?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
tests: avoid using `test_i18ncmp`
[git.git] / pretty.c
blobae0f696d8ec2f4ea3d75df51d2fab8e7b3d83849
1 #include "cache.h"
2 #include "config.h"
3 #include "commit.h"
4 #include "utf8.h"
5 #include "diff.h"
6 #include "revision.h"
7 #include "string-list.h"
8 #include "mailmap.h"
9 #include "log-tree.h"
10 #include "notes.h"
11 #include "color.h"
12 #include "reflog-walk.h"
13 #include "gpg-interface.h"
14 #include "trailer.h"
15
16 /*
17 * The limit for formatting directives, which enable the caller to append
18 * arbitrarily many bytes to the formatted buffer. This includes padding
19 * and wrapping formatters.
20 */
21 #define FORMATTING_LIMIT (16 * 1024)
22
23 static char *user_format;
24 static struct cmt_fmt_map {
25 const char *name;
26 enum cmit_fmt format;
27 int is_tformat;
28 int expand_tabs_in_log;
29 int is_alias;
30 enum date_mode_type default_date_mode_type;
31 const char *user_format;
32 } *commit_formats;
33 static size_t builtin_formats_len;
34 static size_t commit_formats_len;
35 static size_t commit_formats_alloc;
36 static struct cmt_fmt_map *find_commit_format(const char *sought);
37
38 int commit_format_is_empty(enum cmit_fmt fmt)
39 {
40 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
41 }
42
43 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
44 {
45 free(user_format);
46 user_format = xstrdup(cp);
47 if (is_tformat)
48 rev->use_terminator = 1;
49 rev->commit_format = CMIT_FMT_USERFORMAT;
50 }
51
52 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
53 {
54 struct cmt_fmt_map *commit_format = NULL;
55 const char *name;
56 const char *fmt;
57 int i;
58
59 if (!skip_prefix(var, "pretty.", &name))
60 return 0;
61
62 for (i = 0; i < builtin_formats_len; i++) {
63 if (!strcmp(commit_formats[i].name, name))
64 return 0;
65 }
66
67 for (i = builtin_formats_len; i < commit_formats_len; i++) {
68 if (!strcmp(commit_formats[i].name, name)) {
69 commit_format = &commit_formats[i];
70 break;
71 }
72 }
73
74 if (!commit_format) {
75 ALLOC_GROW(commit_formats, commit_formats_len+1,
76 commit_formats_alloc);
77 commit_format = &commit_formats[commit_formats_len];
78 memset(commit_format, 0, sizeof(*commit_format));
79 commit_formats_len++;
80 }
81
82 commit_format->name = xstrdup(name);
83 commit_format->format = CMIT_FMT_USERFORMAT;
84 if (git_config_string(&fmt, var, value))
85 return -1;
86
87 if (skip_prefix(fmt, "format:", &fmt))
88 commit_format->is_tformat = 0;
89 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
90 commit_format->is_tformat = 1;
91 else
92 commit_format->is_alias = 1;
93 commit_format->user_format = fmt;
94
95 return 0;
96 }
97
98 static void setup_commit_formats(void)
99 {
100 struct cmt_fmt_map builtin_formats[] = {
101 { "raw", CMIT_FMT_RAW, 0, 0 },
102 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
103 { "short", CMIT_FMT_SHORT, 0, 0 },
104 { "email", CMIT_FMT_EMAIL, 0, 0 },
105 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
106 { "fuller", CMIT_FMT_FULLER, 0, 8 },
107 { "full", CMIT_FMT_FULL, 0, 8 },
108 { "oneline", CMIT_FMT_ONELINE, 1, 0 },
109 { "reference", CMIT_FMT_USERFORMAT, 1, 0,
110 0, DATE_SHORT, "%C(auto)%h (%s, %ad)" },
111 /*
112 * Please update $__git_log_pretty_formats in
113 * git-completion.bash when you add new formats.
114 */
115 };
116 commit_formats_len = ARRAY_SIZE(builtin_formats);
117 builtin_formats_len = commit_formats_len;
118 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
119 COPY_ARRAY(commit_formats, builtin_formats,
120 ARRAY_SIZE(builtin_formats));
121
122 git_config(git_pretty_formats_config, NULL);
123 }
124
125 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
126 const char *original,
127 int num_redirections)
128 {
129 struct cmt_fmt_map *found = NULL;
130 size_t found_match_len = 0;
131 int i;
132
133 if (num_redirections >= commit_formats_len)
134 die("invalid --pretty format: "
135 "'%s' references an alias which points to itself",
136 original);
137
138 for (i = 0; i < commit_formats_len; i++) {
139 size_t match_len;
140
141 if (!starts_with(commit_formats[i].name, sought))
142 continue;
143
144 match_len = strlen(commit_formats[i].name);
145 if (found == NULL || found_match_len > match_len) {
146 found = &commit_formats[i];
147 found_match_len = match_len;
148 }
149 }
150
151 if (found && found->is_alias) {
152 found = find_commit_format_recursive(found->user_format,
153 original,
154 num_redirections+1);
155 }
156
157 return found;
158 }
159
160 static struct cmt_fmt_map *find_commit_format(const char *sought)
161 {
162 if (!commit_formats)
163 setup_commit_formats();
164
165 return find_commit_format_recursive(sought, sought, 0);
166 }
167
168 void get_commit_format(const char *arg, struct rev_info *rev)
169 {
170 struct cmt_fmt_map *commit_format;
171
172 rev->use_terminator = 0;
173 if (!arg) {
174 rev->commit_format = CMIT_FMT_DEFAULT;
175 return;
176 }
177 if (skip_prefix(arg, "format:", &arg)) {
178 save_user_format(rev, arg, 0);
179 return;
180 }
181
182 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
183 save_user_format(rev, arg, 1);
184 return;
185 }
186
187 commit_format = find_commit_format(arg);
188 if (!commit_format)
189 die("invalid --pretty format: %s", arg);
190
191 rev->commit_format = commit_format->format;
192 rev->use_terminator = commit_format->is_tformat;
193 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
194 if (!rev->date_mode_explicit && commit_format->default_date_mode_type)
195 rev->date_mode.type = commit_format->default_date_mode_type;
196 if (commit_format->format == CMIT_FMT_USERFORMAT) {
197 save_user_format(rev, commit_format->user_format,
198 commit_format->is_tformat);
199 }
200 }
201
202 /*
203 * Generic support for pretty-printing the header
204 */
205 static int get_one_line(const char *msg)
206 {
207 int ret = 0;
208
209 for (;;) {
210 char c = *msg++;
211 if (!c)
212 break;
213 ret++;
214 if (c == '\n')
215 break;
216 }
217 return ret;
218 }
219
220 /* High bit set, or ISO-2022-INT */
221 static int non_ascii(int ch)
222 {
223 return !isascii(ch) || ch == '\033';
224 }
225
226 int has_non_ascii(const char *s)
227 {
228 int ch;
229 if (!s)
230 return 0;
231 while ((ch = *s++) != '\0') {
232 if (non_ascii(ch))
233 return 1;
234 }
235 return 0;
236 }
237
238 static int is_rfc822_special(char ch)
239 {
240 switch (ch) {
241 case '(':
242 case ')':
243 case '<':
244 case '>':
245 case '[':
246 case ']':
247 case ':':
248 case ';':
249 case '@':
250 case ',':
251 case '.':
252 case '"':
253 case '\\':
254 return 1;
255 default:
256 return 0;
257 }
258 }
259
260 static int needs_rfc822_quoting(const char *s, int len)
261 {
262 int i;
263 for (i = 0; i < len; i++)
264 if (is_rfc822_special(s[i]))
265 return 1;
266 return 0;
267 }
268
269 static int last_line_length(struct strbuf *sb)
270 {
271 int i;
272
273 /* How many bytes are already used on the last line? */
274 for (i = sb->len - 1; i >= 0; i--)
275 if (sb->buf[i] == '\n')
276 break;
277 return sb->len - (i + 1);
278 }
279
280 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
281 {
282 int i;
283
284 /* just a guess, we may have to also backslash-quote */
285 strbuf_grow(out, len + 2);
286
287 strbuf_addch(out, '"');
288 for (i = 0; i < len; i++) {
289 switch (s[i]) {
290 case '"':
291 case '\\':
292 strbuf_addch(out, '\\');
293 /* fall through */
294 default:
295 strbuf_addch(out, s[i]);
296 }
297 }
298 strbuf_addch(out, '"');
299 }
300
301 enum rfc2047_type {
302 RFC2047_SUBJECT,
303 RFC2047_ADDRESS
304 };
305
306 static int is_rfc2047_special(char ch, enum rfc2047_type type)
307 {
308 /*
309 * rfc2047, section 4.2:
310 *
311 * 8-bit values which correspond to printable ASCII characters other
312 * than "=", "?", and "_" (underscore), MAY be represented as those
313 * characters. (But see section 5 for restrictions.) In
314 * particular, SPACE and TAB MUST NOT be represented as themselves
315 * within encoded words.
316 */
317
318 /*
319 * rule out non-ASCII characters and non-printable characters (the
320 * non-ASCII check should be redundant as isprint() is not localized
321 * and only knows about ASCII, but be defensive about that)
322 */
323 if (non_ascii(ch) || !isprint(ch))
324 return 1;
325
326 /*
327 * rule out special printable characters (' ' should be the only
328 * whitespace character considered printable, but be defensive and use
329 * isspace())
330 */
331 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
332 return 1;
333
334 /*
335 * rfc2047, section 5.3:
336 *
337 * As a replacement for a 'word' entity within a 'phrase', for example,
338 * one that precedes an address in a From, To, or Cc header. The ABNF
339 * definition for 'phrase' from RFC 822 thus becomes:
340 *
341 * phrase = 1*( encoded-word / word )
342 *
343 * In this case the set of characters that may be used in a "Q"-encoded
344 * 'encoded-word' is restricted to: <upper and lower case ASCII
345 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
346 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
347 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
348 * 'special' by 'linear-white-space'.
349 */
350
351 if (type != RFC2047_ADDRESS)
352 return 0;
353
354 /* '=' and '_' are special cases and have been checked above */
355 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
356 }
357
358 static int needs_rfc2047_encoding(const char *line, int len)
359 {
360 int i;
361
362 for (i = 0; i < len; i++) {
363 int ch = line[i];
364 if (non_ascii(ch) || ch == '\n')
365 return 1;
366 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
367 return 1;
368 }
369
370 return 0;
371 }
372
373 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
374 const char *encoding, enum rfc2047_type type)
375 {
376 static const int max_encoded_length = 76; /* per rfc2047 */
377 int i;
378 int line_len = last_line_length(sb);
379
380 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
381 strbuf_addf(sb, "=?%s?q?", encoding);
382 line_len += strlen(encoding) + 5; /* 5 for =??q? */
383
384 while (len) {
385 /*
386 * RFC 2047, section 5 (3):
387 *
388 * Each 'encoded-word' MUST represent an integral number of
389 * characters. A multi-octet character may not be split across
390 * adjacent 'encoded- word's.
391 */
392 const unsigned char *p = (const unsigned char *)line;
393 int chrlen = mbs_chrlen(&line, &len, encoding);
394 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
395
396 /* "=%02X" * chrlen, or the byte itself */
397 const char *encoded_fmt = is_special ? "=%02X" : "%c";
398 int encoded_len = is_special ? 3 * chrlen : 1;
399
400 /*
401 * According to RFC 2047, we could encode the special character
402 * ' ' (space) with '_' (underscore) for readability. But many
403 * programs do not understand this and just leave the
404 * underscore in place. Thus, we do nothing special here, which
405 * causes ' ' to be encoded as '=20', avoiding this problem.
406 */
407
408 if (line_len + encoded_len + 2 > max_encoded_length) {
409 /* It won't fit with trailing "?=" --- break the line */
410 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
411 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
412 }
413
414 for (i = 0; i < chrlen; i++)
415 strbuf_addf(sb, encoded_fmt, p[i]);
416 line_len += encoded_len;
417 }
418 strbuf_addstr(sb, "?=");
419 }
420
421 const char *show_ident_date(const struct ident_split *ident,
422 const struct date_mode *mode)
423 {
424 timestamp_t date = 0;
425 long tz = 0;
426
427 if (ident->date_begin && ident->date_end)
428 date = parse_timestamp(ident->date_begin, NULL, 10);
429 if (date_overflows(date))
430 date = 0;
431 else {
432 if (ident->tz_begin && ident->tz_end)
433 tz = strtol(ident->tz_begin, NULL, 10);
434 if (tz >= INT_MAX || tz <= INT_MIN)
435 tz = 0;
436 }
437 return show_date(date, tz, mode);
438 }
439
440 void pp_user_info(struct pretty_print_context *pp,
441 const char *what, struct strbuf *sb,
442 const char *line, const char *encoding)
443 {
444 struct ident_split ident;
445 char *line_end;
446 const char *mailbuf, *namebuf;
447 size_t namelen, maillen;
448 int max_length = 78; /* per rfc2822 */
449
450 if (pp->fmt == CMIT_FMT_ONELINE)
451 return;
452
453 line_end = strchrnul(line, '\n');
454 if (split_ident_line(&ident, line, line_end - line))
455 return;
456
457 mailbuf = ident.mail_begin;
458 maillen = ident.mail_end - ident.mail_begin;
459 namebuf = ident.name_begin;
460 namelen = ident.name_end - ident.name_begin;
461
462 if (pp->mailmap)
463 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
464
465 if (cmit_fmt_is_mail(pp->fmt)) {
466 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
467 struct strbuf buf = STRBUF_INIT;
468
469 strbuf_addstr(&buf, "From: ");
470 strbuf_add(&buf, namebuf, namelen);
471 strbuf_addstr(&buf, " <");
472 strbuf_add(&buf, mailbuf, maillen);
473 strbuf_addstr(&buf, ">\n");
474 string_list_append(&pp->in_body_headers,
475 strbuf_detach(&buf, NULL));
476
477 mailbuf = pp->from_ident->mail_begin;
478 maillen = pp->from_ident->mail_end - mailbuf;
479 namebuf = pp->from_ident->name_begin;
480 namelen = pp->from_ident->name_end - namebuf;
481 }
482
483 strbuf_addstr(sb, "From: ");
484 if (pp->encode_email_headers &&
485 needs_rfc2047_encoding(namebuf, namelen)) {
486 add_rfc2047(sb, namebuf, namelen,
487 encoding, RFC2047_ADDRESS);
488 max_length = 76; /* per rfc2047 */
489 } else if (needs_rfc822_quoting(namebuf, namelen)) {
490 struct strbuf quoted = STRBUF_INIT;
491 add_rfc822_quoted(&quoted, namebuf, namelen);
492 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
493 -6, 1, max_length);
494 strbuf_release(&quoted);
495 } else {
496 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
497 -6, 1, max_length);
498 }
499
500 if (max_length <
501 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
502 strbuf_addch(sb, '\n');
503 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
504 } else {
505 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
506 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
507 (int)namelen, namebuf, (int)maillen, mailbuf);
508 }
509
510 switch (pp->fmt) {
511 case CMIT_FMT_MEDIUM:
512 strbuf_addf(sb, "Date: %s\n",
513 show_ident_date(&ident, &pp->date_mode));
514 break;
515 case CMIT_FMT_EMAIL:
516 case CMIT_FMT_MBOXRD:
517 strbuf_addf(sb, "Date: %s\n",
518 show_ident_date(&ident, DATE_MODE(RFC2822)));
519 break;
520 case CMIT_FMT_FULLER:
521 strbuf_addf(sb, "%sDate: %s\n", what,
522 show_ident_date(&ident, &pp->date_mode));
523 break;
524 default:
525 /* notin' */
526 break;
527 }
528 }
529
530 static int is_blank_line(const char *line, int *len_p)
531 {
532 int len = *len_p;
533 while (len && isspace(line[len - 1]))
534 len--;
535 *len_p = len;
536 return !len;
537 }
538
539 const char *skip_blank_lines(const char *msg)
540 {
541 for (;;) {
542 int linelen = get_one_line(msg);
543 int ll = linelen;
544 if (!linelen)
545 break;
546 if (!is_blank_line(msg, &ll))
547 break;
548 msg += linelen;
549 }
550 return msg;
551 }
552
553 static void add_merge_info(const struct pretty_print_context *pp,
554 struct strbuf *sb, const struct commit *commit)
555 {
556 struct commit_list *parent = commit->parents;
557
558 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
559 !parent || !parent->next)
560 return;
561
562 strbuf_addstr(sb, "Merge:");
563
564 while (parent) {
565 struct object_id *oidp = &parent->item->object.oid;
566 strbuf_addch(sb, ' ');
567 if (pp->abbrev)
568 strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
569 else
570 strbuf_addstr(sb, oid_to_hex(oidp));
571 parent = parent->next;
572 }
573 strbuf_addch(sb, '\n');
574 }
575
576 static char *get_header(const char *msg, const char *key)
577 {
578 size_t len;
579 const char *v = find_commit_header(msg, key, &len);
580 return v ? xmemdupz(v, len) : NULL;
581 }
582
583 static char *replace_encoding_header(char *buf, const char *encoding)
584 {
585 struct strbuf tmp = STRBUF_INIT;
586 size_t start, len;
587 char *cp = buf;
588
589 /* guess if there is an encoding header before a \n\n */
590 while (!starts_with(cp, "encoding ")) {
591 cp = strchr(cp, '\n');
592 if (!cp || *++cp == '\n')
593 return buf;
594 }
595 start = cp - buf;
596 cp = strchr(cp, '\n');
597 if (!cp)
598 return buf; /* should not happen but be defensive */
599 len = cp + 1 - (buf + start);
600
601 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
602 if (is_encoding_utf8(encoding)) {
603 /* we have re-coded to UTF-8; drop the header */
604 strbuf_remove(&tmp, start, len);
605 } else {
606 /* just replaces XXXX in 'encoding XXXX\n' */
607 strbuf_splice(&tmp, start + strlen("encoding "),
608 len - strlen("encoding \n"),
609 encoding, strlen(encoding));
610 }
611 return strbuf_detach(&tmp, NULL);
612 }
613
614 const char *repo_logmsg_reencode(struct repository *r,
615 const struct commit *commit,
616 char **commit_encoding,
617 const char *output_encoding)
618 {
619 static const char *utf8 = "UTF-8";
620 const char *use_encoding;
621 char *encoding;
622 const char *msg = repo_get_commit_buffer(r, commit, NULL);
623 char *out;
624
625 if (!output_encoding || !*output_encoding) {
626 if (commit_encoding)
627 *commit_encoding = get_header(msg, "encoding");
628 return msg;
629 }
630 encoding = get_header(msg, "encoding");
631 if (commit_encoding)
632 *commit_encoding = encoding;
633 use_encoding = encoding ? encoding : utf8;
634 if (same_encoding(use_encoding, output_encoding)) {
635 /*
636 * No encoding work to be done. If we have no encoding header
637 * at all, then there's nothing to do, and we can return the
638 * message verbatim (whether newly allocated or not).
639 */
640 if (!encoding)
641 return msg;
642
643 /*
644 * Otherwise, we still want to munge the encoding header in the
645 * result, which will be done by modifying the buffer. If we
646 * are using a fresh copy, we can reuse it. But if we are using
647 * the cached copy from get_commit_buffer, we need to duplicate it
648 * to avoid munging the cached copy.
649 */
650 if (msg == get_cached_commit_buffer(r, commit, NULL))
651 out = xstrdup(msg);
652 else
653 out = (char *)msg;
654 }
655 else {
656 /*
657 * There's actual encoding work to do. Do the reencoding, which
658 * still leaves the header to be replaced in the next step. At
659 * this point, we are done with msg. If we allocated a fresh
660 * copy, we can free it.
661 */
662 out = reencode_string(msg, output_encoding, use_encoding);
663 if (out)
664 repo_unuse_commit_buffer(r, commit, msg);
665 }
666
667 /*
668 * This replacement actually consumes the buffer we hand it, so we do
669 * not have to worry about freeing the old "out" here.
670 */
671 if (out)
672 out = replace_encoding_header(out, output_encoding);
673
674 if (!commit_encoding)
675 free(encoding);
676 /*
677 * If the re-encoding failed, out might be NULL here; in that
678 * case we just return the commit message verbatim.
679 */
680 return out ? out : msg;
681 }
682
683 static int mailmap_name(const char **email, size_t *email_len,
684 const char **name, size_t *name_len)
685 {
686 static struct string_list *mail_map;
687 if (!mail_map) {
688 CALLOC_ARRAY(mail_map, 1);
689 read_mailmap(mail_map);
690 }
691 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
692 }
693
694 static size_t format_person_part(struct strbuf *sb, char part,
695 const char *msg, int len,
696 const struct date_mode *dmode)
697 {
698 /* currently all placeholders have same length */
699 const int placeholder_len = 2;
700 struct ident_split s;
701 const char *name, *mail;
702 size_t maillen, namelen;
703
704 if (split_ident_line(&s, msg, len) < 0)
705 goto skip;
706
707 name = s.name_begin;
708 namelen = s.name_end - s.name_begin;
709 mail = s.mail_begin;
710 maillen = s.mail_end - s.mail_begin;
711
712 if (part == 'N' || part == 'E' || part == 'L') /* mailmap lookup */
713 mailmap_name(&mail, &maillen, &name, &namelen);
714 if (part == 'n' || part == 'N') { /* name */
715 strbuf_add(sb, name, namelen);
716 return placeholder_len;
717 }
718 if (part == 'e' || part == 'E') { /* email */
719 strbuf_add(sb, mail, maillen);
720 return placeholder_len;
721 }
722 if (part == 'l' || part == 'L') { /* local-part */
723 const char *at = memchr(mail, '@', maillen);
724 if (at)
725 maillen = at - mail;
726 strbuf_add(sb, mail, maillen);
727 return placeholder_len;
728 }
729
730 if (!s.date_begin)
731 goto skip;
732
733 if (part == 't') { /* date, UNIX timestamp */
734 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
735 return placeholder_len;
736 }
737
738 switch (part) {
739 case 'd': /* date */
740 strbuf_addstr(sb, show_ident_date(&s, dmode));
741 return placeholder_len;
742 case 'D': /* date, RFC2822 style */
743 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
744 return placeholder_len;
745 case 'r': /* date, relative */
746 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
747 return placeholder_len;
748 case 'i': /* date, ISO 8601-like */
749 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
750 return placeholder_len;
751 case 'I': /* date, ISO 8601 strict */
752 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
753 return placeholder_len;
754 case 's':
755 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
756 return placeholder_len;
757 }
758
759 skip:
760 /*
761 * reading from either a bogus commit, or a reflog entry with
762 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
763 * to compute a valid return value.
764 */
765 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
766 || part == 'D' || part == 'r' || part == 'i')
767 return placeholder_len;
768
769 return 0; /* unknown placeholder */
770 }
771
772 struct chunk {
773 size_t off;
774 size_t len;
775 };
776
777 enum flush_type {
778 no_flush,
779 flush_right,
780 flush_left,
781 flush_left_and_steal,
782 flush_both
783 };
784
785 enum trunc_type {
786 trunc_none,
787 trunc_left,
788 trunc_middle,
789 trunc_right
790 };
791
792 struct format_commit_context {
793 struct repository *repository;
794 const struct commit *commit;
795 const struct pretty_print_context *pretty_ctx;
796 unsigned commit_header_parsed:1;
797 unsigned commit_message_parsed:1;
798 struct signature_check signature_check;
799 enum flush_type flush_type;
800 enum trunc_type truncate;
801 const char *message;
802 char *commit_encoding;
803 size_t width, indent1, indent2;
804 int auto_color;
805 int padding;
806
807 /* These offsets are relative to the start of the commit message. */
808 struct chunk author;
809 struct chunk committer;
810 size_t message_off;
811 size_t subject_off;
812 size_t body_off;
813
814 /* The following ones are relative to the result struct strbuf. */
815 size_t wrap_start;
816 };
817
818 static void parse_commit_header(struct format_commit_context *context)
819 {
820 const char *msg = context->message;
821 int i;
822
823 for (i = 0; msg[i]; i++) {
824 const char *name;
825 int eol;
826 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
827 ; /* do nothing */
828
829 if (i == eol) {
830 break;
831 } else if (skip_prefix(msg + i, "author ", &name)) {
832 context->author.off = name - msg;
833 context->author.len = msg + eol - name;
834 } else if (skip_prefix(msg + i, "committer ", &name)) {
835 context->committer.off = name - msg;
836 context->committer.len = msg + eol - name;
837 }
838 i = eol;
839 }
840 context->message_off = i;
841 context->commit_header_parsed = 1;
842 }
843
844 static int istitlechar(char c)
845 {
846 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
847 (c >= '0' && c <= '9') || c == '.' || c == '_';
848 }
849
850 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
851 {
852 size_t trimlen;
853 size_t start_len = sb->len;
854 int space = 2;
855 int i;
856
857 for (i = 0; i < len; i++) {
858 if (istitlechar(msg[i])) {
859 if (space == 1)
860 strbuf_addch(sb, '-');
861 space = 0;
862 strbuf_addch(sb, msg[i]);
863 if (msg[i] == '.')
864 while (msg[i+1] == '.')
865 i++;
866 } else
867 space |= 1;
868 }
869
870 /* trim any trailing '.' or '-' characters */
871 trimlen = 0;
872 while (sb->len - trimlen > start_len &&
873 (sb->buf[sb->len - 1 - trimlen] == '.'
874 || sb->buf[sb->len - 1 - trimlen] == '-'))
875 trimlen++;
876 strbuf_remove(sb, sb->len - trimlen, trimlen);
877 }
878
879 const char *format_subject(struct strbuf *sb, const char *msg,
880 const char *line_separator)
881 {
882 int first = 1;
883
884 for (;;) {
885 const char *line = msg;
886 int linelen = get_one_line(line);
887
888 msg += linelen;
889 if (!linelen || is_blank_line(line, &linelen))
890 break;
891
892 if (!sb)
893 continue;
894 strbuf_grow(sb, linelen + 2);
895 if (!first)
896 strbuf_addstr(sb, line_separator);
897 strbuf_add(sb, line, linelen);
898 first = 0;
899 }
900 return msg;
901 }
902
903 static void parse_commit_message(struct format_commit_context *c)
904 {
905 const char *msg = c->message + c->message_off;
906 const char *start = c->message;
907
908 msg = skip_blank_lines(msg);
909 c->subject_off = msg - start;
910
911 msg = format_subject(NULL, msg, NULL);
912 msg = skip_blank_lines(msg);
913 c->body_off = msg - start;
914
915 c->commit_message_parsed = 1;
916 }
917
918 static void strbuf_wrap(struct strbuf *sb, size_t pos,
919 size_t width, size_t indent1, size_t indent2)
920 {
921 struct strbuf tmp = STRBUF_INIT;
922
923 if (pos)
924 strbuf_add(&tmp, sb->buf, pos);
925 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
926 cast_size_t_to_int(indent1),
927 cast_size_t_to_int(indent2),
928 cast_size_t_to_int(width));
929 strbuf_swap(&tmp, sb);
930 strbuf_release(&tmp);
931 }
932
933 static void rewrap_message_tail(struct strbuf *sb,
934 struct format_commit_context *c,
935 size_t new_width, size_t new_indent1,
936 size_t new_indent2)
937 {
938 if (c->width == new_width && c->indent1 == new_indent1 &&
939 c->indent2 == new_indent2)
940 return;
941 if (c->wrap_start < sb->len)
942 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
943 c->wrap_start = sb->len;
944 c->width = new_width;
945 c->indent1 = new_indent1;
946 c->indent2 = new_indent2;
947 }
948
949 static int format_reflog_person(struct strbuf *sb,
950 char part,
951 struct reflog_walk_info *log,
952 const struct date_mode *dmode)
953 {
954 const char *ident;
955
956 if (!log)
957 return 2;
958
959 ident = get_reflog_ident(log);
960 if (!ident)
961 return 2;
962
963 return format_person_part(sb, part, ident, strlen(ident), dmode);
964 }
965
966 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
967 const char *placeholder,
968 struct format_commit_context *c)
969 {
970 const char *rest = placeholder;
971 const char *basic_color = NULL;
972
973 if (placeholder[1] == '(') {
974 const char *begin = placeholder + 2;
975 const char *end = strchr(begin, ')');
976 char color[COLOR_MAXLEN];
977
978 if (!end)
979 return 0;
980
981 if (skip_prefix(begin, "auto,", &begin)) {
982 if (!want_color(c->pretty_ctx->color))
983 return end - placeholder + 1;
984 } else if (skip_prefix(begin, "always,", &begin)) {
985 /* nothing to do; we do not respect want_color at all */
986 } else {
987 /* the default is the same as "auto" */
988 if (!want_color(c->pretty_ctx->color))
989 return end - placeholder + 1;
990 }
991
992 if (color_parse_mem(begin, end - begin, color) < 0)
993 die(_("unable to parse --pretty format"));
994 strbuf_addstr(sb, color);
995 return end - placeholder + 1;
996 }
997
998 /*
999 * We handle things like "%C(red)" above; for historical reasons, there
1000 * are a few colors that can be specified without parentheses (and
1001 * they cannot support things like "auto" or "always" at all).
1002 */
1003 if (skip_prefix(placeholder + 1, "red", &rest))
1004 basic_color = GIT_COLOR_RED;
1005 else if (skip_prefix(placeholder + 1, "green", &rest))
1006 basic_color = GIT_COLOR_GREEN;
1007 else if (skip_prefix(placeholder + 1, "blue", &rest))
1008 basic_color = GIT_COLOR_BLUE;
1009 else if (skip_prefix(placeholder + 1, "reset", &rest))
1010 basic_color = GIT_COLOR_RESET;
1011
1012 if (basic_color && want_color(c->pretty_ctx->color))
1013 strbuf_addstr(sb, basic_color);
1014
1015 return rest - placeholder;
1016 }
1017
1018 static size_t parse_padding_placeholder(const char *placeholder,
1019 struct format_commit_context *c)
1020 {
1021 const char *ch = placeholder;
1022 enum flush_type flush_type;
1023 int to_column = 0;
1024
1025 switch (*ch++) {
1026 case '<':
1027 flush_type = flush_right;
1028 break;
1029 case '>':
1030 if (*ch == '<') {
1031 flush_type = flush_both;
1032 ch++;
1033 } else if (*ch == '>') {
1034 flush_type = flush_left_and_steal;
1035 ch++;
1036 } else
1037 flush_type = flush_left;
1038 break;
1039 default:
1040 return 0;
1041 }
1042
1043 /* the next value means "wide enough to that column" */
1044 if (*ch == '|') {
1045 to_column = 1;
1046 ch++;
1047 }
1048
1049 if (*ch == '(') {
1050 const char *start = ch + 1;
1051 const char *end = start + strcspn(start, ",)");
1052 char *next;
1053 int width;
1054 if (!*end || end == start)
1055 return 0;
1056 width = strtol(start, &next, 10);
1057
1058 /*
1059 * We need to limit the amount of padding, or otherwise this
1060 * would allow the user to pad the buffer by arbitrarily many
1061 * bytes and thus cause resource exhaustion.
1062 */
1063 if (width < -FORMATTING_LIMIT || width > FORMATTING_LIMIT)
1064 return 0;
1065
1066 if (next == start || width == 0)
1067 return 0;
1068 if (width < 0) {
1069 if (to_column)
1070 width += term_columns();
1071 if (width < 0)
1072 return 0;
1073 }
1074 c->padding = to_column ? -width : width;
1075 c->flush_type = flush_type;
1076
1077 if (*end == ',') {
1078 start = end + 1;
1079 end = strchr(start, ')');
1080 if (!end || end == start)
1081 return 0;
1082 if (starts_with(start, "trunc)"))
1083 c->truncate = trunc_right;
1084 else if (starts_with(start, "ltrunc)"))
1085 c->truncate = trunc_left;
1086 else if (starts_with(start, "mtrunc)"))
1087 c->truncate = trunc_middle;
1088 else
1089 return 0;
1090 } else
1091 c->truncate = trunc_none;
1092
1093 return end - placeholder + 1;
1094 }
1095 return 0;
1096 }
1097
1098 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1099 const char **end, const char **valuestart,
1100 size_t *valuelen)
1101 {
1102 const char *p;
1103
1104 if (!(skip_prefix(to_parse, candidate, &p)))
1105 return 0;
1106 if (valuestart) {
1107 if (*p == '=') {
1108 *valuestart = p + 1;
1109 *valuelen = strcspn(*valuestart, ",)");
1110 p = *valuestart + *valuelen;
1111 } else {
1112 if (*p != ',' && *p != ')')
1113 return 0;
1114 *valuestart = NULL;
1115 *valuelen = 0;
1116 }
1117 }
1118 if (*p == ',') {
1119 *end = p + 1;
1120 return 1;
1121 }
1122 if (*p == ')') {
1123 *end = p;
1124 return 1;
1125 }
1126 return 0;
1127 }
1128
1129 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1130 const char **end, int *val)
1131 {
1132 const char *argval;
1133 char *strval;
1134 size_t arglen;
1135 int v;
1136
1137 if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1138 return 0;
1139
1140 if (!argval) {
1141 *val = 1;
1142 return 1;
1143 }
1144
1145 strval = xstrndup(argval, arglen);
1146 v = git_parse_maybe_bool(strval);
1147 free(strval);
1148
1149 if (v == -1)
1150 return 0;
1151
1152 *val = v;
1153
1154 return 1;
1155 }
1156
1157 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1158 {
1159 const struct string_list *list = ud;
1160 const struct string_list_item *item;
1161
1162 for_each_string_list_item (item, list) {
1163 if (key->len == (uintptr_t)item->util &&
1164 !strncasecmp(item->string, key->buf, key->len))
1165 return 1;
1166 }
1167 return 0;
1168 }
1169
1170 int format_set_trailers_options(struct process_trailer_options *opts,
1171 struct string_list *filter_list,
1172 struct strbuf *sepbuf,
1173 struct strbuf *kvsepbuf,
1174 const char **arg,
1175 char **invalid_arg)
1176 {
1177 for (;;) {
1178 const char *argval;
1179 size_t arglen;
1180
1181 if (**arg == ')')
1182 break;
1183
1184 if (match_placeholder_arg_value(*arg, "key", arg, &argval, &arglen)) {
1185 uintptr_t len = arglen;
1186
1187 if (!argval)
1188 return -1;
1189
1190 if (len && argval[len - 1] == ':')
1191 len--;
1192 string_list_append(filter_list, argval)->util = (char *)len;
1193
1194 opts->filter = format_trailer_match_cb;
1195 opts->filter_data = filter_list;
1196 opts->only_trailers = 1;
1197 } else if (match_placeholder_arg_value(*arg, "separator", arg, &argval, &arglen)) {
1198 char *fmt;
1199
1200 strbuf_reset(sepbuf);
1201 fmt = xstrndup(argval, arglen);
1202 strbuf_expand(sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1203 free(fmt);
1204 opts->separator = sepbuf;
1205 } else if (match_placeholder_arg_value(*arg, "key_value_separator", arg, &argval, &arglen)) {
1206 char *fmt;
1207
1208 strbuf_reset(kvsepbuf);
1209 fmt = xstrndup(argval, arglen);
1210 strbuf_expand(kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1211 free(fmt);
1212 opts->key_value_separator = kvsepbuf;
1213 } else if (!match_placeholder_bool_arg(*arg, "only", arg, &opts->only_trailers) &&
1214 !match_placeholder_bool_arg(*arg, "unfold", arg, &opts->unfold) &&
1215 !match_placeholder_bool_arg(*arg, "keyonly", arg, &opts->key_only) &&
1216 !match_placeholder_bool_arg(*arg, "valueonly", arg, &opts->value_only)) {
1217 if (invalid_arg) {
1218 size_t len = strcspn(*arg, ",)");
1219 *invalid_arg = xstrndup(*arg, len);
1220 }
1221 return -1;
1222 }
1223 }
1224 return 0;
1225 }
1226
1227 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1228 const char *placeholder,
1229 void *context)
1230 {
1231 struct format_commit_context *c = context;
1232 const struct commit *commit = c->commit;
1233 const char *msg = c->message;
1234 struct commit_list *p;
1235 const char *arg, *eol;
1236 size_t res;
1237 char **slot;
1238
1239 /* these are independent of the commit */
1240 res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1241 if (res)
1242 return res;
1243
1244 switch (placeholder[0]) {
1245 case 'C':
1246 if (starts_with(placeholder + 1, "(auto)")) {
1247 c->auto_color = want_color(c->pretty_ctx->color);
1248 if (c->auto_color && sb->len)
1249 strbuf_addstr(sb, GIT_COLOR_RESET);
1250 return 7; /* consumed 7 bytes, "C(auto)" */
1251 } else {
1252 int ret = parse_color(sb, placeholder, c);
1253 if (ret)
1254 c->auto_color = 0;
1255 /*
1256 * Otherwise, we decided to treat %C<unknown>
1257 * as a literal string, and the previous
1258 * %C(auto) is still valid.
1259 */
1260 return ret;
1261 }
1262 case 'w':
1263 if (placeholder[1] == '(') {
1264 unsigned long width = 0, indent1 = 0, indent2 = 0;
1265 char *next;
1266 const char *start = placeholder + 2;
1267 const char *end = strchr(start, ')');
1268 if (!end)
1269 return 0;
1270 if (end > start) {
1271 width = strtoul(start, &next, 10);
1272 if (*next == ',') {
1273 indent1 = strtoul(next + 1, &next, 10);
1274 if (*next == ',') {
1275 indent2 = strtoul(next + 1,
1276 &next, 10);
1277 }
1278 }
1279 if (*next != ')')
1280 return 0;
1281 }
1282
1283 /*
1284 * We need to limit the format here as it allows the
1285 * user to prepend arbitrarily many bytes to the buffer
1286 * when rewrapping.
1287 */
1288 if (width > FORMATTING_LIMIT ||
1289 indent1 > FORMATTING_LIMIT ||
1290 indent2 > FORMATTING_LIMIT)
1291 return 0;
1292 rewrap_message_tail(sb, c, width, indent1, indent2);
1293 return end - placeholder + 1;
1294 } else
1295 return 0;
1296
1297 case '<':
1298 case '>':
1299 return parse_padding_placeholder(placeholder, c);
1300 }
1301
1302 /* these depend on the commit */
1303 if (!commit->object.parsed)
1304 parse_object(the_repository, &commit->object.oid);
1305
1306 switch (placeholder[0]) {
1307 case 'H': /* commit hash */
1308 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1309 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1310 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1311 return 1;
1312 case 'h': /* abbreviated commit hash */
1313 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1314 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1315 c->pretty_ctx->abbrev);
1316 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1317 return 1;
1318 case 'T': /* tree hash */
1319 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1320 return 1;
1321 case 't': /* abbreviated tree hash */
1322 strbuf_add_unique_abbrev(sb,
1323 get_commit_tree_oid(commit),
1324 c->pretty_ctx->abbrev);
1325 return 1;
1326 case 'P': /* parent hashes */
1327 for (p = commit->parents; p; p = p->next) {
1328 if (p != commit->parents)
1329 strbuf_addch(sb, ' ');
1330 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1331 }
1332 return 1;
1333 case 'p': /* abbreviated parent hashes */
1334 for (p = commit->parents; p; p = p->next) {
1335 if (p != commit->parents)
1336 strbuf_addch(sb, ' ');
1337 strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1338 c->pretty_ctx->abbrev);
1339 }
1340 return 1;
1341 case 'm': /* left/right/bottom */
1342 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1343 return 1;
1344 case 'd':
1345 format_decorations(sb, commit, c->auto_color);
1346 return 1;
1347 case 'D':
1348 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1349 return 1;
1350 case 'S': /* tag/branch like --source */
1351 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1352 return 0;
1353 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1354 if (!(slot && *slot))
1355 return 0;
1356 strbuf_addstr(sb, *slot);
1357 return 1;
1358 case 'g': /* reflog info */
1359 switch(placeholder[1]) {
1360 case 'd': /* reflog selector */
1361 case 'D':
1362 if (c->pretty_ctx->reflog_info)
1363 get_reflog_selector(sb,
1364 c->pretty_ctx->reflog_info,
1365 &c->pretty_ctx->date_mode,
1366 c->pretty_ctx->date_mode_explicit,
1367 (placeholder[1] == 'd'));
1368 return 2;
1369 case 's': /* reflog message */
1370 if (c->pretty_ctx->reflog_info)
1371 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1372 return 2;
1373 case 'n':
1374 case 'N':
1375 case 'e':
1376 case 'E':
1377 return format_reflog_person(sb,
1378 placeholder[1],
1379 c->pretty_ctx->reflog_info,
1380 &c->pretty_ctx->date_mode);
1381 }
1382 return 0; /* unknown %g placeholder */
1383 case 'N':
1384 if (c->pretty_ctx->notes_message) {
1385 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1386 return 1;
1387 }
1388 return 0;
1389 }
1390
1391 if (placeholder[0] == 'G') {
1392 if (!c->signature_check.result)
1393 check_commit_signature(c->commit, &(c->signature_check));
1394 switch (placeholder[1]) {
1395 case 'G':
1396 if (c->signature_check.gpg_output)
1397 strbuf_addstr(sb, c->signature_check.gpg_output);
1398 break;
1399 case '?':
1400 switch (c->signature_check.result) {
1401 case 'G':
1402 switch (c->signature_check.trust_level) {
1403 case TRUST_UNDEFINED:
1404 case TRUST_NEVER:
1405 strbuf_addch(sb, 'U');
1406 break;
1407 default:
1408 strbuf_addch(sb, 'G');
1409 break;
1410 }
1411 break;
1412 case 'B':
1413 case 'E':
1414 case 'N':
1415 case 'X':
1416 case 'Y':
1417 case 'R':
1418 strbuf_addch(sb, c->signature_check.result);
1419 }
1420 break;
1421 case 'S':
1422 if (c->signature_check.signer)
1423 strbuf_addstr(sb, c->signature_check.signer);
1424 break;
1425 case 'K':
1426 if (c->signature_check.key)
1427 strbuf_addstr(sb, c->signature_check.key);
1428 break;
1429 case 'F':
1430 if (c->signature_check.fingerprint)
1431 strbuf_addstr(sb, c->signature_check.fingerprint);
1432 break;
1433 case 'P':
1434 if (c->signature_check.primary_key_fingerprint)
1435 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1436 break;
1437 case 'T':
1438 switch (c->signature_check.trust_level) {
1439 case TRUST_UNDEFINED:
1440 strbuf_addstr(sb, "undefined");
1441 break;
1442 case TRUST_NEVER:
1443 strbuf_addstr(sb, "never");
1444 break;
1445 case TRUST_MARGINAL:
1446 strbuf_addstr(sb, "marginal");
1447 break;
1448 case TRUST_FULLY:
1449 strbuf_addstr(sb, "fully");
1450 break;
1451 case TRUST_ULTIMATE:
1452 strbuf_addstr(sb, "ultimate");
1453 break;
1454 }
1455 break;
1456 default:
1457 return 0;
1458 }
1459 return 2;
1460 }
1461
1462 /* For the rest we have to parse the commit header. */
1463 if (!c->commit_header_parsed) {
1464 msg = c->message =
1465 repo_logmsg_reencode(c->repository, commit,
1466 &c->commit_encoding, "UTF-8");
1467 parse_commit_header(c);
1468 }
1469
1470 switch (placeholder[0]) {
1471 case 'a': /* author ... */
1472 return format_person_part(sb, placeholder[1],
1473 msg + c->author.off, c->author.len,
1474 &c->pretty_ctx->date_mode);
1475 case 'c': /* committer ... */
1476 return format_person_part(sb, placeholder[1],
1477 msg + c->committer.off, c->committer.len,
1478 &c->pretty_ctx->date_mode);
1479 case 'e': /* encoding */
1480 if (c->commit_encoding)
1481 strbuf_addstr(sb, c->commit_encoding);
1482 return 1;
1483 case 'B': /* raw body */
1484 /* message_off is always left at the initial newline */
1485 strbuf_addstr(sb, msg + c->message_off + 1);
1486 return 1;
1487 }
1488
1489 /* Now we need to parse the commit message. */
1490 if (!c->commit_message_parsed)
1491 parse_commit_message(c);
1492
1493 switch (placeholder[0]) {
1494 case 's': /* subject */
1495 format_subject(sb, msg + c->subject_off, " ");
1496 return 1;
1497 case 'f': /* sanitized subject */
1498 eol = strchrnul(msg + c->subject_off, '\n');
1499 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1500 return 1;
1501 case 'b': /* body */
1502 strbuf_addstr(sb, msg + c->body_off);
1503 return 1;
1504 }
1505
1506 if (skip_prefix(placeholder, "(trailers", &arg)) {
1507 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1508 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1509 struct strbuf sepbuf = STRBUF_INIT;
1510 struct strbuf kvsepbuf = STRBUF_INIT;
1511 size_t ret = 0;
1512
1513 opts.no_divider = 1;
1514
1515 if (*arg == ':') {
1516 arg++;
1517 if (format_set_trailers_options(&opts, &filter_list, &sepbuf, &kvsepbuf, &arg, NULL))
1518 goto trailer_out;
1519 }
1520 if (*arg == ')') {
1521 format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1522 ret = arg - placeholder + 1;
1523 }
1524 trailer_out:
1525 string_list_clear(&filter_list, 0);
1526 strbuf_release(&sepbuf);
1527 return ret;
1528 }
1529
1530 return 0; /* unknown placeholder */
1531 }
1532
1533 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1534 const char *placeholder,
1535 struct format_commit_context *c)
1536 {
1537 struct strbuf local_sb = STRBUF_INIT;
1538 size_t total_consumed = 0;
1539 int len, padding = c->padding;
1540
1541 if (padding < 0) {
1542 const char *start = strrchr(sb->buf, '\n');
1543 int occupied;
1544 if (!start)
1545 start = sb->buf;
1546 occupied = utf8_strnwidth(start, strlen(start), 1);
1547 occupied += c->pretty_ctx->graph_width;
1548 padding = (-padding) - occupied;
1549 }
1550 while (1) {
1551 int modifier = *placeholder == 'C';
1552 size_t consumed = format_commit_one(&local_sb, placeholder, c);
1553 total_consumed += consumed;
1554
1555 if (!modifier)
1556 break;
1557
1558 placeholder += consumed;
1559 if (*placeholder != '%')
1560 break;
1561 placeholder++;
1562 total_consumed++;
1563 }
1564 len = utf8_strnwidth(local_sb.buf, local_sb.len, 1);
1565
1566 if (c->flush_type == flush_left_and_steal) {
1567 const char *ch = sb->buf + sb->len - 1;
1568 while (len > padding && ch > sb->buf) {
1569 const char *p;
1570 if (*ch == ' ') {
1571 ch--;
1572 padding++;
1573 continue;
1574 }
1575 /* check for trailing ansi sequences */
1576 if (*ch != 'm')
1577 break;
1578 p = ch - 1;
1579 while (p > sb->buf && ch - p < 10 && *p != '\033')
1580 p--;
1581 if (*p != '\033' ||
1582 ch + 1 - p != display_mode_esc_sequence_len(p))
1583 break;
1584 /*
1585 * got a good ansi sequence, put it back to
1586 * local_sb as we're cutting sb
1587 */
1588 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1589 ch = p - 1;
1590 }
1591 strbuf_setlen(sb, ch + 1 - sb->buf);
1592 c->flush_type = flush_left;
1593 }
1594
1595 if (len > padding) {
1596 switch (c->truncate) {
1597 case trunc_left:
1598 strbuf_utf8_replace(&local_sb,
1599 0, len - (padding - 2),
1600 "..");
1601 break;
1602 case trunc_middle:
1603 strbuf_utf8_replace(&local_sb,
1604 padding / 2 - 1,
1605 len - (padding - 2),
1606 "..");
1607 break;
1608 case trunc_right:
1609 strbuf_utf8_replace(&local_sb,
1610 padding - 2, len - (padding - 2),
1611 "..");
1612 break;
1613 case trunc_none:
1614 break;
1615 }
1616 strbuf_addbuf(sb, &local_sb);
1617 } else {
1618 size_t sb_len = sb->len, offset = 0;
1619 if (c->flush_type == flush_left)
1620 offset = padding - len;
1621 else if (c->flush_type == flush_both)
1622 offset = (padding - len) / 2;
1623 /*
1624 * we calculate padding in columns, now
1625 * convert it back to chars
1626 */
1627 padding = padding - len + local_sb.len;
1628 strbuf_addchars(sb, ' ', padding);
1629 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1630 local_sb.len);
1631 }
1632 strbuf_release(&local_sb);
1633 c->flush_type = no_flush;
1634 return total_consumed;
1635 }
1636
1637 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1638 const char *placeholder,
1639 void *context)
1640 {
1641 size_t consumed, orig_len;
1642 enum {
1643 NO_MAGIC,
1644 ADD_LF_BEFORE_NON_EMPTY,
1645 DEL_LF_BEFORE_EMPTY,
1646 ADD_SP_BEFORE_NON_EMPTY
1647 } magic = NO_MAGIC;
1648
1649 switch (placeholder[0]) {
1650 case '-':
1651 magic = DEL_LF_BEFORE_EMPTY;
1652 break;
1653 case '+':
1654 magic = ADD_LF_BEFORE_NON_EMPTY;
1655 break;
1656 case ' ':
1657 magic = ADD_SP_BEFORE_NON_EMPTY;
1658 break;
1659 default:
1660 break;
1661 }
1662 if (magic != NO_MAGIC) {
1663 placeholder++;
1664
1665 switch (placeholder[0]) {
1666 case 'w':
1667 /*
1668 * `%+w()` cannot ever expand to a non-empty string,
1669 * and it potentially changes the layout of preceding
1670 * contents. We're thus not able to handle the magic in
1671 * this combination and refuse the pattern.
1672 */
1673 return 0;
1674 };
1675 }
1676
1677 orig_len = sb->len;
1678 if (((struct format_commit_context *)context)->flush_type != no_flush)
1679 consumed = format_and_pad_commit(sb, placeholder, context);
1680 else
1681 consumed = format_commit_one(sb, placeholder, context);
1682 if (magic == NO_MAGIC)
1683 return consumed;
1684
1685 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1686 while (sb->len && sb->buf[sb->len - 1] == '\n')
1687 strbuf_setlen(sb, sb->len - 1);
1688 } else if (orig_len != sb->len) {
1689 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1690 strbuf_insertstr(sb, orig_len, "\n");
1691 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1692 strbuf_insertstr(sb, orig_len, " ");
1693 }
1694 return consumed + 1;
1695 }
1696
1697 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1698 void *context)
1699 {
1700 struct userformat_want *w = context;
1701
1702 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1703 placeholder++;
1704
1705 switch (*placeholder) {
1706 case 'N':
1707 w->notes = 1;
1708 break;
1709 case 'S':
1710 w->source = 1;
1711 break;
1712 }
1713 return 0;
1714 }
1715
1716 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1717 {
1718 struct strbuf dummy = STRBUF_INIT;
1719
1720 if (!fmt) {
1721 if (!user_format)
1722 return;
1723 fmt = user_format;
1724 }
1725 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1726 strbuf_release(&dummy);
1727 }
1728
1729 void repo_format_commit_message(struct repository *r,
1730 const struct commit *commit,
1731 const char *format, struct strbuf *sb,
1732 const struct pretty_print_context *pretty_ctx)
1733 {
1734 struct format_commit_context context = {
1735 .repository = r,
1736 .commit = commit,
1737 .pretty_ctx = pretty_ctx,
1738 .wrap_start = sb->len
1739 };
1740 const char *output_enc = pretty_ctx->output_encoding;
1741 const char *utf8 = "UTF-8";
1742
1743 strbuf_expand(sb, format, format_commit_item, &context);
1744 rewrap_message_tail(sb, &context, 0, 0, 0);
1745
1746 /*
1747 * Convert output to an actual output encoding; note that
1748 * format_commit_item() will always use UTF-8, so we don't
1749 * have to bother if that's what the output wants.
1750 */
1751 if (output_enc) {
1752 if (same_encoding(utf8, output_enc))
1753 output_enc = NULL;
1754 } else {
1755 if (context.commit_encoding &&
1756 !same_encoding(context.commit_encoding, utf8))
1757 output_enc = context.commit_encoding;
1758 }
1759
1760 if (output_enc) {
1761 size_t outsz;
1762 char *out = reencode_string_len(sb->buf, sb->len,
1763 output_enc, utf8, &outsz);
1764 if (out)
1765 strbuf_attach(sb, out, outsz, outsz + 1);
1766 }
1767
1768 free(context.commit_encoding);
1769 repo_unuse_commit_buffer(r, commit, context.message);
1770 }
1771
1772 static void pp_header(struct pretty_print_context *pp,
1773 const char *encoding,
1774 const struct commit *commit,
1775 const char **msg_p,
1776 struct strbuf *sb)
1777 {
1778 int parents_shown = 0;
1779
1780 for (;;) {
1781 const char *name, *line = *msg_p;
1782 int linelen = get_one_line(*msg_p);
1783
1784 if (!linelen)
1785 return;
1786 *msg_p += linelen;
1787
1788 if (linelen == 1)
1789 /* End of header */
1790 return;
1791
1792 if (pp->fmt == CMIT_FMT_RAW) {
1793 strbuf_add(sb, line, linelen);
1794 continue;
1795 }
1796
1797 if (starts_with(line, "parent ")) {
1798 if (linelen != the_hash_algo->hexsz + 8)
1799 die("bad parent line in commit");
1800 continue;
1801 }
1802
1803 if (!parents_shown) {
1804 unsigned num = commit_list_count(commit->parents);
1805 /* with enough slop */
1806 strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1807 add_merge_info(pp, sb, commit);
1808 parents_shown = 1;
1809 }
1810
1811 /*
1812 * MEDIUM == DEFAULT shows only author with dates.
1813 * FULL shows both authors but not dates.
1814 * FULLER shows both authors and dates.
1815 */
1816 if (skip_prefix(line, "author ", &name)) {
1817 strbuf_grow(sb, linelen + 80);
1818 pp_user_info(pp, "Author", sb, name, encoding);
1819 }
1820 if (skip_prefix(line, "committer ", &name) &&
1821 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1822 strbuf_grow(sb, linelen + 80);
1823 pp_user_info(pp, "Commit", sb, name, encoding);
1824 }
1825 }
1826 }
1827
1828 void pp_title_line(struct pretty_print_context *pp,
1829 const char **msg_p,
1830 struct strbuf *sb,
1831 const char *encoding,
1832 int need_8bit_cte)
1833 {
1834 static const int max_length = 78; /* per rfc2047 */
1835 struct strbuf title;
1836
1837 strbuf_init(&title, 80);
1838 *msg_p = format_subject(&title, *msg_p,
1839 pp->preserve_subject ? "\n" : " ");
1840
1841 strbuf_grow(sb, title.len + 1024);
1842 if (pp->print_email_subject) {
1843 if (pp->rev)
1844 fmt_output_email_subject(sb, pp->rev);
1845 if (pp->encode_email_headers &&
1846 needs_rfc2047_encoding(title.buf, title.len))
1847 add_rfc2047(sb, title.buf, title.len,
1848 encoding, RFC2047_SUBJECT);
1849 else
1850 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1851 -last_line_length(sb), 1, max_length);
1852 } else {
1853 strbuf_addbuf(sb, &title);
1854 }
1855 strbuf_addch(sb, '\n');
1856
1857 if (need_8bit_cte == 0) {
1858 int i;
1859 for (i = 0; i < pp->in_body_headers.nr; i++) {
1860 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1861 need_8bit_cte = 1;
1862 break;
1863 }
1864 }
1865 }
1866
1867 if (need_8bit_cte > 0) {
1868 const char *header_fmt =
1869 "MIME-Version: 1.0\n"
1870 "Content-Type: text/plain; charset=%s\n"
1871 "Content-Transfer-Encoding: 8bit\n";
1872 strbuf_addf(sb, header_fmt, encoding);
1873 }
1874 if (pp->after_subject) {
1875 strbuf_addstr(sb, pp->after_subject);
1876 }
1877 if (cmit_fmt_is_mail(pp->fmt)) {
1878 strbuf_addch(sb, '\n');
1879 }
1880
1881 if (pp->in_body_headers.nr) {
1882 int i;
1883 for (i = 0; i < pp->in_body_headers.nr; i++) {
1884 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1885 free(pp->in_body_headers.items[i].string);
1886 }
1887 string_list_clear(&pp->in_body_headers, 0);
1888 strbuf_addch(sb, '\n');
1889 }
1890
1891 strbuf_release(&title);
1892 }
1893
1894 static int pp_utf8_width(const char *start, const char *end)
1895 {
1896 int width = 0;
1897 size_t remain = end - start;
1898
1899 while (remain) {
1900 int n = utf8_width(&start, &remain);
1901 if (n < 0 || !start)
1902 return -1;
1903 width += n;
1904 }
1905 return width;
1906 }
1907
1908 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1909 const char *line, int linelen)
1910 {
1911 const char *tab;
1912
1913 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1914 int width = pp_utf8_width(line, tab);
1915
1916 /*
1917 * If it wasn't well-formed utf8, or it
1918 * had characters with badly defined
1919 * width (control characters etc), just
1920 * give up on trying to align things.
1921 */
1922 if (width < 0)
1923 break;
1924
1925 /* Output the data .. */
1926 strbuf_add(sb, line, tab - line);
1927
1928 /* .. and the de-tabified tab */
1929 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1930
1931 /* Skip over the printed part .. */
1932 linelen -= tab + 1 - line;
1933 line = tab + 1;
1934 }
1935
1936 /*
1937 * Print out everything after the last tab without
1938 * worrying about width - there's nothing more to
1939 * align.
1940 */
1941 strbuf_add(sb, line, linelen);
1942 }
1943
1944 /*
1945 * pp_handle_indent() prints out the intendation, and
1946 * the whole line (without the final newline), after
1947 * de-tabifying.
1948 */
1949 static void pp_handle_indent(struct pretty_print_context *pp,
1950 struct strbuf *sb, int indent,
1951 const char *line, int linelen)
1952 {
1953 strbuf_addchars(sb, ' ', indent);
1954 if (pp->expand_tabs_in_log)
1955 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1956 else
1957 strbuf_add(sb, line, linelen);
1958 }
1959
1960 static int is_mboxrd_from(const char *line, int len)
1961 {
1962 /*
1963 * a line matching /^From $/ here would only have len == 4
1964 * at this point because is_empty_line would've trimmed all
1965 * trailing space
1966 */
1967 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1968 }
1969
1970 void pp_remainder(struct pretty_print_context *pp,
1971 const char **msg_p,
1972 struct strbuf *sb,
1973 int indent)
1974 {
1975 int first = 1;
1976 for (;;) {
1977 const char *line = *msg_p;
1978 int linelen = get_one_line(line);
1979 *msg_p += linelen;
1980
1981 if (!linelen)
1982 break;
1983
1984 if (is_blank_line(line, &linelen)) {
1985 if (first)
1986 continue;
1987 if (pp->fmt == CMIT_FMT_SHORT)
1988 break;
1989 }
1990 first = 0;
1991
1992 strbuf_grow(sb, linelen + indent + 20);
1993 if (indent)
1994 pp_handle_indent(pp, sb, indent, line, linelen);
1995 else if (pp->expand_tabs_in_log)
1996 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1997 line, linelen);
1998 else {
1999 if (pp->fmt == CMIT_FMT_MBOXRD &&
2000 is_mboxrd_from(line, linelen))
2001 strbuf_addch(sb, '>');
2002
2003 strbuf_add(sb, line, linelen);
2004 }
2005 strbuf_addch(sb, '\n');
2006 }
2007 }
2008
2009 void pretty_print_commit(struct pretty_print_context *pp,
2010 const struct commit *commit,
2011 struct strbuf *sb)
2012 {
2013 unsigned long beginning_of_body;
2014 int indent = 4;
2015 const char *msg;
2016 const char *reencoded;
2017 const char *encoding;
2018 int need_8bit_cte = pp->need_8bit_cte;
2019
2020 if (pp->fmt == CMIT_FMT_USERFORMAT) {
2021 format_commit_message(commit, user_format, sb, pp);
2022 return;
2023 }
2024
2025 encoding = get_log_output_encoding();
2026 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
2027
2028 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2029 indent = 0;
2030
2031 /*
2032 * We need to check and emit Content-type: to mark it
2033 * as 8-bit if we haven't done so.
2034 */
2035 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
2036 int i, ch, in_body;
2037
2038 for (in_body = i = 0; (ch = msg[i]); i++) {
2039 if (!in_body) {
2040 /* author could be non 7-bit ASCII but
2041 * the log may be so; skip over the
2042 * header part first.
2043 */
2044 if (ch == '\n' && msg[i+1] == '\n')
2045 in_body = 1;
2046 }
2047 else if (non_ascii(ch)) {
2048 need_8bit_cte = 1;
2049 break;
2050 }
2051 }
2052 }
2053
2054 pp_header(pp, encoding, commit, &msg, sb);
2055 if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
2056 strbuf_addch(sb, '\n');
2057 }
2058
2059 /* Skip excess blank lines at the beginning of body, if any... */
2060 msg = skip_blank_lines(msg);
2061
2062 /* These formats treat the title line specially. */
2063 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2064 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2065
2066 beginning_of_body = sb->len;
2067 if (pp->fmt != CMIT_FMT_ONELINE)
2068 pp_remainder(pp, &msg, sb, indent);
2069 strbuf_rtrim(sb);
2070
2071 /* Make sure there is an EOLN for the non-oneline case */
2072 if (pp->fmt != CMIT_FMT_ONELINE)
2073 strbuf_addch(sb, '\n');
2074
2075 /*
2076 * The caller may append additional body text in e-mail
2077 * format. Make sure we did not strip the blank line
2078 * between the header and the body.
2079 */
2080 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2081 strbuf_addch(sb, '\n');
2082
2083 unuse_commit_buffer(commit, reencoded);
2084 }
2085
2086 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2087 struct strbuf *sb)
2088 {
2089 struct pretty_print_context pp = {0};
2090 pp.fmt = fmt;
2091 pretty_print_commit(&pp, commit, sb);
2092 }
The Best VCS Around