?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'rs/code-cleaning'
[git/debian.git] / pretty.c
blob3a1da6fd329efe1723bdb906a907ce4160c4633a
1 #include "cache.h"
2 #include "commit.h"
3 #include "utf8.h"
4 #include "diff.h"
5 #include "revision.h"
6 #include "string-list.h"
7 #include "mailmap.h"
8 #include "log-tree.h"
9 #include "notes.h"
10 #include "color.h"
11 #include "reflog-walk.h"
12 #include "gpg-interface.h"
13
14 static char *user_format;
15 static struct cmt_fmt_map {
16 const char *name;
17 enum cmit_fmt format;
18 int is_tformat;
19 int is_alias;
20 const char *user_format;
21 } *commit_formats;
22 static size_t builtin_formats_len;
23 static size_t commit_formats_len;
24 static size_t commit_formats_alloc;
25 static struct cmt_fmt_map *find_commit_format(const char *sought);
26
27 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
28 {
29 free(user_format);
30 user_format = xstrdup(cp);
31 if (is_tformat)
32 rev->use_terminator = 1;
33 rev->commit_format = CMIT_FMT_USERFORMAT;
34 }
35
36 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
37 {
38 struct cmt_fmt_map *commit_format = NULL;
39 const char *name;
40 const char *fmt;
41 int i;
42
43 if (!skip_prefix(var, "pretty.", &name))
44 return 0;
45
46 for (i = 0; i < builtin_formats_len; i++) {
47 if (!strcmp(commit_formats[i].name, name))
48 return 0;
49 }
50
51 for (i = builtin_formats_len; i < commit_formats_len; i++) {
52 if (!strcmp(commit_formats[i].name, name)) {
53 commit_format = &commit_formats[i];
54 break;
55 }
56 }
57
58 if (!commit_format) {
59 ALLOC_GROW(commit_formats, commit_formats_len+1,
60 commit_formats_alloc);
61 commit_format = &commit_formats[commit_formats_len];
62 memset(commit_format, 0, sizeof(*commit_format));
63 commit_formats_len++;
64 }
65
66 commit_format->name = xstrdup(name);
67 commit_format->format = CMIT_FMT_USERFORMAT;
68 git_config_string(&fmt, var, value);
69 if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
70 commit_format->is_tformat = fmt[0] == 't';
71 fmt = strchr(fmt, ':') + 1;
72 } else if (strchr(fmt, '%'))
73 commit_format->is_tformat = 1;
74 else
75 commit_format->is_alias = 1;
76 commit_format->user_format = fmt;
77
78 return 0;
79 }
80
81 static void setup_commit_formats(void)
82 {
83 struct cmt_fmt_map builtin_formats[] = {
84 { "raw", CMIT_FMT_RAW, 0 },
85 { "medium", CMIT_FMT_MEDIUM, 0 },
86 { "short", CMIT_FMT_SHORT, 0 },
87 { "email", CMIT_FMT_EMAIL, 0 },
88 { "fuller", CMIT_FMT_FULLER, 0 },
89 { "full", CMIT_FMT_FULL, 0 },
90 { "oneline", CMIT_FMT_ONELINE, 1 }
91 };
92 commit_formats_len = ARRAY_SIZE(builtin_formats);
93 builtin_formats_len = commit_formats_len;
94 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
95 memcpy(commit_formats, builtin_formats,
96 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
97
98 git_config(git_pretty_formats_config, NULL);
99 }
100
101 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
102 const char *original,
103 int num_redirections)
104 {
105 struct cmt_fmt_map *found = NULL;
106 size_t found_match_len = 0;
107 int i;
108
109 if (num_redirections >= commit_formats_len)
110 die("invalid --pretty format: "
111 "'%s' references an alias which points to itself",
112 original);
113
114 for (i = 0; i < commit_formats_len; i++) {
115 size_t match_len;
116
117 if (!starts_with(commit_formats[i].name, sought))
118 continue;
119
120 match_len = strlen(commit_formats[i].name);
121 if (found == NULL || found_match_len > match_len) {
122 found = &commit_formats[i];
123 found_match_len = match_len;
124 }
125 }
126
127 if (found && found->is_alias) {
128 found = find_commit_format_recursive(found->user_format,
129 original,
130 num_redirections+1);
131 }
132
133 return found;
134 }
135
136 static struct cmt_fmt_map *find_commit_format(const char *sought)
137 {
138 if (!commit_formats)
139 setup_commit_formats();
140
141 return find_commit_format_recursive(sought, sought, 0);
142 }
143
144 void get_commit_format(const char *arg, struct rev_info *rev)
145 {
146 struct cmt_fmt_map *commit_format;
147
148 rev->use_terminator = 0;
149 if (!arg || !*arg) {
150 rev->commit_format = CMIT_FMT_DEFAULT;
151 return;
152 }
153 if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
154 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
155 return;
156 }
157
158 if (strchr(arg, '%')) {
159 save_user_format(rev, arg, 1);
160 return;
161 }
162
163 commit_format = find_commit_format(arg);
164 if (!commit_format)
165 die("invalid --pretty format: %s", arg);
166
167 rev->commit_format = commit_format->format;
168 rev->use_terminator = commit_format->is_tformat;
169 if (commit_format->format == CMIT_FMT_USERFORMAT) {
170 save_user_format(rev, commit_format->user_format,
171 commit_format->is_tformat);
172 }
173 }
174
175 /*
176 * Generic support for pretty-printing the header
177 */
178 static int get_one_line(const char *msg)
179 {
180 int ret = 0;
181
182 for (;;) {
183 char c = *msg++;
184 if (!c)
185 break;
186 ret++;
187 if (c == '\n')
188 break;
189 }
190 return ret;
191 }
192
193 /* High bit set, or ISO-2022-INT */
194 static int non_ascii(int ch)
195 {
196 return !isascii(ch) || ch == '\033';
197 }
198
199 int has_non_ascii(const char *s)
200 {
201 int ch;
202 if (!s)
203 return 0;
204 while ((ch = *s++) != '\0') {
205 if (non_ascii(ch))
206 return 1;
207 }
208 return 0;
209 }
210
211 static int is_rfc822_special(char ch)
212 {
213 switch (ch) {
214 case '(':
215 case ')':
216 case '<':
217 case '>':
218 case '[':
219 case ']':
220 case ':':
221 case ';':
222 case '@':
223 case ',':
224 case '.':
225 case '"':
226 case '\\':
227 return 1;
228 default:
229 return 0;
230 }
231 }
232
233 static int needs_rfc822_quoting(const char *s, int len)
234 {
235 int i;
236 for (i = 0; i < len; i++)
237 if (is_rfc822_special(s[i]))
238 return 1;
239 return 0;
240 }
241
242 static int last_line_length(struct strbuf *sb)
243 {
244 int i;
245
246 /* How many bytes are already used on the last line? */
247 for (i = sb->len - 1; i >= 0; i--)
248 if (sb->buf[i] == '\n')
249 break;
250 return sb->len - (i + 1);
251 }
252
253 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
254 {
255 int i;
256
257 /* just a guess, we may have to also backslash-quote */
258 strbuf_grow(out, len + 2);
259
260 strbuf_addch(out, '"');
261 for (i = 0; i < len; i++) {
262 switch (s[i]) {
263 case '"':
264 case '\\':
265 strbuf_addch(out, '\\');
266 /* fall through */
267 default:
268 strbuf_addch(out, s[i]);
269 }
270 }
271 strbuf_addch(out, '"');
272 }
273
274 enum rfc2047_type {
275 RFC2047_SUBJECT,
276 RFC2047_ADDRESS
277 };
278
279 static int is_rfc2047_special(char ch, enum rfc2047_type type)
280 {
281 /*
282 * rfc2047, section 4.2:
283 *
284 * 8-bit values which correspond to printable ASCII characters other
285 * than "=", "?", and "_" (underscore), MAY be represented as those
286 * characters. (But see section 5 for restrictions.) In
287 * particular, SPACE and TAB MUST NOT be represented as themselves
288 * within encoded words.
289 */
290
291 /*
292 * rule out non-ASCII characters and non-printable characters (the
293 * non-ASCII check should be redundant as isprint() is not localized
294 * and only knows about ASCII, but be defensive about that)
295 */
296 if (non_ascii(ch) || !isprint(ch))
297 return 1;
298
299 /*
300 * rule out special printable characters (' ' should be the only
301 * whitespace character considered printable, but be defensive and use
302 * isspace())
303 */
304 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
305 return 1;
306
307 /*
308 * rfc2047, section 5.3:
309 *
310 * As a replacement for a 'word' entity within a 'phrase', for example,
311 * one that precedes an address in a From, To, or Cc header. The ABNF
312 * definition for 'phrase' from RFC 822 thus becomes:
313 *
314 * phrase = 1*( encoded-word / word )
315 *
316 * In this case the set of characters that may be used in a "Q"-encoded
317 * 'encoded-word' is restricted to: <upper and lower case ASCII
318 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
319 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
320 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
321 * 'special' by 'linear-white-space'.
322 */
323
324 if (type != RFC2047_ADDRESS)
325 return 0;
326
327 /* '=' and '_' are special cases and have been checked above */
328 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
329 }
330
331 static int needs_rfc2047_encoding(const char *line, int len,
332 enum rfc2047_type type)
333 {
334 int i;
335
336 for (i = 0; i < len; i++) {
337 int ch = line[i];
338 if (non_ascii(ch) || ch == '\n')
339 return 1;
340 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
341 return 1;
342 }
343
344 return 0;
345 }
346
347 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
348 const char *encoding, enum rfc2047_type type)
349 {
350 static const int max_encoded_length = 76; /* per rfc2047 */
351 int i;
352 int line_len = last_line_length(sb);
353
354 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
355 strbuf_addf(sb, "=?%s?q?", encoding);
356 line_len += strlen(encoding) + 5; /* 5 for =??q? */
357
358 while (len) {
359 /*
360 * RFC 2047, section 5 (3):
361 *
362 * Each 'encoded-word' MUST represent an integral number of
363 * characters. A multi-octet character may not be split across
364 * adjacent 'encoded- word's.
365 */
366 const unsigned char *p = (const unsigned char *)line;
367 int chrlen = mbs_chrlen(&line, &len, encoding);
368 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
369
370 /* "=%02X" * chrlen, or the byte itself */
371 const char *encoded_fmt = is_special ? "=%02X" : "%c";
372 int encoded_len = is_special ? 3 * chrlen : 1;
373
374 /*
375 * According to RFC 2047, we could encode the special character
376 * ' ' (space) with '_' (underscore) for readability. But many
377 * programs do not understand this and just leave the
378 * underscore in place. Thus, we do nothing special here, which
379 * causes ' ' to be encoded as '=20', avoiding this problem.
380 */
381
382 if (line_len + encoded_len + 2 > max_encoded_length) {
383 /* It won't fit with trailing "?=" --- break the line */
384 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
385 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
386 }
387
388 for (i = 0; i < chrlen; i++)
389 strbuf_addf(sb, encoded_fmt, p[i]);
390 line_len += encoded_len;
391 }
392 strbuf_addstr(sb, "?=");
393 }
394
395 const char *show_ident_date(const struct ident_split *ident,
396 enum date_mode mode)
397 {
398 unsigned long date = 0;
399 long tz = 0;
400
401 if (ident->date_begin && ident->date_end)
402 date = strtoul(ident->date_begin, NULL, 10);
403 if (date_overflows(date))
404 date = 0;
405 else {
406 if (ident->tz_begin && ident->tz_end)
407 tz = strtol(ident->tz_begin, NULL, 10);
408 if (tz >= INT_MAX || tz <= INT_MIN)
409 tz = 0;
410 }
411 return show_date(date, tz, mode);
412 }
413
414 void pp_user_info(struct pretty_print_context *pp,
415 const char *what, struct strbuf *sb,
416 const char *line, const char *encoding)
417 {
418 struct ident_split ident;
419 char *line_end;
420 const char *mailbuf, *namebuf;
421 size_t namelen, maillen;
422 int max_length = 78; /* per rfc2822 */
423
424 if (pp->fmt == CMIT_FMT_ONELINE)
425 return;
426
427 line_end = strchrnul(line, '\n');
428 if (split_ident_line(&ident, line, line_end - line))
429 return;
430
431 mailbuf = ident.mail_begin;
432 maillen = ident.mail_end - ident.mail_begin;
433 namebuf = ident.name_begin;
434 namelen = ident.name_end - ident.name_begin;
435
436 if (pp->mailmap)
437 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
438
439 if (pp->fmt == CMIT_FMT_EMAIL) {
440 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
441 struct strbuf buf = STRBUF_INIT;
442
443 strbuf_addstr(&buf, "From: ");
444 strbuf_add(&buf, namebuf, namelen);
445 strbuf_addstr(&buf, " <");
446 strbuf_add(&buf, mailbuf, maillen);
447 strbuf_addstr(&buf, ">\n");
448 string_list_append(&pp->in_body_headers,
449 strbuf_detach(&buf, NULL));
450
451 mailbuf = pp->from_ident->mail_begin;
452 maillen = pp->from_ident->mail_end - mailbuf;
453 namebuf = pp->from_ident->name_begin;
454 namelen = pp->from_ident->name_end - namebuf;
455 }
456
457 strbuf_addstr(sb, "From: ");
458 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
459 add_rfc2047(sb, namebuf, namelen,
460 encoding, RFC2047_ADDRESS);
461 max_length = 76; /* per rfc2047 */
462 } else if (needs_rfc822_quoting(namebuf, namelen)) {
463 struct strbuf quoted = STRBUF_INIT;
464 add_rfc822_quoted(&quoted, namebuf, namelen);
465 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
466 -6, 1, max_length);
467 strbuf_release(&quoted);
468 } else {
469 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
470 -6, 1, max_length);
471 }
472
473 if (max_length <
474 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
475 strbuf_addch(sb, '\n');
476 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
477 } else {
478 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
479 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
480 (int)namelen, namebuf, (int)maillen, mailbuf);
481 }
482
483 switch (pp->fmt) {
484 case CMIT_FMT_MEDIUM:
485 strbuf_addf(sb, "Date: %s\n",
486 show_ident_date(&ident, pp->date_mode));
487 break;
488 case CMIT_FMT_EMAIL:
489 strbuf_addf(sb, "Date: %s\n",
490 show_ident_date(&ident, DATE_RFC2822));
491 break;
492 case CMIT_FMT_FULLER:
493 strbuf_addf(sb, "%sDate: %s\n", what,
494 show_ident_date(&ident, pp->date_mode));
495 break;
496 default:
497 /* notin' */
498 break;
499 }
500 }
501
502 static int is_empty_line(const char *line, int *len_p)
503 {
504 int len = *len_p;
505 while (len && isspace(line[len - 1]))
506 len--;
507 *len_p = len;
508 return !len;
509 }
510
511 static const char *skip_empty_lines(const char *msg)
512 {
513 for (;;) {
514 int linelen = get_one_line(msg);
515 int ll = linelen;
516 if (!linelen)
517 break;
518 if (!is_empty_line(msg, &ll))
519 break;
520 msg += linelen;
521 }
522 return msg;
523 }
524
525 static void add_merge_info(const struct pretty_print_context *pp,
526 struct strbuf *sb, const struct commit *commit)
527 {
528 struct commit_list *parent = commit->parents;
529
530 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
531 !parent || !parent->next)
532 return;
533
534 strbuf_addstr(sb, "Merge:");
535
536 while (parent) {
537 struct commit *p = parent->item;
538 const char *hex = NULL;
539 if (pp->abbrev)
540 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
541 if (!hex)
542 hex = sha1_to_hex(p->object.sha1);
543 parent = parent->next;
544
545 strbuf_addf(sb, " %s", hex);
546 }
547 strbuf_addch(sb, '\n');
548 }
549
550 static char *get_header(const struct commit *commit, const char *msg,
551 const char *key)
552 {
553 int key_len = strlen(key);
554 const char *line = msg;
555
556 while (line) {
557 const char *eol = strchrnul(line, '\n'), *next;
558
559 if (line == eol)
560 return NULL;
561 if (!*eol) {
562 warning("malformed commit (header is missing newline): %s",
563 sha1_to_hex(commit->object.sha1));
564 next = NULL;
565 } else
566 next = eol + 1;
567 if (eol - line > key_len &&
568 !strncmp(line, key, key_len) &&
569 line[key_len] == ' ') {
570 return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
571 }
572 line = next;
573 }
574 return 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 (strncmp(cp, "encoding ", strlen("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 *logmsg_reencode(const struct commit *commit,
609 char **commit_encoding,
610 const char *output_encoding)
611 {
612 static const char *utf8 = "UTF-8";
613 const char *use_encoding;
614 char *encoding;
615 const char *msg = get_commit_buffer(commit, NULL);
616 char *out;
617
618 if (!output_encoding || !*output_encoding) {
619 if (commit_encoding)
620 *commit_encoding =
621 get_header(commit, msg, "encoding");
622 return msg;
623 }
624 encoding = get_header(commit, 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(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 unuse_commit_buffer(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 mail_map = xcalloc(1, sizeof(*mail_map));
683 read_mailmap(mail_map, NULL);
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, enum date_mode dmode)
690 {
691 /* currently all placeholders have same length */
692 const int placeholder_len = 2;
693 struct ident_split s;
694 const char *name, *mail;
695 size_t maillen, namelen;
696
697 if (split_ident_line(&s, msg, len) < 0)
698 goto skip;
699
700 name = s.name_begin;
701 namelen = s.name_end - s.name_begin;
702 mail = s.mail_begin;
703 maillen = s.mail_end - s.mail_begin;
704
705 if (part == 'N' || part == 'E') /* mailmap lookup */
706 mailmap_name(&mail, &maillen, &name, &namelen);
707 if (part == 'n' || part == 'N') { /* name */
708 strbuf_add(sb, name, namelen);
709 return placeholder_len;
710 }
711 if (part == 'e' || part == 'E') { /* email */
712 strbuf_add(sb, mail, maillen);
713 return placeholder_len;
714 }
715
716 if (!s.date_begin)
717 goto skip;
718
719 if (part == 't') { /* date, UNIX timestamp */
720 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
721 return placeholder_len;
722 }
723
724 switch (part) {
725 case 'd': /* date */
726 strbuf_addstr(sb, show_ident_date(&s, dmode));
727 return placeholder_len;
728 case 'D': /* date, RFC2822 style */
729 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
730 return placeholder_len;
731 case 'r': /* date, relative */
732 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
733 return placeholder_len;
734 case 'i': /* date, ISO 8601 */
735 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
736 return placeholder_len;
737 }
738
739 skip:
740 /*
741 * reading from either a bogus commit, or a reflog entry with
742 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
743 * to compute a valid return value.
744 */
745 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
746 || part == 'D' || part == 'r' || part == 'i')
747 return placeholder_len;
748
749 return 0; /* unknown placeholder */
750 }
751
752 struct chunk {
753 size_t off;
754 size_t len;
755 };
756
757 enum flush_type {
758 no_flush,
759 flush_right,
760 flush_left,
761 flush_left_and_steal,
762 flush_both
763 };
764
765 enum trunc_type {
766 trunc_none,
767 trunc_left,
768 trunc_middle,
769 trunc_right
770 };
771
772 struct format_commit_context {
773 const struct commit *commit;
774 const struct pretty_print_context *pretty_ctx;
775 unsigned commit_header_parsed:1;
776 unsigned commit_message_parsed:1;
777 struct signature_check signature_check;
778 enum flush_type flush_type;
779 enum trunc_type truncate;
780 const char *message;
781 char *commit_encoding;
782 size_t width, indent1, indent2;
783 int auto_color;
784 int padding;
785
786 /* These offsets are relative to the start of the commit message. */
787 struct chunk author;
788 struct chunk committer;
789 size_t message_off;
790 size_t subject_off;
791 size_t body_off;
792
793 /* The following ones are relative to the result struct strbuf. */
794 struct chunk abbrev_commit_hash;
795 struct chunk abbrev_tree_hash;
796 struct chunk abbrev_parent_hashes;
797 size_t wrap_start;
798 };
799
800 static int add_again(struct strbuf *sb, struct chunk *chunk)
801 {
802 if (chunk->len) {
803 strbuf_adddup(sb, chunk->off, chunk->len);
804 return 1;
805 }
806
807 /*
808 * We haven't seen this chunk before. Our caller is surely
809 * going to add it the hard way now. Remember the most likely
810 * start of the to-be-added chunk: the current end of the
811 * struct strbuf.
812 */
813 chunk->off = sb->len;
814 return 0;
815 }
816
817 static void parse_commit_header(struct format_commit_context *context)
818 {
819 const char *msg = context->message;
820 int i;
821
822 for (i = 0; msg[i]; i++) {
823 int eol;
824 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
825 ; /* do nothing */
826
827 if (i == eol) {
828 break;
829 } else if (starts_with(msg + i, "author ")) {
830 context->author.off = i + 7;
831 context->author.len = eol - i - 7;
832 } else if (starts_with(msg + i, "committer ")) {
833 context->committer.off = i + 10;
834 context->committer.len = eol - i - 10;
835 }
836 i = eol;
837 }
838 context->message_off = i;
839 context->commit_header_parsed = 1;
840 }
841
842 static int istitlechar(char c)
843 {
844 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
845 (c >= '0' && c <= '9') || c == '.' || c == '_';
846 }
847
848 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
849 {
850 size_t trimlen;
851 size_t start_len = sb->len;
852 int space = 2;
853
854 for (; *msg && *msg != '\n'; msg++) {
855 if (istitlechar(*msg)) {
856 if (space == 1)
857 strbuf_addch(sb, '-');
858 space = 0;
859 strbuf_addch(sb, *msg);
860 if (*msg == '.')
861 while (*(msg+1) == '.')
862 msg++;
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_empty_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_empty_lines(msg);
906 c->subject_off = msg - start;
907
908 msg = format_subject(NULL, msg, NULL);
909 msg = skip_empty_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 enum 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 if (placeholder[1] == '(') {
966 const char *begin = placeholder + 2;
967 const char *end = strchr(begin, ')');
968 char color[COLOR_MAXLEN];
969
970 if (!end)
971 return 0;
972 if (starts_with(begin, "auto,")) {
973 if (!want_color(c->pretty_ctx->color))
974 return end - placeholder + 1;
975 begin += 5;
976 }
977 color_parse_mem(begin,
978 end - begin,
979 "--pretty format", color);
980 strbuf_addstr(sb, color);
981 return end - placeholder + 1;
982 }
983 if (starts_with(placeholder + 1, "red")) {
984 strbuf_addstr(sb, GIT_COLOR_RED);
985 return 4;
986 } else if (starts_with(placeholder + 1, "green")) {
987 strbuf_addstr(sb, GIT_COLOR_GREEN);
988 return 6;
989 } else if (starts_with(placeholder + 1, "blue")) {
990 strbuf_addstr(sb, GIT_COLOR_BLUE);
991 return 5;
992 } else if (starts_with(placeholder + 1, "reset")) {
993 strbuf_addstr(sb, GIT_COLOR_RESET);
994 return 6;
995 } else
996 return 0;
997 }
998
999 static size_t parse_padding_placeholder(struct strbuf *sb,
1000 const char *placeholder,
1001 struct format_commit_context *c)
1002 {
1003 const char *ch = placeholder;
1004 enum flush_type flush_type;
1005 int to_column = 0;
1006
1007 switch (*ch++) {
1008 case '<':
1009 flush_type = flush_right;
1010 break;
1011 case '>':
1012 if (*ch == '<') {
1013 flush_type = flush_both;
1014 ch++;
1015 } else if (*ch == '>') {
1016 flush_type = flush_left_and_steal;
1017 ch++;
1018 } else
1019 flush_type = flush_left;
1020 break;
1021 default:
1022 return 0;
1023 }
1024
1025 /* the next value means "wide enough to that column" */
1026 if (*ch == '|') {
1027 to_column = 1;
1028 ch++;
1029 }
1030
1031 if (*ch == '(') {
1032 const char *start = ch + 1;
1033 const char *end = start + strcspn(start, ",)");
1034 char *next;
1035 int width;
1036 if (!end || end == start)
1037 return 0;
1038 width = strtoul(start, &next, 10);
1039 if (next == start || width == 0)
1040 return 0;
1041 c->padding = to_column ? -width : width;
1042 c->flush_type = flush_type;
1043
1044 if (*end == ',') {
1045 start = end + 1;
1046 end = strchr(start, ')');
1047 if (!end || end == start)
1048 return 0;
1049 if (starts_with(start, "trunc)"))
1050 c->truncate = trunc_right;
1051 else if (starts_with(start, "ltrunc)"))
1052 c->truncate = trunc_left;
1053 else if (starts_with(start, "mtrunc)"))
1054 c->truncate = trunc_middle;
1055 else
1056 return 0;
1057 } else
1058 c->truncate = trunc_none;
1059
1060 return end - placeholder + 1;
1061 }
1062 return 0;
1063 }
1064
1065 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1066 const char *placeholder,
1067 void *context)
1068 {
1069 struct format_commit_context *c = context;
1070 const struct commit *commit = c->commit;
1071 const char *msg = c->message;
1072 struct commit_list *p;
1073 int h1, h2;
1074
1075 /* these are independent of the commit */
1076 switch (placeholder[0]) {
1077 case 'C':
1078 if (starts_with(placeholder + 1, "(auto)")) {
1079 c->auto_color = 1;
1080 return 7; /* consumed 7 bytes, "C(auto)" */
1081 } else {
1082 int ret = parse_color(sb, placeholder, c);
1083 if (ret)
1084 c->auto_color = 0;
1085 /*
1086 * Otherwise, we decided to treat %C<unknown>
1087 * as a literal string, and the previous
1088 * %C(auto) is still valid.
1089 */
1090 return ret;
1091 }
1092 case 'n': /* newline */
1093 strbuf_addch(sb, '\n');
1094 return 1;
1095 case 'x':
1096 /* %x00 == NUL, %x0a == LF, etc. */
1097 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1098 h1 <= 16 &&
1099 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1100 h2 <= 16) {
1101 strbuf_addch(sb, (h1<<4)|h2);
1102 return 3;
1103 } else
1104 return 0;
1105 case 'w':
1106 if (placeholder[1] == '(') {
1107 unsigned long width = 0, indent1 = 0, indent2 = 0;
1108 char *next;
1109 const char *start = placeholder + 2;
1110 const char *end = strchr(start, ')');
1111 if (!end)
1112 return 0;
1113 if (end > start) {
1114 width = strtoul(start, &next, 10);
1115 if (*next == ',') {
1116 indent1 = strtoul(next + 1, &next, 10);
1117 if (*next == ',') {
1118 indent2 = strtoul(next + 1,
1119 &next, 10);
1120 }
1121 }
1122 if (*next != ')')
1123 return 0;
1124 }
1125 rewrap_message_tail(sb, c, width, indent1, indent2);
1126 return end - placeholder + 1;
1127 } else
1128 return 0;
1129
1130 case '<':
1131 case '>':
1132 return parse_padding_placeholder(sb, placeholder, c);
1133 }
1134
1135 /* these depend on the commit */
1136 if (!commit->object.parsed)
1137 parse_object(commit->object.sha1);
1138
1139 switch (placeholder[0]) {
1140 case 'H': /* commit hash */
1141 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1142 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1143 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1144 return 1;
1145 case 'h': /* abbreviated commit hash */
1146 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1147 if (add_again(sb, &c->abbrev_commit_hash)) {
1148 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1149 return 1;
1150 }
1151 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1152 c->pretty_ctx->abbrev));
1153 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1154 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1155 return 1;
1156 case 'T': /* tree hash */
1157 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1158 return 1;
1159 case 't': /* abbreviated tree hash */
1160 if (add_again(sb, &c->abbrev_tree_hash))
1161 return 1;
1162 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1163 c->pretty_ctx->abbrev));
1164 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1165 return 1;
1166 case 'P': /* parent hashes */
1167 for (p = commit->parents; p; p = p->next) {
1168 if (p != commit->parents)
1169 strbuf_addch(sb, ' ');
1170 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1171 }
1172 return 1;
1173 case 'p': /* abbreviated parent hashes */
1174 if (add_again(sb, &c->abbrev_parent_hashes))
1175 return 1;
1176 for (p = commit->parents; p; p = p->next) {
1177 if (p != commit->parents)
1178 strbuf_addch(sb, ' ');
1179 strbuf_addstr(sb, find_unique_abbrev(
1180 p->item->object.sha1,
1181 c->pretty_ctx->abbrev));
1182 }
1183 c->abbrev_parent_hashes.len = sb->len -
1184 c->abbrev_parent_hashes.off;
1185 return 1;
1186 case 'm': /* left/right/bottom */
1187 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1188 return 1;
1189 case 'd':
1190 load_ref_decorations(DECORATE_SHORT_REFS);
1191 format_decorations(sb, commit, c->auto_color);
1192 return 1;
1193 case 'g': /* reflog info */
1194 switch(placeholder[1]) {
1195 case 'd': /* reflog selector */
1196 case 'D':
1197 if (c->pretty_ctx->reflog_info)
1198 get_reflog_selector(sb,
1199 c->pretty_ctx->reflog_info,
1200 c->pretty_ctx->date_mode,
1201 c->pretty_ctx->date_mode_explicit,
1202 (placeholder[1] == 'd'));
1203 return 2;
1204 case 's': /* reflog message */
1205 if (c->pretty_ctx->reflog_info)
1206 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1207 return 2;
1208 case 'n':
1209 case 'N':
1210 case 'e':
1211 case 'E':
1212 return format_reflog_person(sb,
1213 placeholder[1],
1214 c->pretty_ctx->reflog_info,
1215 c->pretty_ctx->date_mode);
1216 }
1217 return 0; /* unknown %g placeholder */
1218 case 'N':
1219 if (c->pretty_ctx->notes_message) {
1220 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1221 return 1;
1222 }
1223 return 0;
1224 }
1225
1226 if (placeholder[0] == 'G') {
1227 if (!c->signature_check.result)
1228 check_commit_signature(c->commit, &(c->signature_check));
1229 switch (placeholder[1]) {
1230 case 'G':
1231 if (c->signature_check.gpg_output)
1232 strbuf_addstr(sb, c->signature_check.gpg_output);
1233 break;
1234 case '?':
1235 switch (c->signature_check.result) {
1236 case 'G':
1237 case 'B':
1238 case 'U':
1239 case 'N':
1240 strbuf_addch(sb, c->signature_check.result);
1241 }
1242 break;
1243 case 'S':
1244 if (c->signature_check.signer)
1245 strbuf_addstr(sb, c->signature_check.signer);
1246 break;
1247 case 'K':
1248 if (c->signature_check.key)
1249 strbuf_addstr(sb, c->signature_check.key);
1250 break;
1251 default:
1252 return 0;
1253 }
1254 return 2;
1255 }
1256
1257
1258 /* For the rest we have to parse the commit header. */
1259 if (!c->commit_header_parsed)
1260 parse_commit_header(c);
1261
1262 switch (placeholder[0]) {
1263 case 'a': /* author ... */
1264 return format_person_part(sb, placeholder[1],
1265 msg + c->author.off, c->author.len,
1266 c->pretty_ctx->date_mode);
1267 case 'c': /* committer ... */
1268 return format_person_part(sb, placeholder[1],
1269 msg + c->committer.off, c->committer.len,
1270 c->pretty_ctx->date_mode);
1271 case 'e': /* encoding */
1272 if (c->commit_encoding)
1273 strbuf_addstr(sb, c->commit_encoding);
1274 return 1;
1275 case 'B': /* raw body */
1276 /* message_off is always left at the initial newline */
1277 strbuf_addstr(sb, msg + c->message_off + 1);
1278 return 1;
1279 }
1280
1281 /* Now we need to parse the commit message. */
1282 if (!c->commit_message_parsed)
1283 parse_commit_message(c);
1284
1285 switch (placeholder[0]) {
1286 case 's': /* subject */
1287 format_subject(sb, msg + c->subject_off, " ");
1288 return 1;
1289 case 'f': /* sanitized subject */
1290 format_sanitized_subject(sb, msg + c->subject_off);
1291 return 1;
1292 case 'b': /* body */
1293 strbuf_addstr(sb, msg + c->body_off);
1294 return 1;
1295 }
1296 return 0; /* unknown placeholder */
1297 }
1298
1299 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1300 const char *placeholder,
1301 struct format_commit_context *c)
1302 {
1303 struct strbuf local_sb = STRBUF_INIT;
1304 int total_consumed = 0, len, padding = c->padding;
1305 if (padding < 0) {
1306 const char *start = strrchr(sb->buf, '\n');
1307 int occupied;
1308 if (!start)
1309 start = sb->buf;
1310 occupied = utf8_strnwidth(start, -1, 1);
1311 padding = (-padding) - occupied;
1312 }
1313 while (1) {
1314 int modifier = *placeholder == 'C';
1315 int consumed = format_commit_one(&local_sb, placeholder, c);
1316 total_consumed += consumed;
1317
1318 if (!modifier)
1319 break;
1320
1321 placeholder += consumed;
1322 if (*placeholder != '%')
1323 break;
1324 placeholder++;
1325 total_consumed++;
1326 }
1327 len = utf8_strnwidth(local_sb.buf, -1, 1);
1328
1329 if (c->flush_type == flush_left_and_steal) {
1330 const char *ch = sb->buf + sb->len - 1;
1331 while (len > padding && ch > sb->buf) {
1332 const char *p;
1333 if (*ch == ' ') {
1334 ch--;
1335 padding++;
1336 continue;
1337 }
1338 /* check for trailing ansi sequences */
1339 if (*ch != 'm')
1340 break;
1341 p = ch - 1;
1342 while (ch - p < 10 && *p != '\033')
1343 p--;
1344 if (*p != '\033' ||
1345 ch + 1 - p != display_mode_esc_sequence_len(p))
1346 break;
1347 /*
1348 * got a good ansi sequence, put it back to
1349 * local_sb as we're cutting sb
1350 */
1351 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1352 ch = p - 1;
1353 }
1354 strbuf_setlen(sb, ch + 1 - sb->buf);
1355 c->flush_type = flush_left;
1356 }
1357
1358 if (len > padding) {
1359 switch (c->truncate) {
1360 case trunc_left:
1361 strbuf_utf8_replace(&local_sb,
1362 0, len - (padding - 2),
1363 "..");
1364 break;
1365 case trunc_middle:
1366 strbuf_utf8_replace(&local_sb,
1367 padding / 2 - 1,
1368 len - (padding - 2),
1369 "..");
1370 break;
1371 case trunc_right:
1372 strbuf_utf8_replace(&local_sb,
1373 padding - 2, len - (padding - 2),
1374 "..");
1375 break;
1376 case trunc_none:
1377 break;
1378 }
1379 strbuf_addbuf(sb, &local_sb);
1380 } else {
1381 int sb_len = sb->len, offset = 0;
1382 if (c->flush_type == flush_left)
1383 offset = padding - len;
1384 else if (c->flush_type == flush_both)
1385 offset = (padding - len) / 2;
1386 /*
1387 * we calculate padding in columns, now
1388 * convert it back to chars
1389 */
1390 padding = padding - len + local_sb.len;
1391 strbuf_grow(sb, padding);
1392 strbuf_setlen(sb, sb_len + padding);
1393 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1394 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1395 local_sb.len);
1396 }
1397 strbuf_release(&local_sb);
1398 c->flush_type = no_flush;
1399 return total_consumed;
1400 }
1401
1402 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1403 const char *placeholder,
1404 void *context)
1405 {
1406 int consumed;
1407 size_t orig_len;
1408 enum {
1409 NO_MAGIC,
1410 ADD_LF_BEFORE_NON_EMPTY,
1411 DEL_LF_BEFORE_EMPTY,
1412 ADD_SP_BEFORE_NON_EMPTY
1413 } magic = NO_MAGIC;
1414
1415 switch (placeholder[0]) {
1416 case '-':
1417 magic = DEL_LF_BEFORE_EMPTY;
1418 break;
1419 case '+':
1420 magic = ADD_LF_BEFORE_NON_EMPTY;
1421 break;
1422 case ' ':
1423 magic = ADD_SP_BEFORE_NON_EMPTY;
1424 break;
1425 default:
1426 break;
1427 }
1428 if (magic != NO_MAGIC)
1429 placeholder++;
1430
1431 orig_len = sb->len;
1432 if (((struct format_commit_context *)context)->flush_type != no_flush)
1433 consumed = format_and_pad_commit(sb, placeholder, context);
1434 else
1435 consumed = format_commit_one(sb, placeholder, context);
1436 if (magic == NO_MAGIC)
1437 return consumed;
1438
1439 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1440 while (sb->len && sb->buf[sb->len - 1] == '\n')
1441 strbuf_setlen(sb, sb->len - 1);
1442 } else if (orig_len != sb->len) {
1443 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1444 strbuf_insert(sb, orig_len, "\n", 1);
1445 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1446 strbuf_insert(sb, orig_len, " ", 1);
1447 }
1448 return consumed + 1;
1449 }
1450
1451 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1452 void *context)
1453 {
1454 struct userformat_want *w = context;
1455
1456 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1457 placeholder++;
1458
1459 switch (*placeholder) {
1460 case 'N':
1461 w->notes = 1;
1462 break;
1463 }
1464 return 0;
1465 }
1466
1467 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1468 {
1469 struct strbuf dummy = STRBUF_INIT;
1470
1471 if (!fmt) {
1472 if (!user_format)
1473 return;
1474 fmt = user_format;
1475 }
1476 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1477 strbuf_release(&dummy);
1478 }
1479
1480 void format_commit_message(const struct commit *commit,
1481 const char *format, struct strbuf *sb,
1482 const struct pretty_print_context *pretty_ctx)
1483 {
1484 struct format_commit_context context;
1485 const char *output_enc = pretty_ctx->output_encoding;
1486 const char *utf8 = "UTF-8";
1487
1488 memset(&context, 0, sizeof(context));
1489 context.commit = commit;
1490 context.pretty_ctx = pretty_ctx;
1491 context.wrap_start = sb->len;
1492 /*
1493 * convert a commit message to UTF-8 first
1494 * as far as 'format_commit_item' assumes it in UTF-8
1495 */
1496 context.message = logmsg_reencode(commit,
1497 &context.commit_encoding,
1498 utf8);
1499
1500 strbuf_expand(sb, format, format_commit_item, &context);
1501 rewrap_message_tail(sb, &context, 0, 0, 0);
1502
1503 /* then convert a commit message to an actual output encoding */
1504 if (output_enc) {
1505 if (same_encoding(utf8, output_enc))
1506 output_enc = NULL;
1507 } else {
1508 if (context.commit_encoding &&
1509 !same_encoding(context.commit_encoding, utf8))
1510 output_enc = context.commit_encoding;
1511 }
1512
1513 if (output_enc) {
1514 int outsz;
1515 char *out = reencode_string_len(sb->buf, sb->len,
1516 output_enc, utf8, &outsz);
1517 if (out)
1518 strbuf_attach(sb, out, outsz, outsz + 1);
1519 }
1520
1521 free(context.commit_encoding);
1522 unuse_commit_buffer(commit, context.message);
1523 }
1524
1525 static void pp_header(struct pretty_print_context *pp,
1526 const char *encoding,
1527 const struct commit *commit,
1528 const char **msg_p,
1529 struct strbuf *sb)
1530 {
1531 int parents_shown = 0;
1532
1533 for (;;) {
1534 const char *line = *msg_p;
1535 int linelen = get_one_line(*msg_p);
1536
1537 if (!linelen)
1538 return;
1539 *msg_p += linelen;
1540
1541 if (linelen == 1)
1542 /* End of header */
1543 return;
1544
1545 if (pp->fmt == CMIT_FMT_RAW) {
1546 strbuf_add(sb, line, linelen);
1547 continue;
1548 }
1549
1550 if (starts_with(line, "parent ")) {
1551 if (linelen != 48)
1552 die("bad parent line in commit");
1553 continue;
1554 }
1555
1556 if (!parents_shown) {
1557 unsigned num = commit_list_count(commit->parents);
1558 /* with enough slop */
1559 strbuf_grow(sb, num * 50 + 20);
1560 add_merge_info(pp, sb, commit);
1561 parents_shown = 1;
1562 }
1563
1564 /*
1565 * MEDIUM == DEFAULT shows only author with dates.
1566 * FULL shows both authors but not dates.
1567 * FULLER shows both authors and dates.
1568 */
1569 if (starts_with(line, "author ")) {
1570 strbuf_grow(sb, linelen + 80);
1571 pp_user_info(pp, "Author", sb, line + 7, encoding);
1572 }
1573 if (starts_with(line, "committer ") &&
1574 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1575 strbuf_grow(sb, linelen + 80);
1576 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1577 }
1578 }
1579 }
1580
1581 void pp_title_line(struct pretty_print_context *pp,
1582 const char **msg_p,
1583 struct strbuf *sb,
1584 const char *encoding,
1585 int need_8bit_cte)
1586 {
1587 static const int max_length = 78; /* per rfc2047 */
1588 struct strbuf title;
1589
1590 strbuf_init(&title, 80);
1591 *msg_p = format_subject(&title, *msg_p,
1592 pp->preserve_subject ? "\n" : " ");
1593
1594 strbuf_grow(sb, title.len + 1024);
1595 if (pp->subject) {
1596 strbuf_addstr(sb, pp->subject);
1597 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1598 add_rfc2047(sb, title.buf, title.len,
1599 encoding, RFC2047_SUBJECT);
1600 else
1601 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1602 -last_line_length(sb), 1, max_length);
1603 } else {
1604 strbuf_addbuf(sb, &title);
1605 }
1606 strbuf_addch(sb, '\n');
1607
1608 if (need_8bit_cte == 0) {
1609 int i;
1610 for (i = 0; i < pp->in_body_headers.nr; i++) {
1611 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1612 need_8bit_cte = 1;
1613 break;
1614 }
1615 }
1616 }
1617
1618 if (need_8bit_cte > 0) {
1619 const char *header_fmt =
1620 "MIME-Version: 1.0\n"
1621 "Content-Type: text/plain; charset=%s\n"
1622 "Content-Transfer-Encoding: 8bit\n";
1623 strbuf_addf(sb, header_fmt, encoding);
1624 }
1625 if (pp->after_subject) {
1626 strbuf_addstr(sb, pp->after_subject);
1627 }
1628 if (pp->fmt == CMIT_FMT_EMAIL) {
1629 strbuf_addch(sb, '\n');
1630 }
1631
1632 if (pp->in_body_headers.nr) {
1633 int i;
1634 for (i = 0; i < pp->in_body_headers.nr; i++) {
1635 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1636 free(pp->in_body_headers.items[i].string);
1637 }
1638 string_list_clear(&pp->in_body_headers, 0);
1639 strbuf_addch(sb, '\n');
1640 }
1641
1642 strbuf_release(&title);
1643 }
1644
1645 void pp_remainder(struct pretty_print_context *pp,
1646 const char **msg_p,
1647 struct strbuf *sb,
1648 int indent)
1649 {
1650 int first = 1;
1651 for (;;) {
1652 const char *line = *msg_p;
1653 int linelen = get_one_line(line);
1654 *msg_p += linelen;
1655
1656 if (!linelen)
1657 break;
1658
1659 if (is_empty_line(line, &linelen)) {
1660 if (first)
1661 continue;
1662 if (pp->fmt == CMIT_FMT_SHORT)
1663 break;
1664 }
1665 first = 0;
1666
1667 strbuf_grow(sb, linelen + indent + 20);
1668 if (indent) {
1669 memset(sb->buf + sb->len, ' ', indent);
1670 strbuf_setlen(sb, sb->len + indent);
1671 }
1672 strbuf_add(sb, line, linelen);
1673 strbuf_addch(sb, '\n');
1674 }
1675 }
1676
1677 void pretty_print_commit(struct pretty_print_context *pp,
1678 const struct commit *commit,
1679 struct strbuf *sb)
1680 {
1681 unsigned long beginning_of_body;
1682 int indent = 4;
1683 const char *msg;
1684 const char *reencoded;
1685 const char *encoding;
1686 int need_8bit_cte = pp->need_8bit_cte;
1687
1688 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1689 format_commit_message(commit, user_format, sb, pp);
1690 return;
1691 }
1692
1693 encoding = get_log_output_encoding();
1694 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1695
1696 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1697 indent = 0;
1698
1699 /*
1700 * We need to check and emit Content-type: to mark it
1701 * as 8-bit if we haven't done so.
1702 */
1703 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1704 int i, ch, in_body;
1705
1706 for (in_body = i = 0; (ch = msg[i]); i++) {
1707 if (!in_body) {
1708 /* author could be non 7-bit ASCII but
1709 * the log may be so; skip over the
1710 * header part first.
1711 */
1712 if (ch == '\n' && msg[i+1] == '\n')
1713 in_body = 1;
1714 }
1715 else if (non_ascii(ch)) {
1716 need_8bit_cte = 1;
1717 break;
1718 }
1719 }
1720 }
1721
1722 pp_header(pp, encoding, commit, &msg, sb);
1723 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1724 strbuf_addch(sb, '\n');
1725 }
1726
1727 /* Skip excess blank lines at the beginning of body, if any... */
1728 msg = skip_empty_lines(msg);
1729
1730 /* These formats treat the title line specially. */
1731 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1732 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1733
1734 beginning_of_body = sb->len;
1735 if (pp->fmt != CMIT_FMT_ONELINE)
1736 pp_remainder(pp, &msg, sb, indent);
1737 strbuf_rtrim(sb);
1738
1739 /* Make sure there is an EOLN for the non-oneline case */
1740 if (pp->fmt != CMIT_FMT_ONELINE)
1741 strbuf_addch(sb, '\n');
1742
1743 /*
1744 * The caller may append additional body text in e-mail
1745 * format. Make sure we did not strip the blank line
1746 * between the header and the body.
1747 */
1748 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1749 strbuf_addch(sb, '\n');
1750
1751 unuse_commit_buffer(commit, reencoded);
1752 }
1753
1754 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1755 struct strbuf *sb)
1756 {
1757 struct pretty_print_context pp = {0};
1758 pp.fmt = fmt;
1759 pretty_print_commit(&pp, commit, sb);
1760 }
git for Debian