?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'mg/verify-commit' into maint
[git.git] / pretty.c
blob7b493041814dc8514bde0778497d39868c8455d7
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 int commit_format_is_empty(enum cmit_fmt fmt)
28 {
29 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
30 }
31
32 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
33 {
34 free(user_format);
35 user_format = xstrdup(cp);
36 if (is_tformat)
37 rev->use_terminator = 1;
38 rev->commit_format = CMIT_FMT_USERFORMAT;
39 }
40
41 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
42 {
43 struct cmt_fmt_map *commit_format = NULL;
44 const char *name;
45 const char *fmt;
46 int i;
47
48 if (!skip_prefix(var, "pretty.", &name))
49 return 0;
50
51 for (i = 0; i < builtin_formats_len; i++) {
52 if (!strcmp(commit_formats[i].name, name))
53 return 0;
54 }
55
56 for (i = builtin_formats_len; i < commit_formats_len; i++) {
57 if (!strcmp(commit_formats[i].name, name)) {
58 commit_format = &commit_formats[i];
59 break;
60 }
61 }
62
63 if (!commit_format) {
64 ALLOC_GROW(commit_formats, commit_formats_len+1,
65 commit_formats_alloc);
66 commit_format = &commit_formats[commit_formats_len];
67 memset(commit_format, 0, sizeof(*commit_format));
68 commit_formats_len++;
69 }
70
71 commit_format->name = xstrdup(name);
72 commit_format->format = CMIT_FMT_USERFORMAT;
73 if (git_config_string(&fmt, var, value))
74 return -1;
75
76 if (skip_prefix(fmt, "format:", &fmt))
77 commit_format->is_tformat = 0;
78 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
79 commit_format->is_tformat = 1;
80 else
81 commit_format->is_alias = 1;
82 commit_format->user_format = fmt;
83
84 return 0;
85 }
86
87 static void setup_commit_formats(void)
88 {
89 struct cmt_fmt_map builtin_formats[] = {
90 { "raw", CMIT_FMT_RAW, 0 },
91 { "medium", CMIT_FMT_MEDIUM, 0 },
92 { "short", CMIT_FMT_SHORT, 0 },
93 { "email", CMIT_FMT_EMAIL, 0 },
94 { "fuller", CMIT_FMT_FULLER, 0 },
95 { "full", CMIT_FMT_FULL, 0 },
96 { "oneline", CMIT_FMT_ONELINE, 1 }
97 };
98 commit_formats_len = ARRAY_SIZE(builtin_formats);
99 builtin_formats_len = commit_formats_len;
100 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
101 memcpy(commit_formats, builtin_formats,
102 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
103
104 git_config(git_pretty_formats_config, NULL);
105 }
106
107 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
108 const char *original,
109 int num_redirections)
110 {
111 struct cmt_fmt_map *found = NULL;
112 size_t found_match_len = 0;
113 int i;
114
115 if (num_redirections >= commit_formats_len)
116 die("invalid --pretty format: "
117 "'%s' references an alias which points to itself",
118 original);
119
120 for (i = 0; i < commit_formats_len; i++) {
121 size_t match_len;
122
123 if (!starts_with(commit_formats[i].name, sought))
124 continue;
125
126 match_len = strlen(commit_formats[i].name);
127 if (found == NULL || found_match_len > match_len) {
128 found = &commit_formats[i];
129 found_match_len = match_len;
130 }
131 }
132
133 if (found && found->is_alias) {
134 found = find_commit_format_recursive(found->user_format,
135 original,
136 num_redirections+1);
137 }
138
139 return found;
140 }
141
142 static struct cmt_fmt_map *find_commit_format(const char *sought)
143 {
144 if (!commit_formats)
145 setup_commit_formats();
146
147 return find_commit_format_recursive(sought, sought, 0);
148 }
149
150 void get_commit_format(const char *arg, struct rev_info *rev)
151 {
152 struct cmt_fmt_map *commit_format;
153
154 rev->use_terminator = 0;
155 if (!arg) {
156 rev->commit_format = CMIT_FMT_DEFAULT;
157 return;
158 }
159 if (skip_prefix(arg, "format:", &arg)) {
160 save_user_format(rev, arg, 0);
161 return;
162 }
163
164 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
165 save_user_format(rev, arg, 1);
166 return;
167 }
168
169 commit_format = find_commit_format(arg);
170 if (!commit_format)
171 die("invalid --pretty format: %s", arg);
172
173 rev->commit_format = commit_format->format;
174 rev->use_terminator = commit_format->is_tformat;
175 if (commit_format->format == CMIT_FMT_USERFORMAT) {
176 save_user_format(rev, commit_format->user_format,
177 commit_format->is_tformat);
178 }
179 }
180
181 /*
182 * Generic support for pretty-printing the header
183 */
184 static int get_one_line(const char *msg)
185 {
186 int ret = 0;
187
188 for (;;) {
189 char c = *msg++;
190 if (!c)
191 break;
192 ret++;
193 if (c == '\n')
194 break;
195 }
196 return ret;
197 }
198
199 /* High bit set, or ISO-2022-INT */
200 static int non_ascii(int ch)
201 {
202 return !isascii(ch) || ch == '\033';
203 }
204
205 int has_non_ascii(const char *s)
206 {
207 int ch;
208 if (!s)
209 return 0;
210 while ((ch = *s++) != '\0') {
211 if (non_ascii(ch))
212 return 1;
213 }
214 return 0;
215 }
216
217 static int is_rfc822_special(char ch)
218 {
219 switch (ch) {
220 case '(':
221 case ')':
222 case '<':
223 case '>':
224 case '[':
225 case ']':
226 case ':':
227 case ';':
228 case '@':
229 case ',':
230 case '.':
231 case '"':
232 case '\\':
233 return 1;
234 default:
235 return 0;
236 }
237 }
238
239 static int needs_rfc822_quoting(const char *s, int len)
240 {
241 int i;
242 for (i = 0; i < len; i++)
243 if (is_rfc822_special(s[i]))
244 return 1;
245 return 0;
246 }
247
248 static int last_line_length(struct strbuf *sb)
249 {
250 int i;
251
252 /* How many bytes are already used on the last line? */
253 for (i = sb->len - 1; i >= 0; i--)
254 if (sb->buf[i] == '\n')
255 break;
256 return sb->len - (i + 1);
257 }
258
259 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
260 {
261 int i;
262
263 /* just a guess, we may have to also backslash-quote */
264 strbuf_grow(out, len + 2);
265
266 strbuf_addch(out, '"');
267 for (i = 0; i < len; i++) {
268 switch (s[i]) {
269 case '"':
270 case '\\':
271 strbuf_addch(out, '\\');
272 /* fall through */
273 default:
274 strbuf_addch(out, s[i]);
275 }
276 }
277 strbuf_addch(out, '"');
278 }
279
280 enum rfc2047_type {
281 RFC2047_SUBJECT,
282 RFC2047_ADDRESS
283 };
284
285 static int is_rfc2047_special(char ch, enum rfc2047_type type)
286 {
287 /*
288 * rfc2047, section 4.2:
289 *
290 * 8-bit values which correspond to printable ASCII characters other
291 * than "=", "?", and "_" (underscore), MAY be represented as those
292 * characters. (But see section 5 for restrictions.) In
293 * particular, SPACE and TAB MUST NOT be represented as themselves
294 * within encoded words.
295 */
296
297 /*
298 * rule out non-ASCII characters and non-printable characters (the
299 * non-ASCII check should be redundant as isprint() is not localized
300 * and only knows about ASCII, but be defensive about that)
301 */
302 if (non_ascii(ch) || !isprint(ch))
303 return 1;
304
305 /*
306 * rule out special printable characters (' ' should be the only
307 * whitespace character considered printable, but be defensive and use
308 * isspace())
309 */
310 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
311 return 1;
312
313 /*
314 * rfc2047, section 5.3:
315 *
316 * As a replacement for a 'word' entity within a 'phrase', for example,
317 * one that precedes an address in a From, To, or Cc header. The ABNF
318 * definition for 'phrase' from RFC 822 thus becomes:
319 *
320 * phrase = 1*( encoded-word / word )
321 *
322 * In this case the set of characters that may be used in a "Q"-encoded
323 * 'encoded-word' is restricted to: <upper and lower case ASCII
324 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
325 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
326 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
327 * 'special' by 'linear-white-space'.
328 */
329
330 if (type != RFC2047_ADDRESS)
331 return 0;
332
333 /* '=' and '_' are special cases and have been checked above */
334 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
335 }
336
337 static int needs_rfc2047_encoding(const char *line, int len,
338 enum rfc2047_type type)
339 {
340 int i;
341
342 for (i = 0; i < len; i++) {
343 int ch = line[i];
344 if (non_ascii(ch) || ch == '\n')
345 return 1;
346 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
347 return 1;
348 }
349
350 return 0;
351 }
352
353 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
354 const char *encoding, enum rfc2047_type type)
355 {
356 static const int max_encoded_length = 76; /* per rfc2047 */
357 int i;
358 int line_len = last_line_length(sb);
359
360 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
361 strbuf_addf(sb, "=?%s?q?", encoding);
362 line_len += strlen(encoding) + 5; /* 5 for =??q? */
363
364 while (len) {
365 /*
366 * RFC 2047, section 5 (3):
367 *
368 * Each 'encoded-word' MUST represent an integral number of
369 * characters. A multi-octet character may not be split across
370 * adjacent 'encoded- word's.
371 */
372 const unsigned char *p = (const unsigned char *)line;
373 int chrlen = mbs_chrlen(&line, &len, encoding);
374 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
375
376 /* "=%02X" * chrlen, or the byte itself */
377 const char *encoded_fmt = is_special ? "=%02X" : "%c";
378 int encoded_len = is_special ? 3 * chrlen : 1;
379
380 /*
381 * According to RFC 2047, we could encode the special character
382 * ' ' (space) with '_' (underscore) for readability. But many
383 * programs do not understand this and just leave the
384 * underscore in place. Thus, we do nothing special here, which
385 * causes ' ' to be encoded as '=20', avoiding this problem.
386 */
387
388 if (line_len + encoded_len + 2 > max_encoded_length) {
389 /* It won't fit with trailing "?=" --- break the line */
390 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
391 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
392 }
393
394 for (i = 0; i < chrlen; i++)
395 strbuf_addf(sb, encoded_fmt, p[i]);
396 line_len += encoded_len;
397 }
398 strbuf_addstr(sb, "?=");
399 }
400
401 const char *show_ident_date(const struct ident_split *ident,
402 enum date_mode mode)
403 {
404 unsigned long date = 0;
405 long tz = 0;
406
407 if (ident->date_begin && ident->date_end)
408 date = strtoul(ident->date_begin, NULL, 10);
409 if (date_overflows(date))
410 date = 0;
411 else {
412 if (ident->tz_begin && ident->tz_end)
413 tz = strtol(ident->tz_begin, NULL, 10);
414 if (tz >= INT_MAX || tz <= INT_MIN)
415 tz = 0;
416 }
417 return show_date(date, tz, mode);
418 }
419
420 void pp_user_info(struct pretty_print_context *pp,
421 const char *what, struct strbuf *sb,
422 const char *line, const char *encoding)
423 {
424 struct ident_split ident;
425 char *line_end;
426 const char *mailbuf, *namebuf;
427 size_t namelen, maillen;
428 int max_length = 78; /* per rfc2822 */
429
430 if (pp->fmt == CMIT_FMT_ONELINE)
431 return;
432
433 line_end = strchrnul(line, '\n');
434 if (split_ident_line(&ident, line, line_end - line))
435 return;
436
437 mailbuf = ident.mail_begin;
438 maillen = ident.mail_end - ident.mail_begin;
439 namebuf = ident.name_begin;
440 namelen = ident.name_end - ident.name_begin;
441
442 if (pp->mailmap)
443 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
444
445 if (pp->fmt == CMIT_FMT_EMAIL) {
446 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
447 struct strbuf buf = STRBUF_INIT;
448
449 strbuf_addstr(&buf, "From: ");
450 strbuf_add(&buf, namebuf, namelen);
451 strbuf_addstr(&buf, " <");
452 strbuf_add(&buf, mailbuf, maillen);
453 strbuf_addstr(&buf, ">\n");
454 string_list_append(&pp->in_body_headers,
455 strbuf_detach(&buf, NULL));
456
457 mailbuf = pp->from_ident->mail_begin;
458 maillen = pp->from_ident->mail_end - mailbuf;
459 namebuf = pp->from_ident->name_begin;
460 namelen = pp->from_ident->name_end - namebuf;
461 }
462
463 strbuf_addstr(sb, "From: ");
464 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
465 add_rfc2047(sb, namebuf, namelen,
466 encoding, RFC2047_ADDRESS);
467 max_length = 76; /* per rfc2047 */
468 } else if (needs_rfc822_quoting(namebuf, namelen)) {
469 struct strbuf quoted = STRBUF_INIT;
470 add_rfc822_quoted(&quoted, namebuf, namelen);
471 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
472 -6, 1, max_length);
473 strbuf_release(&quoted);
474 } else {
475 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
476 -6, 1, max_length);
477 }
478
479 if (max_length <
480 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
481 strbuf_addch(sb, '\n');
482 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
483 } else {
484 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
485 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
486 (int)namelen, namebuf, (int)maillen, mailbuf);
487 }
488
489 switch (pp->fmt) {
490 case CMIT_FMT_MEDIUM:
491 strbuf_addf(sb, "Date: %s\n",
492 show_ident_date(&ident, pp->date_mode));
493 break;
494 case CMIT_FMT_EMAIL:
495 strbuf_addf(sb, "Date: %s\n",
496 show_ident_date(&ident, DATE_RFC2822));
497 break;
498 case CMIT_FMT_FULLER:
499 strbuf_addf(sb, "%sDate: %s\n", what,
500 show_ident_date(&ident, pp->date_mode));
501 break;
502 default:
503 /* notin' */
504 break;
505 }
506 }
507
508 static int is_empty_line(const char *line, int *len_p)
509 {
510 int len = *len_p;
511 while (len && isspace(line[len - 1]))
512 len--;
513 *len_p = len;
514 return !len;
515 }
516
517 static const char *skip_empty_lines(const char *msg)
518 {
519 for (;;) {
520 int linelen = get_one_line(msg);
521 int ll = linelen;
522 if (!linelen)
523 break;
524 if (!is_empty_line(msg, &ll))
525 break;
526 msg += linelen;
527 }
528 return msg;
529 }
530
531 static void add_merge_info(const struct pretty_print_context *pp,
532 struct strbuf *sb, const struct commit *commit)
533 {
534 struct commit_list *parent = commit->parents;
535
536 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
537 !parent || !parent->next)
538 return;
539
540 strbuf_addstr(sb, "Merge:");
541
542 while (parent) {
543 struct commit *p = parent->item;
544 const char *hex = NULL;
545 if (pp->abbrev)
546 hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
547 if (!hex)
548 hex = sha1_to_hex(p->object.sha1);
549 parent = parent->next;
550
551 strbuf_addf(sb, " %s", hex);
552 }
553 strbuf_addch(sb, '\n');
554 }
555
556 static char *get_header(const char *msg, const char *key)
557 {
558 size_t len;
559 const char *v = find_commit_header(msg, key, &len);
560 return v ? xmemdupz(v, len) : NULL;
561 }
562
563 static char *replace_encoding_header(char *buf, const char *encoding)
564 {
565 struct strbuf tmp = STRBUF_INIT;
566 size_t start, len;
567 char *cp = buf;
568
569 /* guess if there is an encoding header before a \n\n */
570 while (!starts_with(cp, "encoding ")) {
571 cp = strchr(cp, '\n');
572 if (!cp || *++cp == '\n')
573 return buf;
574 }
575 start = cp - buf;
576 cp = strchr(cp, '\n');
577 if (!cp)
578 return buf; /* should not happen but be defensive */
579 len = cp + 1 - (buf + start);
580
581 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
582 if (is_encoding_utf8(encoding)) {
583 /* we have re-coded to UTF-8; drop the header */
584 strbuf_remove(&tmp, start, len);
585 } else {
586 /* just replaces XXXX in 'encoding XXXX\n' */
587 strbuf_splice(&tmp, start + strlen("encoding "),
588 len - strlen("encoding \n"),
589 encoding, strlen(encoding));
590 }
591 return strbuf_detach(&tmp, NULL);
592 }
593
594 const char *logmsg_reencode(const struct commit *commit,
595 char **commit_encoding,
596 const char *output_encoding)
597 {
598 static const char *utf8 = "UTF-8";
599 const char *use_encoding;
600 char *encoding;
601 const char *msg = get_commit_buffer(commit, NULL);
602 char *out;
603
604 if (!output_encoding || !*output_encoding) {
605 if (commit_encoding)
606 *commit_encoding = get_header(msg, "encoding");
607 return msg;
608 }
609 encoding = get_header(msg, "encoding");
610 if (commit_encoding)
611 *commit_encoding = encoding;
612 use_encoding = encoding ? encoding : utf8;
613 if (same_encoding(use_encoding, output_encoding)) {
614 /*
615 * No encoding work to be done. If we have no encoding header
616 * at all, then there's nothing to do, and we can return the
617 * message verbatim (whether newly allocated or not).
618 */
619 if (!encoding)
620 return msg;
621
622 /*
623 * Otherwise, we still want to munge the encoding header in the
624 * result, which will be done by modifying the buffer. If we
625 * are using a fresh copy, we can reuse it. But if we are using
626 * the cached copy from get_commit_buffer, we need to duplicate it
627 * to avoid munging the cached copy.
628 */
629 if (msg == get_cached_commit_buffer(commit, NULL))
630 out = xstrdup(msg);
631 else
632 out = (char *)msg;
633 }
634 else {
635 /*
636 * There's actual encoding work to do. Do the reencoding, which
637 * still leaves the header to be replaced in the next step. At
638 * this point, we are done with msg. If we allocated a fresh
639 * copy, we can free it.
640 */
641 out = reencode_string(msg, output_encoding, use_encoding);
642 if (out)
643 unuse_commit_buffer(commit, msg);
644 }
645
646 /*
647 * This replacement actually consumes the buffer we hand it, so we do
648 * not have to worry about freeing the old "out" here.
649 */
650 if (out)
651 out = replace_encoding_header(out, output_encoding);
652
653 if (!commit_encoding)
654 free(encoding);
655 /*
656 * If the re-encoding failed, out might be NULL here; in that
657 * case we just return the commit message verbatim.
658 */
659 return out ? out : msg;
660 }
661
662 static int mailmap_name(const char **email, size_t *email_len,
663 const char **name, size_t *name_len)
664 {
665 static struct string_list *mail_map;
666 if (!mail_map) {
667 mail_map = xcalloc(1, sizeof(*mail_map));
668 read_mailmap(mail_map, NULL);
669 }
670 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
671 }
672
673 static size_t format_person_part(struct strbuf *sb, char part,
674 const char *msg, int len, enum date_mode dmode)
675 {
676 /* currently all placeholders have same length */
677 const int placeholder_len = 2;
678 struct ident_split s;
679 const char *name, *mail;
680 size_t maillen, namelen;
681
682 if (split_ident_line(&s, msg, len) < 0)
683 goto skip;
684
685 name = s.name_begin;
686 namelen = s.name_end - s.name_begin;
687 mail = s.mail_begin;
688 maillen = s.mail_end - s.mail_begin;
689
690 if (part == 'N' || part == 'E') /* mailmap lookup */
691 mailmap_name(&mail, &maillen, &name, &namelen);
692 if (part == 'n' || part == 'N') { /* name */
693 strbuf_add(sb, name, namelen);
694 return placeholder_len;
695 }
696 if (part == 'e' || part == 'E') { /* email */
697 strbuf_add(sb, mail, maillen);
698 return placeholder_len;
699 }
700
701 if (!s.date_begin)
702 goto skip;
703
704 if (part == 't') { /* date, UNIX timestamp */
705 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
706 return placeholder_len;
707 }
708
709 switch (part) {
710 case 'd': /* date */
711 strbuf_addstr(sb, show_ident_date(&s, dmode));
712 return placeholder_len;
713 case 'D': /* date, RFC2822 style */
714 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
715 return placeholder_len;
716 case 'r': /* date, relative */
717 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
718 return placeholder_len;
719 case 'i': /* date, ISO 8601-like */
720 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
721 return placeholder_len;
722 case 'I': /* date, ISO 8601 strict */
723 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601_STRICT));
724 return placeholder_len;
725 }
726
727 skip:
728 /*
729 * reading from either a bogus commit, or a reflog entry with
730 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
731 * to compute a valid return value.
732 */
733 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
734 || part == 'D' || part == 'r' || part == 'i')
735 return placeholder_len;
736
737 return 0; /* unknown placeholder */
738 }
739
740 struct chunk {
741 size_t off;
742 size_t len;
743 };
744
745 enum flush_type {
746 no_flush,
747 flush_right,
748 flush_left,
749 flush_left_and_steal,
750 flush_both
751 };
752
753 enum trunc_type {
754 trunc_none,
755 trunc_left,
756 trunc_middle,
757 trunc_right
758 };
759
760 struct format_commit_context {
761 const struct commit *commit;
762 const struct pretty_print_context *pretty_ctx;
763 unsigned commit_header_parsed:1;
764 unsigned commit_message_parsed:1;
765 struct signature_check signature_check;
766 enum flush_type flush_type;
767 enum trunc_type truncate;
768 const char *message;
769 char *commit_encoding;
770 size_t width, indent1, indent2;
771 int auto_color;
772 int padding;
773
774 /* These offsets are relative to the start of the commit message. */
775 struct chunk author;
776 struct chunk committer;
777 size_t message_off;
778 size_t subject_off;
779 size_t body_off;
780
781 /* The following ones are relative to the result struct strbuf. */
782 struct chunk abbrev_commit_hash;
783 struct chunk abbrev_tree_hash;
784 struct chunk abbrev_parent_hashes;
785 size_t wrap_start;
786 };
787
788 static int add_again(struct strbuf *sb, struct chunk *chunk)
789 {
790 if (chunk->len) {
791 strbuf_adddup(sb, chunk->off, chunk->len);
792 return 1;
793 }
794
795 /*
796 * We haven't seen this chunk before. Our caller is surely
797 * going to add it the hard way now. Remember the most likely
798 * start of the to-be-added chunk: the current end of the
799 * struct strbuf.
800 */
801 chunk->off = sb->len;
802 return 0;
803 }
804
805 static void parse_commit_header(struct format_commit_context *context)
806 {
807 const char *msg = context->message;
808 int i;
809
810 for (i = 0; msg[i]; i++) {
811 const char *name;
812 int eol;
813 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
814 ; /* do nothing */
815
816 if (i == eol) {
817 break;
818 } else if (skip_prefix(msg + i, "author ", &name)) {
819 context->author.off = name - msg;
820 context->author.len = msg + eol - name;
821 } else if (skip_prefix(msg + i, "committer ", &name)) {
822 context->committer.off = name - msg;
823 context->committer.len = msg + eol - name;
824 }
825 i = eol;
826 }
827 context->message_off = i;
828 context->commit_header_parsed = 1;
829 }
830
831 static int istitlechar(char c)
832 {
833 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
834 (c >= '0' && c <= '9') || c == '.' || c == '_';
835 }
836
837 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
838 {
839 size_t trimlen;
840 size_t start_len = sb->len;
841 int space = 2;
842
843 for (; *msg && *msg != '\n'; msg++) {
844 if (istitlechar(*msg)) {
845 if (space == 1)
846 strbuf_addch(sb, '-');
847 space = 0;
848 strbuf_addch(sb, *msg);
849 if (*msg == '.')
850 while (*(msg+1) == '.')
851 msg++;
852 } else
853 space |= 1;
854 }
855
856 /* trim any trailing '.' or '-' characters */
857 trimlen = 0;
858 while (sb->len - trimlen > start_len &&
859 (sb->buf[sb->len - 1 - trimlen] == '.'
860 || sb->buf[sb->len - 1 - trimlen] == '-'))
861 trimlen++;
862 strbuf_remove(sb, sb->len - trimlen, trimlen);
863 }
864
865 const char *format_subject(struct strbuf *sb, const char *msg,
866 const char *line_separator)
867 {
868 int first = 1;
869
870 for (;;) {
871 const char *line = msg;
872 int linelen = get_one_line(line);
873
874 msg += linelen;
875 if (!linelen || is_empty_line(line, &linelen))
876 break;
877
878 if (!sb)
879 continue;
880 strbuf_grow(sb, linelen + 2);
881 if (!first)
882 strbuf_addstr(sb, line_separator);
883 strbuf_add(sb, line, linelen);
884 first = 0;
885 }
886 return msg;
887 }
888
889 static void parse_commit_message(struct format_commit_context *c)
890 {
891 const char *msg = c->message + c->message_off;
892 const char *start = c->message;
893
894 msg = skip_empty_lines(msg);
895 c->subject_off = msg - start;
896
897 msg = format_subject(NULL, msg, NULL);
898 msg = skip_empty_lines(msg);
899 c->body_off = msg - start;
900
901 c->commit_message_parsed = 1;
902 }
903
904 static void strbuf_wrap(struct strbuf *sb, size_t pos,
905 size_t width, size_t indent1, size_t indent2)
906 {
907 struct strbuf tmp = STRBUF_INIT;
908
909 if (pos)
910 strbuf_add(&tmp, sb->buf, pos);
911 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
912 (int) indent1, (int) indent2, (int) width);
913 strbuf_swap(&tmp, sb);
914 strbuf_release(&tmp);
915 }
916
917 static void rewrap_message_tail(struct strbuf *sb,
918 struct format_commit_context *c,
919 size_t new_width, size_t new_indent1,
920 size_t new_indent2)
921 {
922 if (c->width == new_width && c->indent1 == new_indent1 &&
923 c->indent2 == new_indent2)
924 return;
925 if (c->wrap_start < sb->len)
926 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
927 c->wrap_start = sb->len;
928 c->width = new_width;
929 c->indent1 = new_indent1;
930 c->indent2 = new_indent2;
931 }
932
933 static int format_reflog_person(struct strbuf *sb,
934 char part,
935 struct reflog_walk_info *log,
936 enum date_mode dmode)
937 {
938 const char *ident;
939
940 if (!log)
941 return 2;
942
943 ident = get_reflog_ident(log);
944 if (!ident)
945 return 2;
946
947 return format_person_part(sb, part, ident, strlen(ident), dmode);
948 }
949
950 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
951 const char *placeholder,
952 struct format_commit_context *c)
953 {
954 const char *rest = placeholder;
955
956 if (placeholder[1] == '(') {
957 const char *begin = placeholder + 2;
958 const char *end = strchr(begin, ')');
959 char color[COLOR_MAXLEN];
960
961 if (!end)
962 return 0;
963 if (skip_prefix(begin, "auto,", &begin)) {
964 if (!want_color(c->pretty_ctx->color))
965 return end - placeholder + 1;
966 }
967 if (color_parse_mem(begin, end - begin, color) < 0)
968 die(_("unable to parse --pretty format"));
969 strbuf_addstr(sb, color);
970 return end - placeholder + 1;
971 }
972 if (skip_prefix(placeholder + 1, "red", &rest))
973 strbuf_addstr(sb, GIT_COLOR_RED);
974 else if (skip_prefix(placeholder + 1, "green", &rest))
975 strbuf_addstr(sb, GIT_COLOR_GREEN);
976 else if (skip_prefix(placeholder + 1, "blue", &rest))
977 strbuf_addstr(sb, GIT_COLOR_BLUE);
978 else if (skip_prefix(placeholder + 1, "reset", &rest))
979 strbuf_addstr(sb, GIT_COLOR_RESET);
980 return rest - placeholder;
981 }
982
983 static size_t parse_padding_placeholder(struct strbuf *sb,
984 const char *placeholder,
985 struct format_commit_context *c)
986 {
987 const char *ch = placeholder;
988 enum flush_type flush_type;
989 int to_column = 0;
990
991 switch (*ch++) {
992 case '<':
993 flush_type = flush_right;
994 break;
995 case '>':
996 if (*ch == '<') {
997 flush_type = flush_both;
998 ch++;
999 } else if (*ch == '>') {
1000 flush_type = flush_left_and_steal;
1001 ch++;
1002 } else
1003 flush_type = flush_left;
1004 break;
1005 default:
1006 return 0;
1007 }
1008
1009 /* the next value means "wide enough to that column" */
1010 if (*ch == '|') {
1011 to_column = 1;
1012 ch++;
1013 }
1014
1015 if (*ch == '(') {
1016 const char *start = ch + 1;
1017 const char *end = start + strcspn(start, ",)");
1018 char *next;
1019 int width;
1020 if (!end || end == start)
1021 return 0;
1022 width = strtoul(start, &next, 10);
1023 if (next == start || width == 0)
1024 return 0;
1025 c->padding = to_column ? -width : width;
1026 c->flush_type = flush_type;
1027
1028 if (*end == ',') {
1029 start = end + 1;
1030 end = strchr(start, ')');
1031 if (!end || end == start)
1032 return 0;
1033 if (starts_with(start, "trunc)"))
1034 c->truncate = trunc_right;
1035 else if (starts_with(start, "ltrunc)"))
1036 c->truncate = trunc_left;
1037 else if (starts_with(start, "mtrunc)"))
1038 c->truncate = trunc_middle;
1039 else
1040 return 0;
1041 } else
1042 c->truncate = trunc_none;
1043
1044 return end - placeholder + 1;
1045 }
1046 return 0;
1047 }
1048
1049 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1050 const char *placeholder,
1051 void *context)
1052 {
1053 struct format_commit_context *c = context;
1054 const struct commit *commit = c->commit;
1055 const char *msg = c->message;
1056 struct commit_list *p;
1057 int h1, h2;
1058
1059 /* these are independent of the commit */
1060 switch (placeholder[0]) {
1061 case 'C':
1062 if (starts_with(placeholder + 1, "(auto)")) {
1063 c->auto_color = 1;
1064 return 7; /* consumed 7 bytes, "C(auto)" */
1065 } else {
1066 int ret = parse_color(sb, placeholder, c);
1067 if (ret)
1068 c->auto_color = 0;
1069 /*
1070 * Otherwise, we decided to treat %C<unknown>
1071 * as a literal string, and the previous
1072 * %C(auto) is still valid.
1073 */
1074 return ret;
1075 }
1076 case 'n': /* newline */
1077 strbuf_addch(sb, '\n');
1078 return 1;
1079 case 'x':
1080 /* %x00 == NUL, %x0a == LF, etc. */
1081 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1082 h1 <= 16 &&
1083 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1084 h2 <= 16) {
1085 strbuf_addch(sb, (h1<<4)|h2);
1086 return 3;
1087 } else
1088 return 0;
1089 case 'w':
1090 if (placeholder[1] == '(') {
1091 unsigned long width = 0, indent1 = 0, indent2 = 0;
1092 char *next;
1093 const char *start = placeholder + 2;
1094 const char *end = strchr(start, ')');
1095 if (!end)
1096 return 0;
1097 if (end > start) {
1098 width = strtoul(start, &next, 10);
1099 if (*next == ',') {
1100 indent1 = strtoul(next + 1, &next, 10);
1101 if (*next == ',') {
1102 indent2 = strtoul(next + 1,
1103 &next, 10);
1104 }
1105 }
1106 if (*next != ')')
1107 return 0;
1108 }
1109 rewrap_message_tail(sb, c, width, indent1, indent2);
1110 return end - placeholder + 1;
1111 } else
1112 return 0;
1113
1114 case '<':
1115 case '>':
1116 return parse_padding_placeholder(sb, placeholder, c);
1117 }
1118
1119 /* these depend on the commit */
1120 if (!commit->object.parsed)
1121 parse_object(commit->object.sha1);
1122
1123 switch (placeholder[0]) {
1124 case 'H': /* commit hash */
1125 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1126 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1127 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1128 return 1;
1129 case 'h': /* abbreviated commit hash */
1130 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1131 if (add_again(sb, &c->abbrev_commit_hash)) {
1132 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1133 return 1;
1134 }
1135 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1136 c->pretty_ctx->abbrev));
1137 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1138 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1139 return 1;
1140 case 'T': /* tree hash */
1141 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1142 return 1;
1143 case 't': /* abbreviated tree hash */
1144 if (add_again(sb, &c->abbrev_tree_hash))
1145 return 1;
1146 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1147 c->pretty_ctx->abbrev));
1148 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1149 return 1;
1150 case 'P': /* parent hashes */
1151 for (p = commit->parents; p; p = p->next) {
1152 if (p != commit->parents)
1153 strbuf_addch(sb, ' ');
1154 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1155 }
1156 return 1;
1157 case 'p': /* abbreviated parent hashes */
1158 if (add_again(sb, &c->abbrev_parent_hashes))
1159 return 1;
1160 for (p = commit->parents; p; p = p->next) {
1161 if (p != commit->parents)
1162 strbuf_addch(sb, ' ');
1163 strbuf_addstr(sb, find_unique_abbrev(
1164 p->item->object.sha1,
1165 c->pretty_ctx->abbrev));
1166 }
1167 c->abbrev_parent_hashes.len = sb->len -
1168 c->abbrev_parent_hashes.off;
1169 return 1;
1170 case 'm': /* left/right/bottom */
1171 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1172 return 1;
1173 case 'd':
1174 load_ref_decorations(DECORATE_SHORT_REFS);
1175 format_decorations(sb, commit, c->auto_color);
1176 return 1;
1177 case 'D':
1178 load_ref_decorations(DECORATE_SHORT_REFS);
1179 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1180 return 1;
1181 case 'g': /* reflog info */
1182 switch(placeholder[1]) {
1183 case 'd': /* reflog selector */
1184 case 'D':
1185 if (c->pretty_ctx->reflog_info)
1186 get_reflog_selector(sb,
1187 c->pretty_ctx->reflog_info,
1188 c->pretty_ctx->date_mode,
1189 c->pretty_ctx->date_mode_explicit,
1190 (placeholder[1] == 'd'));
1191 return 2;
1192 case 's': /* reflog message */
1193 if (c->pretty_ctx->reflog_info)
1194 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1195 return 2;
1196 case 'n':
1197 case 'N':
1198 case 'e':
1199 case 'E':
1200 return format_reflog_person(sb,
1201 placeholder[1],
1202 c->pretty_ctx->reflog_info,
1203 c->pretty_ctx->date_mode);
1204 }
1205 return 0; /* unknown %g placeholder */
1206 case 'N':
1207 if (c->pretty_ctx->notes_message) {
1208 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1209 return 1;
1210 }
1211 return 0;
1212 }
1213
1214 if (placeholder[0] == 'G') {
1215 if (!c->signature_check.result)
1216 check_commit_signature(c->commit, &(c->signature_check));
1217 switch (placeholder[1]) {
1218 case 'G':
1219 if (c->signature_check.gpg_output)
1220 strbuf_addstr(sb, c->signature_check.gpg_output);
1221 break;
1222 case '?':
1223 switch (c->signature_check.result) {
1224 case 'G':
1225 case 'B':
1226 case 'U':
1227 case 'N':
1228 strbuf_addch(sb, c->signature_check.result);
1229 }
1230 break;
1231 case 'S':
1232 if (c->signature_check.signer)
1233 strbuf_addstr(sb, c->signature_check.signer);
1234 break;
1235 case 'K':
1236 if (c->signature_check.key)
1237 strbuf_addstr(sb, c->signature_check.key);
1238 break;
1239 default:
1240 return 0;
1241 }
1242 return 2;
1243 }
1244
1245
1246 /* For the rest we have to parse the commit header. */
1247 if (!c->commit_header_parsed)
1248 parse_commit_header(c);
1249
1250 switch (placeholder[0]) {
1251 case 'a': /* author ... */
1252 return format_person_part(sb, placeholder[1],
1253 msg + c->author.off, c->author.len,
1254 c->pretty_ctx->date_mode);
1255 case 'c': /* committer ... */
1256 return format_person_part(sb, placeholder[1],
1257 msg + c->committer.off, c->committer.len,
1258 c->pretty_ctx->date_mode);
1259 case 'e': /* encoding */
1260 if (c->commit_encoding)
1261 strbuf_addstr(sb, c->commit_encoding);
1262 return 1;
1263 case 'B': /* raw body */
1264 /* message_off is always left at the initial newline */
1265 strbuf_addstr(sb, msg + c->message_off + 1);
1266 return 1;
1267 }
1268
1269 /* Now we need to parse the commit message. */
1270 if (!c->commit_message_parsed)
1271 parse_commit_message(c);
1272
1273 switch (placeholder[0]) {
1274 case 's': /* subject */
1275 format_subject(sb, msg + c->subject_off, " ");
1276 return 1;
1277 case 'f': /* sanitized subject */
1278 format_sanitized_subject(sb, msg + c->subject_off);
1279 return 1;
1280 case 'b': /* body */
1281 strbuf_addstr(sb, msg + c->body_off);
1282 return 1;
1283 }
1284 return 0; /* unknown placeholder */
1285 }
1286
1287 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1288 const char *placeholder,
1289 struct format_commit_context *c)
1290 {
1291 struct strbuf local_sb = STRBUF_INIT;
1292 int total_consumed = 0, len, padding = c->padding;
1293 if (padding < 0) {
1294 const char *start = strrchr(sb->buf, '\n');
1295 int occupied;
1296 if (!start)
1297 start = sb->buf;
1298 occupied = utf8_strnwidth(start, -1, 1);
1299 padding = (-padding) - occupied;
1300 }
1301 while (1) {
1302 int modifier = *placeholder == 'C';
1303 int consumed = format_commit_one(&local_sb, placeholder, c);
1304 total_consumed += consumed;
1305
1306 if (!modifier)
1307 break;
1308
1309 placeholder += consumed;
1310 if (*placeholder != '%')
1311 break;
1312 placeholder++;
1313 total_consumed++;
1314 }
1315 len = utf8_strnwidth(local_sb.buf, -1, 1);
1316
1317 if (c->flush_type == flush_left_and_steal) {
1318 const char *ch = sb->buf + sb->len - 1;
1319 while (len > padding && ch > sb->buf) {
1320 const char *p;
1321 if (*ch == ' ') {
1322 ch--;
1323 padding++;
1324 continue;
1325 }
1326 /* check for trailing ansi sequences */
1327 if (*ch != 'm')
1328 break;
1329 p = ch - 1;
1330 while (ch - p < 10 && *p != '\033')
1331 p--;
1332 if (*p != '\033' ||
1333 ch + 1 - p != display_mode_esc_sequence_len(p))
1334 break;
1335 /*
1336 * got a good ansi sequence, put it back to
1337 * local_sb as we're cutting sb
1338 */
1339 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1340 ch = p - 1;
1341 }
1342 strbuf_setlen(sb, ch + 1 - sb->buf);
1343 c->flush_type = flush_left;
1344 }
1345
1346 if (len > padding) {
1347 switch (c->truncate) {
1348 case trunc_left:
1349 strbuf_utf8_replace(&local_sb,
1350 0, len - (padding - 2),
1351 "..");
1352 break;
1353 case trunc_middle:
1354 strbuf_utf8_replace(&local_sb,
1355 padding / 2 - 1,
1356 len - (padding - 2),
1357 "..");
1358 break;
1359 case trunc_right:
1360 strbuf_utf8_replace(&local_sb,
1361 padding - 2, len - (padding - 2),
1362 "..");
1363 break;
1364 case trunc_none:
1365 break;
1366 }
1367 strbuf_addbuf(sb, &local_sb);
1368 } else {
1369 int sb_len = sb->len, offset = 0;
1370 if (c->flush_type == flush_left)
1371 offset = padding - len;
1372 else if (c->flush_type == flush_both)
1373 offset = (padding - len) / 2;
1374 /*
1375 * we calculate padding in columns, now
1376 * convert it back to chars
1377 */
1378 padding = padding - len + local_sb.len;
1379 strbuf_addchars(sb, ' ', padding);
1380 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1381 local_sb.len);
1382 }
1383 strbuf_release(&local_sb);
1384 c->flush_type = no_flush;
1385 return total_consumed;
1386 }
1387
1388 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1389 const char *placeholder,
1390 void *context)
1391 {
1392 int consumed;
1393 size_t orig_len;
1394 enum {
1395 NO_MAGIC,
1396 ADD_LF_BEFORE_NON_EMPTY,
1397 DEL_LF_BEFORE_EMPTY,
1398 ADD_SP_BEFORE_NON_EMPTY
1399 } magic = NO_MAGIC;
1400
1401 switch (placeholder[0]) {
1402 case '-':
1403 magic = DEL_LF_BEFORE_EMPTY;
1404 break;
1405 case '+':
1406 magic = ADD_LF_BEFORE_NON_EMPTY;
1407 break;
1408 case ' ':
1409 magic = ADD_SP_BEFORE_NON_EMPTY;
1410 break;
1411 default:
1412 break;
1413 }
1414 if (magic != NO_MAGIC)
1415 placeholder++;
1416
1417 orig_len = sb->len;
1418 if (((struct format_commit_context *)context)->flush_type != no_flush)
1419 consumed = format_and_pad_commit(sb, placeholder, context);
1420 else
1421 consumed = format_commit_one(sb, placeholder, context);
1422 if (magic == NO_MAGIC)
1423 return consumed;
1424
1425 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1426 while (sb->len && sb->buf[sb->len - 1] == '\n')
1427 strbuf_setlen(sb, sb->len - 1);
1428 } else if (orig_len != sb->len) {
1429 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1430 strbuf_insert(sb, orig_len, "\n", 1);
1431 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1432 strbuf_insert(sb, orig_len, " ", 1);
1433 }
1434 return consumed + 1;
1435 }
1436
1437 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1438 void *context)
1439 {
1440 struct userformat_want *w = context;
1441
1442 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1443 placeholder++;
1444
1445 switch (*placeholder) {
1446 case 'N':
1447 w->notes = 1;
1448 break;
1449 }
1450 return 0;
1451 }
1452
1453 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1454 {
1455 struct strbuf dummy = STRBUF_INIT;
1456
1457 if (!fmt) {
1458 if (!user_format)
1459 return;
1460 fmt = user_format;
1461 }
1462 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1463 strbuf_release(&dummy);
1464 }
1465
1466 void format_commit_message(const struct commit *commit,
1467 const char *format, struct strbuf *sb,
1468 const struct pretty_print_context *pretty_ctx)
1469 {
1470 struct format_commit_context context;
1471 const char *output_enc = pretty_ctx->output_encoding;
1472 const char *utf8 = "UTF-8";
1473
1474 memset(&context, 0, sizeof(context));
1475 context.commit = commit;
1476 context.pretty_ctx = pretty_ctx;
1477 context.wrap_start = sb->len;
1478 /*
1479 * convert a commit message to UTF-8 first
1480 * as far as 'format_commit_item' assumes it in UTF-8
1481 */
1482 context.message = logmsg_reencode(commit,
1483 &context.commit_encoding,
1484 utf8);
1485
1486 strbuf_expand(sb, format, format_commit_item, &context);
1487 rewrap_message_tail(sb, &context, 0, 0, 0);
1488
1489 /* then convert a commit message to an actual output encoding */
1490 if (output_enc) {
1491 if (same_encoding(utf8, output_enc))
1492 output_enc = NULL;
1493 } else {
1494 if (context.commit_encoding &&
1495 !same_encoding(context.commit_encoding, utf8))
1496 output_enc = context.commit_encoding;
1497 }
1498
1499 if (output_enc) {
1500 int outsz;
1501 char *out = reencode_string_len(sb->buf, sb->len,
1502 output_enc, utf8, &outsz);
1503 if (out)
1504 strbuf_attach(sb, out, outsz, outsz + 1);
1505 }
1506
1507 free(context.commit_encoding);
1508 unuse_commit_buffer(commit, context.message);
1509 }
1510
1511 static void pp_header(struct pretty_print_context *pp,
1512 const char *encoding,
1513 const struct commit *commit,
1514 const char **msg_p,
1515 struct strbuf *sb)
1516 {
1517 int parents_shown = 0;
1518
1519 for (;;) {
1520 const char *name, *line = *msg_p;
1521 int linelen = get_one_line(*msg_p);
1522
1523 if (!linelen)
1524 return;
1525 *msg_p += linelen;
1526
1527 if (linelen == 1)
1528 /* End of header */
1529 return;
1530
1531 if (pp->fmt == CMIT_FMT_RAW) {
1532 strbuf_add(sb, line, linelen);
1533 continue;
1534 }
1535
1536 if (starts_with(line, "parent ")) {
1537 if (linelen != 48)
1538 die("bad parent line in commit");
1539 continue;
1540 }
1541
1542 if (!parents_shown) {
1543 unsigned num = commit_list_count(commit->parents);
1544 /* with enough slop */
1545 strbuf_grow(sb, num * 50 + 20);
1546 add_merge_info(pp, sb, commit);
1547 parents_shown = 1;
1548 }
1549
1550 /*
1551 * MEDIUM == DEFAULT shows only author with dates.
1552 * FULL shows both authors but not dates.
1553 * FULLER shows both authors and dates.
1554 */
1555 if (skip_prefix(line, "author ", &name)) {
1556 strbuf_grow(sb, linelen + 80);
1557 pp_user_info(pp, "Author", sb, name, encoding);
1558 }
1559 if (skip_prefix(line, "committer ", &name) &&
1560 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1561 strbuf_grow(sb, linelen + 80);
1562 pp_user_info(pp, "Commit", sb, name, encoding);
1563 }
1564 }
1565 }
1566
1567 void pp_title_line(struct pretty_print_context *pp,
1568 const char **msg_p,
1569 struct strbuf *sb,
1570 const char *encoding,
1571 int need_8bit_cte)
1572 {
1573 static const int max_length = 78; /* per rfc2047 */
1574 struct strbuf title;
1575
1576 strbuf_init(&title, 80);
1577 *msg_p = format_subject(&title, *msg_p,
1578 pp->preserve_subject ? "\n" : " ");
1579
1580 strbuf_grow(sb, title.len + 1024);
1581 if (pp->subject) {
1582 strbuf_addstr(sb, pp->subject);
1583 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1584 add_rfc2047(sb, title.buf, title.len,
1585 encoding, RFC2047_SUBJECT);
1586 else
1587 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1588 -last_line_length(sb), 1, max_length);
1589 } else {
1590 strbuf_addbuf(sb, &title);
1591 }
1592 strbuf_addch(sb, '\n');
1593
1594 if (need_8bit_cte == 0) {
1595 int i;
1596 for (i = 0; i < pp->in_body_headers.nr; i++) {
1597 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1598 need_8bit_cte = 1;
1599 break;
1600 }
1601 }
1602 }
1603
1604 if (need_8bit_cte > 0) {
1605 const char *header_fmt =
1606 "MIME-Version: 1.0\n"
1607 "Content-Type: text/plain; charset=%s\n"
1608 "Content-Transfer-Encoding: 8bit\n";
1609 strbuf_addf(sb, header_fmt, encoding);
1610 }
1611 if (pp->after_subject) {
1612 strbuf_addstr(sb, pp->after_subject);
1613 }
1614 if (pp->fmt == CMIT_FMT_EMAIL) {
1615 strbuf_addch(sb, '\n');
1616 }
1617
1618 if (pp->in_body_headers.nr) {
1619 int i;
1620 for (i = 0; i < pp->in_body_headers.nr; i++) {
1621 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1622 free(pp->in_body_headers.items[i].string);
1623 }
1624 string_list_clear(&pp->in_body_headers, 0);
1625 strbuf_addch(sb, '\n');
1626 }
1627
1628 strbuf_release(&title);
1629 }
1630
1631 void pp_remainder(struct pretty_print_context *pp,
1632 const char **msg_p,
1633 struct strbuf *sb,
1634 int indent)
1635 {
1636 int first = 1;
1637 for (;;) {
1638 const char *line = *msg_p;
1639 int linelen = get_one_line(line);
1640 *msg_p += linelen;
1641
1642 if (!linelen)
1643 break;
1644
1645 if (is_empty_line(line, &linelen)) {
1646 if (first)
1647 continue;
1648 if (pp->fmt == CMIT_FMT_SHORT)
1649 break;
1650 }
1651 first = 0;
1652
1653 strbuf_grow(sb, linelen + indent + 20);
1654 if (indent)
1655 strbuf_addchars(sb, ' ', indent);
1656 strbuf_add(sb, line, linelen);
1657 strbuf_addch(sb, '\n');
1658 }
1659 }
1660
1661 void pretty_print_commit(struct pretty_print_context *pp,
1662 const struct commit *commit,
1663 struct strbuf *sb)
1664 {
1665 unsigned long beginning_of_body;
1666 int indent = 4;
1667 const char *msg;
1668 const char *reencoded;
1669 const char *encoding;
1670 int need_8bit_cte = pp->need_8bit_cte;
1671
1672 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1673 format_commit_message(commit, user_format, sb, pp);
1674 return;
1675 }
1676
1677 encoding = get_log_output_encoding();
1678 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1679
1680 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1681 indent = 0;
1682
1683 /*
1684 * We need to check and emit Content-type: to mark it
1685 * as 8-bit if we haven't done so.
1686 */
1687 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1688 int i, ch, in_body;
1689
1690 for (in_body = i = 0; (ch = msg[i]); i++) {
1691 if (!in_body) {
1692 /* author could be non 7-bit ASCII but
1693 * the log may be so; skip over the
1694 * header part first.
1695 */
1696 if (ch == '\n' && msg[i+1] == '\n')
1697 in_body = 1;
1698 }
1699 else if (non_ascii(ch)) {
1700 need_8bit_cte = 1;
1701 break;
1702 }
1703 }
1704 }
1705
1706 pp_header(pp, encoding, commit, &msg, sb);
1707 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1708 strbuf_addch(sb, '\n');
1709 }
1710
1711 /* Skip excess blank lines at the beginning of body, if any... */
1712 msg = skip_empty_lines(msg);
1713
1714 /* These formats treat the title line specially. */
1715 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1716 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1717
1718 beginning_of_body = sb->len;
1719 if (pp->fmt != CMIT_FMT_ONELINE)
1720 pp_remainder(pp, &msg, sb, indent);
1721 strbuf_rtrim(sb);
1722
1723 /* Make sure there is an EOLN for the non-oneline case */
1724 if (pp->fmt != CMIT_FMT_ONELINE)
1725 strbuf_addch(sb, '\n');
1726
1727 /*
1728 * The caller may append additional body text in e-mail
1729 * format. Make sure we did not strip the blank line
1730 * between the header and the body.
1731 */
1732 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1733 strbuf_addch(sb, '\n');
1734
1735 unuse_commit_buffer(commit, reencoded);
1736 }
1737
1738 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1739 struct strbuf *sb)
1740 {
1741 struct pretty_print_context pp = {0};
1742 pp.fmt = fmt;
1743 pretty_print_commit(&pp, commit, sb);
1744 }
The Best VCS Around