8ca29e92815608437ccb1fb9909e78d5b4989529
[git/git.git] / pretty.c
1 #include "cache.h"
2 #include "config.h"
3 #include "commit.h"
4 #include "utf8.h"
5 #include "diff.h"
6 #include "revision.h"
7 #include "string-list.h"
8 #include "mailmap.h"
9 #include "log-tree.h"
10 #include "notes.h"
11 #include "color.h"
12 #include "reflog-walk.h"
13 #include "gpg-interface.h"
14 #include "trailer.h"
15
16 static char *user_format;
17 static struct cmt_fmt_map {
18 const char *name;
19 enum cmit_fmt format;
20 int is_tformat;
21 int expand_tabs_in_log;
22 int is_alias;
23 const char *user_format;
24 } *commit_formats;
25 static size_t builtin_formats_len;
26 static size_t commit_formats_len;
27 static size_t commit_formats_alloc;
28 static struct cmt_fmt_map *find_commit_format(const char *sought);
29
30 int commit_format_is_empty(enum cmit_fmt fmt)
31 {
32 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
33 }
34
35 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
36 {
37 free(user_format);
38 user_format = xstrdup(cp);
39 if (is_tformat)
40 rev->use_terminator = 1;
41 rev->commit_format = CMIT_FMT_USERFORMAT;
42 }
43
44 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
45 {
46 struct cmt_fmt_map *commit_format = NULL;
47 const char *name;
48 const char *fmt;
49 int i;
50
51 if (!skip_prefix(var, "pretty.", &name))
52 return 0;
53
54 for (i = 0; i < builtin_formats_len; i++) {
55 if (!strcmp(commit_formats[i].name, name))
56 return 0;
57 }
58
59 for (i = builtin_formats_len; i < commit_formats_len; i++) {
60 if (!strcmp(commit_formats[i].name, name)) {
61 commit_format = &commit_formats[i];
62 break;
63 }
64 }
65
66 if (!commit_format) {
67 ALLOC_GROW(commit_formats, commit_formats_len+1,
68 commit_formats_alloc);
69 commit_format = &commit_formats[commit_formats_len];
70 memset(commit_format, 0, sizeof(*commit_format));
71 commit_formats_len++;
72 }
73
74 commit_format->name = xstrdup(name);
75 commit_format->format = CMIT_FMT_USERFORMAT;
76 if (git_config_string(&fmt, var, value))
77 return -1;
78
79 if (skip_prefix(fmt, "format:", &fmt))
80 commit_format->is_tformat = 0;
81 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
82 commit_format->is_tformat = 1;
83 else
84 commit_format->is_alias = 1;
85 commit_format->user_format = fmt;
86
87 return 0;
88 }
89
90 static void setup_commit_formats(void)
91 {
92 struct cmt_fmt_map builtin_formats[] = {
93 { "raw", CMIT_FMT_RAW, 0, 0 },
94 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
95 { "short", CMIT_FMT_SHORT, 0, 0 },
96 { "email", CMIT_FMT_EMAIL, 0, 0 },
97 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
98 { "fuller", CMIT_FMT_FULLER, 0, 8 },
99 { "full", CMIT_FMT_FULL, 0, 8 },
100 { "oneline", CMIT_FMT_ONELINE, 1, 0 }
101 };
102 commit_formats_len = ARRAY_SIZE(builtin_formats);
103 builtin_formats_len = commit_formats_len;
104 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
105 memcpy(commit_formats, builtin_formats,
106 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
107
108 git_config(git_pretty_formats_config, NULL);
109 }
110
111 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
112 const char *original,
113 int num_redirections)
114 {
115 struct cmt_fmt_map *found = NULL;
116 size_t found_match_len = 0;
117 int i;
118
119 if (num_redirections >= commit_formats_len)
120 die("invalid --pretty format: "
121 "'%s' references an alias which points to itself",
122 original);
123
124 for (i = 0; i < commit_formats_len; i++) {
125 size_t match_len;
126
127 if (!starts_with(commit_formats[i].name, sought))
128 continue;
129
130 match_len = strlen(commit_formats[i].name);
131 if (found == NULL || found_match_len > match_len) {
132 found = &commit_formats[i];
133 found_match_len = match_len;
134 }
135 }
136
137 if (found && found->is_alias) {
138 found = find_commit_format_recursive(found->user_format,
139 original,
140 num_redirections+1);
141 }
142
143 return found;
144 }
145
146 static struct cmt_fmt_map *find_commit_format(const char *sought)
147 {
148 if (!commit_formats)
149 setup_commit_formats();
150
151 return find_commit_format_recursive(sought, sought, 0);
152 }
153
154 void get_commit_format(const char *arg, struct rev_info *rev)
155 {
156 struct cmt_fmt_map *commit_format;
157
158 rev->use_terminator = 0;
159 if (!arg) {
160 rev->commit_format = CMIT_FMT_DEFAULT;
161 return;
162 }
163 if (skip_prefix(arg, "format:", &arg)) {
164 save_user_format(rev, arg, 0);
165 return;
166 }
167
168 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
169 save_user_format(rev, arg, 1);
170 return;
171 }
172
173 commit_format = find_commit_format(arg);
174 if (!commit_format)
175 die("invalid --pretty format: %s", arg);
176
177 rev->commit_format = commit_format->format;
178 rev->use_terminator = commit_format->is_tformat;
179 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
180 if (commit_format->format == CMIT_FMT_USERFORMAT) {
181 save_user_format(rev, commit_format->user_format,
182 commit_format->is_tformat);
183 }
184 }
185
186 /*
187 * Generic support for pretty-printing the header
188 */
189 static int get_one_line(const char *msg)
190 {
191 int ret = 0;
192
193 for (;;) {
194 char c = *msg++;
195 if (!c)
196 break;
197 ret++;
198 if (c == '\n')
199 break;
200 }
201 return ret;
202 }
203
204 /* High bit set, or ISO-2022-INT */
205 static int non_ascii(int ch)
206 {
207 return !isascii(ch) || ch == '\033';
208 }
209
210 int has_non_ascii(const char *s)
211 {
212 int ch;
213 if (!s)
214 return 0;
215 while ((ch = *s++) != '\0') {
216 if (non_ascii(ch))
217 return 1;
218 }
219 return 0;
220 }
221
222 static int is_rfc822_special(char ch)
223 {
224 switch (ch) {
225 case '(':
226 case ')':
227 case '<':
228 case '>':
229 case '[':
230 case ']':
231 case ':':
232 case ';':
233 case '@':
234 case ',':
235 case '.':
236 case '"':
237 case '\\':
238 return 1;
239 default:
240 return 0;
241 }
242 }
243
244 static int needs_rfc822_quoting(const char *s, int len)
245 {
246 int i;
247 for (i = 0; i < len; i++)
248 if (is_rfc822_special(s[i]))
249 return 1;
250 return 0;
251 }
252
253 static int last_line_length(struct strbuf *sb)
254 {
255 int i;
256
257 /* How many bytes are already used on the last line? */
258 for (i = sb->len - 1; i >= 0; i--)
259 if (sb->buf[i] == '\n')
260 break;
261 return sb->len - (i + 1);
262 }
263
264 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
265 {
266 int i;
267
268 /* just a guess, we may have to also backslash-quote */
269 strbuf_grow(out, len + 2);
270
271 strbuf_addch(out, '"');
272 for (i = 0; i < len; i++) {
273 switch (s[i]) {
274 case '"':
275 case '\\':
276 strbuf_addch(out, '\\');
277 /* fall through */
278 default:
279 strbuf_addch(out, s[i]);
280 }
281 }
282 strbuf_addch(out, '"');
283 }
284
285 enum rfc2047_type {
286 RFC2047_SUBJECT,
287 RFC2047_ADDRESS
288 };
289
290 static int is_rfc2047_special(char ch, enum rfc2047_type type)
291 {
292 /*
293 * rfc2047, section 4.2:
294 *
295 * 8-bit values which correspond to printable ASCII characters other
296 * than "=", "?", and "_" (underscore), MAY be represented as those
297 * characters. (But see section 5 for restrictions.) In
298 * particular, SPACE and TAB MUST NOT be represented as themselves
299 * within encoded words.
300 */
301
302 /*
303 * rule out non-ASCII characters and non-printable characters (the
304 * non-ASCII check should be redundant as isprint() is not localized
305 * and only knows about ASCII, but be defensive about that)
306 */
307 if (non_ascii(ch) || !isprint(ch))
308 return 1;
309
310 /*
311 * rule out special printable characters (' ' should be the only
312 * whitespace character considered printable, but be defensive and use
313 * isspace())
314 */
315 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
316 return 1;
317
318 /*
319 * rfc2047, section 5.3:
320 *
321 * As a replacement for a 'word' entity within a 'phrase', for example,
322 * one that precedes an address in a From, To, or Cc header. The ABNF
323 * definition for 'phrase' from RFC 822 thus becomes:
324 *
325 * phrase = 1*( encoded-word / word )
326 *
327 * In this case the set of characters that may be used in a "Q"-encoded
328 * 'encoded-word' is restricted to: <upper and lower case ASCII
329 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
330 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
331 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
332 * 'special' by 'linear-white-space'.
333 */
334
335 if (type != RFC2047_ADDRESS)
336 return 0;
337
338 /* '=' and '_' are special cases and have been checked above */
339 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
340 }
341
342 static int needs_rfc2047_encoding(const char *line, int len,
343 enum rfc2047_type type)
344 {
345 int i;
346
347 for (i = 0; i < len; i++) {
348 int ch = line[i];
349 if (non_ascii(ch) || ch == '\n')
350 return 1;
351 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
352 return 1;
353 }
354
355 return 0;
356 }
357
358 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
359 const char *encoding, enum rfc2047_type type)
360 {
361 static const int max_encoded_length = 76; /* per rfc2047 */
362 int i;
363 int line_len = last_line_length(sb);
364
365 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
366 strbuf_addf(sb, "=?%s?q?", encoding);
367 line_len += strlen(encoding) + 5; /* 5 for =??q? */
368
369 while (len) {
370 /*
371 * RFC 2047, section 5 (3):
372 *
373 * Each 'encoded-word' MUST represent an integral number of
374 * characters. A multi-octet character may not be split across
375 * adjacent 'encoded- word's.
376 */
377 const unsigned char *p = (const unsigned char *)line;
378 int chrlen = mbs_chrlen(&line, &len, encoding);
379 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
380
381 /* "=%02X" * chrlen, or the byte itself */
382 const char *encoded_fmt = is_special ? "=%02X" : "%c";
383 int encoded_len = is_special ? 3 * chrlen : 1;
384
385 /*
386 * According to RFC 2047, we could encode the special character
387 * ' ' (space) with '_' (underscore) for readability. But many
388 * programs do not understand this and just leave the
389 * underscore in place. Thus, we do nothing special here, which
390 * causes ' ' to be encoded as '=20', avoiding this problem.
391 */
392
393 if (line_len + encoded_len + 2 > max_encoded_length) {
394 /* It won't fit with trailing "?=" --- break the line */
395 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
396 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
397 }
398
399 for (i = 0; i < chrlen; i++)
400 strbuf_addf(sb, encoded_fmt, p[i]);
401 line_len += encoded_len;
402 }
403 strbuf_addstr(sb, "?=");
404 }
405
406 const char *show_ident_date(const struct ident_split *ident,
407 const struct date_mode *mode)
408 {
409 timestamp_t date = 0;
410 long tz = 0;
411
412 if (ident->date_begin && ident->date_end)
413 date = parse_timestamp(ident->date_begin, NULL, 10);
414 if (date_overflows(date))
415 date = 0;
416 else {
417 if (ident->tz_begin && ident->tz_end)
418 tz = strtol(ident->tz_begin, NULL, 10);
419 if (tz >= INT_MAX || tz <= INT_MIN)
420 tz = 0;
421 }
422 return show_date(date, tz, mode);
423 }
424
425 void pp_user_info(struct pretty_print_context *pp,
426 const char *what, struct strbuf *sb,
427 const char *line, const char *encoding)
428 {
429 struct ident_split ident;
430 char *line_end;
431 const char *mailbuf, *namebuf;
432 size_t namelen, maillen;
433 int max_length = 78; /* per rfc2822 */
434
435 if (pp->fmt == CMIT_FMT_ONELINE)
436 return;
437
438 line_end = strchrnul(line, '\n');
439 if (split_ident_line(&ident, line, line_end - line))
440 return;
441
442 mailbuf = ident.mail_begin;
443 maillen = ident.mail_end - ident.mail_begin;
444 namebuf = ident.name_begin;
445 namelen = ident.name_end - ident.name_begin;
446
447 if (pp->mailmap)
448 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
449
450 if (cmit_fmt_is_mail(pp->fmt)) {
451 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
452 struct strbuf buf = STRBUF_INIT;
453
454 strbuf_addstr(&buf, "From: ");
455 strbuf_add(&buf, namebuf, namelen);
456 strbuf_addstr(&buf, " <");
457 strbuf_add(&buf, mailbuf, maillen);
458 strbuf_addstr(&buf, ">\n");
459 string_list_append(&pp->in_body_headers,
460 strbuf_detach(&buf, NULL));
461
462 mailbuf = pp->from_ident->mail_begin;
463 maillen = pp->from_ident->mail_end - mailbuf;
464 namebuf = pp->from_ident->name_begin;
465 namelen = pp->from_ident->name_end - namebuf;
466 }
467
468 strbuf_addstr(sb, "From: ");
469 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
470 add_rfc2047(sb, namebuf, namelen,
471 encoding, RFC2047_ADDRESS);
472 max_length = 76; /* per rfc2047 */
473 } else if (needs_rfc822_quoting(namebuf, namelen)) {
474 struct strbuf quoted = STRBUF_INIT;
475 add_rfc822_quoted(&quoted, namebuf, namelen);
476 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
477 -6, 1, max_length);
478 strbuf_release(&quoted);
479 } else {
480 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
481 -6, 1, max_length);
482 }
483
484 if (max_length <
485 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
486 strbuf_addch(sb, '\n');
487 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
488 } else {
489 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
490 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
491 (int)namelen, namebuf, (int)maillen, mailbuf);
492 }
493
494 switch (pp->fmt) {
495 case CMIT_FMT_MEDIUM:
496 strbuf_addf(sb, "Date: %s\n",
497 show_ident_date(&ident, &pp->date_mode));
498 break;
499 case CMIT_FMT_EMAIL:
500 case CMIT_FMT_MBOXRD:
501 strbuf_addf(sb, "Date: %s\n",
502 show_ident_date(&ident, DATE_MODE(RFC2822)));
503 break;
504 case CMIT_FMT_FULLER:
505 strbuf_addf(sb, "%sDate: %s\n", what,
506 show_ident_date(&ident, &pp->date_mode));
507 break;
508 default:
509 /* notin' */
510 break;
511 }
512 }
513
514 static int is_blank_line(const char *line, int *len_p)
515 {
516 int len = *len_p;
517 while (len && isspace(line[len - 1]))
518 len--;
519 *len_p = len;
520 return !len;
521 }
522
523 const char *skip_blank_lines(const char *msg)
524 {
525 for (;;) {
526 int linelen = get_one_line(msg);
527 int ll = linelen;
528 if (!linelen)
529 break;
530 if (!is_blank_line(msg, &ll))
531 break;
532 msg += linelen;
533 }
534 return msg;
535 }
536
537 static void add_merge_info(const struct pretty_print_context *pp,
538 struct strbuf *sb, const struct commit *commit)
539 {
540 struct commit_list *parent = commit->parents;
541
542 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
543 !parent || !parent->next)
544 return;
545
546 strbuf_addstr(sb, "Merge:");
547
548 while (parent) {
549 struct object_id *oidp = &parent->item->object.oid;
550 strbuf_addch(sb, ' ');
551 if (pp->abbrev)
552 strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
553 else
554 strbuf_addstr(sb, oid_to_hex(oidp));
555 parent = parent->next;
556 }
557 strbuf_addch(sb, '\n');
558 }
559
560 static char *get_header(const char *msg, const char *key)
561 {
562 size_t len;
563 const char *v = find_commit_header(msg, key, &len);
564 return v ? xmemdupz(v, len) : NULL;
565 }
566
567 static char *replace_encoding_header(char *buf, const char *encoding)
568 {
569 struct strbuf tmp = STRBUF_INIT;
570 size_t start, len;
571 char *cp = buf;
572
573 /* guess if there is an encoding header before a \n\n */
574 while (!starts_with(cp, "encoding ")) {
575 cp = strchr(cp, '\n');
576 if (!cp || *++cp == '\n')
577 return buf;
578 }
579 start = cp - buf;
580 cp = strchr(cp, '\n');
581 if (!cp)
582 return buf; /* should not happen but be defensive */
583 len = cp + 1 - (buf + start);
584
585 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
586 if (is_encoding_utf8(encoding)) {
587 /* we have re-coded to UTF-8; drop the header */
588 strbuf_remove(&tmp, start, len);
589 } else {
590 /* just replaces XXXX in 'encoding XXXX\n' */
591 strbuf_splice(&tmp, start + strlen("encoding "),
592 len - strlen("encoding \n"),
593 encoding, strlen(encoding));
594 }
595 return strbuf_detach(&tmp, NULL);
596 }
597
598 const char *logmsg_reencode(const struct commit *commit,
599 char **commit_encoding,
600 const char *output_encoding)
601 {
602 static const char *utf8 = "UTF-8";
603 const char *use_encoding;
604 char *encoding;
605 const char *msg = get_commit_buffer(commit, NULL);
606 char *out;
607
608 if (!output_encoding || !*output_encoding) {
609 if (commit_encoding)
610 *commit_encoding = get_header(msg, "encoding");
611 return msg;
612 }
613 encoding = get_header(msg, "encoding");
614 if (commit_encoding)
615 *commit_encoding = encoding;
616 use_encoding = encoding ? encoding : utf8;
617 if (same_encoding(use_encoding, output_encoding)) {
618 /*
619 * No encoding work to be done. If we have no encoding header
620 * at all, then there's nothing to do, and we can return the
621 * message verbatim (whether newly allocated or not).
622 */
623 if (!encoding)
624 return msg;
625
626 /*
627 * Otherwise, we still want to munge the encoding header in the
628 * result, which will be done by modifying the buffer. If we
629 * are using a fresh copy, we can reuse it. But if we are using
630 * the cached copy from get_commit_buffer, we need to duplicate it
631 * to avoid munging the cached copy.
632 */
633 if (msg == get_cached_commit_buffer(the_repository, commit, NULL))
634 out = xstrdup(msg);
635 else
636 out = (char *)msg;
637 }
638 else {
639 /*
640 * There's actual encoding work to do. Do the reencoding, which
641 * still leaves the header to be replaced in the next step. At
642 * this point, we are done with msg. If we allocated a fresh
643 * copy, we can free it.
644 */
645 out = reencode_string(msg, output_encoding, use_encoding);
646 if (out)
647 unuse_commit_buffer(commit, msg);
648 }
649
650 /*
651 * This replacement actually consumes the buffer we hand it, so we do
652 * not have to worry about freeing the old "out" here.
653 */
654 if (out)
655 out = replace_encoding_header(out, output_encoding);
656
657 if (!commit_encoding)
658 free(encoding);
659 /*
660 * If the re-encoding failed, out might be NULL here; in that
661 * case we just return the commit message verbatim.
662 */
663 return out ? out : msg;
664 }
665
666 static int mailmap_name(const char **email, size_t *email_len,
667 const char **name, size_t *name_len)
668 {
669 static struct string_list *mail_map;
670 if (!mail_map) {
671 mail_map = xcalloc(1, sizeof(*mail_map));
672 read_mailmap(mail_map, NULL);
673 }
674 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
675 }
676
677 static size_t format_person_part(struct strbuf *sb, char part,
678 const char *msg, int len,
679 const struct date_mode *dmode)
680 {
681 /* currently all placeholders have same length */
682 const int placeholder_len = 2;
683 struct ident_split s;
684 const char *name, *mail;
685 size_t maillen, namelen;
686
687 if (split_ident_line(&s, msg, len) < 0)
688 goto skip;
689
690 name = s.name_begin;
691 namelen = s.name_end - s.name_begin;
692 mail = s.mail_begin;
693 maillen = s.mail_end - s.mail_begin;
694
695 if (part == 'N' || part == 'E') /* mailmap lookup */
696 mailmap_name(&mail, &maillen, &name, &namelen);
697 if (part == 'n' || part == 'N') { /* name */
698 strbuf_add(sb, name, namelen);
699 return placeholder_len;
700 }
701 if (part == 'e' || part == 'E') { /* email */
702 strbuf_add(sb, mail, maillen);
703 return placeholder_len;
704 }
705
706 if (!s.date_begin)
707 goto skip;
708
709 if (part == 't') { /* date, UNIX timestamp */
710 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
711 return placeholder_len;
712 }
713
714 switch (part) {
715 case 'd': /* date */
716 strbuf_addstr(sb, show_ident_date(&s, dmode));
717 return placeholder_len;
718 case 'D': /* date, RFC2822 style */
719 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
720 return placeholder_len;
721 case 'r': /* date, relative */
722 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
723 return placeholder_len;
724 case 'i': /* date, ISO 8601-like */
725 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
726 return placeholder_len;
727 case 'I': /* date, ISO 8601 strict */
728 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
729 return placeholder_len;
730 }
731
732 skip:
733 /*
734 * reading from either a bogus commit, or a reflog entry with
735 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
736 * to compute a valid return value.
737 */
738 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
739 || part == 'D' || part == 'r' || part == 'i')
740 return placeholder_len;
741
742 return 0; /* unknown placeholder */
743 }
744
745 struct chunk {
746 size_t off;
747 size_t len;
748 };
749
750 enum flush_type {
751 no_flush,
752 flush_right,
753 flush_left,
754 flush_left_and_steal,
755 flush_both
756 };
757
758 enum trunc_type {
759 trunc_none,
760 trunc_left,
761 trunc_middle,
762 trunc_right
763 };
764
765 struct format_commit_context {
766 const struct commit *commit;
767 const struct pretty_print_context *pretty_ctx;
768 unsigned commit_header_parsed:1;
769 unsigned commit_message_parsed:1;
770 struct signature_check signature_check;
771 enum flush_type flush_type;
772 enum trunc_type truncate;
773 const char *message;
774 char *commit_encoding;
775 size_t width, indent1, indent2;
776 int auto_color;
777 int padding;
778
779 /* These offsets are relative to the start of the commit message. */
780 struct chunk author;
781 struct chunk committer;
782 size_t message_off;
783 size_t subject_off;
784 size_t body_off;
785
786 /* The following ones are relative to the result struct strbuf. */
787 size_t wrap_start;
788 };
789
790 static void parse_commit_header(struct format_commit_context *context)
791 {
792 const char *msg = context->message;
793 int i;
794
795 for (i = 0; msg[i]; i++) {
796 const char *name;
797 int eol;
798 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
799 ; /* do nothing */
800
801 if (i == eol) {
802 break;
803 } else if (skip_prefix(msg + i, "author ", &name)) {
804 context->author.off = name - msg;
805 context->author.len = msg + eol - name;
806 } else if (skip_prefix(msg + i, "committer ", &name)) {
807 context->committer.off = name - msg;
808 context->committer.len = msg + eol - name;
809 }
810 i = eol;
811 }
812 context->message_off = i;
813 context->commit_header_parsed = 1;
814 }
815
816 static int istitlechar(char c)
817 {
818 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
819 (c >= '0' && c <= '9') || c == '.' || c == '_';
820 }
821
822 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
823 {
824 size_t trimlen;
825 size_t start_len = sb->len;
826 int space = 2;
827
828 for (; *msg && *msg != '\n'; msg++) {
829 if (istitlechar(*msg)) {
830 if (space == 1)
831 strbuf_addch(sb, '-');
832 space = 0;
833 strbuf_addch(sb, *msg);
834 if (*msg == '.')
835 while (*(msg+1) == '.')
836 msg++;
837 } else
838 space |= 1;
839 }
840
841 /* trim any trailing '.' or '-' characters */
842 trimlen = 0;
843 while (sb->len - trimlen > start_len &&
844 (sb->buf[sb->len - 1 - trimlen] == '.'
845 || sb->buf[sb->len - 1 - trimlen] == '-'))
846 trimlen++;
847 strbuf_remove(sb, sb->len - trimlen, trimlen);
848 }
849
850 const char *format_subject(struct strbuf *sb, const char *msg,
851 const char *line_separator)
852 {
853 int first = 1;
854
855 for (;;) {
856 const char *line = msg;
857 int linelen = get_one_line(line);
858
859 msg += linelen;
860 if (!linelen || is_blank_line(line, &linelen))
861 break;
862
863 if (!sb)
864 continue;
865 strbuf_grow(sb, linelen + 2);
866 if (!first)
867 strbuf_addstr(sb, line_separator);
868 strbuf_add(sb, line, linelen);
869 first = 0;
870 }
871 return msg;
872 }
873
874 static void parse_commit_message(struct format_commit_context *c)
875 {
876 const char *msg = c->message + c->message_off;
877 const char *start = c->message;
878
879 msg = skip_blank_lines(msg);
880 c->subject_off = msg - start;
881
882 msg = format_subject(NULL, msg, NULL);
883 msg = skip_blank_lines(msg);
884 c->body_off = msg - start;
885
886 c->commit_message_parsed = 1;
887 }
888
889 static void strbuf_wrap(struct strbuf *sb, size_t pos,
890 size_t width, size_t indent1, size_t indent2)
891 {
892 struct strbuf tmp = STRBUF_INIT;
893
894 if (pos)
895 strbuf_add(&tmp, sb->buf, pos);
896 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
897 (int) indent1, (int) indent2, (int) width);
898 strbuf_swap(&tmp, sb);
899 strbuf_release(&tmp);
900 }
901
902 static void rewrap_message_tail(struct strbuf *sb,
903 struct format_commit_context *c,
904 size_t new_width, size_t new_indent1,
905 size_t new_indent2)
906 {
907 if (c->width == new_width && c->indent1 == new_indent1 &&
908 c->indent2 == new_indent2)
909 return;
910 if (c->wrap_start < sb->len)
911 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
912 c->wrap_start = sb->len;
913 c->width = new_width;
914 c->indent1 = new_indent1;
915 c->indent2 = new_indent2;
916 }
917
918 static int format_reflog_person(struct strbuf *sb,
919 char part,
920 struct reflog_walk_info *log,
921 const struct date_mode *dmode)
922 {
923 const char *ident;
924
925 if (!log)
926 return 2;
927
928 ident = get_reflog_ident(log);
929 if (!ident)
930 return 2;
931
932 return format_person_part(sb, part, ident, strlen(ident), dmode);
933 }
934
935 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
936 const char *placeholder,
937 struct format_commit_context *c)
938 {
939 const char *rest = placeholder;
940 const char *basic_color = NULL;
941
942 if (placeholder[1] == '(') {
943 const char *begin = placeholder + 2;
944 const char *end = strchr(begin, ')');
945 char color[COLOR_MAXLEN];
946
947 if (!end)
948 return 0;
949
950 if (skip_prefix(begin, "auto,", &begin)) {
951 if (!want_color(c->pretty_ctx->color))
952 return end - placeholder + 1;
953 } else if (skip_prefix(begin, "always,", &begin)) {
954 /* nothing to do; we do not respect want_color at all */
955 } else {
956 /* the default is the same as "auto" */
957 if (!want_color(c->pretty_ctx->color))
958 return end - placeholder + 1;
959 }
960
961 if (color_parse_mem(begin, end - begin, color) < 0)
962 die(_("unable to parse --pretty format"));
963 strbuf_addstr(sb, color);
964 return end - placeholder + 1;
965 }
966
967 /*
968 * We handle things like "%C(red)" above; for historical reasons, there
969 * are a few colors that can be specified without parentheses (and
970 * they cannot support things like "auto" or "always" at all).
971 */
972 if (skip_prefix(placeholder + 1, "red", &rest))
973 basic_color = GIT_COLOR_RED;
974 else if (skip_prefix(placeholder + 1, "green", &rest))
975 basic_color = GIT_COLOR_GREEN;
976 else if (skip_prefix(placeholder + 1, "blue", &rest))
977 basic_color = GIT_COLOR_BLUE;
978 else if (skip_prefix(placeholder + 1, "reset", &rest))
979 basic_color = GIT_COLOR_RESET;
980
981 if (basic_color && want_color(c->pretty_ctx->color))
982 strbuf_addstr(sb, basic_color);
983
984 return rest - placeholder;
985 }
986
987 static size_t parse_padding_placeholder(struct strbuf *sb,
988 const char *placeholder,
989 struct format_commit_context *c)
990 {
991 const char *ch = placeholder;
992 enum flush_type flush_type;
993 int to_column = 0;
994
995 switch (*ch++) {
996 case '<':
997 flush_type = flush_right;
998 break;
999 case '>':
1000 if (*ch == '<') {
1001 flush_type = flush_both;
1002 ch++;
1003 } else if (*ch == '>') {
1004 flush_type = flush_left_and_steal;
1005 ch++;
1006 } else
1007 flush_type = flush_left;
1008 break;
1009 default:
1010 return 0;
1011 }
1012
1013 /* the next value means "wide enough to that column" */
1014 if (*ch == '|') {
1015 to_column = 1;
1016 ch++;
1017 }
1018
1019 if (*ch == '(') {
1020 const char *start = ch + 1;
1021 const char *end = start + strcspn(start, ",)");
1022 char *next;
1023 int width;
1024 if (!end || end == start)
1025 return 0;
1026 width = strtol(start, &next, 10);
1027 if (next == start || width == 0)
1028 return 0;
1029 if (width < 0) {
1030 if (to_column)
1031 width += term_columns();
1032 if (width < 0)
1033 return 0;
1034 }
1035 c->padding = to_column ? -width : width;
1036 c->flush_type = flush_type;
1037
1038 if (*end == ',') {
1039 start = end + 1;
1040 end = strchr(start, ')');
1041 if (!end || end == start)
1042 return 0;
1043 if (starts_with(start, "trunc)"))
1044 c->truncate = trunc_right;
1045 else if (starts_with(start, "ltrunc)"))
1046 c->truncate = trunc_left;
1047 else if (starts_with(start, "mtrunc)"))
1048 c->truncate = trunc_middle;
1049 else
1050 return 0;
1051 } else
1052 c->truncate = trunc_none;
1053
1054 return end - placeholder + 1;
1055 }
1056 return 0;
1057 }
1058
1059 static int match_placeholder_arg(const char *to_parse, const char *candidate,
1060 const char **end)
1061 {
1062 const char *p;
1063
1064 if (!(skip_prefix(to_parse, candidate, &p)))
1065 return 0;
1066 if (*p == ',') {
1067 *end = p + 1;
1068 return 1;
1069 }
1070 if (*p == ')') {
1071 *end = p;
1072 return 1;
1073 }
1074 return 0;
1075 }
1076
1077 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1078 const char *placeholder,
1079 void *context)
1080 {
1081 struct format_commit_context *c = context;
1082 const struct commit *commit = c->commit;
1083 const char *msg = c->message;
1084 struct commit_list *p;
1085 const char *arg;
1086 int ch;
1087
1088 /* these are independent of the commit */
1089 switch (placeholder[0]) {
1090 case 'C':
1091 if (starts_with(placeholder + 1, "(auto)")) {
1092 c->auto_color = want_color(c->pretty_ctx->color);
1093 if (c->auto_color && sb->len)
1094 strbuf_addstr(sb, GIT_COLOR_RESET);
1095 return 7; /* consumed 7 bytes, "C(auto)" */
1096 } else {
1097 int ret = parse_color(sb, placeholder, c);
1098 if (ret)
1099 c->auto_color = 0;
1100 /*
1101 * Otherwise, we decided to treat %C<unknown>
1102 * as a literal string, and the previous
1103 * %C(auto) is still valid.
1104 */
1105 return ret;
1106 }
1107 case 'n': /* newline */
1108 strbuf_addch(sb, '\n');
1109 return 1;
1110 case 'x':
1111 /* %x00 == NUL, %x0a == LF, etc. */
1112 ch = hex2chr(placeholder + 1);
1113 if (ch < 0)
1114 return 0;
1115 strbuf_addch(sb, ch);
1116 return 3;
1117 case 'w':
1118 if (placeholder[1] == '(') {
1119 unsigned long width = 0, indent1 = 0, indent2 = 0;
1120 char *next;
1121 const char *start = placeholder + 2;
1122 const char *end = strchr(start, ')');
1123 if (!end)
1124 return 0;
1125 if (end > start) {
1126 width = strtoul(start, &next, 10);
1127 if (*next == ',') {
1128 indent1 = strtoul(next + 1, &next, 10);
1129 if (*next == ',') {
1130 indent2 = strtoul(next + 1,
1131 &next, 10);
1132 }
1133 }
1134 if (*next != ')')
1135 return 0;
1136 }
1137 rewrap_message_tail(sb, c, width, indent1, indent2);
1138 return end - placeholder + 1;
1139 } else
1140 return 0;
1141
1142 case '<':
1143 case '>':
1144 return parse_padding_placeholder(sb, placeholder, c);
1145 }
1146
1147 /* these depend on the commit */
1148 if (!commit->object.parsed)
1149 parse_object(the_repository, &commit->object.oid);
1150
1151 switch (placeholder[0]) {
1152 case 'H': /* commit hash */
1153 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1154 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1155 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1156 return 1;
1157 case 'h': /* abbreviated commit hash */
1158 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1159 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1160 c->pretty_ctx->abbrev);
1161 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1162 return 1;
1163 case 'T': /* tree hash */
1164 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1165 return 1;
1166 case 't': /* abbreviated tree hash */
1167 strbuf_add_unique_abbrev(sb,
1168 get_commit_tree_oid(commit),
1169 c->pretty_ctx->abbrev);
1170 return 1;
1171 case 'P': /* parent hashes */
1172 for (p = commit->parents; p; p = p->next) {
1173 if (p != commit->parents)
1174 strbuf_addch(sb, ' ');
1175 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1176 }
1177 return 1;
1178 case 'p': /* abbreviated parent hashes */
1179 for (p = commit->parents; p; p = p->next) {
1180 if (p != commit->parents)
1181 strbuf_addch(sb, ' ');
1182 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1183 c->pretty_ctx->abbrev);
1184 }
1185 return 1;
1186 case 'm': /* left/right/bottom */
1187 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1188 return 1;
1189 case 'd':
1190 load_ref_decorations(NULL, DECORATE_SHORT_REFS);
1191 format_decorations(sb, commit, c->auto_color);
1192 return 1;
1193 case 'D':
1194 load_ref_decorations(NULL, DECORATE_SHORT_REFS);
1195 format_decorations_extended(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 'E':
1243 case 'U':
1244 case 'N':
1245 case 'X':
1246 case 'Y':
1247 case 'R':
1248 strbuf_addch(sb, c->signature_check.result);
1249 }
1250 break;
1251 case 'S':
1252 if (c->signature_check.signer)
1253 strbuf_addstr(sb, c->signature_check.signer);
1254 break;
1255 case 'K':
1256 if (c->signature_check.key)
1257 strbuf_addstr(sb, c->signature_check.key);
1258 break;
1259 default:
1260 return 0;
1261 }
1262 return 2;
1263 }
1264
1265
1266 /* For the rest we have to parse the commit header. */
1267 if (!c->commit_header_parsed)
1268 parse_commit_header(c);
1269
1270 switch (placeholder[0]) {
1271 case 'a': /* author ... */
1272 return format_person_part(sb, placeholder[1],
1273 msg + c->author.off, c->author.len,
1274 &c->pretty_ctx->date_mode);
1275 case 'c': /* committer ... */
1276 return format_person_part(sb, placeholder[1],
1277 msg + c->committer.off, c->committer.len,
1278 &c->pretty_ctx->date_mode);
1279 case 'e': /* encoding */
1280 if (c->commit_encoding)
1281 strbuf_addstr(sb, c->commit_encoding);
1282 return 1;
1283 case 'B': /* raw body */
1284 /* message_off is always left at the initial newline */
1285 strbuf_addstr(sb, msg + c->message_off + 1);
1286 return 1;
1287 }
1288
1289 /* Now we need to parse the commit message. */
1290 if (!c->commit_message_parsed)
1291 parse_commit_message(c);
1292
1293 switch (placeholder[0]) {
1294 case 's': /* subject */
1295 format_subject(sb, msg + c->subject_off, " ");
1296 return 1;
1297 case 'f': /* sanitized subject */
1298 format_sanitized_subject(sb, msg + c->subject_off);
1299 return 1;
1300 case 'b': /* body */
1301 strbuf_addstr(sb, msg + c->body_off);
1302 return 1;
1303 }
1304
1305 if (skip_prefix(placeholder, "(trailers", &arg)) {
1306 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1307
1308 opts.no_divider = 1;
1309
1310 if (*arg == ':') {
1311 arg++;
1312 for (;;) {
1313 if (match_placeholder_arg(arg, "only", &arg))
1314 opts.only_trailers = 1;
1315 else if (match_placeholder_arg(arg, "unfold", &arg))
1316 opts.unfold = 1;
1317 else
1318 break;
1319 }
1320 }
1321 if (*arg == ')') {
1322 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1323 return arg - placeholder + 1;
1324 }
1325 }
1326
1327 return 0; /* unknown placeholder */
1328 }
1329
1330 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1331 const char *placeholder,
1332 struct format_commit_context *c)
1333 {
1334 struct strbuf local_sb = STRBUF_INIT;
1335 int total_consumed = 0, len, padding = c->padding;
1336 if (padding < 0) {
1337 const char *start = strrchr(sb->buf, '\n');
1338 int occupied;
1339 if (!start)
1340 start = sb->buf;
1341 occupied = utf8_strnwidth(start, -1, 1);
1342 occupied += c->pretty_ctx->graph_width;
1343 padding = (-padding) - occupied;
1344 }
1345 while (1) {
1346 int modifier = *placeholder == 'C';
1347 int consumed = format_commit_one(&local_sb, placeholder, c);
1348 total_consumed += consumed;
1349
1350 if (!modifier)
1351 break;
1352
1353 placeholder += consumed;
1354 if (*placeholder != '%')
1355 break;
1356 placeholder++;
1357 total_consumed++;
1358 }
1359 len = utf8_strnwidth(local_sb.buf, -1, 1);
1360
1361 if (c->flush_type == flush_left_and_steal) {
1362 const char *ch = sb->buf + sb->len - 1;
1363 while (len > padding && ch > sb->buf) {
1364 const char *p;
1365 if (*ch == ' ') {
1366 ch--;
1367 padding++;
1368 continue;
1369 }
1370 /* check for trailing ansi sequences */
1371 if (*ch != 'm')
1372 break;
1373 p = ch - 1;
1374 while (ch - p < 10 && *p != '\033')
1375 p--;
1376 if (*p != '\033' ||
1377 ch + 1 - p != display_mode_esc_sequence_len(p))
1378 break;
1379 /*
1380 * got a good ansi sequence, put it back to
1381 * local_sb as we're cutting sb
1382 */
1383 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1384 ch = p - 1;
1385 }
1386 strbuf_setlen(sb, ch + 1 - sb->buf);
1387 c->flush_type = flush_left;
1388 }
1389
1390 if (len > padding) {
1391 switch (c->truncate) {
1392 case trunc_left:
1393 strbuf_utf8_replace(&local_sb,
1394 0, len - (padding - 2),
1395 "..");
1396 break;
1397 case trunc_middle:
1398 strbuf_utf8_replace(&local_sb,
1399 padding / 2 - 1,
1400 len - (padding - 2),
1401 "..");
1402 break;
1403 case trunc_right:
1404 strbuf_utf8_replace(&local_sb,
1405 padding - 2, len - (padding - 2),
1406 "..");
1407 break;
1408 case trunc_none:
1409 break;
1410 }
1411 strbuf_addbuf(sb, &local_sb);
1412 } else {
1413 int sb_len = sb->len, offset = 0;
1414 if (c->flush_type == flush_left)
1415 offset = padding - len;
1416 else if (c->flush_type == flush_both)
1417 offset = (padding - len) / 2;
1418 /*
1419 * we calculate padding in columns, now
1420 * convert it back to chars
1421 */
1422 padding = padding - len + local_sb.len;
1423 strbuf_addchars(sb, ' ', padding);
1424 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1425 local_sb.len);
1426 }
1427 strbuf_release(&local_sb);
1428 c->flush_type = no_flush;
1429 return total_consumed;
1430 }
1431
1432 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1433 const char *placeholder,
1434 void *context)
1435 {
1436 int consumed;
1437 size_t orig_len;
1438 enum {
1439 NO_MAGIC,
1440 ADD_LF_BEFORE_NON_EMPTY,
1441 DEL_LF_BEFORE_EMPTY,
1442 ADD_SP_BEFORE_NON_EMPTY
1443 } magic = NO_MAGIC;
1444
1445 switch (placeholder[0]) {
1446 case '-':
1447 magic = DEL_LF_BEFORE_EMPTY;
1448 break;
1449 case '+':
1450 magic = ADD_LF_BEFORE_NON_EMPTY;
1451 break;
1452 case ' ':
1453 magic = ADD_SP_BEFORE_NON_EMPTY;
1454 break;
1455 default:
1456 break;
1457 }
1458 if (magic != NO_MAGIC)
1459 placeholder++;
1460
1461 orig_len = sb->len;
1462 if (((struct format_commit_context *)context)->flush_type != no_flush)
1463 consumed = format_and_pad_commit(sb, placeholder, context);
1464 else
1465 consumed = format_commit_one(sb, placeholder, context);
1466 if (magic == NO_MAGIC)
1467 return consumed;
1468
1469 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1470 while (sb->len && sb->buf[sb->len - 1] == '\n')
1471 strbuf_setlen(sb, sb->len - 1);
1472 } else if (orig_len != sb->len) {
1473 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1474 strbuf_insert(sb, orig_len, "\n", 1);
1475 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1476 strbuf_insert(sb, orig_len, " ", 1);
1477 }
1478 return consumed + 1;
1479 }
1480
1481 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1482 void *context)
1483 {
1484 struct userformat_want *w = context;
1485
1486 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1487 placeholder++;
1488
1489 switch (*placeholder) {
1490 case 'N':
1491 w->notes = 1;
1492 break;
1493 }
1494 return 0;
1495 }
1496
1497 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1498 {
1499 struct strbuf dummy = STRBUF_INIT;
1500
1501 if (!fmt) {
1502 if (!user_format)
1503 return;
1504 fmt = user_format;
1505 }
1506 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1507 strbuf_release(&dummy);
1508 }
1509
1510 void format_commit_message(const struct commit *commit,
1511 const char *format, struct strbuf *sb,
1512 const struct pretty_print_context *pretty_ctx)
1513 {
1514 struct format_commit_context context;
1515 const char *output_enc = pretty_ctx->output_encoding;
1516 const char *utf8 = "UTF-8";
1517
1518 memset(&context, 0, sizeof(context));
1519 context.commit = commit;
1520 context.pretty_ctx = pretty_ctx;
1521 context.wrap_start = sb->len;
1522 /*
1523 * convert a commit message to UTF-8 first
1524 * as far as 'format_commit_item' assumes it in UTF-8
1525 */
1526 context.message = logmsg_reencode(commit,
1527 &context.commit_encoding,
1528 utf8);
1529
1530 strbuf_expand(sb, format, format_commit_item, &context);
1531 rewrap_message_tail(sb, &context, 0, 0, 0);
1532
1533 /* then convert a commit message to an actual output encoding */
1534 if (output_enc) {
1535 if (same_encoding(utf8, output_enc))
1536 output_enc = NULL;
1537 } else {
1538 if (context.commit_encoding &&
1539 !same_encoding(context.commit_encoding, utf8))
1540 output_enc = context.commit_encoding;
1541 }
1542
1543 if (output_enc) {
1544 size_t outsz;
1545 char *out = reencode_string_len(sb->buf, sb->len,
1546 output_enc, utf8, &outsz);
1547 if (out)
1548 strbuf_attach(sb, out, outsz, outsz + 1);
1549 }
1550
1551 free(context.commit_encoding);
1552 unuse_commit_buffer(commit, context.message);
1553 }
1554
1555 static void pp_header(struct pretty_print_context *pp,
1556 const char *encoding,
1557 const struct commit *commit,
1558 const char **msg_p,
1559 struct strbuf *sb)
1560 {
1561 int parents_shown = 0;
1562
1563 for (;;) {
1564 const char *name, *line = *msg_p;
1565 int linelen = get_one_line(*msg_p);
1566
1567 if (!linelen)
1568 return;
1569 *msg_p += linelen;
1570
1571 if (linelen == 1)
1572 /* End of header */
1573 return;
1574
1575 if (pp->fmt == CMIT_FMT_RAW) {
1576 strbuf_add(sb, line, linelen);
1577 continue;
1578 }
1579
1580 if (starts_with(line, "parent ")) {
1581 if (linelen != the_hash_algo->hexsz + 8)
1582 die("bad parent line in commit");
1583 continue;
1584 }
1585
1586 if (!parents_shown) {
1587 unsigned num = commit_list_count(commit->parents);
1588 /* with enough slop */
1589 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1590 add_merge_info(pp, sb, commit);
1591 parents_shown = 1;
1592 }
1593
1594 /*
1595 * MEDIUM == DEFAULT shows only author with dates.
1596 * FULL shows both authors but not dates.
1597 * FULLER shows both authors and dates.
1598 */
1599 if (skip_prefix(line, "author ", &name)) {
1600 strbuf_grow(sb, linelen + 80);
1601 pp_user_info(pp, "Author", sb, name, encoding);
1602 }
1603 if (skip_prefix(line, "committer ", &name) &&
1604 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1605 strbuf_grow(sb, linelen + 80);
1606 pp_user_info(pp, "Commit", sb, name, encoding);
1607 }
1608 }
1609 }
1610
1611 void pp_title_line(struct pretty_print_context *pp,
1612 const char **msg_p,
1613 struct strbuf *sb,
1614 const char *encoding,
1615 int need_8bit_cte)
1616 {
1617 static const int max_length = 78; /* per rfc2047 */
1618 struct strbuf title;
1619
1620 strbuf_init(&title, 80);
1621 *msg_p = format_subject(&title, *msg_p,
1622 pp->preserve_subject ? "\n" : " ");
1623
1624 strbuf_grow(sb, title.len + 1024);
1625 if (pp->print_email_subject) {
1626 if (pp->rev)
1627 fmt_output_email_subject(sb, pp->rev);
1628 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1629 add_rfc2047(sb, title.buf, title.len,
1630 encoding, RFC2047_SUBJECT);
1631 else
1632 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1633 -last_line_length(sb), 1, max_length);
1634 } else {
1635 strbuf_addbuf(sb, &title);
1636 }
1637 strbuf_addch(sb, '\n');
1638
1639 if (need_8bit_cte == 0) {
1640 int i;
1641 for (i = 0; i < pp->in_body_headers.nr; i++) {
1642 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1643 need_8bit_cte = 1;
1644 break;
1645 }
1646 }
1647 }
1648
1649 if (need_8bit_cte > 0) {
1650 const char *header_fmt =
1651 "MIME-Version: 1.0\n"
1652 "Content-Type: text/plain; charset=%s\n"
1653 "Content-Transfer-Encoding: 8bit\n";
1654 strbuf_addf(sb, header_fmt, encoding);
1655 }
1656 if (pp->after_subject) {
1657 strbuf_addstr(sb, pp->after_subject);
1658 }
1659 if (cmit_fmt_is_mail(pp->fmt)) {
1660 strbuf_addch(sb, '\n');
1661 }
1662
1663 if (pp->in_body_headers.nr) {
1664 int i;
1665 for (i = 0; i < pp->in_body_headers.nr; i++) {
1666 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1667 free(pp->in_body_headers.items[i].string);
1668 }
1669 string_list_clear(&pp->in_body_headers, 0);
1670 strbuf_addch(sb, '\n');
1671 }
1672
1673 strbuf_release(&title);
1674 }
1675
1676 static int pp_utf8_width(const char *start, const char *end)
1677 {
1678 int width = 0;
1679 size_t remain = end - start;
1680
1681 while (remain) {
1682 int n = utf8_width(&start, &remain);
1683 if (n < 0 || !start)
1684 return -1;
1685 width += n;
1686 }
1687 return width;
1688 }
1689
1690 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1691 const char *line, int linelen)
1692 {
1693 const char *tab;
1694
1695 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1696 int width = pp_utf8_width(line, tab);
1697
1698 /*
1699 * If it wasn't well-formed utf8, or it
1700 * had characters with badly defined
1701 * width (control characters etc), just
1702 * give up on trying to align things.
1703 */
1704 if (width < 0)
1705 break;
1706
1707 /* Output the data .. */
1708 strbuf_add(sb, line, tab - line);
1709
1710 /* .. and the de-tabified tab */
1711 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1712
1713 /* Skip over the printed part .. */
1714 linelen -= tab + 1 - line;
1715 line = tab + 1;
1716 }
1717
1718 /*
1719 * Print out everything after the last tab without
1720 * worrying about width - there's nothing more to
1721 * align.
1722 */
1723 strbuf_add(sb, line, linelen);
1724 }
1725
1726 /*
1727 * pp_handle_indent() prints out the intendation, and
1728 * the whole line (without the final newline), after
1729 * de-tabifying.
1730 */
1731 static void pp_handle_indent(struct pretty_print_context *pp,
1732 struct strbuf *sb, int indent,
1733 const char *line, int linelen)
1734 {
1735 strbuf_addchars(sb, ' ', indent);
1736 if (pp->expand_tabs_in_log)
1737 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1738 else
1739 strbuf_add(sb, line, linelen);
1740 }
1741
1742 static int is_mboxrd_from(const char *line, int len)
1743 {
1744 /*
1745 * a line matching /^From $/ here would only have len == 4
1746 * at this point because is_empty_line would've trimmed all
1747 * trailing space
1748 */
1749 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1750 }
1751
1752 void pp_remainder(struct pretty_print_context *pp,
1753 const char **msg_p,
1754 struct strbuf *sb,
1755 int indent)
1756 {
1757 int first = 1;
1758 for (;;) {
1759 const char *line = *msg_p;
1760 int linelen = get_one_line(line);
1761 *msg_p += linelen;
1762
1763 if (!linelen)
1764 break;
1765
1766 if (is_blank_line(line, &linelen)) {
1767 if (first)
1768 continue;
1769 if (pp->fmt == CMIT_FMT_SHORT)
1770 break;
1771 }
1772 first = 0;
1773
1774 strbuf_grow(sb, linelen + indent + 20);
1775 if (indent)
1776 pp_handle_indent(pp, sb, indent, line, linelen);
1777 else if (pp->expand_tabs_in_log)
1778 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1779 line, linelen);
1780 else {
1781 if (pp->fmt == CMIT_FMT_MBOXRD &&
1782 is_mboxrd_from(line, linelen))
1783 strbuf_addch(sb, '>');
1784
1785 strbuf_add(sb, line, linelen);
1786 }
1787 strbuf_addch(sb, '\n');
1788 }
1789 }
1790
1791 void pretty_print_commit(struct pretty_print_context *pp,
1792 const struct commit *commit,
1793 struct strbuf *sb)
1794 {
1795 unsigned long beginning_of_body;
1796 int indent = 4;
1797 const char *msg;
1798 const char *reencoded;
1799 const char *encoding;
1800 int need_8bit_cte = pp->need_8bit_cte;
1801
1802 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1803 format_commit_message(commit, user_format, sb, pp);
1804 return;
1805 }
1806
1807 encoding = get_log_output_encoding();
1808 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1809
1810 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1811 indent = 0;
1812
1813 /*
1814 * We need to check and emit Content-type: to mark it
1815 * as 8-bit if we haven't done so.
1816 */
1817 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1818 int i, ch, in_body;
1819
1820 for (in_body = i = 0; (ch = msg[i]); i++) {
1821 if (!in_body) {
1822 /* author could be non 7-bit ASCII but
1823 * the log may be so; skip over the
1824 * header part first.
1825 */
1826 if (ch == '\n' && msg[i+1] == '\n')
1827 in_body = 1;
1828 }
1829 else if (non_ascii(ch)) {
1830 need_8bit_cte = 1;
1831 break;
1832 }
1833 }
1834 }
1835
1836 pp_header(pp, encoding, commit, &msg, sb);
1837 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
1838 strbuf_addch(sb, '\n');
1839 }
1840
1841 /* Skip excess blank lines at the beginning of body, if any... */
1842 msg = skip_blank_lines(msg);
1843
1844 /* These formats treat the title line specially. */
1845 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1846 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1847
1848 beginning_of_body = sb->len;
1849 if (pp->fmt != CMIT_FMT_ONELINE)
1850 pp_remainder(pp, &msg, sb, indent);
1851 strbuf_rtrim(sb);
1852
1853 /* Make sure there is an EOLN for the non-oneline case */
1854 if (pp->fmt != CMIT_FMT_ONELINE)
1855 strbuf_addch(sb, '\n');
1856
1857 /*
1858 * The caller may append additional body text in e-mail
1859 * format. Make sure we did not strip the blank line
1860 * between the header and the body.
1861 */
1862 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1863 strbuf_addch(sb, '\n');
1864
1865 unuse_commit_buffer(commit, reencoded);
1866 }
1867
1868 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1869 struct strbuf *sb)
1870 {
1871 struct pretty_print_context pp = {0};
1872 pp.fmt = fmt;
1873 pretty_print_commit(&pp, commit, sb);
1874 }