?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
pickaxe -S: slightly optimize contains()
[git/debian.git] / pretty.c
blobe5b33ba034bd600b343ea146b0c38f992d1b462e
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 #include "run-command.h"
16
17 static char *user_format;
18 static struct cmt_fmt_map {
19 const char *name;
20 enum cmit_fmt format;
21 int is_tformat;
22 int expand_tabs_in_log;
23 int is_alias;
24 enum date_mode_type default_date_mode_type;
25 const char *user_format;
26 } *commit_formats;
27 static size_t builtin_formats_len;
28 static size_t commit_formats_len;
29 static size_t commit_formats_alloc;
30 static struct cmt_fmt_map *find_commit_format(const char *sought);
31
32 int commit_format_is_empty(enum cmit_fmt fmt)
33 {
34 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
35 }
36
37 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
38 {
39 free(user_format);
40 user_format = xstrdup(cp);
41 if (is_tformat)
42 rev->use_terminator = 1;
43 rev->commit_format = CMIT_FMT_USERFORMAT;
44 }
45
46 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
47 {
48 struct cmt_fmt_map *commit_format = NULL;
49 const char *name;
50 const char *fmt;
51 int i;
52
53 if (!skip_prefix(var, "pretty.", &name))
54 return 0;
55
56 for (i = 0; i < builtin_formats_len; i++) {
57 if (!strcmp(commit_formats[i].name, name))
58 return 0;
59 }
60
61 for (i = builtin_formats_len; i < commit_formats_len; i++) {
62 if (!strcmp(commit_formats[i].name, name)) {
63 commit_format = &commit_formats[i];
64 break;
65 }
66 }
67
68 if (!commit_format) {
69 ALLOC_GROW(commit_formats, commit_formats_len+1,
70 commit_formats_alloc);
71 commit_format = &commit_formats[commit_formats_len];
72 memset(commit_format, 0, sizeof(*commit_format));
73 commit_formats_len++;
74 }
75
76 commit_format->name = xstrdup(name);
77 commit_format->format = CMIT_FMT_USERFORMAT;
78 if (git_config_string(&fmt, var, value))
79 return -1;
80
81 if (skip_prefix(fmt, "format:", &fmt))
82 commit_format->is_tformat = 0;
83 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
84 commit_format->is_tformat = 1;
85 else
86 commit_format->is_alias = 1;
87 commit_format->user_format = fmt;
88
89 return 0;
90 }
91
92 static void setup_commit_formats(void)
93 {
94 struct cmt_fmt_map builtin_formats[] = {
95 { "raw", CMIT_FMT_RAW, 0, 0 },
96 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
97 { "short", CMIT_FMT_SHORT, 0, 0 },
98 { "email", CMIT_FMT_EMAIL, 0, 0 },
99 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
100 { "fuller", CMIT_FMT_FULLER, 0, 8 },
101 { "full", CMIT_FMT_FULL, 0, 8 },
102 { "oneline", CMIT_FMT_ONELINE, 1, 0 },
103 { "reference", CMIT_FMT_USERFORMAT, 1, 0,
104 0, DATE_SHORT, "%C(auto)%h (%s, %ad)" },
105 /*
106 * Please update $__git_log_pretty_formats in
107 * git-completion.bash when you add new formats.
108 */
109 };
110 commit_formats_len = ARRAY_SIZE(builtin_formats);
111 builtin_formats_len = commit_formats_len;
112 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
113 COPY_ARRAY(commit_formats, builtin_formats,
114 ARRAY_SIZE(builtin_formats));
115
116 git_config(git_pretty_formats_config, NULL);
117 }
118
119 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
120 const char *original,
121 int num_redirections)
122 {
123 struct cmt_fmt_map *found = NULL;
124 size_t found_match_len = 0;
125 int i;
126
127 if (num_redirections >= commit_formats_len)
128 die("invalid --pretty format: "
129 "'%s' references an alias which points to itself",
130 original);
131
132 for (i = 0; i < commit_formats_len; i++) {
133 size_t match_len;
134
135 if (!starts_with(commit_formats[i].name, sought))
136 continue;
137
138 match_len = strlen(commit_formats[i].name);
139 if (found == NULL || found_match_len > match_len) {
140 found = &commit_formats[i];
141 found_match_len = match_len;
142 }
143 }
144
145 if (found && found->is_alias) {
146 found = find_commit_format_recursive(found->user_format,
147 original,
148 num_redirections+1);
149 }
150
151 return found;
152 }
153
154 static struct cmt_fmt_map *find_commit_format(const char *sought)
155 {
156 if (!commit_formats)
157 setup_commit_formats();
158
159 return find_commit_format_recursive(sought, sought, 0);
160 }
161
162 void get_commit_format(const char *arg, struct rev_info *rev)
163 {
164 struct cmt_fmt_map *commit_format;
165
166 rev->use_terminator = 0;
167 if (!arg) {
168 rev->commit_format = CMIT_FMT_DEFAULT;
169 return;
170 }
171 if (skip_prefix(arg, "format:", &arg)) {
172 save_user_format(rev, arg, 0);
173 return;
174 }
175
176 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
177 save_user_format(rev, arg, 1);
178 return;
179 }
180
181 commit_format = find_commit_format(arg);
182 if (!commit_format)
183 die("invalid --pretty format: %s", arg);
184
185 rev->commit_format = commit_format->format;
186 rev->use_terminator = commit_format->is_tformat;
187 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
188 if (!rev->date_mode_explicit && commit_format->default_date_mode_type)
189 rev->date_mode.type = commit_format->default_date_mode_type;
190 if (commit_format->format == CMIT_FMT_USERFORMAT) {
191 save_user_format(rev, commit_format->user_format,
192 commit_format->is_tformat);
193 }
194 }
195
196 /*
197 * Generic support for pretty-printing the header
198 */
199 static int get_one_line(const char *msg)
200 {
201 int ret = 0;
202
203 for (;;) {
204 char c = *msg++;
205 if (!c)
206 break;
207 ret++;
208 if (c == '\n')
209 break;
210 }
211 return ret;
212 }
213
214 /* High bit set, or ISO-2022-INT */
215 static int non_ascii(int ch)
216 {
217 return !isascii(ch) || ch == '\033';
218 }
219
220 int has_non_ascii(const char *s)
221 {
222 int ch;
223 if (!s)
224 return 0;
225 while ((ch = *s++) != '\0') {
226 if (non_ascii(ch))
227 return 1;
228 }
229 return 0;
230 }
231
232 static int is_rfc822_special(char ch)
233 {
234 switch (ch) {
235 case '(':
236 case ')':
237 case '<':
238 case '>':
239 case '[':
240 case ']':
241 case ':':
242 case ';':
243 case '@':
244 case ',':
245 case '.':
246 case '"':
247 case '\\':
248 return 1;
249 default:
250 return 0;
251 }
252 }
253
254 static int needs_rfc822_quoting(const char *s, int len)
255 {
256 int i;
257 for (i = 0; i < len; i++)
258 if (is_rfc822_special(s[i]))
259 return 1;
260 return 0;
261 }
262
263 static int last_line_length(struct strbuf *sb)
264 {
265 int i;
266
267 /* How many bytes are already used on the last line? */
268 for (i = sb->len - 1; i >= 0; i--)
269 if (sb->buf[i] == '\n')
270 break;
271 return sb->len - (i + 1);
272 }
273
274 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
275 {
276 int i;
277
278 /* just a guess, we may have to also backslash-quote */
279 strbuf_grow(out, len + 2);
280
281 strbuf_addch(out, '"');
282 for (i = 0; i < len; i++) {
283 switch (s[i]) {
284 case '"':
285 case '\\':
286 strbuf_addch(out, '\\');
287 /* fall through */
288 default:
289 strbuf_addch(out, s[i]);
290 }
291 }
292 strbuf_addch(out, '"');
293 }
294
295 enum rfc2047_type {
296 RFC2047_SUBJECT,
297 RFC2047_ADDRESS
298 };
299
300 static int is_rfc2047_special(char ch, enum rfc2047_type type)
301 {
302 /*
303 * rfc2047, section 4.2:
304 *
305 * 8-bit values which correspond to printable ASCII characters other
306 * than "=", "?", and "_" (underscore), MAY be represented as those
307 * characters. (But see section 5 for restrictions.) In
308 * particular, SPACE and TAB MUST NOT be represented as themselves
309 * within encoded words.
310 */
311
312 /*
313 * rule out non-ASCII characters and non-printable characters (the
314 * non-ASCII check should be redundant as isprint() is not localized
315 * and only knows about ASCII, but be defensive about that)
316 */
317 if (non_ascii(ch) || !isprint(ch))
318 return 1;
319
320 /*
321 * rule out special printable characters (' ' should be the only
322 * whitespace character considered printable, but be defensive and use
323 * isspace())
324 */
325 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
326 return 1;
327
328 /*
329 * rfc2047, section 5.3:
330 *
331 * As a replacement for a 'word' entity within a 'phrase', for example,
332 * one that precedes an address in a From, To, or Cc header. The ABNF
333 * definition for 'phrase' from RFC 822 thus becomes:
334 *
335 * phrase = 1*( encoded-word / word )
336 *
337 * In this case the set of characters that may be used in a "Q"-encoded
338 * 'encoded-word' is restricted to: <upper and lower case ASCII
339 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
340 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
341 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
342 * 'special' by 'linear-white-space'.
343 */
344
345 if (type != RFC2047_ADDRESS)
346 return 0;
347
348 /* '=' and '_' are special cases and have been checked above */
349 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
350 }
351
352 static int needs_rfc2047_encoding(const char *line, int len)
353 {
354 int i;
355
356 for (i = 0; i < len; i++) {
357 int ch = line[i];
358 if (non_ascii(ch) || ch == '\n')
359 return 1;
360 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
361 return 1;
362 }
363
364 return 0;
365 }
366
367 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
368 const char *encoding, enum rfc2047_type type)
369 {
370 static const int max_encoded_length = 76; /* per rfc2047 */
371 int i;
372 int line_len = last_line_length(sb);
373
374 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
375 strbuf_addf(sb, "=?%s?q?", encoding);
376 line_len += strlen(encoding) + 5; /* 5 for =??q? */
377
378 while (len) {
379 /*
380 * RFC 2047, section 5 (3):
381 *
382 * Each 'encoded-word' MUST represent an integral number of
383 * characters. A multi-octet character may not be split across
384 * adjacent 'encoded- word's.
385 */
386 const unsigned char *p = (const unsigned char *)line;
387 int chrlen = mbs_chrlen(&line, &len, encoding);
388 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
389
390 /* "=%02X" * chrlen, or the byte itself */
391 const char *encoded_fmt = is_special ? "=%02X" : "%c";
392 int encoded_len = is_special ? 3 * chrlen : 1;
393
394 /*
395 * According to RFC 2047, we could encode the special character
396 * ' ' (space) with '_' (underscore) for readability. But many
397 * programs do not understand this and just leave the
398 * underscore in place. Thus, we do nothing special here, which
399 * causes ' ' to be encoded as '=20', avoiding this problem.
400 */
401
402 if (line_len + encoded_len + 2 > max_encoded_length) {
403 /* It won't fit with trailing "?=" --- break the line */
404 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
405 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
406 }
407
408 for (i = 0; i < chrlen; i++)
409 strbuf_addf(sb, encoded_fmt, p[i]);
410 line_len += encoded_len;
411 }
412 strbuf_addstr(sb, "?=");
413 }
414
415 const char *show_ident_date(const struct ident_split *ident,
416 const struct date_mode *mode)
417 {
418 timestamp_t date = 0;
419 long tz = 0;
420
421 if (ident->date_begin && ident->date_end)
422 date = parse_timestamp(ident->date_begin, NULL, 10);
423 if (date_overflows(date))
424 date = 0;
425 else {
426 if (ident->tz_begin && ident->tz_end)
427 tz = strtol(ident->tz_begin, NULL, 10);
428 if (tz >= INT_MAX || tz <= INT_MIN)
429 tz = 0;
430 }
431 return show_date(date, tz, mode);
432 }
433
434 void pp_user_info(struct pretty_print_context *pp,
435 const char *what, struct strbuf *sb,
436 const char *line, const char *encoding)
437 {
438 struct ident_split ident;
439 char *line_end;
440 const char *mailbuf, *namebuf;
441 size_t namelen, maillen;
442 int max_length = 78; /* per rfc2822 */
443
444 if (pp->fmt == CMIT_FMT_ONELINE)
445 return;
446
447 line_end = strchrnul(line, '\n');
448 if (split_ident_line(&ident, line, line_end - line))
449 return;
450
451 mailbuf = ident.mail_begin;
452 maillen = ident.mail_end - ident.mail_begin;
453 namebuf = ident.name_begin;
454 namelen = ident.name_end - ident.name_begin;
455
456 if (pp->mailmap)
457 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
458
459 if (cmit_fmt_is_mail(pp->fmt)) {
460 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
461 struct strbuf buf = STRBUF_INIT;
462
463 strbuf_addstr(&buf, "From: ");
464 strbuf_add(&buf, namebuf, namelen);
465 strbuf_addstr(&buf, " <");
466 strbuf_add(&buf, mailbuf, maillen);
467 strbuf_addstr(&buf, ">\n");
468 string_list_append(&pp->in_body_headers,
469 strbuf_detach(&buf, NULL));
470
471 mailbuf = pp->from_ident->mail_begin;
472 maillen = pp->from_ident->mail_end - mailbuf;
473 namebuf = pp->from_ident->name_begin;
474 namelen = pp->from_ident->name_end - namebuf;
475 }
476
477 strbuf_addstr(sb, "From: ");
478 if (pp->encode_email_headers &&
479 needs_rfc2047_encoding(namebuf, namelen)) {
480 add_rfc2047(sb, namebuf, namelen,
481 encoding, RFC2047_ADDRESS);
482 max_length = 76; /* per rfc2047 */
483 } else if (needs_rfc822_quoting(namebuf, namelen)) {
484 struct strbuf quoted = STRBUF_INIT;
485 add_rfc822_quoted(&quoted, namebuf, namelen);
486 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
487 -6, 1, max_length);
488 strbuf_release(&quoted);
489 } else {
490 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
491 -6, 1, max_length);
492 }
493
494 if (max_length <
495 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
496 strbuf_addch(sb, '\n');
497 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
498 } else {
499 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
500 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
501 (int)namelen, namebuf, (int)maillen, mailbuf);
502 }
503
504 switch (pp->fmt) {
505 case CMIT_FMT_MEDIUM:
506 strbuf_addf(sb, "Date: %s\n",
507 show_ident_date(&ident, &pp->date_mode));
508 break;
509 case CMIT_FMT_EMAIL:
510 case CMIT_FMT_MBOXRD:
511 strbuf_addf(sb, "Date: %s\n",
512 show_ident_date(&ident, DATE_MODE(RFC2822)));
513 break;
514 case CMIT_FMT_FULLER:
515 strbuf_addf(sb, "%sDate: %s\n", what,
516 show_ident_date(&ident, &pp->date_mode));
517 break;
518 default:
519 /* notin' */
520 break;
521 }
522 }
523
524 static int is_blank_line(const char *line, int *len_p)
525 {
526 int len = *len_p;
527 while (len && isspace(line[len - 1]))
528 len--;
529 *len_p = len;
530 return !len;
531 }
532
533 const char *skip_blank_lines(const char *msg)
534 {
535 for (;;) {
536 int linelen = get_one_line(msg);
537 int ll = linelen;
538 if (!linelen)
539 break;
540 if (!is_blank_line(msg, &ll))
541 break;
542 msg += linelen;
543 }
544 return msg;
545 }
546
547 static void add_merge_info(const struct pretty_print_context *pp,
548 struct strbuf *sb, const struct commit *commit)
549 {
550 struct commit_list *parent = commit->parents;
551
552 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
553 !parent || !parent->next)
554 return;
555
556 strbuf_addstr(sb, "Merge:");
557
558 while (parent) {
559 struct object_id *oidp = &parent->item->object.oid;
560 strbuf_addch(sb, ' ');
561 if (pp->abbrev)
562 strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
563 else
564 strbuf_addstr(sb, oid_to_hex(oidp));
565 parent = parent->next;
566 }
567 strbuf_addch(sb, '\n');
568 }
569
570 static char *get_header(const char *msg, const char *key)
571 {
572 size_t len;
573 const char *v = find_commit_header(msg, key, &len);
574 return v ? xmemdupz(v, len) : NULL;
575 }
576
577 static char *replace_encoding_header(char *buf, const char *encoding)
578 {
579 struct strbuf tmp = STRBUF_INIT;
580 size_t start, len;
581 char *cp = buf;
582
583 /* guess if there is an encoding header before a \n\n */
584 while (!starts_with(cp, "encoding ")) {
585 cp = strchr(cp, '\n');
586 if (!cp || *++cp == '\n')
587 return buf;
588 }
589 start = cp - buf;
590 cp = strchr(cp, '\n');
591 if (!cp)
592 return buf; /* should not happen but be defensive */
593 len = cp + 1 - (buf + start);
594
595 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
596 if (is_encoding_utf8(encoding)) {
597 /* we have re-coded to UTF-8; drop the header */
598 strbuf_remove(&tmp, start, len);
599 } else {
600 /* just replaces XXXX in 'encoding XXXX\n' */
601 strbuf_splice(&tmp, start + strlen("encoding "),
602 len - strlen("encoding \n"),
603 encoding, strlen(encoding));
604 }
605 return strbuf_detach(&tmp, NULL);
606 }
607
608 const char *repo_logmsg_reencode(struct repository *r,
609 const struct commit *commit,
610 char **commit_encoding,
611 const char *output_encoding)
612 {
613 static const char *utf8 = "UTF-8";
614 const char *use_encoding;
615 char *encoding;
616 const char *msg = repo_get_commit_buffer(r, commit, NULL);
617 char *out;
618
619 if (!output_encoding || !*output_encoding) {
620 if (commit_encoding)
621 *commit_encoding = get_header(msg, "encoding");
622 return msg;
623 }
624 encoding = get_header(msg, "encoding");
625 if (commit_encoding)
626 *commit_encoding = encoding;
627 use_encoding = encoding ? encoding : utf8;
628 if (same_encoding(use_encoding, output_encoding)) {
629 /*
630 * No encoding work to be done. If we have no encoding header
631 * at all, then there's nothing to do, and we can return the
632 * message verbatim (whether newly allocated or not).
633 */
634 if (!encoding)
635 return msg;
636
637 /*
638 * Otherwise, we still want to munge the encoding header in the
639 * result, which will be done by modifying the buffer. If we
640 * are using a fresh copy, we can reuse it. But if we are using
641 * the cached copy from get_commit_buffer, we need to duplicate it
642 * to avoid munging the cached copy.
643 */
644 if (msg == get_cached_commit_buffer(r, commit, NULL))
645 out = xstrdup(msg);
646 else
647 out = (char *)msg;
648 }
649 else {
650 /*
651 * There's actual encoding work to do. Do the reencoding, which
652 * still leaves the header to be replaced in the next step. At
653 * this point, we are done with msg. If we allocated a fresh
654 * copy, we can free it.
655 */
656 out = reencode_string(msg, output_encoding, use_encoding);
657 if (out)
658 repo_unuse_commit_buffer(r, commit, msg);
659 }
660
661 /*
662 * This replacement actually consumes the buffer we hand it, so we do
663 * not have to worry about freeing the old "out" here.
664 */
665 if (out)
666 out = replace_encoding_header(out, output_encoding);
667
668 if (!commit_encoding)
669 free(encoding);
670 /*
671 * If the re-encoding failed, out might be NULL here; in that
672 * case we just return the commit message verbatim.
673 */
674 return out ? out : msg;
675 }
676
677 static int mailmap_name(const char **email, size_t *email_len,
678 const char **name, size_t *name_len)
679 {
680 static struct string_list *mail_map;
681 if (!mail_map) {
682 CALLOC_ARRAY(mail_map, 1);
683 read_mailmap(mail_map);
684 }
685 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
686 }
687
688 static size_t format_person_part(struct strbuf *sb, char part,
689 const char *msg, int len,
690 const struct date_mode *dmode)
691 {
692 /* currently all placeholders have same length */
693 const int placeholder_len = 2;
694 struct ident_split s;
695 const char *name, *mail;
696 size_t maillen, namelen;
697
698 if (split_ident_line(&s, msg, len) < 0)
699 goto skip;
700
701 name = s.name_begin;
702 namelen = s.name_end - s.name_begin;
703 mail = s.mail_begin;
704 maillen = s.mail_end - s.mail_begin;
705
706 if (part == 'N' || part == 'E' || part == 'L') /* mailmap lookup */
707 mailmap_name(&mail, &maillen, &name, &namelen);
708 if (part == 'n' || part == 'N') { /* name */
709 strbuf_add(sb, name, namelen);
710 return placeholder_len;
711 }
712 if (part == 'e' || part == 'E') { /* email */
713 strbuf_add(sb, mail, maillen);
714 return placeholder_len;
715 }
716 if (part == 'l' || part == 'L') { /* local-part */
717 const char *at = memchr(mail, '@', maillen);
718 if (at)
719 maillen = at - mail;
720 strbuf_add(sb, mail, maillen);
721 return placeholder_len;
722 }
723
724 if (!s.date_begin)
725 goto skip;
726
727 if (part == 't') { /* date, UNIX timestamp */
728 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
729 return placeholder_len;
730 }
731
732 switch (part) {
733 case 'd': /* date */
734 strbuf_addstr(sb, show_ident_date(&s, dmode));
735 return placeholder_len;
736 case 'D': /* date, RFC2822 style */
737 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
738 return placeholder_len;
739 case 'r': /* date, relative */
740 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
741 return placeholder_len;
742 case 'i': /* date, ISO 8601-like */
743 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
744 return placeholder_len;
745 case 'I': /* date, ISO 8601 strict */
746 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
747 return placeholder_len;
748 case 's':
749 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
750 return placeholder_len;
751 }
752
753 skip:
754 /*
755 * reading from either a bogus commit, or a reflog entry with
756 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
757 * to compute a valid return value.
758 */
759 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
760 || part == 'D' || part == 'r' || part == 'i')
761 return placeholder_len;
762
763 return 0; /* unknown placeholder */
764 }
765
766 struct chunk {
767 size_t off;
768 size_t len;
769 };
770
771 enum flush_type {
772 no_flush,
773 flush_right,
774 flush_left,
775 flush_left_and_steal,
776 flush_both
777 };
778
779 enum trunc_type {
780 trunc_none,
781 trunc_left,
782 trunc_middle,
783 trunc_right
784 };
785
786 struct format_commit_context {
787 struct repository *repository;
788 const struct commit *commit;
789 const struct pretty_print_context *pretty_ctx;
790 unsigned commit_header_parsed:1;
791 unsigned commit_message_parsed:1;
792 struct signature_check signature_check;
793 enum flush_type flush_type;
794 enum trunc_type truncate;
795 const char *message;
796 char *commit_encoding;
797 size_t width, indent1, indent2;
798 int auto_color;
799 int padding;
800
801 /* These offsets are relative to the start of the commit message. */
802 struct chunk author;
803 struct chunk committer;
804 size_t message_off;
805 size_t subject_off;
806 size_t body_off;
807
808 /* The following ones are relative to the result struct strbuf. */
809 size_t wrap_start;
810 };
811
812 static void parse_commit_header(struct format_commit_context *context)
813 {
814 const char *msg = context->message;
815 int i;
816
817 for (i = 0; msg[i]; i++) {
818 const char *name;
819 int eol;
820 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
821 ; /* do nothing */
822
823 if (i == eol) {
824 break;
825 } else if (skip_prefix(msg + i, "author ", &name)) {
826 context->author.off = name - msg;
827 context->author.len = msg + eol - name;
828 } else if (skip_prefix(msg + i, "committer ", &name)) {
829 context->committer.off = name - msg;
830 context->committer.len = msg + eol - name;
831 }
832 i = eol;
833 }
834 context->message_off = i;
835 context->commit_header_parsed = 1;
836 }
837
838 static int istitlechar(char c)
839 {
840 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
841 (c >= '0' && c <= '9') || c == '.' || c == '_';
842 }
843
844 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
845 {
846 size_t trimlen;
847 size_t start_len = sb->len;
848 int space = 2;
849 int i;
850
851 for (i = 0; i < len; i++) {
852 if (istitlechar(msg[i])) {
853 if (space == 1)
854 strbuf_addch(sb, '-');
855 space = 0;
856 strbuf_addch(sb, msg[i]);
857 if (msg[i] == '.')
858 while (msg[i+1] == '.')
859 i++;
860 } else
861 space |= 1;
862 }
863
864 /* trim any trailing '.' or '-' characters */
865 trimlen = 0;
866 while (sb->len - trimlen > start_len &&
867 (sb->buf[sb->len - 1 - trimlen] == '.'
868 || sb->buf[sb->len - 1 - trimlen] == '-'))
869 trimlen++;
870 strbuf_remove(sb, sb->len - trimlen, trimlen);
871 }
872
873 const char *format_subject(struct strbuf *sb, const char *msg,
874 const char *line_separator)
875 {
876 int first = 1;
877
878 for (;;) {
879 const char *line = msg;
880 int linelen = get_one_line(line);
881
882 msg += linelen;
883 if (!linelen || is_blank_line(line, &linelen))
884 break;
885
886 if (!sb)
887 continue;
888 strbuf_grow(sb, linelen + 2);
889 if (!first)
890 strbuf_addstr(sb, line_separator);
891 strbuf_add(sb, line, linelen);
892 first = 0;
893 }
894 return msg;
895 }
896
897 static void parse_commit_message(struct format_commit_context *c)
898 {
899 const char *msg = c->message + c->message_off;
900 const char *start = c->message;
901
902 msg = skip_blank_lines(msg);
903 c->subject_off = msg - start;
904
905 msg = format_subject(NULL, msg, NULL);
906 msg = skip_blank_lines(msg);
907 c->body_off = msg - start;
908
909 c->commit_message_parsed = 1;
910 }
911
912 static void strbuf_wrap(struct strbuf *sb, size_t pos,
913 size_t width, size_t indent1, size_t indent2)
914 {
915 struct strbuf tmp = STRBUF_INIT;
916
917 if (pos)
918 strbuf_add(&tmp, sb->buf, pos);
919 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
920 (int) indent1, (int) indent2, (int) width);
921 strbuf_swap(&tmp, sb);
922 strbuf_release(&tmp);
923 }
924
925 static void rewrap_message_tail(struct strbuf *sb,
926 struct format_commit_context *c,
927 size_t new_width, size_t new_indent1,
928 size_t new_indent2)
929 {
930 if (c->width == new_width && c->indent1 == new_indent1 &&
931 c->indent2 == new_indent2)
932 return;
933 if (c->wrap_start < sb->len)
934 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
935 c->wrap_start = sb->len;
936 c->width = new_width;
937 c->indent1 = new_indent1;
938 c->indent2 = new_indent2;
939 }
940
941 static int format_reflog_person(struct strbuf *sb,
942 char part,
943 struct reflog_walk_info *log,
944 const struct date_mode *dmode)
945 {
946 const char *ident;
947
948 if (!log)
949 return 2;
950
951 ident = get_reflog_ident(log);
952 if (!ident)
953 return 2;
954
955 return format_person_part(sb, part, ident, strlen(ident), dmode);
956 }
957
958 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
959 const char *placeholder,
960 struct format_commit_context *c)
961 {
962 const char *rest = placeholder;
963 const char *basic_color = NULL;
964
965 if (placeholder[1] == '(') {
966 const char *begin = placeholder + 2;
967 const char *end = strchr(begin, ')');
968 char color[COLOR_MAXLEN];
969
970 if (!end)
971 return 0;
972
973 if (skip_prefix(begin, "auto,", &begin)) {
974 if (!want_color(c->pretty_ctx->color))
975 return end - placeholder + 1;
976 } else if (skip_prefix(begin, "always,", &begin)) {
977 /* nothing to do; we do not respect want_color at all */
978 } else {
979 /* the default is the same as "auto" */
980 if (!want_color(c->pretty_ctx->color))
981 return end - placeholder + 1;
982 }
983
984 if (color_parse_mem(begin, end - begin, color) < 0)
985 die(_("unable to parse --pretty format"));
986 strbuf_addstr(sb, color);
987 return end - placeholder + 1;
988 }
989
990 /*
991 * We handle things like "%C(red)" above; for historical reasons, there
992 * are a few colors that can be specified without parentheses (and
993 * they cannot support things like "auto" or "always" at all).
994 */
995 if (skip_prefix(placeholder + 1, "red", &rest))
996 basic_color = GIT_COLOR_RED;
997 else if (skip_prefix(placeholder + 1, "green", &rest))
998 basic_color = GIT_COLOR_GREEN;
999 else if (skip_prefix(placeholder + 1, "blue", &rest))
1000 basic_color = GIT_COLOR_BLUE;
1001 else if (skip_prefix(placeholder + 1, "reset", &rest))
1002 basic_color = GIT_COLOR_RESET;
1003
1004 if (basic_color && want_color(c->pretty_ctx->color))
1005 strbuf_addstr(sb, basic_color);
1006
1007 return rest - placeholder;
1008 }
1009
1010 static size_t parse_padding_placeholder(const char *placeholder,
1011 struct format_commit_context *c)
1012 {
1013 const char *ch = placeholder;
1014 enum flush_type flush_type;
1015 int to_column = 0;
1016
1017 switch (*ch++) {
1018 case '<':
1019 flush_type = flush_right;
1020 break;
1021 case '>':
1022 if (*ch == '<') {
1023 flush_type = flush_both;
1024 ch++;
1025 } else if (*ch == '>') {
1026 flush_type = flush_left_and_steal;
1027 ch++;
1028 } else
1029 flush_type = flush_left;
1030 break;
1031 default:
1032 return 0;
1033 }
1034
1035 /* the next value means "wide enough to that column" */
1036 if (*ch == '|') {
1037 to_column = 1;
1038 ch++;
1039 }
1040
1041 if (*ch == '(') {
1042 const char *start = ch + 1;
1043 const char *end = start + strcspn(start, ",)");
1044 char *next;
1045 int width;
1046 if (!end || end == start)
1047 return 0;
1048 width = strtol(start, &next, 10);
1049 if (next == start || width == 0)
1050 return 0;
1051 if (width < 0) {
1052 if (to_column)
1053 width += term_columns();
1054 if (width < 0)
1055 return 0;
1056 }
1057 c->padding = to_column ? -width : width;
1058 c->flush_type = flush_type;
1059
1060 if (*end == ',') {
1061 start = end + 1;
1062 end = strchr(start, ')');
1063 if (!end || end == start)
1064 return 0;
1065 if (starts_with(start, "trunc)"))
1066 c->truncate = trunc_right;
1067 else if (starts_with(start, "ltrunc)"))
1068 c->truncate = trunc_left;
1069 else if (starts_with(start, "mtrunc)"))
1070 c->truncate = trunc_middle;
1071 else
1072 return 0;
1073 } else
1074 c->truncate = trunc_none;
1075
1076 return end - placeholder + 1;
1077 }
1078 return 0;
1079 }
1080
1081 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1082 const char **end, const char **valuestart,
1083 size_t *valuelen)
1084 {
1085 const char *p;
1086
1087 if (!(skip_prefix(to_parse, candidate, &p)))
1088 return 0;
1089 if (valuestart) {
1090 if (*p == '=') {
1091 *valuestart = p + 1;
1092 *valuelen = strcspn(*valuestart, ",)");
1093 p = *valuestart + *valuelen;
1094 } else {
1095 if (*p != ',' && *p != ')')
1096 return 0;
1097 *valuestart = NULL;
1098 *valuelen = 0;
1099 }
1100 }
1101 if (*p == ',') {
1102 *end = p + 1;
1103 return 1;
1104 }
1105 if (*p == ')') {
1106 *end = p;
1107 return 1;
1108 }
1109 return 0;
1110 }
1111
1112 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1113 const char **end, int *val)
1114 {
1115 const char *argval;
1116 char *strval;
1117 size_t arglen;
1118 int v;
1119
1120 if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1121 return 0;
1122
1123 if (!argval) {
1124 *val = 1;
1125 return 1;
1126 }
1127
1128 strval = xstrndup(argval, arglen);
1129 v = git_parse_maybe_bool(strval);
1130 free(strval);
1131
1132 if (v == -1)
1133 return 0;
1134
1135 *val = v;
1136
1137 return 1;
1138 }
1139
1140 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1141 {
1142 const struct string_list *list = ud;
1143 const struct string_list_item *item;
1144
1145 for_each_string_list_item (item, list) {
1146 if (key->len == (uintptr_t)item->util &&
1147 !strncasecmp(item->string, key->buf, key->len))
1148 return 1;
1149 }
1150 return 0;
1151 }
1152
1153 int format_set_trailers_options(struct process_trailer_options *opts,
1154 struct string_list *filter_list,
1155 struct strbuf *sepbuf,
1156 struct strbuf *kvsepbuf,
1157 const char **arg,
1158 char **invalid_arg)
1159 {
1160 for (;;) {
1161 const char *argval;
1162 size_t arglen;
1163
1164 if (**arg == ')')
1165 break;
1166
1167 if (match_placeholder_arg_value(*arg, "key", arg, &argval, &arglen)) {
1168 uintptr_t len = arglen;
1169
1170 if (!argval)
1171 return -1;
1172
1173 if (len && argval[len - 1] == ':')
1174 len--;
1175 string_list_append(filter_list, argval)->util = (char *)len;
1176
1177 opts->filter = format_trailer_match_cb;
1178 opts->filter_data = filter_list;
1179 opts->only_trailers = 1;
1180 } else if (match_placeholder_arg_value(*arg, "separator", arg, &argval, &arglen)) {
1181 char *fmt;
1182
1183 strbuf_reset(sepbuf);
1184 fmt = xstrndup(argval, arglen);
1185 strbuf_expand(sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1186 free(fmt);
1187 opts->separator = sepbuf;
1188 } else if (match_placeholder_arg_value(*arg, "key_value_separator", arg, &argval, &arglen)) {
1189 char *fmt;
1190
1191 strbuf_reset(kvsepbuf);
1192 fmt = xstrndup(argval, arglen);
1193 strbuf_expand(kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1194 free(fmt);
1195 opts->key_value_separator = kvsepbuf;
1196 } else if (!match_placeholder_bool_arg(*arg, "only", arg, &opts->only_trailers) &&
1197 !match_placeholder_bool_arg(*arg, "unfold", arg, &opts->unfold) &&
1198 !match_placeholder_bool_arg(*arg, "keyonly", arg, &opts->key_only) &&
1199 !match_placeholder_bool_arg(*arg, "valueonly", arg, &opts->value_only)) {
1200 if (invalid_arg) {
1201 size_t len = strcspn(*arg, ",)");
1202 *invalid_arg = xstrndup(*arg, len);
1203 }
1204 return -1;
1205 }
1206 }
1207 return 0;
1208 }
1209
1210 static size_t parse_describe_args(const char *start, struct strvec *args)
1211 {
1212 const char *options[] = { "match", "exclude" };
1213 const char *arg = start;
1214
1215 for (;;) {
1216 const char *matched = NULL;
1217 const char *argval;
1218 size_t arglen = 0;
1219 int i;
1220
1221 for (i = 0; i < ARRAY_SIZE(options); i++) {
1222 if (match_placeholder_arg_value(arg, options[i], &arg,
1223 &argval, &arglen)) {
1224 matched = options[i];
1225 break;
1226 }
1227 }
1228 if (!matched)
1229 break;
1230
1231 if (!arglen)
1232 return 0;
1233 strvec_pushf(args, "--%s=%.*s", matched, (int)arglen, argval);
1234 }
1235 return arg - start;
1236 }
1237
1238 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1239 const char *placeholder,
1240 void *context)
1241 {
1242 struct format_commit_context *c = context;
1243 const struct commit *commit = c->commit;
1244 const char *msg = c->message;
1245 struct commit_list *p;
1246 const char *arg, *eol;
1247 size_t res;
1248 char **slot;
1249
1250 /* these are independent of the commit */
1251 res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1252 if (res)
1253 return res;
1254
1255 switch (placeholder[0]) {
1256 case 'C':
1257 if (starts_with(placeholder + 1, "(auto)")) {
1258 c->auto_color = want_color(c->pretty_ctx->color);
1259 if (c->auto_color && sb->len)
1260 strbuf_addstr(sb, GIT_COLOR_RESET);
1261 return 7; /* consumed 7 bytes, "C(auto)" */
1262 } else {
1263 int ret = parse_color(sb, placeholder, c);
1264 if (ret)
1265 c->auto_color = 0;
1266 /*
1267 * Otherwise, we decided to treat %C<unknown>
1268 * as a literal string, and the previous
1269 * %C(auto) is still valid.
1270 */
1271 return ret;
1272 }
1273 case 'w':
1274 if (placeholder[1] == '(') {
1275 unsigned long width = 0, indent1 = 0, indent2 = 0;
1276 char *next;
1277 const char *start = placeholder + 2;
1278 const char *end = strchr(start, ')');
1279 if (!end)
1280 return 0;
1281 if (end > start) {
1282 width = strtoul(start, &next, 10);
1283 if (*next == ',') {
1284 indent1 = strtoul(next + 1, &next, 10);
1285 if (*next == ',') {
1286 indent2 = strtoul(next + 1,
1287 &next, 10);
1288 }
1289 }
1290 if (*next != ')')
1291 return 0;
1292 }
1293 rewrap_message_tail(sb, c, width, indent1, indent2);
1294 return end - placeholder + 1;
1295 } else
1296 return 0;
1297
1298 case '<':
1299 case '>':
1300 return parse_padding_placeholder(placeholder, c);
1301 }
1302
1303 if (skip_prefix(placeholder, "(describe", &arg)) {
1304 struct child_process cmd = CHILD_PROCESS_INIT;
1305 struct strbuf out = STRBUF_INIT;
1306 struct strbuf err = STRBUF_INIT;
1307 struct pretty_print_describe_status *describe_status;
1308
1309 describe_status = c->pretty_ctx->describe_status;
1310 if (describe_status) {
1311 if (!describe_status->max_invocations)
1312 return 0;
1313 describe_status->max_invocations--;
1314 }
1315
1316 cmd.git_cmd = 1;
1317 strvec_push(&cmd.args, "describe");
1318
1319 if (*arg == ':') {
1320 arg++;
1321 arg += parse_describe_args(arg, &cmd.args);
1322 }
1323
1324 if (*arg != ')') {
1325 child_process_clear(&cmd);
1326 return 0;
1327 }
1328
1329 strvec_push(&cmd.args, oid_to_hex(&commit->object.oid));
1330 pipe_command(&cmd, NULL, 0, &out, 0, &err, 0);
1331 strbuf_rtrim(&out);
1332 strbuf_addbuf(sb, &out);
1333 strbuf_release(&out);
1334 strbuf_release(&err);
1335 return arg - placeholder + 1;
1336 }
1337
1338 /* these depend on the commit */
1339 if (!commit->object.parsed)
1340 parse_object(the_repository, &commit->object.oid);
1341
1342 switch (placeholder[0]) {
1343 case 'H': /* commit hash */
1344 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1345 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1346 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1347 return 1;
1348 case 'h': /* abbreviated commit hash */
1349 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1350 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1351 c->pretty_ctx->abbrev);
1352 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1353 return 1;
1354 case 'T': /* tree hash */
1355 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1356 return 1;
1357 case 't': /* abbreviated tree hash */
1358 strbuf_add_unique_abbrev(sb,
1359 get_commit_tree_oid(commit),
1360 c->pretty_ctx->abbrev);
1361 return 1;
1362 case 'P': /* parent hashes */
1363 for (p = commit->parents; p; p = p->next) {
1364 if (p != commit->parents)
1365 strbuf_addch(sb, ' ');
1366 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1367 }
1368 return 1;
1369 case 'p': /* abbreviated parent hashes */
1370 for (p = commit->parents; p; p = p->next) {
1371 if (p != commit->parents)
1372 strbuf_addch(sb, ' ');
1373 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1374 c->pretty_ctx->abbrev);
1375 }
1376 return 1;
1377 case 'm': /* left/right/bottom */
1378 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1379 return 1;
1380 case 'd':
1381 format_decorations(sb, commit, c->auto_color);
1382 return 1;
1383 case 'D':
1384 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1385 return 1;
1386 case 'S': /* tag/branch like --source */
1387 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1388 return 0;
1389 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1390 if (!(slot && *slot))
1391 return 0;
1392 strbuf_addstr(sb, *slot);
1393 return 1;
1394 case 'g': /* reflog info */
1395 switch(placeholder[1]) {
1396 case 'd': /* reflog selector */
1397 case 'D':
1398 if (c->pretty_ctx->reflog_info)
1399 get_reflog_selector(sb,
1400 c->pretty_ctx->reflog_info,
1401 &c->pretty_ctx->date_mode,
1402 c->pretty_ctx->date_mode_explicit,
1403 (placeholder[1] == 'd'));
1404 return 2;
1405 case 's': /* reflog message */
1406 if (c->pretty_ctx->reflog_info)
1407 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1408 return 2;
1409 case 'n':
1410 case 'N':
1411 case 'e':
1412 case 'E':
1413 return format_reflog_person(sb,
1414 placeholder[1],
1415 c->pretty_ctx->reflog_info,
1416 &c->pretty_ctx->date_mode);
1417 }
1418 return 0; /* unknown %g placeholder */
1419 case 'N':
1420 if (c->pretty_ctx->notes_message) {
1421 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1422 return 1;
1423 }
1424 return 0;
1425 }
1426
1427 if (placeholder[0] == 'G') {
1428 if (!c->signature_check.result)
1429 check_commit_signature(c->commit, &(c->signature_check));
1430 switch (placeholder[1]) {
1431 case 'G':
1432 if (c->signature_check.gpg_output)
1433 strbuf_addstr(sb, c->signature_check.gpg_output);
1434 break;
1435 case '?':
1436 switch (c->signature_check.result) {
1437 case 'G':
1438 switch (c->signature_check.trust_level) {
1439 case TRUST_UNDEFINED:
1440 case TRUST_NEVER:
1441 strbuf_addch(sb, 'U');
1442 break;
1443 default:
1444 strbuf_addch(sb, 'G');
1445 break;
1446 }
1447 break;
1448 case 'B':
1449 case 'E':
1450 case 'N':
1451 case 'X':
1452 case 'Y':
1453 case 'R':
1454 strbuf_addch(sb, c->signature_check.result);
1455 }
1456 break;
1457 case 'S':
1458 if (c->signature_check.signer)
1459 strbuf_addstr(sb, c->signature_check.signer);
1460 break;
1461 case 'K':
1462 if (c->signature_check.key)
1463 strbuf_addstr(sb, c->signature_check.key);
1464 break;
1465 case 'F':
1466 if (c->signature_check.fingerprint)
1467 strbuf_addstr(sb, c->signature_check.fingerprint);
1468 break;
1469 case 'P':
1470 if (c->signature_check.primary_key_fingerprint)
1471 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1472 break;
1473 case 'T':
1474 switch (c->signature_check.trust_level) {
1475 case TRUST_UNDEFINED:
1476 strbuf_addstr(sb, "undefined");
1477 break;
1478 case TRUST_NEVER:
1479 strbuf_addstr(sb, "never");
1480 break;
1481 case TRUST_MARGINAL:
1482 strbuf_addstr(sb, "marginal");
1483 break;
1484 case TRUST_FULLY:
1485 strbuf_addstr(sb, "fully");
1486 break;
1487 case TRUST_ULTIMATE:
1488 strbuf_addstr(sb, "ultimate");
1489 break;
1490 }
1491 break;
1492 default:
1493 return 0;
1494 }
1495 return 2;
1496 }
1497
1498 /* For the rest we have to parse the commit header. */
1499 if (!c->commit_header_parsed) {
1500 msg = c->message =
1501 repo_logmsg_reencode(c->repository, commit,
1502 &c->commit_encoding, "UTF-8");
1503 parse_commit_header(c);
1504 }
1505
1506 switch (placeholder[0]) {
1507 case 'a': /* author ... */
1508 return format_person_part(sb, placeholder[1],
1509 msg + c->author.off, c->author.len,
1510 &c->pretty_ctx->date_mode);
1511 case 'c': /* committer ... */
1512 return format_person_part(sb, placeholder[1],
1513 msg + c->committer.off, c->committer.len,
1514 &c->pretty_ctx->date_mode);
1515 case 'e': /* encoding */
1516 if (c->commit_encoding)
1517 strbuf_addstr(sb, c->commit_encoding);
1518 return 1;
1519 case 'B': /* raw body */
1520 /* message_off is always left at the initial newline */
1521 strbuf_addstr(sb, msg + c->message_off + 1);
1522 return 1;
1523 }
1524
1525 /* Now we need to parse the commit message. */
1526 if (!c->commit_message_parsed)
1527 parse_commit_message(c);
1528
1529 switch (placeholder[0]) {
1530 case 's': /* subject */
1531 format_subject(sb, msg + c->subject_off, " ");
1532 return 1;
1533 case 'f': /* sanitized subject */
1534 eol = strchrnul(msg + c->subject_off, '\n');
1535 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1536 return 1;
1537 case 'b': /* body */
1538 strbuf_addstr(sb, msg + c->body_off);
1539 return 1;
1540 }
1541
1542 if (skip_prefix(placeholder, "(trailers", &arg)) {
1543 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1544 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1545 struct strbuf sepbuf = STRBUF_INIT;
1546 struct strbuf kvsepbuf = STRBUF_INIT;
1547 size_t ret = 0;
1548
1549 opts.no_divider = 1;
1550
1551 if (*arg == ':') {
1552 arg++;
1553 if (format_set_trailers_options(&opts, &filter_list, &sepbuf, &kvsepbuf, &arg, NULL))
1554 goto trailer_out;
1555 }
1556 if (*arg == ')') {
1557 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1558 ret = arg - placeholder + 1;
1559 }
1560 trailer_out:
1561 string_list_clear(&filter_list, 0);
1562 strbuf_release(&sepbuf);
1563 return ret;
1564 }
1565
1566 return 0; /* unknown placeholder */
1567 }
1568
1569 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1570 const char *placeholder,
1571 struct format_commit_context *c)
1572 {
1573 struct strbuf local_sb = STRBUF_INIT;
1574 int total_consumed = 0, len, padding = c->padding;
1575 if (padding < 0) {
1576 const char *start = strrchr(sb->buf, '\n');
1577 int occupied;
1578 if (!start)
1579 start = sb->buf;
1580 occupied = utf8_strnwidth(start, -1, 1);
1581 occupied += c->pretty_ctx->graph_width;
1582 padding = (-padding) - occupied;
1583 }
1584 while (1) {
1585 int modifier = *placeholder == 'C';
1586 int consumed = format_commit_one(&local_sb, placeholder, c);
1587 total_consumed += consumed;
1588
1589 if (!modifier)
1590 break;
1591
1592 placeholder += consumed;
1593 if (*placeholder != '%')
1594 break;
1595 placeholder++;
1596 total_consumed++;
1597 }
1598 len = utf8_strnwidth(local_sb.buf, -1, 1);
1599
1600 if (c->flush_type == flush_left_and_steal) {
1601 const char *ch = sb->buf + sb->len - 1;
1602 while (len > padding && ch > sb->buf) {
1603 const char *p;
1604 if (*ch == ' ') {
1605 ch--;
1606 padding++;
1607 continue;
1608 }
1609 /* check for trailing ansi sequences */
1610 if (*ch != 'm')
1611 break;
1612 p = ch - 1;
1613 while (ch - p < 10 && *p != '\033')
1614 p--;
1615 if (*p != '\033' ||
1616 ch + 1 - p != display_mode_esc_sequence_len(p))
1617 break;
1618 /*
1619 * got a good ansi sequence, put it back to
1620 * local_sb as we're cutting sb
1621 */
1622 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1623 ch = p - 1;
1624 }
1625 strbuf_setlen(sb, ch + 1 - sb->buf);
1626 c->flush_type = flush_left;
1627 }
1628
1629 if (len > padding) {
1630 switch (c->truncate) {
1631 case trunc_left:
1632 strbuf_utf8_replace(&local_sb,
1633 0, len - (padding - 2),
1634 "..");
1635 break;
1636 case trunc_middle:
1637 strbuf_utf8_replace(&local_sb,
1638 padding / 2 - 1,
1639 len - (padding - 2),
1640 "..");
1641 break;
1642 case trunc_right:
1643 strbuf_utf8_replace(&local_sb,
1644 padding - 2, len - (padding - 2),
1645 "..");
1646 break;
1647 case trunc_none:
1648 break;
1649 }
1650 strbuf_addbuf(sb, &local_sb);
1651 } else {
1652 int sb_len = sb->len, offset = 0;
1653 if (c->flush_type == flush_left)
1654 offset = padding - len;
1655 else if (c->flush_type == flush_both)
1656 offset = (padding - len) / 2;
1657 /*
1658 * we calculate padding in columns, now
1659 * convert it back to chars
1660 */
1661 padding = padding - len + local_sb.len;
1662 strbuf_addchars(sb, ' ', padding);
1663 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1664 local_sb.len);
1665 }
1666 strbuf_release(&local_sb);
1667 c->flush_type = no_flush;
1668 return total_consumed;
1669 }
1670
1671 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1672 const char *placeholder,
1673 void *context)
1674 {
1675 int consumed;
1676 size_t orig_len;
1677 enum {
1678 NO_MAGIC,
1679 ADD_LF_BEFORE_NON_EMPTY,
1680 DEL_LF_BEFORE_EMPTY,
1681 ADD_SP_BEFORE_NON_EMPTY
1682 } magic = NO_MAGIC;
1683
1684 switch (placeholder[0]) {
1685 case '-':
1686 magic = DEL_LF_BEFORE_EMPTY;
1687 break;
1688 case '+':
1689 magic = ADD_LF_BEFORE_NON_EMPTY;
1690 break;
1691 case ' ':
1692 magic = ADD_SP_BEFORE_NON_EMPTY;
1693 break;
1694 default:
1695 break;
1696 }
1697 if (magic != NO_MAGIC)
1698 placeholder++;
1699
1700 orig_len = sb->len;
1701 if (((struct format_commit_context *)context)->flush_type != no_flush)
1702 consumed = format_and_pad_commit(sb, placeholder, context);
1703 else
1704 consumed = format_commit_one(sb, placeholder, context);
1705 if (magic == NO_MAGIC)
1706 return consumed;
1707
1708 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1709 while (sb->len && sb->buf[sb->len - 1] == '\n')
1710 strbuf_setlen(sb, sb->len - 1);
1711 } else if (orig_len != sb->len) {
1712 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1713 strbuf_insertstr(sb, orig_len, "\n");
1714 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1715 strbuf_insertstr(sb, orig_len, " ");
1716 }
1717 return consumed + 1;
1718 }
1719
1720 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1721 void *context)
1722 {
1723 struct userformat_want *w = context;
1724
1725 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1726 placeholder++;
1727
1728 switch (*placeholder) {
1729 case 'N':
1730 w->notes = 1;
1731 break;
1732 case 'S':
1733 w->source = 1;
1734 break;
1735 }
1736 return 0;
1737 }
1738
1739 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1740 {
1741 struct strbuf dummy = STRBUF_INIT;
1742
1743 if (!fmt) {
1744 if (!user_format)
1745 return;
1746 fmt = user_format;
1747 }
1748 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1749 strbuf_release(&dummy);
1750 }
1751
1752 void repo_format_commit_message(struct repository *r,
1753 const struct commit *commit,
1754 const char *format, struct strbuf *sb,
1755 const struct pretty_print_context *pretty_ctx)
1756 {
1757 struct format_commit_context context = {
1758 .repository = r,
1759 .commit = commit,
1760 .pretty_ctx = pretty_ctx,
1761 .wrap_start = sb->len
1762 };
1763 const char *output_enc = pretty_ctx->output_encoding;
1764 const char *utf8 = "UTF-8";
1765
1766 strbuf_expand(sb, format, format_commit_item, &context);
1767 rewrap_message_tail(sb, &context, 0, 0, 0);
1768
1769 /*
1770 * Convert output to an actual output encoding; note that
1771 * format_commit_item() will always use UTF-8, so we don't
1772 * have to bother if that's what the output wants.
1773 */
1774 if (output_enc) {
1775 if (same_encoding(utf8, output_enc))
1776 output_enc = NULL;
1777 } else {
1778 if (context.commit_encoding &&
1779 !same_encoding(context.commit_encoding, utf8))
1780 output_enc = context.commit_encoding;
1781 }
1782
1783 if (output_enc) {
1784 size_t outsz;
1785 char *out = reencode_string_len(sb->buf, sb->len,
1786 output_enc, utf8, &outsz);
1787 if (out)
1788 strbuf_attach(sb, out, outsz, outsz + 1);
1789 }
1790
1791 free(context.commit_encoding);
1792 repo_unuse_commit_buffer(r, commit, context.message);
1793 }
1794
1795 static void pp_header(struct pretty_print_context *pp,
1796 const char *encoding,
1797 const struct commit *commit,
1798 const char **msg_p,
1799 struct strbuf *sb)
1800 {
1801 int parents_shown = 0;
1802
1803 for (;;) {
1804 const char *name, *line = *msg_p;
1805 int linelen = get_one_line(*msg_p);
1806
1807 if (!linelen)
1808 return;
1809 *msg_p += linelen;
1810
1811 if (linelen == 1)
1812 /* End of header */
1813 return;
1814
1815 if (pp->fmt == CMIT_FMT_RAW) {
1816 strbuf_add(sb, line, linelen);
1817 continue;
1818 }
1819
1820 if (starts_with(line, "parent ")) {
1821 if (linelen != the_hash_algo->hexsz + 8)
1822 die("bad parent line in commit");
1823 continue;
1824 }
1825
1826 if (!parents_shown) {
1827 unsigned num = commit_list_count(commit->parents);
1828 /* with enough slop */
1829 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1830 add_merge_info(pp, sb, commit);
1831 parents_shown = 1;
1832 }
1833
1834 /*
1835 * MEDIUM == DEFAULT shows only author with dates.
1836 * FULL shows both authors but not dates.
1837 * FULLER shows both authors and dates.
1838 */
1839 if (skip_prefix(line, "author ", &name)) {
1840 strbuf_grow(sb, linelen + 80);
1841 pp_user_info(pp, "Author", sb, name, encoding);
1842 }
1843 if (skip_prefix(line, "committer ", &name) &&
1844 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1845 strbuf_grow(sb, linelen + 80);
1846 pp_user_info(pp, "Commit", sb, name, encoding);
1847 }
1848 }
1849 }
1850
1851 void pp_title_line(struct pretty_print_context *pp,
1852 const char **msg_p,
1853 struct strbuf *sb,
1854 const char *encoding,
1855 int need_8bit_cte)
1856 {
1857 static const int max_length = 78; /* per rfc2047 */
1858 struct strbuf title;
1859
1860 strbuf_init(&title, 80);
1861 *msg_p = format_subject(&title, *msg_p,
1862 pp->preserve_subject ? "\n" : " ");
1863
1864 strbuf_grow(sb, title.len + 1024);
1865 if (pp->print_email_subject) {
1866 if (pp->rev)
1867 fmt_output_email_subject(sb, pp->rev);
1868 if (pp->encode_email_headers &&
1869 needs_rfc2047_encoding(title.buf, title.len))
1870 add_rfc2047(sb, title.buf, title.len,
1871 encoding, RFC2047_SUBJECT);
1872 else
1873 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1874 -last_line_length(sb), 1, max_length);
1875 } else {
1876 strbuf_addbuf(sb, &title);
1877 }
1878 strbuf_addch(sb, '\n');
1879
1880 if (need_8bit_cte == 0) {
1881 int i;
1882 for (i = 0; i < pp->in_body_headers.nr; i++) {
1883 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1884 need_8bit_cte = 1;
1885 break;
1886 }
1887 }
1888 }
1889
1890 if (need_8bit_cte > 0) {
1891 const char *header_fmt =
1892 "MIME-Version: 1.0\n"
1893 "Content-Type: text/plain; charset=%s\n"
1894 "Content-Transfer-Encoding: 8bit\n";
1895 strbuf_addf(sb, header_fmt, encoding);
1896 }
1897 if (pp->after_subject) {
1898 strbuf_addstr(sb, pp->after_subject);
1899 }
1900 if (cmit_fmt_is_mail(pp->fmt)) {
1901 strbuf_addch(sb, '\n');
1902 }
1903
1904 if (pp->in_body_headers.nr) {
1905 int i;
1906 for (i = 0; i < pp->in_body_headers.nr; i++) {
1907 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1908 free(pp->in_body_headers.items[i].string);
1909 }
1910 string_list_clear(&pp->in_body_headers, 0);
1911 strbuf_addch(sb, '\n');
1912 }
1913
1914 strbuf_release(&title);
1915 }
1916
1917 static int pp_utf8_width(const char *start, const char *end)
1918 {
1919 int width = 0;
1920 size_t remain = end - start;
1921
1922 while (remain) {
1923 int n = utf8_width(&start, &remain);
1924 if (n < 0 || !start)
1925 return -1;
1926 width += n;
1927 }
1928 return width;
1929 }
1930
1931 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1932 const char *line, int linelen)
1933 {
1934 const char *tab;
1935
1936 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1937 int width = pp_utf8_width(line, tab);
1938
1939 /*
1940 * If it wasn't well-formed utf8, or it
1941 * had characters with badly defined
1942 * width (control characters etc), just
1943 * give up on trying to align things.
1944 */
1945 if (width < 0)
1946 break;
1947
1948 /* Output the data .. */
1949 strbuf_add(sb, line, tab - line);
1950
1951 /* .. and the de-tabified tab */
1952 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1953
1954 /* Skip over the printed part .. */
1955 linelen -= tab + 1 - line;
1956 line = tab + 1;
1957 }
1958
1959 /*
1960 * Print out everything after the last tab without
1961 * worrying about width - there's nothing more to
1962 * align.
1963 */
1964 strbuf_add(sb, line, linelen);
1965 }
1966
1967 /*
1968 * pp_handle_indent() prints out the intendation, and
1969 * the whole line (without the final newline), after
1970 * de-tabifying.
1971 */
1972 static void pp_handle_indent(struct pretty_print_context *pp,
1973 struct strbuf *sb, int indent,
1974 const char *line, int linelen)
1975 {
1976 strbuf_addchars(sb, ' ', indent);
1977 if (pp->expand_tabs_in_log)
1978 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1979 else
1980 strbuf_add(sb, line, linelen);
1981 }
1982
1983 static int is_mboxrd_from(const char *line, int len)
1984 {
1985 /*
1986 * a line matching /^From $/ here would only have len == 4
1987 * at this point because is_empty_line would've trimmed all
1988 * trailing space
1989 */
1990 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1991 }
1992
1993 void pp_remainder(struct pretty_print_context *pp,
1994 const char **msg_p,
1995 struct strbuf *sb,
1996 int indent)
1997 {
1998 int first = 1;
1999 for (;;) {
2000 const char *line = *msg_p;
2001 int linelen = get_one_line(line);
2002 *msg_p += linelen;
2003
2004 if (!linelen)
2005 break;
2006
2007 if (is_blank_line(line, &linelen)) {
2008 if (first)
2009 continue;
2010 if (pp->fmt == CMIT_FMT_SHORT)
2011 break;
2012 }
2013 first = 0;
2014
2015 strbuf_grow(sb, linelen + indent + 20);
2016 if (indent)
2017 pp_handle_indent(pp, sb, indent, line, linelen);
2018 else if (pp->expand_tabs_in_log)
2019 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
2020 line, linelen);
2021 else {
2022 if (pp->fmt == CMIT_FMT_MBOXRD &&
2023 is_mboxrd_from(line, linelen))
2024 strbuf_addch(sb, '>');
2025
2026 strbuf_add(sb, line, linelen);
2027 }
2028 strbuf_addch(sb, '\n');
2029 }
2030 }
2031
2032 void pretty_print_commit(struct pretty_print_context *pp,
2033 const struct commit *commit,
2034 struct strbuf *sb)
2035 {
2036 unsigned long beginning_of_body;
2037 int indent = 4;
2038 const char *msg;
2039 const char *reencoded;
2040 const char *encoding;
2041 int need_8bit_cte = pp->need_8bit_cte;
2042
2043 if (pp->fmt == CMIT_FMT_USERFORMAT) {
2044 format_commit_message(commit, user_format, sb, pp);
2045 return;
2046 }
2047
2048 encoding = get_log_output_encoding();
2049 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
2050
2051 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2052 indent = 0;
2053
2054 /*
2055 * We need to check and emit Content-type: to mark it
2056 * as 8-bit if we haven't done so.
2057 */
2058 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
2059 int i, ch, in_body;
2060
2061 for (in_body = i = 0; (ch = msg[i]); i++) {
2062 if (!in_body) {
2063 /* author could be non 7-bit ASCII but
2064 * the log may be so; skip over the
2065 * header part first.
2066 */
2067 if (ch == '\n' && msg[i+1] == '\n')
2068 in_body = 1;
2069 }
2070 else if (non_ascii(ch)) {
2071 need_8bit_cte = 1;
2072 break;
2073 }
2074 }
2075 }
2076
2077 pp_header(pp, encoding, commit, &msg, sb);
2078 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
2079 strbuf_addch(sb, '\n');
2080 }
2081
2082 /* Skip excess blank lines at the beginning of body, if any... */
2083 msg = skip_blank_lines(msg);
2084
2085 /* These formats treat the title line specially. */
2086 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2087 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2088
2089 beginning_of_body = sb->len;
2090 if (pp->fmt != CMIT_FMT_ONELINE)
2091 pp_remainder(pp, &msg, sb, indent);
2092 strbuf_rtrim(sb);
2093
2094 /* Make sure there is an EOLN for the non-oneline case */
2095 if (pp->fmt != CMIT_FMT_ONELINE)
2096 strbuf_addch(sb, '\n');
2097
2098 /*
2099 * The caller may append additional body text in e-mail
2100 * format. Make sure we did not strip the blank line
2101 * between the header and the body.
2102 */
2103 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2104 strbuf_addch(sb, '\n');
2105
2106 unuse_commit_buffer(commit, reencoded);
2107 }
2108
2109 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2110 struct strbuf *sb)
2111 {
2112 struct pretty_print_context pp = {0};
2113 pp.fmt = fmt;
2114 pretty_print_commit(&pp, commit, sb);
2115 }
git for Debian