commit: prepare logmsg_reencode to handle arbitrary repositories
[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 *repo_logmsg_reencode(struct repository *r,
599 const struct commit *commit,
600 char **commit_encoding,
601 const char *output_encoding)
602 {
603 static const char *utf8 = "UTF-8";
604 const char *use_encoding;
605 char *encoding;
606 const char *msg = repo_get_commit_buffer(r, commit, NULL);
607 char *out;
608
609 if (!output_encoding || !*output_encoding) {
610 if (commit_encoding)
611 *commit_encoding = get_header(msg, "encoding");
612 return msg;
613 }
614 encoding = get_header(msg, "encoding");
615 if (commit_encoding)
616 *commit_encoding = encoding;
617 use_encoding = encoding ? encoding : utf8;
618 if (same_encoding(use_encoding, output_encoding)) {
619 /*
620 * No encoding work to be done. If we have no encoding header
621 * at all, then there's nothing to do, and we can return the
622 * message verbatim (whether newly allocated or not).
623 */
624 if (!encoding)
625 return msg;
626
627 /*
628 * Otherwise, we still want to munge the encoding header in the
629 * result, which will be done by modifying the buffer. If we
630 * are using a fresh copy, we can reuse it. But if we are using
631 * the cached copy from get_commit_buffer, we need to duplicate it
632 * to avoid munging the cached copy.
633 */
634 if (msg == get_cached_commit_buffer(r, commit, NULL))
635 out = xstrdup(msg);
636 else
637 out = (char *)msg;
638 }
639 else {
640 /*
641 * There's actual encoding work to do. Do the reencoding, which
642 * still leaves the header to be replaced in the next step. At
643 * this point, we are done with msg. If we allocated a fresh
644 * copy, we can free it.
645 */
646 out = reencode_string(msg, output_encoding, use_encoding);
647 if (out)
648 repo_unuse_commit_buffer(r, commit, msg);
649 }
650
651 /*
652 * This replacement actually consumes the buffer we hand it, so we do
653 * not have to worry about freeing the old "out" here.
654 */
655 if (out)
656 out = replace_encoding_header(out, output_encoding);
657
658 if (!commit_encoding)
659 free(encoding);
660 /*
661 * If the re-encoding failed, out might be NULL here; in that
662 * case we just return the commit message verbatim.
663 */
664 return out ? out : msg;
665 }
666
667 static int mailmap_name(const char **email, size_t *email_len,
668 const char **name, size_t *name_len)
669 {
670 static struct string_list *mail_map;
671 if (!mail_map) {
672 mail_map = xcalloc(1, sizeof(*mail_map));
673 read_mailmap(mail_map, NULL);
674 }
675 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
676 }
677
678 static size_t format_person_part(struct strbuf *sb, char part,
679 const char *msg, int len,
680 const struct date_mode *dmode)
681 {
682 /* currently all placeholders have same length */
683 const int placeholder_len = 2;
684 struct ident_split s;
685 const char *name, *mail;
686 size_t maillen, namelen;
687
688 if (split_ident_line(&s, msg, len) < 0)
689 goto skip;
690
691 name = s.name_begin;
692 namelen = s.name_end - s.name_begin;
693 mail = s.mail_begin;
694 maillen = s.mail_end - s.mail_begin;
695
696 if (part == 'N' || part == 'E') /* mailmap lookup */
697 mailmap_name(&mail, &maillen, &name, &namelen);
698 if (part == 'n' || part == 'N') { /* name */
699 strbuf_add(sb, name, namelen);
700 return placeholder_len;
701 }
702 if (part == 'e' || part == 'E') { /* email */
703 strbuf_add(sb, mail, maillen);
704 return placeholder_len;
705 }
706
707 if (!s.date_begin)
708 goto skip;
709
710 if (part == 't') { /* date, UNIX timestamp */
711 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
712 return placeholder_len;
713 }
714
715 switch (part) {
716 case 'd': /* date */
717 strbuf_addstr(sb, show_ident_date(&s, dmode));
718 return placeholder_len;
719 case 'D': /* date, RFC2822 style */
720 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
721 return placeholder_len;
722 case 'r': /* date, relative */
723 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
724 return placeholder_len;
725 case 'i': /* date, ISO 8601-like */
726 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
727 return placeholder_len;
728 case 'I': /* date, ISO 8601 strict */
729 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
730 return placeholder_len;
731 }
732
733 skip:
734 /*
735 * reading from either a bogus commit, or a reflog entry with
736 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
737 * to compute a valid return value.
738 */
739 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
740 || part == 'D' || part == 'r' || part == 'i')
741 return placeholder_len;
742
743 return 0; /* unknown placeholder */
744 }
745
746 struct chunk {
747 size_t off;
748 size_t len;
749 };
750
751 enum flush_type {
752 no_flush,
753 flush_right,
754 flush_left,
755 flush_left_and_steal,
756 flush_both
757 };
758
759 enum trunc_type {
760 trunc_none,
761 trunc_left,
762 trunc_middle,
763 trunc_right
764 };
765
766 struct format_commit_context {
767 const struct commit *commit;
768 const struct pretty_print_context *pretty_ctx;
769 unsigned commit_header_parsed:1;
770 unsigned commit_message_parsed:1;
771 struct signature_check signature_check;
772 enum flush_type flush_type;
773 enum trunc_type truncate;
774 const char *message;
775 char *commit_encoding;
776 size_t width, indent1, indent2;
777 int auto_color;
778 int padding;
779
780 /* These offsets are relative to the start of the commit message. */
781 struct chunk author;
782 struct chunk committer;
783 size_t message_off;
784 size_t subject_off;
785 size_t body_off;
786
787 /* The following ones are relative to the result struct strbuf. */
788 size_t wrap_start;
789 };
790
791 static void parse_commit_header(struct format_commit_context *context)
792 {
793 const char *msg = context->message;
794 int i;
795
796 for (i = 0; msg[i]; i++) {
797 const char *name;
798 int eol;
799 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
800 ; /* do nothing */
801
802 if (i == eol) {
803 break;
804 } else if (skip_prefix(msg + i, "author ", &name)) {
805 context->author.off = name - msg;
806 context->author.len = msg + eol - name;
807 } else if (skip_prefix(msg + i, "committer ", &name)) {
808 context->committer.off = name - msg;
809 context->committer.len = msg + eol - name;
810 }
811 i = eol;
812 }
813 context->message_off = i;
814 context->commit_header_parsed = 1;
815 }
816
817 static int istitlechar(char c)
818 {
819 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
820 (c >= '0' && c <= '9') || c == '.' || c == '_';
821 }
822
823 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
824 {
825 size_t trimlen;
826 size_t start_len = sb->len;
827 int space = 2;
828
829 for (; *msg && *msg != '\n'; msg++) {
830 if (istitlechar(*msg)) {
831 if (space == 1)
832 strbuf_addch(sb, '-');
833 space = 0;
834 strbuf_addch(sb, *msg);
835 if (*msg == '.')
836 while (*(msg+1) == '.')
837 msg++;
838 } else
839 space |= 1;
840 }
841
842 /* trim any trailing '.' or '-' characters */
843 trimlen = 0;
844 while (sb->len - trimlen > start_len &&
845 (sb->buf[sb->len - 1 - trimlen] == '.'
846 || sb->buf[sb->len - 1 - trimlen] == '-'))
847 trimlen++;
848 strbuf_remove(sb, sb->len - trimlen, trimlen);
849 }
850
851 const char *format_subject(struct strbuf *sb, const char *msg,
852 const char *line_separator)
853 {
854 int first = 1;
855
856 for (;;) {
857 const char *line = msg;
858 int linelen = get_one_line(line);
859
860 msg += linelen;
861 if (!linelen || is_blank_line(line, &linelen))
862 break;
863
864 if (!sb)
865 continue;
866 strbuf_grow(sb, linelen + 2);
867 if (!first)
868 strbuf_addstr(sb, line_separator);
869 strbuf_add(sb, line, linelen);
870 first = 0;
871 }
872 return msg;
873 }
874
875 static void parse_commit_message(struct format_commit_context *c)
876 {
877 const char *msg = c->message + c->message_off;
878 const char *start = c->message;
879
880 msg = skip_blank_lines(msg);
881 c->subject_off = msg - start;
882
883 msg = format_subject(NULL, msg, NULL);
884 msg = skip_blank_lines(msg);
885 c->body_off = msg - start;
886
887 c->commit_message_parsed = 1;
888 }
889
890 static void strbuf_wrap(struct strbuf *sb, size_t pos,
891 size_t width, size_t indent1, size_t indent2)
892 {
893 struct strbuf tmp = STRBUF_INIT;
894
895 if (pos)
896 strbuf_add(&tmp, sb->buf, pos);
897 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
898 (int) indent1, (int) indent2, (int) width);
899 strbuf_swap(&tmp, sb);
900 strbuf_release(&tmp);
901 }
902
903 static void rewrap_message_tail(struct strbuf *sb,
904 struct format_commit_context *c,
905 size_t new_width, size_t new_indent1,
906 size_t new_indent2)
907 {
908 if (c->width == new_width && c->indent1 == new_indent1 &&
909 c->indent2 == new_indent2)
910 return;
911 if (c->wrap_start < sb->len)
912 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
913 c->wrap_start = sb->len;
914 c->width = new_width;
915 c->indent1 = new_indent1;
916 c->indent2 = new_indent2;
917 }
918
919 static int format_reflog_person(struct strbuf *sb,
920 char part,
921 struct reflog_walk_info *log,
922 const struct date_mode *dmode)
923 {
924 const char *ident;
925
926 if (!log)
927 return 2;
928
929 ident = get_reflog_ident(log);
930 if (!ident)
931 return 2;
932
933 return format_person_part(sb, part, ident, strlen(ident), dmode);
934 }
935
936 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
937 const char *placeholder,
938 struct format_commit_context *c)
939 {
940 const char *rest = placeholder;
941 const char *basic_color = NULL;
942
943 if (placeholder[1] == '(') {
944 const char *begin = placeholder + 2;
945 const char *end = strchr(begin, ')');
946 char color[COLOR_MAXLEN];
947
948 if (!end)
949 return 0;
950
951 if (skip_prefix(begin, "auto,", &begin)) {
952 if (!want_color(c->pretty_ctx->color))
953 return end - placeholder + 1;
954 } else if (skip_prefix(begin, "always,", &begin)) {
955 /* nothing to do; we do not respect want_color at all */
956 } else {
957 /* the default is the same as "auto" */
958 if (!want_color(c->pretty_ctx->color))
959 return end - placeholder + 1;
960 }
961
962 if (color_parse_mem(begin, end - begin, color) < 0)
963 die(_("unable to parse --pretty format"));
964 strbuf_addstr(sb, color);
965 return end - placeholder + 1;
966 }
967
968 /*
969 * We handle things like "%C(red)" above; for historical reasons, there
970 * are a few colors that can be specified without parentheses (and
971 * they cannot support things like "auto" or "always" at all).
972 */
973 if (skip_prefix(placeholder + 1, "red", &rest))
974 basic_color = GIT_COLOR_RED;
975 else if (skip_prefix(placeholder + 1, "green", &rest))
976 basic_color = GIT_COLOR_GREEN;
977 else if (skip_prefix(placeholder + 1, "blue", &rest))
978 basic_color = GIT_COLOR_BLUE;
979 else if (skip_prefix(placeholder + 1, "reset", &rest))
980 basic_color = GIT_COLOR_RESET;
981
982 if (basic_color && want_color(c->pretty_ctx->color))
983 strbuf_addstr(sb, basic_color);
984
985 return rest - placeholder;
986 }
987
988 static size_t parse_padding_placeholder(struct strbuf *sb,
989 const char *placeholder,
990 struct format_commit_context *c)
991 {
992 const char *ch = placeholder;
993 enum flush_type flush_type;
994 int to_column = 0;
995
996 switch (*ch++) {
997 case '<':
998 flush_type = flush_right;
999 break;
1000 case '>':
1001 if (*ch == '<') {
1002 flush_type = flush_both;
1003 ch++;
1004 } else if (*ch == '>') {
1005 flush_type = flush_left_and_steal;
1006 ch++;
1007 } else
1008 flush_type = flush_left;
1009 break;
1010 default:
1011 return 0;
1012 }
1013
1014 /* the next value means "wide enough to that column" */
1015 if (*ch == '|') {
1016 to_column = 1;
1017 ch++;
1018 }
1019
1020 if (*ch == '(') {
1021 const char *start = ch + 1;
1022 const char *end = start + strcspn(start, ",)");
1023 char *next;
1024 int width;
1025 if (!end || end == start)
1026 return 0;
1027 width = strtol(start, &next, 10);
1028 if (next == start || width == 0)
1029 return 0;
1030 if (width < 0) {
1031 if (to_column)
1032 width += term_columns();
1033 if (width < 0)
1034 return 0;
1035 }
1036 c->padding = to_column ? -width : width;
1037 c->flush_type = flush_type;
1038
1039 if (*end == ',') {
1040 start = end + 1;
1041 end = strchr(start, ')');
1042 if (!end || end == start)
1043 return 0;
1044 if (starts_with(start, "trunc)"))
1045 c->truncate = trunc_right;
1046 else if (starts_with(start, "ltrunc)"))
1047 c->truncate = trunc_left;
1048 else if (starts_with(start, "mtrunc)"))
1049 c->truncate = trunc_middle;
1050 else
1051 return 0;
1052 } else
1053 c->truncate = trunc_none;
1054
1055 return end - placeholder + 1;
1056 }
1057 return 0;
1058 }
1059
1060 static int match_placeholder_arg(const char *to_parse, const char *candidate,
1061 const char **end)
1062 {
1063 const char *p;
1064
1065 if (!(skip_prefix(to_parse, candidate, &p)))
1066 return 0;
1067 if (*p == ',') {
1068 *end = p + 1;
1069 return 1;
1070 }
1071 if (*p == ')') {
1072 *end = p;
1073 return 1;
1074 }
1075 return 0;
1076 }
1077
1078 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1079 const char *placeholder,
1080 void *context)
1081 {
1082 struct format_commit_context *c = context;
1083 const struct commit *commit = c->commit;
1084 const char *msg = c->message;
1085 struct commit_list *p;
1086 const char *arg;
1087 int ch;
1088
1089 /* these are independent of the commit */
1090 switch (placeholder[0]) {
1091 case 'C':
1092 if (starts_with(placeholder + 1, "(auto)")) {
1093 c->auto_color = want_color(c->pretty_ctx->color);
1094 if (c->auto_color && sb->len)
1095 strbuf_addstr(sb, GIT_COLOR_RESET);
1096 return 7; /* consumed 7 bytes, "C(auto)" */
1097 } else {
1098 int ret = parse_color(sb, placeholder, c);
1099 if (ret)
1100 c->auto_color = 0;
1101 /*
1102 * Otherwise, we decided to treat %C<unknown>
1103 * as a literal string, and the previous
1104 * %C(auto) is still valid.
1105 */
1106 return ret;
1107 }
1108 case 'n': /* newline */
1109 strbuf_addch(sb, '\n');
1110 return 1;
1111 case 'x':
1112 /* %x00 == NUL, %x0a == LF, etc. */
1113 ch = hex2chr(placeholder + 1);
1114 if (ch < 0)
1115 return 0;
1116 strbuf_addch(sb, ch);
1117 return 3;
1118 case 'w':
1119 if (placeholder[1] == '(') {
1120 unsigned long width = 0, indent1 = 0, indent2 = 0;
1121 char *next;
1122 const char *start = placeholder + 2;
1123 const char *end = strchr(start, ')');
1124 if (!end)
1125 return 0;
1126 if (end > start) {
1127 width = strtoul(start, &next, 10);
1128 if (*next == ',') {
1129 indent1 = strtoul(next + 1, &next, 10);
1130 if (*next == ',') {
1131 indent2 = strtoul(next + 1,
1132 &next, 10);
1133 }
1134 }
1135 if (*next != ')')
1136 return 0;
1137 }
1138 rewrap_message_tail(sb, c, width, indent1, indent2);
1139 return end - placeholder + 1;
1140 } else
1141 return 0;
1142
1143 case '<':
1144 case '>':
1145 return parse_padding_placeholder(sb, placeholder, c);
1146 }
1147
1148 /* these depend on the commit */
1149 if (!commit->object.parsed)
1150 parse_object(the_repository, &commit->object.oid);
1151
1152 switch (placeholder[0]) {
1153 case 'H': /* commit hash */
1154 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1155 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1156 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1157 return 1;
1158 case 'h': /* abbreviated commit hash */
1159 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1160 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1161 c->pretty_ctx->abbrev);
1162 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1163 return 1;
1164 case 'T': /* tree hash */
1165 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1166 return 1;
1167 case 't': /* abbreviated tree hash */
1168 strbuf_add_unique_abbrev(sb,
1169 get_commit_tree_oid(commit),
1170 c->pretty_ctx->abbrev);
1171 return 1;
1172 case 'P': /* parent hashes */
1173 for (p = commit->parents; p; p = p->next) {
1174 if (p != commit->parents)
1175 strbuf_addch(sb, ' ');
1176 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1177 }
1178 return 1;
1179 case 'p': /* abbreviated parent hashes */
1180 for (p = commit->parents; p; p = p->next) {
1181 if (p != commit->parents)
1182 strbuf_addch(sb, ' ');
1183 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1184 c->pretty_ctx->abbrev);
1185 }
1186 return 1;
1187 case 'm': /* left/right/bottom */
1188 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1189 return 1;
1190 case 'd':
1191 load_ref_decorations(NULL, DECORATE_SHORT_REFS);
1192 format_decorations(sb, commit, c->auto_color);
1193 return 1;
1194 case 'D':
1195 load_ref_decorations(NULL, DECORATE_SHORT_REFS);
1196 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1197 return 1;
1198 case 'g': /* reflog info */
1199 switch(placeholder[1]) {
1200 case 'd': /* reflog selector */
1201 case 'D':
1202 if (c->pretty_ctx->reflog_info)
1203 get_reflog_selector(sb,
1204 c->pretty_ctx->reflog_info,
1205 &c->pretty_ctx->date_mode,
1206 c->pretty_ctx->date_mode_explicit,
1207 (placeholder[1] == 'd'));
1208 return 2;
1209 case 's': /* reflog message */
1210 if (c->pretty_ctx->reflog_info)
1211 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1212 return 2;
1213 case 'n':
1214 case 'N':
1215 case 'e':
1216 case 'E':
1217 return format_reflog_person(sb,
1218 placeholder[1],
1219 c->pretty_ctx->reflog_info,
1220 &c->pretty_ctx->date_mode);
1221 }
1222 return 0; /* unknown %g placeholder */
1223 case 'N':
1224 if (c->pretty_ctx->notes_message) {
1225 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1226 return 1;
1227 }
1228 return 0;
1229 }
1230
1231 if (placeholder[0] == 'G') {
1232 if (!c->signature_check.result)
1233 check_commit_signature(c->commit, &(c->signature_check));
1234 switch (placeholder[1]) {
1235 case 'G':
1236 if (c->signature_check.gpg_output)
1237 strbuf_addstr(sb, c->signature_check.gpg_output);
1238 break;
1239 case '?':
1240 switch (c->signature_check.result) {
1241 case 'G':
1242 case 'B':
1243 case 'E':
1244 case 'U':
1245 case 'N':
1246 case 'X':
1247 case 'Y':
1248 case 'R':
1249 strbuf_addch(sb, c->signature_check.result);
1250 }
1251 break;
1252 case 'S':
1253 if (c->signature_check.signer)
1254 strbuf_addstr(sb, c->signature_check.signer);
1255 break;
1256 case 'K':
1257 if (c->signature_check.key)
1258 strbuf_addstr(sb, c->signature_check.key);
1259 break;
1260 default:
1261 return 0;
1262 }
1263 return 2;
1264 }
1265
1266
1267 /* For the rest we have to parse the commit header. */
1268 if (!c->commit_header_parsed)
1269 parse_commit_header(c);
1270
1271 switch (placeholder[0]) {
1272 case 'a': /* author ... */
1273 return format_person_part(sb, placeholder[1],
1274 msg + c->author.off, c->author.len,
1275 &c->pretty_ctx->date_mode);
1276 case 'c': /* committer ... */
1277 return format_person_part(sb, placeholder[1],
1278 msg + c->committer.off, c->committer.len,
1279 &c->pretty_ctx->date_mode);
1280 case 'e': /* encoding */
1281 if (c->commit_encoding)
1282 strbuf_addstr(sb, c->commit_encoding);
1283 return 1;
1284 case 'B': /* raw body */
1285 /* message_off is always left at the initial newline */
1286 strbuf_addstr(sb, msg + c->message_off + 1);
1287 return 1;
1288 }
1289
1290 /* Now we need to parse the commit message. */
1291 if (!c->commit_message_parsed)
1292 parse_commit_message(c);
1293
1294 switch (placeholder[0]) {
1295 case 's': /* subject */
1296 format_subject(sb, msg + c->subject_off, " ");
1297 return 1;
1298 case 'f': /* sanitized subject */
1299 format_sanitized_subject(sb, msg + c->subject_off);
1300 return 1;
1301 case 'b': /* body */
1302 strbuf_addstr(sb, msg + c->body_off);
1303 return 1;
1304 }
1305
1306 if (skip_prefix(placeholder, "(trailers", &arg)) {
1307 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1308
1309 opts.no_divider = 1;
1310
1311 if (*arg == ':') {
1312 arg++;
1313 for (;;) {
1314 if (match_placeholder_arg(arg, "only", &arg))
1315 opts.only_trailers = 1;
1316 else if (match_placeholder_arg(arg, "unfold", &arg))
1317 opts.unfold = 1;
1318 else
1319 break;
1320 }
1321 }
1322 if (*arg == ')') {
1323 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1324 return arg - placeholder + 1;
1325 }
1326 }
1327
1328 return 0; /* unknown placeholder */
1329 }
1330
1331 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1332 const char *placeholder,
1333 struct format_commit_context *c)
1334 {
1335 struct strbuf local_sb = STRBUF_INIT;
1336 int total_consumed = 0, len, padding = c->padding;
1337 if (padding < 0) {
1338 const char *start = strrchr(sb->buf, '\n');
1339 int occupied;
1340 if (!start)
1341 start = sb->buf;
1342 occupied = utf8_strnwidth(start, -1, 1);
1343 occupied += c->pretty_ctx->graph_width;
1344 padding = (-padding) - occupied;
1345 }
1346 while (1) {
1347 int modifier = *placeholder == 'C';
1348 int consumed = format_commit_one(&local_sb, placeholder, c);
1349 total_consumed += consumed;
1350
1351 if (!modifier)
1352 break;
1353
1354 placeholder += consumed;
1355 if (*placeholder != '%')
1356 break;
1357 placeholder++;
1358 total_consumed++;
1359 }
1360 len = utf8_strnwidth(local_sb.buf, -1, 1);
1361
1362 if (c->flush_type == flush_left_and_steal) {
1363 const char *ch = sb->buf + sb->len - 1;
1364 while (len > padding && ch > sb->buf) {
1365 const char *p;
1366 if (*ch == ' ') {
1367 ch--;
1368 padding++;
1369 continue;
1370 }
1371 /* check for trailing ansi sequences */
1372 if (*ch != 'm')
1373 break;
1374 p = ch - 1;
1375 while (ch - p < 10 && *p != '\033')
1376 p--;
1377 if (*p != '\033' ||
1378 ch + 1 - p != display_mode_esc_sequence_len(p))
1379 break;
1380 /*
1381 * got a good ansi sequence, put it back to
1382 * local_sb as we're cutting sb
1383 */
1384 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1385 ch = p - 1;
1386 }
1387 strbuf_setlen(sb, ch + 1 - sb->buf);
1388 c->flush_type = flush_left;
1389 }
1390
1391 if (len > padding) {
1392 switch (c->truncate) {
1393 case trunc_left:
1394 strbuf_utf8_replace(&local_sb,
1395 0, len - (padding - 2),
1396 "..");
1397 break;
1398 case trunc_middle:
1399 strbuf_utf8_replace(&local_sb,
1400 padding / 2 - 1,
1401 len - (padding - 2),
1402 "..");
1403 break;
1404 case trunc_right:
1405 strbuf_utf8_replace(&local_sb,
1406 padding - 2, len - (padding - 2),
1407 "..");
1408 break;
1409 case trunc_none:
1410 break;
1411 }
1412 strbuf_addbuf(sb, &local_sb);
1413 } else {
1414 int sb_len = sb->len, offset = 0;
1415 if (c->flush_type == flush_left)
1416 offset = padding - len;
1417 else if (c->flush_type == flush_both)
1418 offset = (padding - len) / 2;
1419 /*
1420 * we calculate padding in columns, now
1421 * convert it back to chars
1422 */
1423 padding = padding - len + local_sb.len;
1424 strbuf_addchars(sb, ' ', padding);
1425 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1426 local_sb.len);
1427 }
1428 strbuf_release(&local_sb);
1429 c->flush_type = no_flush;
1430 return total_consumed;
1431 }
1432
1433 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1434 const char *placeholder,
1435 void *context)
1436 {
1437 int consumed;
1438 size_t orig_len;
1439 enum {
1440 NO_MAGIC,
1441 ADD_LF_BEFORE_NON_EMPTY,
1442 DEL_LF_BEFORE_EMPTY,
1443 ADD_SP_BEFORE_NON_EMPTY
1444 } magic = NO_MAGIC;
1445
1446 switch (placeholder[0]) {
1447 case '-':
1448 magic = DEL_LF_BEFORE_EMPTY;
1449 break;
1450 case '+':
1451 magic = ADD_LF_BEFORE_NON_EMPTY;
1452 break;
1453 case ' ':
1454 magic = ADD_SP_BEFORE_NON_EMPTY;
1455 break;
1456 default:
1457 break;
1458 }
1459 if (magic != NO_MAGIC)
1460 placeholder++;
1461
1462 orig_len = sb->len;
1463 if (((struct format_commit_context *)context)->flush_type != no_flush)
1464 consumed = format_and_pad_commit(sb, placeholder, context);
1465 else
1466 consumed = format_commit_one(sb, placeholder, context);
1467 if (magic == NO_MAGIC)
1468 return consumed;
1469
1470 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1471 while (sb->len && sb->buf[sb->len - 1] == '\n')
1472 strbuf_setlen(sb, sb->len - 1);
1473 } else if (orig_len != sb->len) {
1474 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1475 strbuf_insert(sb, orig_len, "\n", 1);
1476 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1477 strbuf_insert(sb, orig_len, " ", 1);
1478 }
1479 return consumed + 1;
1480 }
1481
1482 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1483 void *context)
1484 {
1485 struct userformat_want *w = context;
1486
1487 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1488 placeholder++;
1489
1490 switch (*placeholder) {
1491 case 'N':
1492 w->notes = 1;
1493 break;
1494 }
1495 return 0;
1496 }
1497
1498 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1499 {
1500 struct strbuf dummy = STRBUF_INIT;
1501
1502 if (!fmt) {
1503 if (!user_format)
1504 return;
1505 fmt = user_format;
1506 }
1507 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1508 strbuf_release(&dummy);
1509 }
1510
1511 void format_commit_message(const struct commit *commit,
1512 const char *format, struct strbuf *sb,
1513 const struct pretty_print_context *pretty_ctx)
1514 {
1515 struct format_commit_context context;
1516 const char *output_enc = pretty_ctx->output_encoding;
1517 const char *utf8 = "UTF-8";
1518
1519 memset(&context, 0, sizeof(context));
1520 context.commit = commit;
1521 context.pretty_ctx = pretty_ctx;
1522 context.wrap_start = sb->len;
1523 /*
1524 * convert a commit message to UTF-8 first
1525 * as far as 'format_commit_item' assumes it in UTF-8
1526 */
1527 context.message = logmsg_reencode(commit,
1528 &context.commit_encoding,
1529 utf8);
1530
1531 strbuf_expand(sb, format, format_commit_item, &context);
1532 rewrap_message_tail(sb, &context, 0, 0, 0);
1533
1534 /* then convert a commit message to an actual output encoding */
1535 if (output_enc) {
1536 if (same_encoding(utf8, output_enc))
1537 output_enc = NULL;
1538 } else {
1539 if (context.commit_encoding &&
1540 !same_encoding(context.commit_encoding, utf8))
1541 output_enc = context.commit_encoding;
1542 }
1543
1544 if (output_enc) {
1545 size_t outsz;
1546 char *out = reencode_string_len(sb->buf, sb->len,
1547 output_enc, utf8, &outsz);
1548 if (out)
1549 strbuf_attach(sb, out, outsz, outsz + 1);
1550 }
1551
1552 free(context.commit_encoding);
1553 unuse_commit_buffer(commit, context.message);
1554 }
1555
1556 static void pp_header(struct pretty_print_context *pp,
1557 const char *encoding,
1558 const struct commit *commit,
1559 const char **msg_p,
1560 struct strbuf *sb)
1561 {
1562 int parents_shown = 0;
1563
1564 for (;;) {
1565 const char *name, *line = *msg_p;
1566 int linelen = get_one_line(*msg_p);
1567
1568 if (!linelen)
1569 return;
1570 *msg_p += linelen;
1571
1572 if (linelen == 1)
1573 /* End of header */
1574 return;
1575
1576 if (pp->fmt == CMIT_FMT_RAW) {
1577 strbuf_add(sb, line, linelen);
1578 continue;
1579 }
1580
1581 if (starts_with(line, "parent ")) {
1582 if (linelen != the_hash_algo->hexsz + 8)
1583 die("bad parent line in commit");
1584 continue;
1585 }
1586
1587 if (!parents_shown) {
1588 unsigned num = commit_list_count(commit->parents);
1589 /* with enough slop */
1590 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1591 add_merge_info(pp, sb, commit);
1592 parents_shown = 1;
1593 }
1594
1595 /*
1596 * MEDIUM == DEFAULT shows only author with dates.
1597 * FULL shows both authors but not dates.
1598 * FULLER shows both authors and dates.
1599 */
1600 if (skip_prefix(line, "author ", &name)) {
1601 strbuf_grow(sb, linelen + 80);
1602 pp_user_info(pp, "Author", sb, name, encoding);
1603 }
1604 if (skip_prefix(line, "committer ", &name) &&
1605 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1606 strbuf_grow(sb, linelen + 80);
1607 pp_user_info(pp, "Commit", sb, name, encoding);
1608 }
1609 }
1610 }
1611
1612 void pp_title_line(struct pretty_print_context *pp,
1613 const char **msg_p,
1614 struct strbuf *sb,
1615 const char *encoding,
1616 int need_8bit_cte)
1617 {
1618 static const int max_length = 78; /* per rfc2047 */
1619 struct strbuf title;
1620
1621 strbuf_init(&title, 80);
1622 *msg_p = format_subject(&title, *msg_p,
1623 pp->preserve_subject ? "\n" : " ");
1624
1625 strbuf_grow(sb, title.len + 1024);
1626 if (pp->print_email_subject) {
1627 if (pp->rev)
1628 fmt_output_email_subject(sb, pp->rev);
1629 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1630 add_rfc2047(sb, title.buf, title.len,
1631 encoding, RFC2047_SUBJECT);
1632 else
1633 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1634 -last_line_length(sb), 1, max_length);
1635 } else {
1636 strbuf_addbuf(sb, &title);
1637 }
1638 strbuf_addch(sb, '\n');
1639
1640 if (need_8bit_cte == 0) {
1641 int i;
1642 for (i = 0; i < pp->in_body_headers.nr; i++) {
1643 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1644 need_8bit_cte = 1;
1645 break;
1646 }
1647 }
1648 }
1649
1650 if (need_8bit_cte > 0) {
1651 const char *header_fmt =
1652 "MIME-Version: 1.0\n"
1653 "Content-Type: text/plain; charset=%s\n"
1654 "Content-Transfer-Encoding: 8bit\n";
1655 strbuf_addf(sb, header_fmt, encoding);
1656 }
1657 if (pp->after_subject) {
1658 strbuf_addstr(sb, pp->after_subject);
1659 }
1660 if (cmit_fmt_is_mail(pp->fmt)) {
1661 strbuf_addch(sb, '\n');
1662 }
1663
1664 if (pp->in_body_headers.nr) {
1665 int i;
1666 for (i = 0; i < pp->in_body_headers.nr; i++) {
1667 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1668 free(pp->in_body_headers.items[i].string);
1669 }
1670 string_list_clear(&pp->in_body_headers, 0);
1671 strbuf_addch(sb, '\n');
1672 }
1673
1674 strbuf_release(&title);
1675 }
1676
1677 static int pp_utf8_width(const char *start, const char *end)
1678 {
1679 int width = 0;
1680 size_t remain = end - start;
1681
1682 while (remain) {
1683 int n = utf8_width(&start, &remain);
1684 if (n < 0 || !start)
1685 return -1;
1686 width += n;
1687 }
1688 return width;
1689 }
1690
1691 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1692 const char *line, int linelen)
1693 {
1694 const char *tab;
1695
1696 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1697 int width = pp_utf8_width(line, tab);
1698
1699 /*
1700 * If it wasn't well-formed utf8, or it
1701 * had characters with badly defined
1702 * width (control characters etc), just
1703 * give up on trying to align things.
1704 */
1705 if (width < 0)
1706 break;
1707
1708 /* Output the data .. */
1709 strbuf_add(sb, line, tab - line);
1710
1711 /* .. and the de-tabified tab */
1712 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1713
1714 /* Skip over the printed part .. */
1715 linelen -= tab + 1 - line;
1716 line = tab + 1;
1717 }
1718
1719 /*
1720 * Print out everything after the last tab without
1721 * worrying about width - there's nothing more to
1722 * align.
1723 */
1724 strbuf_add(sb, line, linelen);
1725 }
1726
1727 /*
1728 * pp_handle_indent() prints out the intendation, and
1729 * the whole line (without the final newline), after
1730 * de-tabifying.
1731 */
1732 static void pp_handle_indent(struct pretty_print_context *pp,
1733 struct strbuf *sb, int indent,
1734 const char *line, int linelen)
1735 {
1736 strbuf_addchars(sb, ' ', indent);
1737 if (pp->expand_tabs_in_log)
1738 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1739 else
1740 strbuf_add(sb, line, linelen);
1741 }
1742
1743 static int is_mboxrd_from(const char *line, int len)
1744 {
1745 /*
1746 * a line matching /^From $/ here would only have len == 4
1747 * at this point because is_empty_line would've trimmed all
1748 * trailing space
1749 */
1750 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1751 }
1752
1753 void pp_remainder(struct pretty_print_context *pp,
1754 const char **msg_p,
1755 struct strbuf *sb,
1756 int indent)
1757 {
1758 int first = 1;
1759 for (;;) {
1760 const char *line = *msg_p;
1761 int linelen = get_one_line(line);
1762 *msg_p += linelen;
1763
1764 if (!linelen)
1765 break;
1766
1767 if (is_blank_line(line, &linelen)) {
1768 if (first)
1769 continue;
1770 if (pp->fmt == CMIT_FMT_SHORT)
1771 break;
1772 }
1773 first = 0;
1774
1775 strbuf_grow(sb, linelen + indent + 20);
1776 if (indent)
1777 pp_handle_indent(pp, sb, indent, line, linelen);
1778 else if (pp->expand_tabs_in_log)
1779 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1780 line, linelen);
1781 else {
1782 if (pp->fmt == CMIT_FMT_MBOXRD &&
1783 is_mboxrd_from(line, linelen))
1784 strbuf_addch(sb, '>');
1785
1786 strbuf_add(sb, line, linelen);
1787 }
1788 strbuf_addch(sb, '\n');
1789 }
1790 }
1791
1792 void pretty_print_commit(struct pretty_print_context *pp,
1793 const struct commit *commit,
1794 struct strbuf *sb)
1795 {
1796 unsigned long beginning_of_body;
1797 int indent = 4;
1798 const char *msg;
1799 const char *reencoded;
1800 const char *encoding;
1801 int need_8bit_cte = pp->need_8bit_cte;
1802
1803 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1804 format_commit_message(commit, user_format, sb, pp);
1805 return;
1806 }
1807
1808 encoding = get_log_output_encoding();
1809 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1810
1811 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1812 indent = 0;
1813
1814 /*
1815 * We need to check and emit Content-type: to mark it
1816 * as 8-bit if we haven't done so.
1817 */
1818 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1819 int i, ch, in_body;
1820
1821 for (in_body = i = 0; (ch = msg[i]); i++) {
1822 if (!in_body) {
1823 /* author could be non 7-bit ASCII but
1824 * the log may be so; skip over the
1825 * header part first.
1826 */
1827 if (ch == '\n' && msg[i+1] == '\n')
1828 in_body = 1;
1829 }
1830 else if (non_ascii(ch)) {
1831 need_8bit_cte = 1;
1832 break;
1833 }
1834 }
1835 }
1836
1837 pp_header(pp, encoding, commit, &msg, sb);
1838 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
1839 strbuf_addch(sb, '\n');
1840 }
1841
1842 /* Skip excess blank lines at the beginning of body, if any... */
1843 msg = skip_blank_lines(msg);
1844
1845 /* These formats treat the title line specially. */
1846 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1847 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1848
1849 beginning_of_body = sb->len;
1850 if (pp->fmt != CMIT_FMT_ONELINE)
1851 pp_remainder(pp, &msg, sb, indent);
1852 strbuf_rtrim(sb);
1853
1854 /* Make sure there is an EOLN for the non-oneline case */
1855 if (pp->fmt != CMIT_FMT_ONELINE)
1856 strbuf_addch(sb, '\n');
1857
1858 /*
1859 * The caller may append additional body text in e-mail
1860 * format. Make sure we did not strip the blank line
1861 * between the header and the body.
1862 */
1863 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1864 strbuf_addch(sb, '\n');
1865
1866 unuse_commit_buffer(commit, reencoded);
1867 }
1868
1869 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1870 struct strbuf *sb)
1871 {
1872 struct pretty_print_context pp = {0};
1873 pp.fmt = fmt;
1874 pretty_print_commit(&pp, commit, sb);
1875 }