?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ci: run `make sparse` as part of the GitHub workflow
[git.git] / pretty.c
blobb1ecd039cef29ecc77edd894c500dfadf49058c8
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 'h': /* date, human */
749 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(HUMAN)));
750 return placeholder_len;
751 case 's':
752 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
753 return placeholder_len;
754 }
755
756 skip:
757 /*
758 * reading from either a bogus commit, or a reflog entry with
759 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
760 * to compute a valid return value.
761 */
762 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
763 || part == 'D' || part == 'r' || part == 'i')
764 return placeholder_len;
765
766 return 0; /* unknown placeholder */
767 }
768
769 struct chunk {
770 size_t off;
771 size_t len;
772 };
773
774 enum flush_type {
775 no_flush,
776 flush_right,
777 flush_left,
778 flush_left_and_steal,
779 flush_both
780 };
781
782 enum trunc_type {
783 trunc_none,
784 trunc_left,
785 trunc_middle,
786 trunc_right
787 };
788
789 struct format_commit_context {
790 struct repository *repository;
791 const struct commit *commit;
792 const struct pretty_print_context *pretty_ctx;
793 unsigned commit_header_parsed:1;
794 unsigned commit_message_parsed:1;
795 struct signature_check signature_check;
796 enum flush_type flush_type;
797 enum trunc_type truncate;
798 const char *message;
799 char *commit_encoding;
800 size_t width, indent1, indent2;
801 int auto_color;
802 int padding;
803
804 /* These offsets are relative to the start of the commit message. */
805 struct chunk author;
806 struct chunk committer;
807 size_t message_off;
808 size_t subject_off;
809 size_t body_off;
810
811 /* The following ones are relative to the result struct strbuf. */
812 size_t wrap_start;
813 };
814
815 static void parse_commit_header(struct format_commit_context *context)
816 {
817 const char *msg = context->message;
818 int i;
819
820 for (i = 0; msg[i]; i++) {
821 const char *name;
822 int eol;
823 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
824 ; /* do nothing */
825
826 if (i == eol) {
827 break;
828 } else if (skip_prefix(msg + i, "author ", &name)) {
829 context->author.off = name - msg;
830 context->author.len = msg + eol - name;
831 } else if (skip_prefix(msg + i, "committer ", &name)) {
832 context->committer.off = name - msg;
833 context->committer.len = msg + eol - name;
834 }
835 i = eol;
836 }
837 context->message_off = i;
838 context->commit_header_parsed = 1;
839 }
840
841 static int istitlechar(char c)
842 {
843 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
844 (c >= '0' && c <= '9') || c == '.' || c == '_';
845 }
846
847 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
848 {
849 size_t trimlen;
850 size_t start_len = sb->len;
851 int space = 2;
852 int i;
853
854 for (i = 0; i < len; i++) {
855 if (istitlechar(msg[i])) {
856 if (space == 1)
857 strbuf_addch(sb, '-');
858 space = 0;
859 strbuf_addch(sb, msg[i]);
860 if (msg[i] == '.')
861 while (msg[i+1] == '.')
862 i++;
863 } else
864 space |= 1;
865 }
866
867 /* trim any trailing '.' or '-' characters */
868 trimlen = 0;
869 while (sb->len - trimlen > start_len &&
870 (sb->buf[sb->len - 1 - trimlen] == '.'
871 || sb->buf[sb->len - 1 - trimlen] == '-'))
872 trimlen++;
873 strbuf_remove(sb, sb->len - trimlen, trimlen);
874 }
875
876 const char *format_subject(struct strbuf *sb, const char *msg,
877 const char *line_separator)
878 {
879 int first = 1;
880
881 for (;;) {
882 const char *line = msg;
883 int linelen = get_one_line(line);
884
885 msg += linelen;
886 if (!linelen || is_blank_line(line, &linelen))
887 break;
888
889 if (!sb)
890 continue;
891 strbuf_grow(sb, linelen + 2);
892 if (!first)
893 strbuf_addstr(sb, line_separator);
894 strbuf_add(sb, line, linelen);
895 first = 0;
896 }
897 return msg;
898 }
899
900 static void parse_commit_message(struct format_commit_context *c)
901 {
902 const char *msg = c->message + c->message_off;
903 const char *start = c->message;
904
905 msg = skip_blank_lines(msg);
906 c->subject_off = msg - start;
907
908 msg = format_subject(NULL, msg, NULL);
909 msg = skip_blank_lines(msg);
910 c->body_off = msg - start;
911
912 c->commit_message_parsed = 1;
913 }
914
915 static void strbuf_wrap(struct strbuf *sb, size_t pos,
916 size_t width, size_t indent1, size_t indent2)
917 {
918 struct strbuf tmp = STRBUF_INIT;
919
920 if (pos)
921 strbuf_add(&tmp, sb->buf, pos);
922 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
923 (int) indent1, (int) indent2, (int) width);
924 strbuf_swap(&tmp, sb);
925 strbuf_release(&tmp);
926 }
927
928 static void rewrap_message_tail(struct strbuf *sb,
929 struct format_commit_context *c,
930 size_t new_width, size_t new_indent1,
931 size_t new_indent2)
932 {
933 if (c->width == new_width && c->indent1 == new_indent1 &&
934 c->indent2 == new_indent2)
935 return;
936 if (c->wrap_start < sb->len)
937 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
938 c->wrap_start = sb->len;
939 c->width = new_width;
940 c->indent1 = new_indent1;
941 c->indent2 = new_indent2;
942 }
943
944 static int format_reflog_person(struct strbuf *sb,
945 char part,
946 struct reflog_walk_info *log,
947 const struct date_mode *dmode)
948 {
949 const char *ident;
950
951 if (!log)
952 return 2;
953
954 ident = get_reflog_ident(log);
955 if (!ident)
956 return 2;
957
958 return format_person_part(sb, part, ident, strlen(ident), dmode);
959 }
960
961 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
962 const char *placeholder,
963 struct format_commit_context *c)
964 {
965 const char *rest = placeholder;
966 const char *basic_color = NULL;
967
968 if (placeholder[1] == '(') {
969 const char *begin = placeholder + 2;
970 const char *end = strchr(begin, ')');
971 char color[COLOR_MAXLEN];
972
973 if (!end)
974 return 0;
975
976 if (skip_prefix(begin, "auto,", &begin)) {
977 if (!want_color(c->pretty_ctx->color))
978 return end - placeholder + 1;
979 } else if (skip_prefix(begin, "always,", &begin)) {
980 /* nothing to do; we do not respect want_color at all */
981 } else {
982 /* the default is the same as "auto" */
983 if (!want_color(c->pretty_ctx->color))
984 return end - placeholder + 1;
985 }
986
987 if (color_parse_mem(begin, end - begin, color) < 0)
988 die(_("unable to parse --pretty format"));
989 strbuf_addstr(sb, color);
990 return end - placeholder + 1;
991 }
992
993 /*
994 * We handle things like "%C(red)" above; for historical reasons, there
995 * are a few colors that can be specified without parentheses (and
996 * they cannot support things like "auto" or "always" at all).
997 */
998 if (skip_prefix(placeholder + 1, "red", &rest))
999 basic_color = GIT_COLOR_RED;
1000 else if (skip_prefix(placeholder + 1, "green", &rest))
1001 basic_color = GIT_COLOR_GREEN;
1002 else if (skip_prefix(placeholder + 1, "blue", &rest))
1003 basic_color = GIT_COLOR_BLUE;
1004 else if (skip_prefix(placeholder + 1, "reset", &rest))
1005 basic_color = GIT_COLOR_RESET;
1006
1007 if (basic_color && want_color(c->pretty_ctx->color))
1008 strbuf_addstr(sb, basic_color);
1009
1010 return rest - placeholder;
1011 }
1012
1013 static size_t parse_padding_placeholder(const char *placeholder,
1014 struct format_commit_context *c)
1015 {
1016 const char *ch = placeholder;
1017 enum flush_type flush_type;
1018 int to_column = 0;
1019
1020 switch (*ch++) {
1021 case '<':
1022 flush_type = flush_right;
1023 break;
1024 case '>':
1025 if (*ch == '<') {
1026 flush_type = flush_both;
1027 ch++;
1028 } else if (*ch == '>') {
1029 flush_type = flush_left_and_steal;
1030 ch++;
1031 } else
1032 flush_type = flush_left;
1033 break;
1034 default:
1035 return 0;
1036 }
1037
1038 /* the next value means "wide enough to that column" */
1039 if (*ch == '|') {
1040 to_column = 1;
1041 ch++;
1042 }
1043
1044 if (*ch == '(') {
1045 const char *start = ch + 1;
1046 const char *end = start + strcspn(start, ",)");
1047 char *next;
1048 int width;
1049 if (!end || end == start)
1050 return 0;
1051 width = strtol(start, &next, 10);
1052 if (next == start || width == 0)
1053 return 0;
1054 if (width < 0) {
1055 if (to_column)
1056 width += term_columns();
1057 if (width < 0)
1058 return 0;
1059 }
1060 c->padding = to_column ? -width : width;
1061 c->flush_type = flush_type;
1062
1063 if (*end == ',') {
1064 start = end + 1;
1065 end = strchr(start, ')');
1066 if (!end || end == start)
1067 return 0;
1068 if (starts_with(start, "trunc)"))
1069 c->truncate = trunc_right;
1070 else if (starts_with(start, "ltrunc)"))
1071 c->truncate = trunc_left;
1072 else if (starts_with(start, "mtrunc)"))
1073 c->truncate = trunc_middle;
1074 else
1075 return 0;
1076 } else
1077 c->truncate = trunc_none;
1078
1079 return end - placeholder + 1;
1080 }
1081 return 0;
1082 }
1083
1084 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1085 const char **end, const char **valuestart,
1086 size_t *valuelen)
1087 {
1088 const char *p;
1089
1090 if (!(skip_prefix(to_parse, candidate, &p)))
1091 return 0;
1092 if (valuestart) {
1093 if (*p == '=') {
1094 *valuestart = p + 1;
1095 *valuelen = strcspn(*valuestart, ",)");
1096 p = *valuestart + *valuelen;
1097 } else {
1098 if (*p != ',' && *p != ')')
1099 return 0;
1100 *valuestart = NULL;
1101 *valuelen = 0;
1102 }
1103 }
1104 if (*p == ',') {
1105 *end = p + 1;
1106 return 1;
1107 }
1108 if (*p == ')') {
1109 *end = p;
1110 return 1;
1111 }
1112 return 0;
1113 }
1114
1115 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1116 const char **end, int *val)
1117 {
1118 const char *argval;
1119 char *strval;
1120 size_t arglen;
1121 int v;
1122
1123 if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1124 return 0;
1125
1126 if (!argval) {
1127 *val = 1;
1128 return 1;
1129 }
1130
1131 strval = xstrndup(argval, arglen);
1132 v = git_parse_maybe_bool(strval);
1133 free(strval);
1134
1135 if (v == -1)
1136 return 0;
1137
1138 *val = v;
1139
1140 return 1;
1141 }
1142
1143 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1144 {
1145 const struct string_list *list = ud;
1146 const struct string_list_item *item;
1147
1148 for_each_string_list_item (item, list) {
1149 if (key->len == (uintptr_t)item->util &&
1150 !strncasecmp(item->string, key->buf, key->len))
1151 return 1;
1152 }
1153 return 0;
1154 }
1155
1156 int format_set_trailers_options(struct process_trailer_options *opts,
1157 struct string_list *filter_list,
1158 struct strbuf *sepbuf,
1159 struct strbuf *kvsepbuf,
1160 const char **arg,
1161 char **invalid_arg)
1162 {
1163 for (;;) {
1164 const char *argval;
1165 size_t arglen;
1166
1167 if (**arg == ')')
1168 break;
1169
1170 if (match_placeholder_arg_value(*arg, "key", arg, &argval, &arglen)) {
1171 uintptr_t len = arglen;
1172
1173 if (!argval)
1174 return -1;
1175
1176 if (len && argval[len - 1] == ':')
1177 len--;
1178 string_list_append(filter_list, argval)->util = (char *)len;
1179
1180 opts->filter = format_trailer_match_cb;
1181 opts->filter_data = filter_list;
1182 opts->only_trailers = 1;
1183 } else if (match_placeholder_arg_value(*arg, "separator", arg, &argval, &arglen)) {
1184 char *fmt;
1185
1186 strbuf_reset(sepbuf);
1187 fmt = xstrndup(argval, arglen);
1188 strbuf_expand(sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1189 free(fmt);
1190 opts->separator = sepbuf;
1191 } else if (match_placeholder_arg_value(*arg, "key_value_separator", arg, &argval, &arglen)) {
1192 char *fmt;
1193
1194 strbuf_reset(kvsepbuf);
1195 fmt = xstrndup(argval, arglen);
1196 strbuf_expand(kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1197 free(fmt);
1198 opts->key_value_separator = kvsepbuf;
1199 } else if (!match_placeholder_bool_arg(*arg, "only", arg, &opts->only_trailers) &&
1200 !match_placeholder_bool_arg(*arg, "unfold", arg, &opts->unfold) &&
1201 !match_placeholder_bool_arg(*arg, "keyonly", arg, &opts->key_only) &&
1202 !match_placeholder_bool_arg(*arg, "valueonly", arg, &opts->value_only)) {
1203 if (invalid_arg) {
1204 size_t len = strcspn(*arg, ",)");
1205 *invalid_arg = xstrndup(*arg, len);
1206 }
1207 return -1;
1208 }
1209 }
1210 return 0;
1211 }
1212
1213 static size_t parse_describe_args(const char *start, struct strvec *args)
1214 {
1215 const char *options[] = { "match", "exclude" };
1216 const char *arg = start;
1217
1218 for (;;) {
1219 const char *matched = NULL;
1220 const char *argval;
1221 size_t arglen = 0;
1222 int i;
1223
1224 for (i = 0; i < ARRAY_SIZE(options); i++) {
1225 if (match_placeholder_arg_value(arg, options[i], &arg,
1226 &argval, &arglen)) {
1227 matched = options[i];
1228 break;
1229 }
1230 }
1231 if (!matched)
1232 break;
1233
1234 if (!arglen)
1235 return 0;
1236 strvec_pushf(args, "--%s=%.*s", matched, (int)arglen, argval);
1237 }
1238 return arg - start;
1239 }
1240
1241 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1242 const char *placeholder,
1243 void *context)
1244 {
1245 struct format_commit_context *c = context;
1246 const struct commit *commit = c->commit;
1247 const char *msg = c->message;
1248 struct commit_list *p;
1249 const char *arg, *eol;
1250 size_t res;
1251 char **slot;
1252
1253 /* these are independent of the commit */
1254 res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1255 if (res)
1256 return res;
1257
1258 switch (placeholder[0]) {
1259 case 'C':
1260 if (starts_with(placeholder + 1, "(auto)")) {
1261 c->auto_color = want_color(c->pretty_ctx->color);
1262 if (c->auto_color && sb->len)
1263 strbuf_addstr(sb, GIT_COLOR_RESET);
1264 return 7; /* consumed 7 bytes, "C(auto)" */
1265 } else {
1266 int ret = parse_color(sb, placeholder, c);
1267 if (ret)
1268 c->auto_color = 0;
1269 /*
1270 * Otherwise, we decided to treat %C<unknown>
1271 * as a literal string, and the previous
1272 * %C(auto) is still valid.
1273 */
1274 return ret;
1275 }
1276 case 'w':
1277 if (placeholder[1] == '(') {
1278 unsigned long width = 0, indent1 = 0, indent2 = 0;
1279 char *next;
1280 const char *start = placeholder + 2;
1281 const char *end = strchr(start, ')');
1282 if (!end)
1283 return 0;
1284 if (end > start) {
1285 width = strtoul(start, &next, 10);
1286 if (*next == ',') {
1287 indent1 = strtoul(next + 1, &next, 10);
1288 if (*next == ',') {
1289 indent2 = strtoul(next + 1,
1290 &next, 10);
1291 }
1292 }
1293 if (*next != ')')
1294 return 0;
1295 }
1296 rewrap_message_tail(sb, c, width, indent1, indent2);
1297 return end - placeholder + 1;
1298 } else
1299 return 0;
1300
1301 case '<':
1302 case '>':
1303 return parse_padding_placeholder(placeholder, c);
1304 }
1305
1306 if (skip_prefix(placeholder, "(describe", &arg)) {
1307 struct child_process cmd = CHILD_PROCESS_INIT;
1308 struct strbuf out = STRBUF_INIT;
1309 struct strbuf err = STRBUF_INIT;
1310 struct pretty_print_describe_status *describe_status;
1311
1312 describe_status = c->pretty_ctx->describe_status;
1313 if (describe_status) {
1314 if (!describe_status->max_invocations)
1315 return 0;
1316 describe_status->max_invocations--;
1317 }
1318
1319 cmd.git_cmd = 1;
1320 strvec_push(&cmd.args, "describe");
1321
1322 if (*arg == ':') {
1323 arg++;
1324 arg += parse_describe_args(arg, &cmd.args);
1325 }
1326
1327 if (*arg != ')') {
1328 child_process_clear(&cmd);
1329 return 0;
1330 }
1331
1332 strvec_push(&cmd.args, oid_to_hex(&commit->object.oid));
1333 pipe_command(&cmd, NULL, 0, &out, 0, &err, 0);
1334 strbuf_rtrim(&out);
1335 strbuf_addbuf(sb, &out);
1336 strbuf_release(&out);
1337 strbuf_release(&err);
1338 return arg - placeholder + 1;
1339 }
1340
1341 /* these depend on the commit */
1342 if (!commit->object.parsed)
1343 parse_object(the_repository, &commit->object.oid);
1344
1345 switch (placeholder[0]) {
1346 case 'H': /* commit hash */
1347 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1348 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1349 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1350 return 1;
1351 case 'h': /* abbreviated commit hash */
1352 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1353 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1354 c->pretty_ctx->abbrev);
1355 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1356 return 1;
1357 case 'T': /* tree hash */
1358 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1359 return 1;
1360 case 't': /* abbreviated tree hash */
1361 strbuf_add_unique_abbrev(sb,
1362 get_commit_tree_oid(commit),
1363 c->pretty_ctx->abbrev);
1364 return 1;
1365 case 'P': /* parent hashes */
1366 for (p = commit->parents; p; p = p->next) {
1367 if (p != commit->parents)
1368 strbuf_addch(sb, ' ');
1369 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1370 }
1371 return 1;
1372 case 'p': /* abbreviated parent hashes */
1373 for (p = commit->parents; p; p = p->next) {
1374 if (p != commit->parents)
1375 strbuf_addch(sb, ' ');
1376 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1377 c->pretty_ctx->abbrev);
1378 }
1379 return 1;
1380 case 'm': /* left/right/bottom */
1381 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1382 return 1;
1383 case 'd':
1384 format_decorations(sb, commit, c->auto_color);
1385 return 1;
1386 case 'D':
1387 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1388 return 1;
1389 case 'S': /* tag/branch like --source */
1390 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1391 return 0;
1392 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1393 if (!(slot && *slot))
1394 return 0;
1395 strbuf_addstr(sb, *slot);
1396 return 1;
1397 case 'g': /* reflog info */
1398 switch(placeholder[1]) {
1399 case 'd': /* reflog selector */
1400 case 'D':
1401 if (c->pretty_ctx->reflog_info)
1402 get_reflog_selector(sb,
1403 c->pretty_ctx->reflog_info,
1404 &c->pretty_ctx->date_mode,
1405 c->pretty_ctx->date_mode_explicit,
1406 (placeholder[1] == 'd'));
1407 return 2;
1408 case 's': /* reflog message */
1409 if (c->pretty_ctx->reflog_info)
1410 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1411 return 2;
1412 case 'n':
1413 case 'N':
1414 case 'e':
1415 case 'E':
1416 return format_reflog_person(sb,
1417 placeholder[1],
1418 c->pretty_ctx->reflog_info,
1419 &c->pretty_ctx->date_mode);
1420 }
1421 return 0; /* unknown %g placeholder */
1422 case 'N':
1423 if (c->pretty_ctx->notes_message) {
1424 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1425 return 1;
1426 }
1427 return 0;
1428 }
1429
1430 if (placeholder[0] == 'G') {
1431 if (!c->signature_check.result)
1432 check_commit_signature(c->commit, &(c->signature_check));
1433 switch (placeholder[1]) {
1434 case 'G':
1435 if (c->signature_check.gpg_output)
1436 strbuf_addstr(sb, c->signature_check.gpg_output);
1437 break;
1438 case '?':
1439 switch (c->signature_check.result) {
1440 case 'G':
1441 switch (c->signature_check.trust_level) {
1442 case TRUST_UNDEFINED:
1443 case TRUST_NEVER:
1444 strbuf_addch(sb, 'U');
1445 break;
1446 default:
1447 strbuf_addch(sb, 'G');
1448 break;
1449 }
1450 break;
1451 case 'B':
1452 case 'E':
1453 case 'N':
1454 case 'X':
1455 case 'Y':
1456 case 'R':
1457 strbuf_addch(sb, c->signature_check.result);
1458 }
1459 break;
1460 case 'S':
1461 if (c->signature_check.signer)
1462 strbuf_addstr(sb, c->signature_check.signer);
1463 break;
1464 case 'K':
1465 if (c->signature_check.key)
1466 strbuf_addstr(sb, c->signature_check.key);
1467 break;
1468 case 'F':
1469 if (c->signature_check.fingerprint)
1470 strbuf_addstr(sb, c->signature_check.fingerprint);
1471 break;
1472 case 'P':
1473 if (c->signature_check.primary_key_fingerprint)
1474 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1475 break;
1476 case 'T':
1477 switch (c->signature_check.trust_level) {
1478 case TRUST_UNDEFINED:
1479 strbuf_addstr(sb, "undefined");
1480 break;
1481 case TRUST_NEVER:
1482 strbuf_addstr(sb, "never");
1483 break;
1484 case TRUST_MARGINAL:
1485 strbuf_addstr(sb, "marginal");
1486 break;
1487 case TRUST_FULLY:
1488 strbuf_addstr(sb, "fully");
1489 break;
1490 case TRUST_ULTIMATE:
1491 strbuf_addstr(sb, "ultimate");
1492 break;
1493 }
1494 break;
1495 default:
1496 return 0;
1497 }
1498 return 2;
1499 }
1500
1501 /* For the rest we have to parse the commit header. */
1502 if (!c->commit_header_parsed) {
1503 msg = c->message =
1504 repo_logmsg_reencode(c->repository, commit,
1505 &c->commit_encoding, "UTF-8");
1506 parse_commit_header(c);
1507 }
1508
1509 switch (placeholder[0]) {
1510 case 'a': /* author ... */
1511 return format_person_part(sb, placeholder[1],
1512 msg + c->author.off, c->author.len,
1513 &c->pretty_ctx->date_mode);
1514 case 'c': /* committer ... */
1515 return format_person_part(sb, placeholder[1],
1516 msg + c->committer.off, c->committer.len,
1517 &c->pretty_ctx->date_mode);
1518 case 'e': /* encoding */
1519 if (c->commit_encoding)
1520 strbuf_addstr(sb, c->commit_encoding);
1521 return 1;
1522 case 'B': /* raw body */
1523 /* message_off is always left at the initial newline */
1524 strbuf_addstr(sb, msg + c->message_off + 1);
1525 return 1;
1526 }
1527
1528 /* Now we need to parse the commit message. */
1529 if (!c->commit_message_parsed)
1530 parse_commit_message(c);
1531
1532 switch (placeholder[0]) {
1533 case 's': /* subject */
1534 format_subject(sb, msg + c->subject_off, " ");
1535 return 1;
1536 case 'f': /* sanitized subject */
1537 eol = strchrnul(msg + c->subject_off, '\n');
1538 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1539 return 1;
1540 case 'b': /* body */
1541 strbuf_addstr(sb, msg + c->body_off);
1542 return 1;
1543 }
1544
1545 if (skip_prefix(placeholder, "(trailers", &arg)) {
1546 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1547 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1548 struct strbuf sepbuf = STRBUF_INIT;
1549 struct strbuf kvsepbuf = STRBUF_INIT;
1550 size_t ret = 0;
1551
1552 opts.no_divider = 1;
1553
1554 if (*arg == ':') {
1555 arg++;
1556 if (format_set_trailers_options(&opts, &filter_list, &sepbuf, &kvsepbuf, &arg, NULL))
1557 goto trailer_out;
1558 }
1559 if (*arg == ')') {
1560 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1561 ret = arg - placeholder + 1;
1562 }
1563 trailer_out:
1564 string_list_clear(&filter_list, 0);
1565 strbuf_release(&sepbuf);
1566 return ret;
1567 }
1568
1569 return 0; /* unknown placeholder */
1570 }
1571
1572 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1573 const char *placeholder,
1574 struct format_commit_context *c)
1575 {
1576 struct strbuf local_sb = STRBUF_INIT;
1577 int total_consumed = 0, len, padding = c->padding;
1578 if (padding < 0) {
1579 const char *start = strrchr(sb->buf, '\n');
1580 int occupied;
1581 if (!start)
1582 start = sb->buf;
1583 occupied = utf8_strnwidth(start, -1, 1);
1584 occupied += c->pretty_ctx->graph_width;
1585 padding = (-padding) - occupied;
1586 }
1587 while (1) {
1588 int modifier = *placeholder == 'C';
1589 int consumed = format_commit_one(&local_sb, placeholder, c);
1590 total_consumed += consumed;
1591
1592 if (!modifier)
1593 break;
1594
1595 placeholder += consumed;
1596 if (*placeholder != '%')
1597 break;
1598 placeholder++;
1599 total_consumed++;
1600 }
1601 len = utf8_strnwidth(local_sb.buf, -1, 1);
1602
1603 if (c->flush_type == flush_left_and_steal) {
1604 const char *ch = sb->buf + sb->len - 1;
1605 while (len > padding && ch > sb->buf) {
1606 const char *p;
1607 if (*ch == ' ') {
1608 ch--;
1609 padding++;
1610 continue;
1611 }
1612 /* check for trailing ansi sequences */
1613 if (*ch != 'm')
1614 break;
1615 p = ch - 1;
1616 while (ch - p < 10 && *p != '\033')
1617 p--;
1618 if (*p != '\033' ||
1619 ch + 1 - p != display_mode_esc_sequence_len(p))
1620 break;
1621 /*
1622 * got a good ansi sequence, put it back to
1623 * local_sb as we're cutting sb
1624 */
1625 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1626 ch = p - 1;
1627 }
1628 strbuf_setlen(sb, ch + 1 - sb->buf);
1629 c->flush_type = flush_left;
1630 }
1631
1632 if (len > padding) {
1633 switch (c->truncate) {
1634 case trunc_left:
1635 strbuf_utf8_replace(&local_sb,
1636 0, len - (padding - 2),
1637 "..");
1638 break;
1639 case trunc_middle:
1640 strbuf_utf8_replace(&local_sb,
1641 padding / 2 - 1,
1642 len - (padding - 2),
1643 "..");
1644 break;
1645 case trunc_right:
1646 strbuf_utf8_replace(&local_sb,
1647 padding - 2, len - (padding - 2),
1648 "..");
1649 break;
1650 case trunc_none:
1651 break;
1652 }
1653 strbuf_addbuf(sb, &local_sb);
1654 } else {
1655 int sb_len = sb->len, offset = 0;
1656 if (c->flush_type == flush_left)
1657 offset = padding - len;
1658 else if (c->flush_type == flush_both)
1659 offset = (padding - len) / 2;
1660 /*
1661 * we calculate padding in columns, now
1662 * convert it back to chars
1663 */
1664 padding = padding - len + local_sb.len;
1665 strbuf_addchars(sb, ' ', padding);
1666 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1667 local_sb.len);
1668 }
1669 strbuf_release(&local_sb);
1670 c->flush_type = no_flush;
1671 return total_consumed;
1672 }
1673
1674 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1675 const char *placeholder,
1676 void *context)
1677 {
1678 int consumed;
1679 size_t orig_len;
1680 enum {
1681 NO_MAGIC,
1682 ADD_LF_BEFORE_NON_EMPTY,
1683 DEL_LF_BEFORE_EMPTY,
1684 ADD_SP_BEFORE_NON_EMPTY
1685 } magic = NO_MAGIC;
1686
1687 switch (placeholder[0]) {
1688 case '-':
1689 magic = DEL_LF_BEFORE_EMPTY;
1690 break;
1691 case '+':
1692 magic = ADD_LF_BEFORE_NON_EMPTY;
1693 break;
1694 case ' ':
1695 magic = ADD_SP_BEFORE_NON_EMPTY;
1696 break;
1697 default:
1698 break;
1699 }
1700 if (magic != NO_MAGIC)
1701 placeholder++;
1702
1703 orig_len = sb->len;
1704 if (((struct format_commit_context *)context)->flush_type != no_flush)
1705 consumed = format_and_pad_commit(sb, placeholder, context);
1706 else
1707 consumed = format_commit_one(sb, placeholder, context);
1708 if (magic == NO_MAGIC)
1709 return consumed;
1710
1711 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1712 while (sb->len && sb->buf[sb->len - 1] == '\n')
1713 strbuf_setlen(sb, sb->len - 1);
1714 } else if (orig_len != sb->len) {
1715 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1716 strbuf_insertstr(sb, orig_len, "\n");
1717 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1718 strbuf_insertstr(sb, orig_len, " ");
1719 }
1720 return consumed + 1;
1721 }
1722
1723 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1724 void *context)
1725 {
1726 struct userformat_want *w = context;
1727
1728 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1729 placeholder++;
1730
1731 switch (*placeholder) {
1732 case 'N':
1733 w->notes = 1;
1734 break;
1735 case 'S':
1736 w->source = 1;
1737 break;
1738 }
1739 return 0;
1740 }
1741
1742 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1743 {
1744 struct strbuf dummy = STRBUF_INIT;
1745
1746 if (!fmt) {
1747 if (!user_format)
1748 return;
1749 fmt = user_format;
1750 }
1751 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1752 strbuf_release(&dummy);
1753 }
1754
1755 void repo_format_commit_message(struct repository *r,
1756 const struct commit *commit,
1757 const char *format, struct strbuf *sb,
1758 const struct pretty_print_context *pretty_ctx)
1759 {
1760 struct format_commit_context context = {
1761 .repository = r,
1762 .commit = commit,
1763 .pretty_ctx = pretty_ctx,
1764 .wrap_start = sb->len
1765 };
1766 const char *output_enc = pretty_ctx->output_encoding;
1767 const char *utf8 = "UTF-8";
1768
1769 strbuf_expand(sb, format, format_commit_item, &context);
1770 rewrap_message_tail(sb, &context, 0, 0, 0);
1771
1772 /*
1773 * Convert output to an actual output encoding; note that
1774 * format_commit_item() will always use UTF-8, so we don't
1775 * have to bother if that's what the output wants.
1776 */
1777 if (output_enc) {
1778 if (same_encoding(utf8, output_enc))
1779 output_enc = NULL;
1780 } else {
1781 if (context.commit_encoding &&
1782 !same_encoding(context.commit_encoding, utf8))
1783 output_enc = context.commit_encoding;
1784 }
1785
1786 if (output_enc) {
1787 size_t outsz;
1788 char *out = reencode_string_len(sb->buf, sb->len,
1789 output_enc, utf8, &outsz);
1790 if (out)
1791 strbuf_attach(sb, out, outsz, outsz + 1);
1792 }
1793
1794 free(context.commit_encoding);
1795 repo_unuse_commit_buffer(r, commit, context.message);
1796 }
1797
1798 static void pp_header(struct pretty_print_context *pp,
1799 const char *encoding,
1800 const struct commit *commit,
1801 const char **msg_p,
1802 struct strbuf *sb)
1803 {
1804 int parents_shown = 0;
1805
1806 for (;;) {
1807 const char *name, *line = *msg_p;
1808 int linelen = get_one_line(*msg_p);
1809
1810 if (!linelen)
1811 return;
1812 *msg_p += linelen;
1813
1814 if (linelen == 1)
1815 /* End of header */
1816 return;
1817
1818 if (pp->fmt == CMIT_FMT_RAW) {
1819 strbuf_add(sb, line, linelen);
1820 continue;
1821 }
1822
1823 if (starts_with(line, "parent ")) {
1824 if (linelen != the_hash_algo->hexsz + 8)
1825 die("bad parent line in commit");
1826 continue;
1827 }
1828
1829 if (!parents_shown) {
1830 unsigned num = commit_list_count(commit->parents);
1831 /* with enough slop */
1832 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1833 add_merge_info(pp, sb, commit);
1834 parents_shown = 1;
1835 }
1836
1837 /*
1838 * MEDIUM == DEFAULT shows only author with dates.
1839 * FULL shows both authors but not dates.
1840 * FULLER shows both authors and dates.
1841 */
1842 if (skip_prefix(line, "author ", &name)) {
1843 strbuf_grow(sb, linelen + 80);
1844 pp_user_info(pp, "Author", sb, name, encoding);
1845 }
1846 if (skip_prefix(line, "committer ", &name) &&
1847 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1848 strbuf_grow(sb, linelen + 80);
1849 pp_user_info(pp, "Commit", sb, name, encoding);
1850 }
1851 }
1852 }
1853
1854 void pp_title_line(struct pretty_print_context *pp,
1855 const char **msg_p,
1856 struct strbuf *sb,
1857 const char *encoding,
1858 int need_8bit_cte)
1859 {
1860 static const int max_length = 78; /* per rfc2047 */
1861 struct strbuf title;
1862
1863 strbuf_init(&title, 80);
1864 *msg_p = format_subject(&title, *msg_p,
1865 pp->preserve_subject ? "\n" : " ");
1866
1867 strbuf_grow(sb, title.len + 1024);
1868 if (pp->print_email_subject) {
1869 if (pp->rev)
1870 fmt_output_email_subject(sb, pp->rev);
1871 if (pp->encode_email_headers &&
1872 needs_rfc2047_encoding(title.buf, title.len))
1873 add_rfc2047(sb, title.buf, title.len,
1874 encoding, RFC2047_SUBJECT);
1875 else
1876 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1877 -last_line_length(sb), 1, max_length);
1878 } else {
1879 strbuf_addbuf(sb, &title);
1880 }
1881 strbuf_addch(sb, '\n');
1882
1883 if (need_8bit_cte == 0) {
1884 int i;
1885 for (i = 0; i < pp->in_body_headers.nr; i++) {
1886 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1887 need_8bit_cte = 1;
1888 break;
1889 }
1890 }
1891 }
1892
1893 if (need_8bit_cte > 0) {
1894 const char *header_fmt =
1895 "MIME-Version: 1.0\n"
1896 "Content-Type: text/plain; charset=%s\n"
1897 "Content-Transfer-Encoding: 8bit\n";
1898 strbuf_addf(sb, header_fmt, encoding);
1899 }
1900 if (pp->after_subject) {
1901 strbuf_addstr(sb, pp->after_subject);
1902 }
1903 if (cmit_fmt_is_mail(pp->fmt)) {
1904 strbuf_addch(sb, '\n');
1905 }
1906
1907 if (pp->in_body_headers.nr) {
1908 int i;
1909 for (i = 0; i < pp->in_body_headers.nr; i++) {
1910 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1911 free(pp->in_body_headers.items[i].string);
1912 }
1913 string_list_clear(&pp->in_body_headers, 0);
1914 strbuf_addch(sb, '\n');
1915 }
1916
1917 strbuf_release(&title);
1918 }
1919
1920 static int pp_utf8_width(const char *start, const char *end)
1921 {
1922 int width = 0;
1923 size_t remain = end - start;
1924
1925 while (remain) {
1926 int n = utf8_width(&start, &remain);
1927 if (n < 0 || !start)
1928 return -1;
1929 width += n;
1930 }
1931 return width;
1932 }
1933
1934 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1935 const char *line, int linelen)
1936 {
1937 const char *tab;
1938
1939 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1940 int width = pp_utf8_width(line, tab);
1941
1942 /*
1943 * If it wasn't well-formed utf8, or it
1944 * had characters with badly defined
1945 * width (control characters etc), just
1946 * give up on trying to align things.
1947 */
1948 if (width < 0)
1949 break;
1950
1951 /* Output the data .. */
1952 strbuf_add(sb, line, tab - line);
1953
1954 /* .. and the de-tabified tab */
1955 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1956
1957 /* Skip over the printed part .. */
1958 linelen -= tab + 1 - line;
1959 line = tab + 1;
1960 }
1961
1962 /*
1963 * Print out everything after the last tab without
1964 * worrying about width - there's nothing more to
1965 * align.
1966 */
1967 strbuf_add(sb, line, linelen);
1968 }
1969
1970 /*
1971 * pp_handle_indent() prints out the intendation, and
1972 * the whole line (without the final newline), after
1973 * de-tabifying.
1974 */
1975 static void pp_handle_indent(struct pretty_print_context *pp,
1976 struct strbuf *sb, int indent,
1977 const char *line, int linelen)
1978 {
1979 strbuf_addchars(sb, ' ', indent);
1980 if (pp->expand_tabs_in_log)
1981 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1982 else
1983 strbuf_add(sb, line, linelen);
1984 }
1985
1986 static int is_mboxrd_from(const char *line, int len)
1987 {
1988 /*
1989 * a line matching /^From $/ here would only have len == 4
1990 * at this point because is_empty_line would've trimmed all
1991 * trailing space
1992 */
1993 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1994 }
1995
1996 void pp_remainder(struct pretty_print_context *pp,
1997 const char **msg_p,
1998 struct strbuf *sb,
1999 int indent)
2000 {
2001 int first = 1;
2002 for (;;) {
2003 const char *line = *msg_p;
2004 int linelen = get_one_line(line);
2005 *msg_p += linelen;
2006
2007 if (!linelen)
2008 break;
2009
2010 if (is_blank_line(line, &linelen)) {
2011 if (first)
2012 continue;
2013 if (pp->fmt == CMIT_FMT_SHORT)
2014 break;
2015 }
2016 first = 0;
2017
2018 strbuf_grow(sb, linelen + indent + 20);
2019 if (indent)
2020 pp_handle_indent(pp, sb, indent, line, linelen);
2021 else if (pp->expand_tabs_in_log)
2022 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
2023 line, linelen);
2024 else {
2025 if (pp->fmt == CMIT_FMT_MBOXRD &&
2026 is_mboxrd_from(line, linelen))
2027 strbuf_addch(sb, '>');
2028
2029 strbuf_add(sb, line, linelen);
2030 }
2031 strbuf_addch(sb, '\n');
2032 }
2033 }
2034
2035 void pretty_print_commit(struct pretty_print_context *pp,
2036 const struct commit *commit,
2037 struct strbuf *sb)
2038 {
2039 unsigned long beginning_of_body;
2040 int indent = 4;
2041 const char *msg;
2042 const char *reencoded;
2043 const char *encoding;
2044 int need_8bit_cte = pp->need_8bit_cte;
2045
2046 if (pp->fmt == CMIT_FMT_USERFORMAT) {
2047 format_commit_message(commit, user_format, sb, pp);
2048 return;
2049 }
2050
2051 encoding = get_log_output_encoding();
2052 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
2053
2054 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2055 indent = 0;
2056
2057 /*
2058 * We need to check and emit Content-type: to mark it
2059 * as 8-bit if we haven't done so.
2060 */
2061 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
2062 int i, ch, in_body;
2063
2064 for (in_body = i = 0; (ch = msg[i]); i++) {
2065 if (!in_body) {
2066 /* author could be non 7-bit ASCII but
2067 * the log may be so; skip over the
2068 * header part first.
2069 */
2070 if (ch == '\n' && msg[i+1] == '\n')
2071 in_body = 1;
2072 }
2073 else if (non_ascii(ch)) {
2074 need_8bit_cte = 1;
2075 break;
2076 }
2077 }
2078 }
2079
2080 pp_header(pp, encoding, commit, &msg, sb);
2081 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
2082 strbuf_addch(sb, '\n');
2083 }
2084
2085 /* Skip excess blank lines at the beginning of body, if any... */
2086 msg = skip_blank_lines(msg);
2087
2088 /* These formats treat the title line specially. */
2089 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2090 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2091
2092 beginning_of_body = sb->len;
2093 if (pp->fmt != CMIT_FMT_ONELINE)
2094 pp_remainder(pp, &msg, sb, indent);
2095 strbuf_rtrim(sb);
2096
2097 /* Make sure there is an EOLN for the non-oneline case */
2098 if (pp->fmt != CMIT_FMT_ONELINE)
2099 strbuf_addch(sb, '\n');
2100
2101 /*
2102 * The caller may append additional body text in e-mail
2103 * format. Make sure we did not strip the blank line
2104 * between the header and the body.
2105 */
2106 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2107 strbuf_addch(sb, '\n');
2108
2109 unuse_commit_buffer(commit, reencoded);
2110 }
2111
2112 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2113 struct strbuf *sb)
2114 {
2115 struct pretty_print_context pp = {0};
2116 pp.fmt = fmt;
2117 pretty_print_commit(&pp, commit, sb);
2118 }
The Best VCS Around