Type: text/plain; charset=%s\n"
1626 "Content-Transfer-Encoding: 8bit\n";
1627 strbuf_addf(sb, header_fmt, encoding);
1628 }
1629 if (pp->after_subject) {
1630 strbuf_addstr(sb, pp->after_subject);
1631 }
1632 if (pp->fmt == CMIT_FMT_EMAIL) {
1633 strbuf_addch(sb, '\n');
1634 }
1635
1636 if (pp->in_body_headers.nr) {
1637 int i;
1638 for (i = 0; i < pp->in_body_headers.nr; i++) {
1639 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1640 free(pp->in_body_headers.items[i].string);
1641 }
1642 string_list_clear(&pp->in_body_headers, 0);
1643 strbuf_addch(sb, '\n');
1644 }
1645
1646 strbuf_release(&title);
1647 }
1648
1649 void pp_remainder(struct pretty_print_context *pp,
1650 const char **msg_p,
1651 struct strbuf *sb,
1652 int indent)
1653 {
1654 int first = 1;
1655 for (;;) {
1656 const char *line = *msg_p;
1657 int linelen = get_one_line(line);
1658 *msg_p += linelen;
1659
1660 if (!linelen)
1661 break;
1662
1663 if (is_empty_line(line, &linelen)) {
1664 if (first)
1665 continue;
1666 if (pp->fmt == CMIT_FMT_SHORT)
1667 break;
1668 }
1669 first = 0;
1670
1671 strbuf_grow(sb, linelen + indent + 20);
1672 if (indent) {
1673 memset(sb->buf + sb->len, ' ', indent);
1674 strbuf_setlen(sb, sb->len + indent);
1675 }
1676 strbuf_add(sb, line, linelen);
1677 strbuf_addch(sb, '\n');
1678 }
1679 }
1680
1681 void pretty_print_commit(struct pretty_print_context *pp,
1682 const struct commit *commit,
1683 struct strbuf *sb)
1684 {
1685 unsigned long beginning_of_body;
1686 int indent = 4;
1687 const char *msg;
1688 const char *reencoded;
1689 const char *encoding;
1690 int need_8bit_cte = pp->need_8bit_cte;
1691
1692 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1693 format_commit_message(commit, user_format, sb, pp);
1694 return;
1695 }
1696
1697 encoding = get_log_output_encoding();
1698 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1699
1700 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1701 indent = 0;
1702
1703 /*
1704 * We need to check and emit Content-type: to mark it
1705 * as 8-bit if we haven't done so.
1706 */
1707 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1708 int i, ch, in_body;
1709
1710 for (in_body = i = 0; (ch = msg[i]); i++) {
1711 if (!in_body) {
1712 /* author could be non 7-bit ASCII but
1713 * the log may be so; skip over the
1714 * header part first.
1715 */
1716 if (ch == '\n' && msg[i+1] == '\n')
1717 in_body = 1;
1718 }
1719 else if (non_ascii(ch)) {
1720 need_8bit_cte = 1;
1721 break;
1722 }
1723 }
1724 }
1725
1726 pp_header(pp, encoding, commit, &msg, sb);
1727 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1728 strbuf_addch(sb, '\n');
1729 }
1730
1731 /* Skip excess blank lines at the beginning of body, if any... */
1732 msg = skip_empty_lines(msg);
1733
1734 /* These formats treat the title line specially. */
1735 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1736 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1737
1738 beginning_of_body = sb->len;
1739 if (pp->fmt != CMIT_FMT_ONELINE)
1740 pp_remainder(pp, &msg, sb, indent);
1741 strbuf_rtrim(sb);
1742
1743 /* Make sure there is an EOLN for the non-oneline case */
1744 if (pp->fmt != CMIT_FMT_ONELINE)
1745 strbuf_addch(sb, '\n');
1746
1747 /*
1748 * The caller may append additional body text in e-mail
1749 * format. Make sure we did not strip the blank line
1750 * between the header and the body.
1751 */
1752 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1753 strbuf_addch(sb, '\n');
1754
1755 unuse_commit_buffer(commit, reencoded);
1756 }
1757
1758 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1759 struct strbuf *sb)
1760 {
1761 struct pretty_print_context pp = {0};
1762 pp.fmt = fmt;
1763 pretty_print_commit(&pp, commit, sb);
1764 }
Gitweb development nest