object: convert parse_object* to take struct object_id
[git/git.git] / pretty.c
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 #include "trailer.h"
14
15 static char *user_format;
16 static struct cmt_fmt_map {
17 const char *name;
18 enum cmit_fmt format;
19 int is_tformat;
20 int expand_tabs_in_log;
21 int is_alias;
22 const char *user_format;
23 } *commit_formats;
24 static size_t builtin_formats_len;
25 static size_t commit_formats_len;
26 static size_t commit_formats_alloc;
27 static struct cmt_fmt_map *find_commit_format(const char *sought);
28
29 int commit_format_is_empty(enum cmit_fmt fmt)
30 {
31 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
32 }
33
34 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
35 {
36 free(user_format);
37 user_format = xstrdup(cp);
38 if (is_tformat)
39 rev->use_terminator = 1;
40 rev->commit_format = CMIT_FMT_USERFORMAT;
41 }
42
43 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
44 {
45 struct cmt_fmt_map *commit_format = NULL;
46 const char *name;
47 const char *fmt;
48 int i;
49
50 if (!skip_prefix(var, "pretty.", &name))
51 return 0;
52
53 for (i = 0; i < builtin_formats_len; i++) {
54 if (!strcmp(commit_formats[i].name, name))
55 return 0;
56 }
57
58 for (i = builtin_formats_len; i < commit_formats_len; i++) {
59 if (!strcmp(commit_formats[i].name, name)) {
60 commit_format = &commit_formats[i];
61 break;
62 }
63 }
64
65 if (!commit_format) {
66 ALLOC_GROW(commit_formats, commit_formats_len+1,
67 commit_formats_alloc);
68 commit_format = &commit_formats[commit_formats_len];
69 memset(commit_format, 0, sizeof(*commit_format));
70 commit_formats_len++;
71 }
72
73 commit_format->name = xstrdup(name);
74 commit_format->format = CMIT_FMT_USERFORMAT;
75 if (git_config_string(&fmt, var, value))
76 return -1;
77
78 if (skip_prefix(fmt, "format:", &fmt))
79 commit_format->is_tformat = 0;
80 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
81 commit_format->is_tformat = 1;
82 else
83 commit_format->is_alias = 1;
84 commit_format->user_format = fmt;
85
86 return 0;
87 }
88
89 static void setup_commit_formats(void)
90 {
91 struct cmt_fmt_map builtin_formats[] = {
92 { "raw", CMIT_FMT_RAW, 0, 0 },
93 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
94 { "short", CMIT_FMT_SHORT, 0, 0 },
95 { "email", CMIT_FMT_EMAIL, 0, 0 },
96 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
97 { "fuller", CMIT_FMT_FULLER, 0, 8 },
98 { "full", CMIT_FMT_FULL, 0, 8 },
99 { "oneline", CMIT_FMT_ONELINE, 1, 0 }
100 };
101 commit_formats_len = ARRAY_SIZE(builtin_formats);
102 builtin_formats_len = commit_formats_len;
103 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
104 memcpy(commit_formats, builtin_formats,
105 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
106
107 git_config(git_pretty_formats_config, NULL);
108 }
109
110 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
111 const char *original,
112 int num_redirections)
113 {
114 struct cmt_fmt_map *found = NULL;
115 size_t found_match_len = 0;
116 int i;
117
118 if (num_redirections >= commit_formats_len)
119 die("invalid --pretty format: "
120 "'%s' references an alias which points to itself",
121 original);
122
123 for (i = 0; i < commit_formats_len; i++) {
124 size_t match_len;
125
126 if (!starts_with(commit_formats[i].name, sought))
127 continue;
128
129 match_len = strlen(commit_formats[i].name);
130 if (found == NULL || found_match_len > match_len) {
131 found = &commit_formats[i];
132 found_match_len = match_len;
133 }
134 }
135
136 if (found && found->is_alias) {
137 found = find_commit_format_recursive(found->user_format,
138 original,
139 num_redirections+1);
140 }
141
142 return found;
143 }
144
145 static struct cmt_fmt_map *find_commit_format(const char *sought)
146 {
147 if (!commit_formats)
148 setup_commit_formats();
149
150 return find_commit_format_recursive(sought, sought, 0);
151 }
152
153 void get_commit_format(const char *arg, struct rev_info *rev)
154 {
155 struct cmt_fmt_map *commit_format;
156
157 rev->use_terminator = 0;
158 if (!arg) {
159 rev->commit_format = CMIT_FMT_DEFAULT;
160 return;
161 }
162 if (skip_prefix(arg, "format:", &arg)) {
163 save_user_format(rev, arg, 0);
164 return;
165 }
166
167 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
168 save_user_format(rev, arg, 1);
169 return;
170 }
171
172 commit_format = find_commit_format(arg);
173 if (!commit_format)
174 die("invalid --pretty format: %s", arg);
175
176 rev->commit_format = commit_format->format;
177 rev->use_terminator = commit_format->is_tformat;
178 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
179 if (commit_format->format == CMIT_FMT_USERFORMAT) {
180 save_user_format(rev, commit_format->user_format,
181 commit_format->is_tformat);
182 }
183 }
184
185 /*
186 * Generic support for pretty-printing the header
187 */
188 static int get_one_line(const char *msg)
189 {
190 int ret = 0;
191
192 for (;;) {
193 char c = *msg++;
194 if (!c)
195 break;
196 ret++;
197 if (c == '\n')
198 break;
199 }
200 return ret;
201 }
202
203 /* High bit set, or ISO-2022-INT */
204 static int non_ascii(int ch)
205 {
206 return !isascii(ch) || ch == '\033';
207 }
208
209 int has_non_ascii(const char *s)
210 {
211 int ch;
212 if (!s)
213 return 0;
214 while ((ch = *s++) != '\0') {
215 if (non_ascii(ch))
216 return 1;
217 }
218 return 0;
219 }
220
221 static int is_rfc822_special(char ch)
222 {
223 switch (ch) {
224 case '(':
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 return 1;
238 default:
239 return 0;
240 }
241 }
242
243 static int needs_rfc822_quoting(const char *s, int len)
244 {
245 int i;
246 for (i = 0; i < len; i++)
247 if (is_rfc822_special(s[i]))
248 return 1;
249 return 0;
250 }
251
252 static int last_line_length(struct strbuf *sb)
253 {
254 int i;
255
256 /* How many bytes are already used on the last line? */
257 for (i = sb->len - 1; i >= 0; i--)
258 if (sb->buf[i] == '\n')
259 break;
260 return sb->len - (i + 1);
261 }
262
263 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
264 {
265 int i;
266
267 /* just a guess, we may have to also backslash-quote */
268 strbuf_grow(out, len + 2);
269
270 strbuf_addch(out, '"');
271 for (i = 0; i < len; i++) {
272 switch (s[i]) {
273 case '"':
274 case '\\':
275 strbuf_addch(out, '\\');
276 /* fall through */
277 default:
278 strbuf_addch(out, s[i]);
279 }
280 }
281 strbuf_addch(out, '"');
282 }
283
284 enum rfc2047_type {
285 RFC2047_SUBJECT,
286 RFC2047_ADDRESS
287 };
288
289 static int is_rfc2047_special(char ch, enum rfc2047_type type)
290 {
291 /*
292 * rfc2047, section 4.2:
293 *
294 * 8-bit values which correspond to printable ASCII characters other
295 * than "=", "?", and "_" (underscore), MAY be represented as those
296 * characters. (But see section 5 for restrictions.) In
297 * particular, SPACE and TAB MUST NOT be represented as themselves
298 * within encoded words.
299 */
300
301 /*
302 * rule out non-ASCII characters and non-printable characters (the
303 * non-ASCII check should be redundant as isprint() is not localized
304 * and only knows about ASCII, but be defensive about that)
305 */
306 if (non_ascii(ch) || !isprint(ch))
307 return 1;
308
309 /*
310 * rule out special printable characters (' ' should be the only
311 * whitespace character considered printable, but be defensive and use
312 * isspace())
313 */
314 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
315 return 1;
316
317 /*
318 * rfc2047, section 5.3:
319 *
320 * As a replacement for a 'word' entity within a 'phrase', for example,
321 * one that precedes an address in a From, To, or Cc header. The ABNF
322 * definition for 'phrase' from RFC 822 thus becomes:
323 *
324 * phrase = 1*( encoded-word / word )
325 *
326 * In this case the set of characters that may be used in a "Q"-encoded
327 * 'encoded-word' is restricted to: <upper and lower case ASCII
328 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
329 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
330 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
331 * 'special' by 'linear-white-space'.
332 */
333
334 if (type != RFC2047_ADDRESS)
335 return 0;
336
337 /* '=' and '_' are special cases and have been checked above */
338 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
339 }
340
341 static int needs_rfc2047_encoding(const char *line, int len,
342 enum rfc2047_type type)
343 {
344 int i;
345
346 for (i = 0; i < len; i++) {
347 int ch = line[i];
348 if (non_ascii(ch) || ch == '\n')
349 return 1;
350 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
351 return 1;
352 }
353
354 return 0;
355 }
356
357 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
358 const char *encoding, enum rfc2047_type type)
359 {
360 static const int max_encoded_length = 76; /* per rfc2047 */
361 int i;
362 int line_len = last_line_length(sb);
363
364 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
365 strbuf_addf(sb, "=?%s?q?", encoding);
366 line_len += strlen(encoding) + 5; /* 5 for =??q? */
367
368 while (len) {
369 /*
370 * RFC 2047, section 5 (3):
371 *
372 * Each 'encoded-word' MUST represent an integral number of
373 * characters. A multi-octet character may not be split across
374 * adjacent 'encoded- word's.
375 */
376 const unsigned char *p = (const unsigned char *)line;
377 int chrlen = mbs_chrlen(&line, &len, encoding);
378 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
379
380 /* "=%02X" * chrlen, or the byte itself */
381 const char *encoded_fmt = is_special ? "=%02X" : "%c";
382 int encoded_len = is_special ? 3 * chrlen : 1;
383
384 /*
385 * According to RFC 2047, we could encode the special character
386 * ' ' (space) with '_' (underscore) for readability. But many
387 * programs do not understand this and just leave the
388 * underscore in place. Thus, we do nothing special here, which
389 * causes ' ' to be encoded as '=20', avoiding this problem.
390 */
391
392 if (line_len + encoded_len + 2 > max_encoded_length) {
393 /* It won't fit with trailing "?=" --- break the line */
394 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
395 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
396 }
397
398 for (i = 0; i < chrlen; i++)
399 strbuf_addf(sb, encoded_fmt, p[i]);
400 line_len += encoded_len;
401 }
402 strbuf_addstr(sb, "?=");
403 }
404
405 const char *show_ident_date(const struct ident_split *ident,
406 const struct date_mode *mode)
407 {
408 unsigned long date = 0;
409 long tz = 0;
410
411 if (ident->date_begin && ident->date_end)
412 date = strtoul(ident->date_begin, NULL, 10);
413 if (date_overflows(date))
414 date = 0;
415 else {
416 if (ident->tz_begin && ident->tz_end)
417 tz = strtol(ident->tz_begin, NULL, 10);
418 if (tz >= INT_MAX || tz <= INT_MIN)
419 tz = 0;
420 }
421 return show_date(date, tz, mode);
422 }
423
424 void pp_user_info(struct pretty_print_context *pp,
425 const char *what, struct strbuf *sb,
426 const char *line, const char *encoding)
427 {
428 struct ident_split ident;
429 char *line_end;
430 const char *mailbuf, *namebuf;
431 size_t namelen, maillen;
432 int max_length = 78; /* per rfc2822 */
433
434 if (pp->fmt == CMIT_FMT_ONELINE)
435 return;
436
437 line_end = strchrnul(line, '\n');
438 if (split_ident_line(&ident, line, line_end - line))
439 return;
440
441 mailbuf = ident.mail_begin;
442 maillen = ident.mail_end - ident.mail_begin;
443 namebuf = ident.name_begin;
444 namelen = ident.name_end - ident.name_begin;
445
446 if (pp->mailmap)
447 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
448
449 if (cmit_fmt_is_mail(pp->fmt)) {
450 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
451 struct strbuf buf = STRBUF_INIT;
452
453 strbuf_addstr(&buf, "From: ");
454 strbuf_add(&buf, namebuf, namelen);
455 strbuf_addstr(&buf, " <");
456 strbuf_add(&buf, mailbuf, maillen);
457 strbuf_addstr(&buf, ">\n");
458 string_list_append(&pp->in_body_headers,
459 strbuf_detach(&buf, NULL));
460
461 mailbuf = pp->from_ident->mail_begin;
462 maillen = pp->from_ident->mail_end - mailbuf;
463 namebuf = pp->from_ident->name_begin;
464 namelen = pp->from_ident->name_end - namebuf;
465 }
466
467 strbuf_addstr(sb, "From: ");
468 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
469 add_rfc2047(sb, namebuf, namelen,
470 encoding, RFC2047_ADDRESS);
471 max_length = 76; /* per rfc2047 */
472 } else if (needs_rfc822_quoting(namebuf, namelen)) {
473 struct strbuf quoted = STRBUF_INIT;
474 add_rfc822_quoted(&quoted, namebuf, namelen);
475 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
476 -6, 1, max_length);
477 strbuf_release(&quoted);
478 } else {
479 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
480 -6, 1, max_length);
481 }
482
483 if (max_length <
484 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
485 strbuf_addch(sb, '\n');
486 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
487 } else {
488 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
489 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
490 (int)namelen, namebuf, (int)maillen, mailbuf);
491 }
492
493 switch (pp->fmt) {
494 case CMIT_FMT_MEDIUM:
495 strbuf_addf(sb, "Date: %s\n",
496 show_ident_date(&ident, &pp->date_mode));
497 break;
498 case CMIT_FMT_EMAIL:
499 case CMIT_FMT_MBOXRD:
500 strbuf_addf(sb, "Date: %s\n",
501 show_ident_date(&ident, DATE_MODE(RFC2822)));
502 break;
503 case CMIT_FMT_FULLER:
504 strbuf_addf(sb, "%sDate: %s\n", what,
505 show_ident_date(&ident, &pp->date_mode));
506 break;
507 default:
508 /* notin' */
509 break;
510 }
511 }
512
513 static int is_blank_line(const char *line, int *len_p)
514 {
515 int len = *len_p;
516 while (len && isspace(line[len - 1]))
517 len--;
518 *len_p = len;
519 return !len;
520 }
521
522 const char *skip_blank_lines(const char *msg)
523 {
524 for (;;) {
525 int linelen = get_one_line(msg);
526 int ll = linelen;
527 if (!linelen)
528 break;
529 if (!is_blank_line(msg, &ll))
530 break;
531 msg += linelen;
532 }
533 return msg;
534 }
535
536 static void add_merge_info(const struct pretty_print_context *pp,
537 struct strbuf *sb, const struct commit *commit)
538 {
539 struct commit_list *parent = commit->parents;
540
541 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
542 !parent || !parent->next)
543 return;
544
545 strbuf_addstr(sb, "Merge:");
546
547 while (parent) {
548 struct object_id *oidp = &parent->item->object.oid;
549 strbuf_addch(sb, ' ');
550 if (pp->abbrev)
551 strbuf_add_unique_abbrev(sb, oidp->hash, pp->abbrev);
552 else
553 strbuf_addstr(sb, oid_to_hex(oidp));
554 parent = parent->next;
555 }
556 strbuf_addch(sb, '\n');
557 }
558
559 static char *get_header(const char *msg, const char *key)
560 {
561 size_t len;
562 const char *v = find_commit_header(msg, key, &len);
563 return v ? xmemdupz(v, len) : NULL;
564 }
565
566 static char *replace_encoding_header(char *buf, const char *encoding)
567 {
568 struct strbuf tmp = STRBUF_INIT;
569 size_t start, len;
570 char *cp = buf;
571
572 /* guess if there is an encoding header before a \n\n */
573 while (!starts_with(cp, "encoding ")) {
574 cp = strchr(cp, '\n');
575 if (!cp || *++cp == '\n')
576 return buf;
577 }
578 start = cp - buf;
579 cp = strchr(cp, '\n');
580 if (!cp)
581 return buf; /* should not happen but be defensive */
582 len = cp + 1 - (buf + start);
583
584 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
585 if (is_encoding_utf8(encoding)) {
586 /* we have re-coded to UTF-8; drop the header */
587 strbuf_remove(&tmp, start, len);
588 } else {
589 /* just replaces XXXX in 'encoding XXXX\n' */
590 strbuf_splice(&tmp, start + strlen("encoding "),
591 len - strlen("encoding \n"),
592 encoding, strlen(encoding));
593 }
594 return strbuf_detach(&tmp, NULL);
595 }
596
597 const char *logmsg_reencode(const struct commit *commit,
598 char **commit_encoding,
599 const char *output_encoding)
600 {
601 static const char *utf8 = "UTF-8";
602 const char *use_encoding;
603 char *encoding;
604 const char *msg = get_commit_buffer(commit, NULL);
605 char *out;
606
607 if (!output_encoding || !*output_encoding) {
608 if (commit_encoding)
609 *commit_encoding = get_header(msg, "encoding");
610 return msg;
611 }
612 encoding = get_header(msg, "encoding");
613 if (commit_encoding)
614 *commit_encoding = encoding;
615 use_encoding = encoding ? encoding : utf8;
616 if (same_encoding(use_encoding, output_encoding)) {
617 /*
618 * No encoding work to be done. If we have no encoding header
619 * at all, then there's nothing to do, and we can return the
620 * message verbatim (whether newly allocated or not).
621 */
622 if (!encoding)
623 return msg;
624
625 /*
626 * Otherwise, we still want to munge the encoding header in the
627 * result, which will be done by modifying the buffer. If we
628 * are using a fresh copy, we can reuse it. But if we are using
629 * the cached copy from get_commit_buffer, we need to duplicate it
630 * to avoid munging the cached copy.
631 */
632 if (msg == get_cached_commit_buffer(commit, NULL))
633 out = xstrdup(msg);
634 else
635 out = (char *)msg;
636 }
637 else {
638 /*
639 * There's actual encoding work to do. Do the reencoding, which
640 * still leaves the header to be replaced in the next step. At
641 * this point, we are done with msg. If we allocated a fresh
642 * copy, we can free it.
643 */
644 out = reencode_string(msg, output_encoding, use_encoding);
645 if (out)
646 unuse_commit_buffer(commit, msg);
647 }
648
649 /*
650 * This replacement actually consumes the buffer we hand it, so we do
651 * not have to worry about freeing the old "out" here.
652 */
653 if (out)
654 out = replace_encoding_header(out, output_encoding);
655
656 if (!commit_encoding)
657 free(encoding);
658 /*
659 * If the re-encoding failed, out might be NULL here; in that
660 * case we just return the commit message verbatim.
661 */
662 return out ? out : msg;
663 }
664
665 static int mailmap_name(const char **email, size_t *email_len,
666 const char **name, size_t *name_len)
667 {
668 static struct string_list *mail_map;
669 if (!mail_map) {
670 mail_map = xcalloc(1, sizeof(*mail_map));
671 read_mailmap(mail_map, NULL);
672 }
673 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
674 }
675
676 static size_t format_person_part(struct strbuf *sb, char part,
677 const char *msg, int len,
678 const struct date_mode *dmode)
679 {
680 /* currently all placeholders have same length */
681 const int placeholder_len = 2;
682 struct ident_split s;
683 const char *name, *mail;
684 size_t maillen, namelen;
685
686 if (split_ident_line(&s, msg, len) < 0)
687 goto skip;
688
689 name = s.name_begin;
690 namelen = s.name_end - s.name_begin;
691 mail = s.mail_begin;
692 maillen = s.mail_end - s.mail_begin;
693
694 if (part == 'N' || part == 'E') /* mailmap lookup */
695 mailmap_name(&mail, &maillen, &name, &namelen);
696 if (part == 'n' || part == 'N') { /* name */
697 strbuf_add(sb, name, namelen);
698 return placeholder_len;
699 }
700 if (part == 'e' || part == 'E') { /* email */
701 strbuf_add(sb, mail, maillen);
702 return placeholder_len;
703 }
704
705 if (!s.date_begin)
706 goto skip;
707
708 if (part == 't') { /* date, UNIX timestamp */
709 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
710 return placeholder_len;
711 }
712
713 switch (part) {
714 case 'd': /* date */
715 strbuf_addstr(sb, show_ident_date(&s, dmode));
716 return placeholder_len;
717 case 'D': /* date, RFC2822 style */
718 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
719 return placeholder_len;
720 case 'r': /* date, relative */
721 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
722 return placeholder_len;
723 case 'i': /* date, ISO 8601-like */
724 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
725 return placeholder_len;
726 case 'I': /* date, ISO 8601 strict */
727 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
728 return placeholder_len;
729 }
730
731 skip:
732 /*
733 * reading from either a bogus commit, or a reflog entry with
734 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
735 * to compute a valid return value.
736 */
737 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
738 || part == 'D' || part == 'r' || part == 'i')
739 return placeholder_len;
740
741 return 0; /* unknown placeholder */
742 }
743
744 struct chunk {
745 size_t off;
746 size_t len;
747 };
748
749 enum flush_type {
750 no_flush,
751 flush_right,
752 flush_left,
753 flush_left_and_steal,
754 flush_both
755 };
756
757 enum trunc_type {
758 trunc_none,
759 trunc_left,
760 trunc_middle,
761 trunc_right
762 };
763
764 struct format_commit_context {
765 const struct commit *commit;
766 const struct pretty_print_context *pretty_ctx;
767 unsigned commit_header_parsed:1;
768 unsigned commit_message_parsed:1;
769 struct signature_check signature_check;
770 enum flush_type flush_type;
771 enum trunc_type truncate;
772 const char *message;
773 char *commit_encoding;
774 size_t width, indent1, indent2;
775 int auto_color;
776 int padding;
777
778 /* These offsets are relative to the start of the commit message. */
779 struct chunk author;
780 struct chunk committer;
781 size_t message_off;
782 size_t subject_off;
783 size_t body_off;
784
785 /* The following ones are relative to the result struct strbuf. */
786 struct chunk abbrev_commit_hash;
787 struct chunk abbrev_tree_hash;
788 struct chunk abbrev_parent_hashes;
789 size_t wrap_start;
790 };
791
792 static int add_again(struct strbuf *sb, struct chunk *chunk)
793 {
794 if (chunk->len) {
795 strbuf_adddup(sb, chunk->off, chunk->len);
796 return 1;
797 }
798
799 /*
800 * We haven't seen this chunk before. Our caller is surely
801 * going to add it the hard way now. Remember the most likely
802 * start of the to-be-added chunk: the current end of the
803 * struct strbuf.
804 */
805 chunk->off = sb->len;
806 return 0;
807 }
808
809 static void parse_commit_header(struct format_commit_context *context)
810 {
811 const char *msg = context->message;
812 int i;
813
814 for (i = 0; msg[i]; i++) {
815 const char *name;
816 int eol;
817 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
818 ; /* do nothing */
819
820 if (i == eol) {
821 break;
822 } else if (skip_prefix(msg + i, "author ", &name)) {
823 context->author.off = name - msg;
824 context->author.len = msg + eol - name;
825 } else if (skip_prefix(msg + i, "committer ", &name)) {
826 context->committer.off = name - msg;
827 context->committer.len = msg + eol - name;
828 }
829 i = eol;
830 }
831 context->message_off = i;
832 context->commit_header_parsed = 1;
833 }
834
835 static int istitlechar(char c)
836 {
837 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
838 (c >= '0' && c <= '9') || c == '.' || c == '_';
839 }
840
841 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
842 {
843 size_t trimlen;
844 size_t start_len = sb->len;
845 int space = 2;
846
847 for (; *msg && *msg != '\n'; msg++) {
848 if (istitlechar(*msg)) {
849 if (space == 1)
850 strbuf_addch(sb, '-');
851 space = 0;
852 strbuf_addch(sb, *msg);
853 if (*msg == '.')
854 while (*(msg+1) == '.')
855 msg++;
856 } else
857 space |= 1;
858 }
859
860 /* trim any trailing '.' or '-' characters */
861 trimlen = 0;
862 while (sb->len - trimlen > start_len &&
863 (sb->buf[sb->len - 1 - trimlen] == '.'
864 || sb->buf[sb->len - 1 - trimlen] == '-'))
865 trimlen++;
866 strbuf_remove(sb, sb->len - trimlen, trimlen);
867 }
868
869 const char *format_subject(struct strbuf *sb, const char *msg,
870 const char *line_separator)
871 {
872 int first = 1;
873
874 for (;;) {
875 const char *line = msg;
876 int linelen = get_one_line(line);
877
878 msg += linelen;
879 if (!linelen || is_blank_line(line, &linelen))
880 break;
881
882 if (!sb)
883 continue;
884 strbuf_grow(sb, linelen + 2);
885 if (!first)
886 strbuf_addstr(sb, line_separator);
887 strbuf_add(sb, line, linelen);
888 first = 0;
889 }
890 return msg;
891 }
892
893 static void format_trailers(struct strbuf *sb, const char *msg)
894 {
895 struct trailer_info info;
896
897 trailer_info_get(&info, msg);
898 strbuf_add(sb, info.trailer_start,
899 info.trailer_end - info.trailer_start);
900 trailer_info_release(&info);
901 }
902
903 static void parse_commit_message(struct format_commit_context *c)
904 {
905 const char *msg = c->message + c->message_off;
906 const char *start = c->message;
907
908 msg = skip_blank_lines(msg);
909 c->subject_off = msg - start;
910
911 msg = format_subject(NULL, msg, NULL);
912 msg = skip_blank_lines(msg);
913 c->body_off = msg - start;
914
915 c->commit_message_parsed = 1;
916 }
917
918 static void strbuf_wrap(struct strbuf *sb, size_t pos,
919 size_t width, size_t indent1, size_t indent2)
920 {
921 struct strbuf tmp = STRBUF_INIT;
922
923 if (pos)
924 strbuf_add(&tmp, sb->buf, pos);
925 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
926 (int) indent1, (int) indent2, (int) width);
927 strbuf_swap(&tmp, sb);
928 strbuf_release(&tmp);
929 }
930
931 static void rewrap_message_tail(struct strbuf *sb,
932 struct format_commit_context *c,
933 size_t new_width, size_t new_indent1,
934 size_t new_indent2)
935 {
936 if (c->width == new_width && c->indent1 == new_indent1 &&
937 c->indent2 == new_indent2)
938 return;
939 if (c->wrap_start < sb->len)
940 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
941 c->wrap_start = sb->len;
942 c->width = new_width;
943 c->indent1 = new_indent1;
944 c->indent2 = new_indent2;
945 }
946
947 static int format_reflog_person(struct strbuf *sb,
948 char part,
949 struct reflog_walk_info *log,
950 const struct date_mode *dmode)
951 {
952 const char *ident;
953
954 if (!log)
955 return 2;
956
957 ident = get_reflog_ident(log);
958 if (!ident)
959 return 2;
960
961 return format_person_part(sb, part, ident, strlen(ident), dmode);
962 }
963
964 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
965 const char *placeholder,
966 struct format_commit_context *c)
967 {
968 const char *rest = placeholder;
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 (skip_prefix(begin, "auto,", &begin)) {
978 if (!want_color(c->pretty_ctx->color))
979 return end - placeholder + 1;
980 }
981 if (color_parse_mem(begin, end - begin, color) < 0)
982 die(_("unable to parse --pretty format"));
983 strbuf_addstr(sb, color);
984 return end - placeholder + 1;
985 }
986 if (skip_prefix(placeholder + 1, "red", &rest))
987 strbuf_addstr(sb, GIT_COLOR_RED);
988 else if (skip_prefix(placeholder + 1, "green", &rest))
989 strbuf_addstr(sb, GIT_COLOR_GREEN);
990 else if (skip_prefix(placeholder + 1, "blue", &rest))
991 strbuf_addstr(sb, GIT_COLOR_BLUE);
992 else if (skip_prefix(placeholder + 1, "reset", &rest))
993 strbuf_addstr(sb, GIT_COLOR_RESET);
994 return rest - placeholder;
995 }
996
997 static size_t parse_padding_placeholder(struct strbuf *sb,
998 const char *placeholder,
999 struct format_commit_context *c)
1000 {
1001 const char *ch = placeholder;
1002 enum flush_type flush_type;
1003 int to_column = 0;
1004
1005 switch (*ch++) {
1006 case '<':
1007 flush_type = flush_right;
1008 break;
1009 case '>':
1010 if (*ch == '<') {
1011 flush_type = flush_both;
1012 ch++;
1013 } else if (*ch == '>') {
1014 flush_type = flush_left_and_steal;
1015 ch++;
1016 } else
1017 flush_type = flush_left;
1018 break;
1019 default:
1020 return 0;
1021 }
1022
1023 /* the next value means "wide enough to that column" */
1024 if (*ch == '|') {
1025 to_column = 1;
1026 ch++;
1027 }
1028
1029 if (*ch == '(') {
1030 const char *start = ch + 1;
1031 const char *end = start + strcspn(start, ",)");
1032 char *next;
1033 int width;
1034 if (!end || end == start)
1035 return 0;
1036 width = strtol(start, &next, 10);
1037 if (next == start || width == 0)
1038 return 0;
1039 if (width < 0) {
1040 if (to_column)
1041 width += term_columns();
1042 if (width < 0)
1043 return 0;
1044 }
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 ch;
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 = want_color(c->pretty_ctx->color);
1084 if (c->auto_color && sb->len)
1085 strbuf_addstr(sb, GIT_COLOR_RESET);
1086 return 7; /* consumed 7 bytes, "C(auto)" */
1087 } else {
1088 int ret = parse_color(sb, placeholder, c);
1089 if (ret)
1090 c->auto_color = 0;
1091 /*
1092 * Otherwise, we decided to treat %C<unknown>
1093 * as a literal string, and the previous
1094 * %C(auto) is still valid.
1095 */
1096 return ret;
1097 }
1098 case 'n': /* newline */
1099 strbuf_addch(sb, '\n');
1100 return 1;
1101 case 'x':
1102 /* %x00 == NUL, %x0a == LF, etc. */
1103 ch = hex2chr(placeholder + 1);
1104 if (ch < 0)
1105 return 0;
1106 strbuf_addch(sb, ch);
1107 return 3;
1108 case 'w':
1109 if (placeholder[1] == '(') {
1110 unsigned long width = 0, indent1 = 0, indent2 = 0;
1111 char *next;
1112 const char *start = placeholder + 2;
1113 const char *end = strchr(start, ')');
1114 if (!end)
1115 return 0;
1116 if (end > start) {
1117 width = strtoul(start, &next, 10);
1118 if (*next == ',') {
1119 indent1 = strtoul(next + 1, &next, 10);
1120 if (*next == ',') {
1121 indent2 = strtoul(next + 1,
1122 &next, 10);
1123 }
1124 }
1125 if (*next != ')')
1126 return 0;
1127 }
1128 rewrap_message_tail(sb, c, width, indent1, indent2);
1129 return end - placeholder + 1;
1130 } else
1131 return 0;
1132
1133 case '<':
1134 case '>':
1135 return parse_padding_placeholder(sb, placeholder, c);
1136 }
1137
1138 /* these depend on the commit */
1139 if (!commit->object.parsed)
1140 parse_object(&commit->object.oid);
1141
1142 switch (placeholder[0]) {
1143 case 'H': /* commit hash */
1144 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1145 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1146 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1147 return 1;
1148 case 'h': /* abbreviated commit hash */
1149 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1150 if (add_again(sb, &c->abbrev_commit_hash)) {
1151 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1152 return 1;
1153 }
1154 strbuf_add_unique_abbrev(sb, commit->object.oid.hash,
1155 c->pretty_ctx->abbrev);
1156 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1157 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1158 return 1;
1159 case 'T': /* tree hash */
1160 strbuf_addstr(sb, oid_to_hex(&commit->tree->object.oid));
1161 return 1;
1162 case 't': /* abbreviated tree hash */
1163 if (add_again(sb, &c->abbrev_tree_hash))
1164 return 1;
1165 strbuf_add_unique_abbrev(sb, commit->tree->object.oid.hash,
1166 c->pretty_ctx->abbrev);
1167 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1168 return 1;
1169 case 'P': /* parent hashes */
1170 for (p = commit->parents; p; p = p->next) {
1171 if (p != commit->parents)
1172 strbuf_addch(sb, ' ');
1173 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1174 }
1175 return 1;
1176 case 'p': /* abbreviated parent hashes */
1177 if (add_again(sb, &c->abbrev_parent_hashes))
1178 return 1;
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.hash,
1183 c->pretty_ctx->abbrev);
1184 }
1185 c->abbrev_parent_hashes.len = sb->len -
1186 c->abbrev_parent_hashes.off;
1187 return 1;
1188 case 'm': /* left/right/bottom */
1189 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1190 return 1;
1191 case 'd':
1192 load_ref_decorations(DECORATE_SHORT_REFS);
1193 format_decorations(sb, commit, c->auto_color);
1194 return 1;
1195 case 'D':
1196 load_ref_decorations(DECORATE_SHORT_REFS);
1197 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1198 return 1;
1199 case 'g': /* reflog info */
1200 switch(placeholder[1]) {
1201 case 'd': /* reflog selector */
1202 case 'D':
1203 if (c->pretty_ctx->reflog_info)
1204 get_reflog_selector(sb,
1205 c->pretty_ctx->reflog_info,
1206 &c->pretty_ctx->date_mode,
1207 c->pretty_ctx->date_mode_explicit,
1208 (placeholder[1] == 'd'));
1209 return 2;
1210 case 's': /* reflog message */
1211 if (c->pretty_ctx->reflog_info)
1212 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1213 return 2;
1214 case 'n':
1215 case 'N':
1216 case 'e':
1217 case 'E':
1218 return format_reflog_person(sb,
1219 placeholder[1],
1220 c->pretty_ctx->reflog_info,
1221 &c->pretty_ctx->date_mode);
1222 }
1223 return 0; /* unknown %g placeholder */
1224 case 'N':
1225 if (c->pretty_ctx->notes_message) {
1226 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1227 return 1;
1228 }
1229 return 0;
1230 }
1231
1232 if (placeholder[0] == 'G') {
1233 if (!c->signature_check.result)
1234 check_commit_signature(c->commit, &(c->signature_check));
1235 switch (placeholder[1]) {
1236 case 'G':
1237 if (c->signature_check.gpg_output)
1238 strbuf_addstr(sb, c->signature_check.gpg_output);
1239 break;
1240 case '?':
1241 switch (c->signature_check.result) {
1242 case 'G':
1243 case 'B':
1244 case 'E':
1245 case 'U':
1246 case 'N':
1247 case 'X':
1248 case 'Y':
1249 case 'R':
1250 strbuf_addch(sb, c->signature_check.result);
1251 }
1252 break;
1253 case 'S':
1254 if (c->signature_check.signer)
1255 strbuf_addstr(sb, c->signature_check.signer);
1256 break;
1257 case 'K':
1258 if (c->signature_check.key)
1259 strbuf_addstr(sb, c->signature_check.key);
1260 break;
1261 default:
1262 return 0;
1263 }
1264 return 2;
1265 }
1266
1267
1268 /* For the rest we have to parse the commit header. */
1269 if (!c->commit_header_parsed)
1270 parse_commit_header(c);
1271
1272 switch (placeholder[0]) {
1273 case 'a': /* author ... */
1274 return format_person_part(sb, placeholder[1],
1275 msg + c->author.off, c->author.len,
1276 &c->pretty_ctx->date_mode);
1277 case 'c': /* committer ... */
1278 return format_person_part(sb, placeholder[1],
1279 msg + c->committer.off, c->committer.len,
1280 &c->pretty_ctx->date_mode);
1281 case 'e': /* encoding */
1282 if (c->commit_encoding)
1283 strbuf_addstr(sb, c->commit_encoding);
1284 return 1;
1285 case 'B': /* raw body */
1286 /* message_off is always left at the initial newline */
1287 strbuf_addstr(sb, msg + c->message_off + 1);
1288 return 1;
1289 }
1290
1291 /* Now we need to parse the commit message. */
1292 if (!c->commit_message_parsed)
1293 parse_commit_message(c);
1294
1295 switch (placeholder[0]) {
1296 case 's': /* subject */
1297 format_subject(sb, msg + c->subject_off, " ");
1298 return 1;
1299 case 'f': /* sanitized subject */
1300 format_sanitized_subject(sb, msg + c->subject_off);
1301 return 1;
1302 case 'b': /* body */
1303 strbuf_addstr(sb, msg + c->body_off);
1304 return 1;
1305 }
1306
1307 if (starts_with(placeholder, "(trailers)")) {
1308 format_trailers(sb, msg + c->subject_off);
1309 return strlen("(trailers)");
1310 }
1311
1312 return 0; /* unknown placeholder */
1313 }
1314
1315 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1316 const char *placeholder,
1317 struct format_commit_context *c)
1318 {
1319 struct strbuf local_sb = STRBUF_INIT;
1320 int total_consumed = 0, len, padding = c->padding;
1321 if (padding < 0) {
1322 const char *start = strrchr(sb->buf, '\n');
1323 int occupied;
1324 if (!start)
1325 start = sb->buf;
1326 occupied = utf8_strnwidth(start, -1, 1);
1327 occupied += c->pretty_ctx->graph_width;
1328 padding = (-padding) - occupied;
1329 }
1330 while (1) {
1331 int modifier = *placeholder == 'C';
1332 int consumed = format_commit_one(&local_sb, placeholder, c);
1333 total_consumed += consumed;
1334
1335 if (!modifier)
1336 break;
1337
1338 placeholder += consumed;
1339 if (*placeholder != '%')
1340 break;
1341 placeholder++;
1342 total_consumed++;
1343 }
1344 len = utf8_strnwidth(local_sb.buf, -1, 1);
1345
1346 if (c->flush_type == flush_left_and_steal) {
1347 const char *ch = sb->buf + sb->len - 1;
1348 while (len > padding && ch > sb->buf) {
1349 const char *p;
1350 if (*ch == ' ') {
1351 ch--;
1352 padding++;
1353 continue;
1354 }
1355 /* check for trailing ansi sequences */
1356 if (*ch != 'm')
1357 break;
1358 p = ch - 1;
1359 while (ch - p < 10 && *p != '\033')
1360 p--;
1361 if (*p != '\033' ||
1362 ch + 1 - p != display_mode_esc_sequence_len(p))
1363 break;
1364 /*
1365 * got a good ansi sequence, put it back to
1366 * local_sb as we're cutting sb
1367 */
1368 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1369 ch = p - 1;
1370 }
1371 strbuf_setlen(sb, ch + 1 - sb->buf);
1372 c->flush_type = flush_left;
1373 }
1374
1375 if (len > padding) {
1376 switch (c->truncate) {
1377 case trunc_left:
1378 strbuf_utf8_replace(&local_sb,
1379 0, len - (padding - 2),
1380 "..");
1381 break;
1382 case trunc_middle:
1383 strbuf_utf8_replace(&local_sb,
1384 padding / 2 - 1,
1385 len - (padding - 2),
1386 "..");
1387 break;
1388 case trunc_right:
1389 strbuf_utf8_replace(&local_sb,
1390 padding - 2, len - (padding - 2),
1391 "..");
1392 break;
1393 case trunc_none:
1394 break;
1395 }
1396 strbuf_addbuf(sb, &local_sb);
1397 } else {
1398 int sb_len = sb->len, offset = 0;
1399 if (c->flush_type == flush_left)
1400 offset = padding - len;
1401 else if (c->flush_type == flush_both)
1402 offset = (padding - len) / 2;
1403 /*
1404 * we calculate padding in columns, now
1405 * convert it back to chars
1406 */
1407 padding = padding - len + local_sb.len;
1408 strbuf_addchars(sb, ' ', padding);
1409 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1410 local_sb.len);
1411 }
1412 strbuf_release(&local_sb);
1413 c->flush_type = no_flush;
1414 return total_consumed;
1415 }
1416
1417 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1418 const char *placeholder,
1419 void *context)
1420 {
1421 int consumed;
1422 size_t orig_len;
1423 enum {
1424 NO_MAGIC,
1425 ADD_LF_BEFORE_NON_EMPTY,
1426 DEL_LF_BEFORE_EMPTY,
1427 ADD_SP_BEFORE_NON_EMPTY
1428 } magic = NO_MAGIC;
1429
1430 switch (placeholder[0]) {
1431 case '-':
1432 magic = DEL_LF_BEFORE_EMPTY;
1433 break;
1434 case '+':
1435 magic = ADD_LF_BEFORE_NON_EMPTY;
1436 break;
1437 case ' ':
1438 magic = ADD_SP_BEFORE_NON_EMPTY;
1439 break;
1440 default:
1441 break;
1442 }
1443 if (magic != NO_MAGIC)
1444 placeholder++;
1445
1446 orig_len = sb->len;
1447 if (((struct format_commit_context *)context)->flush_type != no_flush)
1448 consumed = format_and_pad_commit(sb, placeholder, context);
1449 else
1450 consumed = format_commit_one(sb, placeholder, context);
1451 if (magic == NO_MAGIC)
1452 return consumed;
1453
1454 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1455 while (sb->len && sb->buf[sb->len - 1] == '\n')
1456 strbuf_setlen(sb, sb->len - 1);
1457 } else if (orig_len != sb->len) {
1458 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1459 strbuf_insert(sb, orig_len, "\n", 1);
1460 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1461 strbuf_insert(sb, orig_len, " ", 1);
1462 }
1463 return consumed + 1;
1464 }
1465
1466 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1467 void *context)
1468 {
1469 struct userformat_want *w = context;
1470
1471 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1472 placeholder++;
1473
1474 switch (*placeholder) {
1475 case 'N':
1476 w->notes = 1;
1477 break;
1478 }
1479 return 0;
1480 }
1481
1482 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1483 {
1484 struct strbuf dummy = STRBUF_INIT;
1485
1486 if (!fmt) {
1487 if (!user_format)
1488 return;
1489 fmt = user_format;
1490 }
1491 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1492 strbuf_release(&dummy);
1493 }
1494
1495 void format_commit_message(const struct commit *commit,
1496 const char *format, struct strbuf *sb,
1497 const struct pretty_print_context *pretty_ctx)
1498 {
1499 struct format_commit_context context;
1500 const char *output_enc = pretty_ctx->output_encoding;
1501 const char *utf8 = "UTF-8";
1502
1503 memset(&context, 0, sizeof(context));
1504 context.commit = commit;
1505 context.pretty_ctx = pretty_ctx;
1506 context.wrap_start = sb->len;
1507 /*
1508 * convert a commit message to UTF-8 first
1509 * as far as 'format_commit_item' assumes it in UTF-8
1510 */
1511 context.message = logmsg_reencode(commit,
1512 &context.commit_encoding,
1513 utf8);
1514
1515 strbuf_expand(sb, format, format_commit_item, &context);
1516 rewrap_message_tail(sb, &context, 0, 0, 0);
1517
1518 /* then convert a commit message to an actual output encoding */
1519 if (output_enc) {
1520 if (same_encoding(utf8, output_enc))
1521 output_enc = NULL;
1522 } else {
1523 if (context.commit_encoding &&
1524 !same_encoding(context.commit_encoding, utf8))
1525 output_enc = context.commit_encoding;
1526 }
1527
1528 if (output_enc) {
1529 int outsz;
1530 char *out = reencode_string_len(sb->buf, sb->len,
1531 output_enc, utf8, &outsz);
1532 if (out)
1533 strbuf_attach(sb, out, outsz, outsz + 1);
1534 }
1535
1536 free(context.commit_encoding);
1537 unuse_commit_buffer(commit, context.message);
1538 }
1539
1540 static void pp_header(struct pretty_print_context *pp,
1541 const char *encoding,
1542 const struct commit *commit,
1543 const char **msg_p,
1544 struct strbuf *sb)
1545 {
1546 int parents_shown = 0;
1547
1548 for (;;) {
1549 const char *name, *line = *msg_p;
1550 int linelen = get_one_line(*msg_p);
1551
1552 if (!linelen)
1553 return;
1554 *msg_p += linelen;
1555
1556 if (linelen == 1)
1557 /* End of header */
1558 return;
1559
1560 if (pp->fmt == CMIT_FMT_RAW) {
1561 strbuf_add(sb, line, linelen);
1562 continue;
1563 }
1564
1565 if (starts_with(line, "parent ")) {
1566 if (linelen != 48)
1567 die("bad parent line in commit");
1568 continue;
1569 }
1570
1571 if (!parents_shown) {
1572 unsigned num = commit_list_count(commit->parents);
1573 /* with enough slop */
1574 strbuf_grow(sb, num * 50 + 20);
1575 add_merge_info(pp, sb, commit);
1576 parents_shown = 1;
1577 }
1578
1579 /*
1580 * MEDIUM == DEFAULT shows only author with dates.
1581 * FULL shows both authors but not dates.
1582 * FULLER shows both authors and dates.
1583 */
1584 if (skip_prefix(line, "author ", &name)) {
1585 strbuf_grow(sb, linelen + 80);
1586 pp_user_info(pp, "Author", sb, name, encoding);
1587 }
1588 if (skip_prefix(line, "committer ", &name) &&
1589 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1590 strbuf_grow(sb, linelen + 80);
1591 pp_user_info(pp, "Commit", sb, name, encoding);
1592 }
1593 }
1594 }
1595
1596 void pp_title_line(struct pretty_print_context *pp,
1597 const char **msg_p,
1598 struct strbuf *sb,
1599 const char *encoding,
1600 int need_8bit_cte)
1601 {
1602 static const int max_length = 78; /* per rfc2047 */
1603 struct strbuf title;
1604
1605 strbuf_init(&title, 80);
1606 *msg_p = format_subject(&title, *msg_p,
1607 pp->preserve_subject ? "\n" : " ");
1608
1609 strbuf_grow(sb, title.len + 1024);
1610 if (pp->print_email_subject) {
1611 if (pp->rev)
1612 fmt_output_email_subject(sb, pp->rev);
1613 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1614 add_rfc2047(sb, title.buf, title.len,
1615 encoding, RFC2047_SUBJECT);
1616 else
1617 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1618 -last_line_length(sb), 1, max_length);
1619 } else {
1620 strbuf_addbuf(sb, &title);
1621 }
1622 strbuf_addch(sb, '\n');
1623
1624 if (need_8bit_cte == 0) {
1625 int i;
1626 for (i = 0; i < pp->in_body_headers.nr; i++) {
1627 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1628 need_8bit_cte = 1;
1629 break;
1630 }
1631 }
1632 }
1633
1634 if (need_8bit_cte > 0) {
1635 const char *header_fmt =
1636 "MIME-Version: 1.0\n"
1637 "Content-Type: text/plain; charset=%s\n"
1638 "Content-Transfer-Encoding: 8bit\n";
1639 strbuf_addf(sb, header_fmt, encoding);
1640 }
1641 if (pp->after_subject) {
1642 strbuf_addstr(sb, pp->after_subject);
1643 }
1644 if (cmit_fmt_is_mail(pp->fmt)) {
1645 strbuf_addch(sb, '\n');
1646 }
1647
1648 if (pp->in_body_headers.nr) {
1649 int i;
1650 for (i = 0; i < pp->in_body_headers.nr; i++) {
1651 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1652 free(pp->in_body_headers.items[i].string);
1653 }
1654 string_list_clear(&pp->in_body_headers, 0);
1655 strbuf_addch(sb, '\n');
1656 }
1657
1658 strbuf_release(&title);
1659 }
1660
1661 static int pp_utf8_width(const char *start, const char *end)
1662 {
1663 int width = 0;
1664 size_t remain = end - start;
1665
1666 while (remain) {
1667 int n = utf8_width(&start, &remain);
1668 if (n < 0 || !start)
1669 return -1;
1670 width += n;
1671 }
1672 return width;
1673 }
1674
1675 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1676 const char *line, int linelen)
1677 {
1678 const char *tab;
1679
1680 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1681 int width = pp_utf8_width(line, tab);
1682
1683 /*
1684 * If it wasn't well-formed utf8, or it
1685 * had characters with badly defined
1686 * width (control characters etc), just
1687 * give up on trying to align things.
1688 */
1689 if (width < 0)
1690 break;
1691
1692 /* Output the data .. */
1693 strbuf_add(sb, line, tab - line);
1694
1695 /* .. and the de-tabified tab */
1696 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1697
1698 /* Skip over the printed part .. */
1699 linelen -= tab + 1 - line;
1700 line = tab + 1;
1701 }
1702
1703 /*
1704 * Print out everything after the last tab without
1705 * worrying about width - there's nothing more to
1706 * align.
1707 */
1708 strbuf_add(sb, line, linelen);
1709 }
1710
1711 /*
1712 * pp_handle_indent() prints out the intendation, and
1713 * the whole line (without the final newline), after
1714 * de-tabifying.
1715 */
1716 static void pp_handle_indent(struct pretty_print_context *pp,
1717 struct strbuf *sb, int indent,
1718 const char *line, int linelen)
1719 {
1720 strbuf_addchars(sb, ' ', indent);
1721 if (pp->expand_tabs_in_log)
1722 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1723 else
1724 strbuf_add(sb, line, linelen);
1725 }
1726
1727 static int is_mboxrd_from(const char *line, int len)
1728 {
1729 /*
1730 * a line matching /^From $/ here would only have len == 4
1731 * at this point because is_empty_line would've trimmed all
1732 * trailing space
1733 */
1734 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1735 }
1736
1737 void pp_remainder(struct pretty_print_context *pp,
1738 const char **msg_p,
1739 struct strbuf *sb,
1740 int indent)
1741 {
1742 int first = 1;
1743 for (;;) {
1744 const char *line = *msg_p;
1745 int linelen = get_one_line(line);
1746 *msg_p += linelen;
1747
1748 if (!linelen)
1749 break;
1750
1751 if (is_blank_line(line, &linelen)) {
1752 if (first)
1753 continue;
1754 if (pp->fmt == CMIT_FMT_SHORT)
1755 break;
1756 }
1757 first = 0;
1758
1759 strbuf_grow(sb, linelen + indent + 20);
1760 if (indent)
1761 pp_handle_indent(pp, sb, indent, line, linelen);
1762 else if (pp->expand_tabs_in_log)
1763 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1764 line, linelen);
1765 else {
1766 if (pp->fmt == CMIT_FMT_MBOXRD &&
1767 is_mboxrd_from(line, linelen))
1768 strbuf_addch(sb, '>');
1769
1770 strbuf_add(sb, line, linelen);
1771 }
1772 strbuf_addch(sb, '\n');
1773 }
1774 }
1775
1776 void pretty_print_commit(struct pretty_print_context *pp,
1777 const struct commit *commit,
1778 struct strbuf *sb)
1779 {
1780 unsigned long beginning_of_body;
1781 int indent = 4;
1782 const char *msg;
1783 const char *reencoded;
1784 const char *encoding;
1785 int need_8bit_cte = pp->need_8bit_cte;
1786
1787 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1788 format_commit_message(commit, user_format, sb, pp);
1789 return;
1790 }
1791
1792 encoding = get_log_output_encoding();
1793 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1794
1795 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1796 indent = 0;
1797
1798 /*
1799 * We need to check and emit Content-type: to mark it
1800 * as 8-bit if we haven't done so.
1801 */
1802 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1803 int i, ch, in_body;
1804
1805 for (in_body = i = 0; (ch = msg[i]); i++) {
1806 if (!in_body) {
1807 /* author could be non 7-bit ASCII but
1808 * the log may be so; skip over the
1809 * header part first.
1810 */
1811 if (ch == '\n' && msg[i+1] == '\n')
1812 in_body = 1;
1813 }
1814 else if (non_ascii(ch)) {
1815 need_8bit_cte = 1;
1816 break;
1817 }
1818 }
1819 }
1820
1821 pp_header(pp, encoding, commit, &msg, sb);
1822 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
1823 strbuf_addch(sb, '\n');
1824 }
1825
1826 /* Skip excess blank lines at the beginning of body, if any... */
1827 msg = skip_blank_lines(msg);
1828
1829 /* These formats treat the title line specially. */
1830 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1831 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1832
1833 beginning_of_body = sb->len;
1834 if (pp->fmt != CMIT_FMT_ONELINE)
1835 pp_remainder(pp, &msg, sb, indent);
1836 strbuf_rtrim(sb);
1837
1838 /* Make sure there is an EOLN for the non-oneline case */
1839 if (pp->fmt != CMIT_FMT_ONELINE)
1840 strbuf_addch(sb, '\n');
1841
1842 /*
1843 * The caller may append additional body text in e-mail
1844 * format. Make sure we did not strip the blank line
1845 * between the header and the body.
1846 */
1847 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1848 strbuf_addch(sb, '\n');
1849
1850 unuse_commit_buffer(commit, reencoded);
1851 }
1852
1853 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1854 struct strbuf *sb)
1855 {
1856 struct pretty_print_context pp = {0};
1857 pp.fmt = fmt;
1858 pretty_print_commit(&pp, commit, sb);
1859 }