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
14static char *user_format;
15static 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;
22static size_t builtin_formats_len;
23static size_t commit_formats_len;
24static size_t commit_formats_alloc;
25static struct cmt_fmt_map *find_commit_format(const char *sought);
26
27static 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
36static 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
81static 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
101static 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
136static 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
144void 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 */
178static 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 */
194static int non_ascii(int ch)
195{
196 return !isascii(ch) || ch == '\033';
197}
198
199int 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
211static 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
233static 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
242static 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
253static 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
274enum rfc2047_type {
275 RFC2047_SUBJECT,
276 RFC2047_ADDRESS
277};
278
279static 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
331static 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
347static 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
395const 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
414void 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("ed, namebuf, namelen);
465 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
466 -6, 1, max_length);
467 strbuf_release("ed);
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
502static 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
511static 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
525static 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
550static 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
577static 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
608const 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
677static 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
688static 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-like */
735 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
736 return placeholder_len;
737 case 'I': /* date, ISO 8601 strict */
738 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601_STRICT));
739 return placeholder_len;
740 }
741
742skip:
743 /*
744 * reading from either a bogus commit, or a reflog entry with
745 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
746 * to compute a valid return value.
747 */
748 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
749 || part == 'D' || part == 'r' || part == 'i')
750 return placeholder_len;
751
752 return 0; /* unknown placeholder */
753}
754
755struct chunk {
756 size_t off;
757 size_t len;
758};
759
760enum flush_type {
761 no_flush,
762 flush_right,
763 flush_left,
764 flush_left_and_steal,
765 flush_both
766};
767
768enum trunc_type {
769 trunc_none,
770 trunc_left,
771 trunc_middle,
772 trunc_right
773};
774
775struct format_commit_context {
776 const struct commit *commit;
777 const struct pretty_print_context *pretty_ctx;
778 unsigned commit_header_parsed:1;
779 unsigned commit_message_parsed:1;
780 struct signature_check signature_check;
781 enum flush_type flush_type;
782 enum trunc_type truncate;
783 const char *message;
784 char *commit_encoding;
785 size_t width, indent1, indent2;
786 int auto_color;
787 int padding;
788
789 /* These offsets are relative to the start of the commit message. */
790 struct chunk author;
791 struct chunk committer;
792 size_t message_off;
793 size_t subject_off;
794 size_t body_off;
795
796 /* The following ones are relative to the result struct strbuf. */
797 struct chunk abbrev_commit_hash;
798 struct chunk abbrev_tree_hash;
799 struct chunk abbrev_parent_hashes;
800 size_t wrap_start;
801};
802
803static int add_again(struct strbuf *sb, struct chunk *chunk)
804{
805 if (chunk->len) {
806 strbuf_adddup(sb, chunk->off, chunk->len);
807 return 1;
808 }
809
810 /*
811 * We haven't seen this chunk before. Our caller is surely
812 * going to add it the hard way now. Remember the most likely
813 * start of the to-be-added chunk: the current end of the
814 * struct strbuf.
815 */
816 chunk->off = sb->len;
817 return 0;
818}
819
820static void parse_commit_header(struct format_commit_context *context)
821{
822 const char *msg = context->message;
823 int i;
824
825 for (i = 0; msg[i]; i++) {
826 int eol;
827 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
828 ; /* do nothing */
829
830 if (i == eol) {
831 break;
832 } else if (starts_with(msg + i, "author ")) {
833 context->author.off = i + 7;
834 context->author.len = eol - i - 7;
835 } else if (starts_with(msg + i, "committer ")) {
836 context->committer.off = i + 10;
837 context->committer.len = eol - i - 10;
838 }
839 i = eol;
840 }
841 context->message_off = i;
842 context->commit_header_parsed = 1;
843}
844
845static int istitlechar(char c)
846{
847 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
848 (c >= '0' && c <= '9') || c == '.' || c == '_';
849}
850
851static void format_sanitized_subject(struct strbuf *sb, const char *msg)
852{
853 size_t trimlen;
854 size_t start_len = sb->len;
855 int space = 2;
856
857 for (; *msg && *msg != '\n'; msg++) {
858 if (istitlechar(*msg)) {
859 if (space == 1)
860 strbuf_addch(sb, '-');
861 space = 0;
862 strbuf_addch(sb, *msg);
863 if (*msg == '.')
864 while (*(msg+1) == '.')
865 msg++;
866 } else
867 space |= 1;
868 }
869
870 /* trim any trailing '.' or '-' characters */
871 trimlen = 0;
872 while (sb->len - trimlen > start_len &&
873 (sb->buf[sb->len - 1 - trimlen] == '.'
874 || sb->buf[sb->len - 1 - trimlen] == '-'))
875 trimlen++;
876 strbuf_remove(sb, sb->len - trimlen, trimlen);
877}
878
879const char *format_subject(struct strbuf *sb, const char *msg,
880 const char *line_separator)
881{
882 int first = 1;
883
884 for (;;) {
885 const char *line = msg;
886 int linelen = get_one_line(line);
887
888 msg += linelen;
889 if (!linelen || is_empty_line(line, &linelen))
890 break;
891
892 if (!sb)
893 continue;
894 strbuf_grow(sb, linelen + 2);
895 if (!first)
896 strbuf_addstr(sb, line_separator);
897 strbuf_add(sb, line, linelen);
898 first = 0;
899 }
900 return msg;
901}
902
903static void parse_commit_message(struct format_commit_context *c)
904{
905 const char *msg = c->message + c->message_off;
906 const char *start = c->message;
907
908 msg = skip_empty_lines(msg);
909 c->subject_off = msg - start;
910
911 msg = format_subject(NULL, msg, NULL);
912 msg = skip_empty_lines(msg);
913 c->body_off = msg - start;
914
915 c->commit_message_parsed = 1;
916}
917
918static void strbuf_wrap(struct strbuf *sb, size_t pos,
919 size_t width, size_t indent1, size_t indent2)
920{
921 struct strbuf tmp = STRBUF_INIT;
922
923 if (pos)
924 strbuf_add(&tmp, sb->buf, pos);
925 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
926 (int) indent1, (int) indent2, (int) width);
927 strbuf_swap(&tmp, sb);
928 strbuf_release(&tmp);
929}
930
931static void rewrap_message_tail(struct strbuf *sb,
932 struct format_commit_context *c,
933 size_t new_width, size_t new_indent1,
934 size_t new_indent2)
935{
936 if (c->width == new_width && c->indent1 == new_indent1 &&
937 c->indent2 == new_indent2)
938 return;
939 if (c->wrap_start < sb->len)
940 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
941 c->wrap_start = sb->len;
942 c->width = new_width;
943 c->indent1 = new_indent1;
944 c->indent2 = new_indent2;
945}
946
947static int format_reflog_person(struct strbuf *sb,
948 char part,
949 struct reflog_walk_info *log,
950 enum date_mode dmode)
951{
952 const char *ident;
953
954 if (!log)
955 return 2;
956
957 ident = get_reflog_ident(log);
958 if (!ident)
959 return 2;
960
961 return format_person_part(sb, part, ident, strlen(ident), dmode);
962}
963
964static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
965 const char *placeholder,
966 struct format_commit_context *c)
967{
968 if (placeholder[1] == '(') {
969 const char *begin = placeholder + 2;
970 const char *end = strchr(begin, ')');
971 char color[COLOR_MAXLEN];
972
973 if (!end)
974 return 0;
975 if (starts_with(begin, "auto,")) {
976 if (!want_color(c->pretty_ctx->color))
977 return end - placeholder + 1;
978 begin += 5;
979 }
980 color_parse_mem(begin,
981 end - begin,
982 "--pretty format", color);
983 strbuf_addstr(sb, color);
984 return end - placeholder + 1;
985 }
986 if (starts_with(placeholder + 1, "red")) {
987 strbuf_addstr(sb, GIT_COLOR_RED);
988 return 4;
989 } else if (starts_with(placeholder + 1, "green")) {
990 strbuf_addstr(sb, GIT_COLOR_GREEN);
991 return 6;
992 } else if (starts_with(placeholder + 1, "blue")) {
993 strbuf_addstr(sb, GIT_COLOR_BLUE);
994 return 5;
995 } else if (starts_with(placeholder + 1, "reset")) {
996 strbuf_addstr(sb, GIT_COLOR_RESET);
997 return 6;
998 } else
999 return 0;
1000}
1001
1002static size_t parse_padding_placeholder(struct strbuf *sb,
1003 const char *placeholder,
1004 struct format_commit_context *c)
1005{
1006 const char *ch = placeholder;
1007 enum flush_type flush_type;
1008 int to_column = 0;
1009
1010 switch (*ch++) {
1011 case '<':
1012 flush_type = flush_right;
1013 break;
1014 case '>':
1015 if (*ch == '<') {
1016 flush_type = flush_both;
1017 ch++;
1018 } else if (*ch == '>') {
1019 flush_type = flush_left_and_steal;
1020 ch++;
1021 } else
1022 flush_type = flush_left;
1023 break;
1024 default:
1025 return 0;
1026 }
1027
1028 /* the next value means "wide enough to that column" */
1029 if (*ch == '|') {
1030 to_column = 1;
1031 ch++;
1032 }
1033
1034 if (*ch == '(') {
1035 const char *start = ch + 1;
1036 const char *end = start + strcspn(start, ",)");
1037 char *next;
1038 int width;
1039 if (!end || end == start)
1040 return 0;
1041 width = strtoul(start, &next, 10);
1042 if (next == start || width == 0)
1043 return 0;
1044 c->padding = to_column ? -width : width;
1045 c->flush_type = flush_type;
1046
1047 if (*end == ',') {
1048 start = end + 1;
1049 end = strchr(start, ')');
1050 if (!end || end == start)
1051 return 0;
1052 if (starts_with(start, "trunc)"))
1053 c->truncate = trunc_right;
1054 else if (starts_with(start, "ltrunc)"))
1055 c->truncate = trunc_left;
1056 else if (starts_with(start, "mtrunc)"))
1057 c->truncate = trunc_middle;
1058 else
1059 return 0;
1060 } else
1061 c->truncate = trunc_none;
1062
1063 return end - placeholder + 1;
1064 }
1065 return 0;
1066}
1067
1068static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1069 const char *placeholder,
1070 void *context)
1071{
1072 struct format_commit_context *c = context;
1073 const struct commit *commit = c->commit;
1074 const char *msg = c->message;
1075 struct commit_list *p;
1076 int h1, h2;
1077
1078 /* these are independent of the commit */
1079 switch (placeholder[0]) {
1080 case 'C':
1081 if (starts_with(placeholder + 1, "(auto)")) {
1082 c->auto_color = 1;
1083 return 7; /* consumed 7 bytes, "C(auto)" */
1084 } else {
1085 int ret = parse_color(sb, placeholder, c);
1086 if (ret)
1087 c->auto_color = 0;
1088 /*
1089 * Otherwise, we decided to treat %C<unknown>
1090 * as a literal string, and the previous
1091 * %C(auto) is still valid.
1092 */
1093 return ret;
1094 }
1095 case 'n': /* newline */
1096 strbuf_addch(sb, '\n');
1097 return 1;
1098 case 'x':
1099 /* %x00 == NUL, %x0a == LF, etc. */
1100 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1101 h1 <= 16 &&
1102 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1103 h2 <= 16) {
1104 strbuf_addch(sb, (h1<<4)|h2);
1105 return 3;
1106 } else
1107 return 0;
1108 case 'w':
1109 if (placeholder[1] == '(') {
1110 unsigned long width = 0, indent1 = 0, indent2 = 0;
1111 char *next;
1112 const char *start = placeholder + 2;
1113 const char *end = strchr(start, ')');
1114 if (!end)
1115 return 0;
1116 if (end > start) {
1117 width = strtoul(start, &next, 10);
1118 if (*next == ',') {
1119 indent1 = strtoul(next + 1, &next, 10);
1120 if (*next == ',') {
1121 indent2 = strtoul(next + 1,
1122 &next, 10);
1123 }
1124 }
1125 if (*next != ')')
1126 return 0;
1127 }
1128 rewrap_message_tail(sb, c, width, indent1, indent2);
1129 return end - placeholder + 1;
1130 } else
1131 return 0;
1132
1133 case '<':
1134 case '>':
1135 return parse_padding_placeholder(sb, placeholder, c);
1136 }
1137
1138 /* these depend on the commit */
1139 if (!commit->object.parsed)
1140 parse_object(commit->object.sha1);
1141
1142 switch (placeholder[0]) {
1143 case 'H': /* commit hash */
1144 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1145 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1146 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1147 return 1;
1148 case 'h': /* abbreviated commit hash */
1149 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1150 if (add_again(sb, &c->abbrev_commit_hash)) {
1151 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1152 return 1;
1153 }
1154 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1155 c->pretty_ctx->abbrev));
1156 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1157 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1158 return 1;
1159 case 'T': /* tree hash */
1160 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1161 return 1;
1162 case 't': /* abbreviated tree hash */
1163 if (add_again(sb, &c->abbrev_tree_hash))
1164 return 1;
1165 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1166 c->pretty_ctx->abbrev));
1167 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1168 return 1;
1169 case 'P': /* parent hashes */
1170 for (p = commit->parents; p; p = p->next) {
1171 if (p != commit->parents)
1172 strbuf_addch(sb, ' ');
1173 strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1174 }
1175 return 1;
1176 case 'p': /* abbreviated parent hashes */
1177 if (add_again(sb, &c->abbrev_parent_hashes))
1178 return 1;
1179 for (p = commit->parents; p; p = p->next) {
1180 if (p != commit->parents)
1181 strbuf_addch(sb, ' ');
1182 strbuf_addstr(sb, find_unique_abbrev(
1183 p->item->object.sha1,
1184 c->pretty_ctx->abbrev));
1185 }
1186 c->abbrev_parent_hashes.len = sb->len -
1187 c->abbrev_parent_hashes.off;
1188 return 1;
1189 case 'm': /* left/right/bottom */
1190 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1191 return 1;
1192 case 'd':
1193 load_ref_decorations(DECORATE_SHORT_REFS);
1194 format_decorations(sb, commit, c->auto_color);
1195 return 1;
1196 case 'g': /* reflog info */
1197 switch(placeholder[1]) {
1198 case 'd': /* reflog selector */
1199 case 'D':
1200 if (c->pretty_ctx->reflog_info)
1201 get_reflog_selector(sb,
1202 c->pretty_ctx->reflog_info,
1203 c->pretty_ctx->date_mode,
1204 c->pretty_ctx->date_mode_explicit,
1205 (placeholder[1] == 'd'));
1206 return 2;
1207 case 's': /* reflog message */
1208 if (c->pretty_ctx->reflog_info)
1209 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1210 return 2;
1211 case 'n':
1212 case 'N':
1213 case 'e':
1214 case 'E':
1215 return format_reflog_person(sb,
1216 placeholder[1],
1217 c->pretty_ctx->reflog_info,
1218 c->pretty_ctx->date_mode);
1219 }
1220 return 0; /* unknown %g placeholder */
1221 case 'N':
1222 if (c->pretty_ctx->notes_message) {
1223 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1224 return 1;
1225 }
1226 return 0;
1227 }
1228
1229 if (placeholder[0] == 'G') {
1230 if (!c->signature_check.result)
1231 check_commit_signature(c->commit, &(c->signature_check));
1232 switch (placeholder[1]) {
1233 case 'G':
1234 if (c->signature_check.gpg_output)
1235 strbuf_addstr(sb, c->signature_check.gpg_output);
1236 break;
1237 case '?':
1238 switch (c->signature_check.result) {
1239 case 'G':
1240 case 'B':
1241 case 'U':
1242 case 'N':
1243 strbuf_addch(sb, c->signature_check.result);
1244 }
1245 break;
1246 case 'S':
1247 if (c->signature_check.signer)
1248 strbuf_addstr(sb, c->signature_check.signer);
1249 break;
1250 case 'K':
1251 if (c->signature_check.key)
1252 strbuf_addstr(sb, c->signature_check.key);
1253 break;
1254 default:
1255 return 0;
1256 }
1257 return 2;
1258 }
1259
1260
1261 /* For the rest we have to parse the commit header. */
1262 if (!c->commit_header_parsed)
1263 parse_commit_header(c);
1264
1265 switch (placeholder[0]) {
1266 case 'a': /* author ... */
1267 return format_person_part(sb, placeholder[1],
1268 msg + c->author.off, c->author.len,
1269 c->pretty_ctx->date_mode);
1270 case 'c': /* committer ... */
1271 return format_person_part(sb, placeholder[1],
1272 msg + c->committer.off, c->committer.len,
1273 c->pretty_ctx->date_mode);
1274 case 'e': /* encoding */
1275 if (c->commit_encoding)
1276 strbuf_addstr(sb, c->commit_encoding);
1277 return 1;
1278 case 'B': /* raw body */
1279 /* message_off is always left at the initial newline */
1280 strbuf_addstr(sb, msg + c->message_off + 1);
1281 return 1;
1282 }
1283
1284 /* Now we need to parse the commit message. */
1285 if (!c->commit_message_parsed)
1286 parse_commit_message(c);
1287
1288 switch (placeholder[0]) {
1289 case 's': /* subject */
1290 format_subject(sb, msg + c->subject_off, " ");
1291 return 1;
1292 case 'f': /* sanitized subject */
1293 format_sanitized_subject(sb, msg + c->subject_off);
1294 return 1;
1295 case 'b': /* body */
1296 strbuf_addstr(sb, msg + c->body_off);
1297 return 1;
1298 }
1299 return 0; /* unknown placeholder */
1300}
1301
1302static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1303 const char *placeholder,
1304 struct format_commit_context *c)
1305{
1306 struct strbuf local_sb = STRBUF_INIT;
1307 int total_consumed = 0, len, padding = c->padding;
1308 if (padding < 0) {
1309 const char *start = strrchr(sb->buf, '\n');
1310 int occupied;
1311 if (!start)
1312 start = sb->buf;
1313 occupied = utf8_strnwidth(start, -1, 1);
1314 padding = (-padding) - occupied;
1315 }
1316 while (1) {
1317 int modifier = *placeholder == 'C';
1318 int consumed = format_commit_one(&local_sb, placeholder, c);
1319 total_consumed += consumed;
1320
1321 if (!modifier)
1322 break;
1323
1324 placeholder += consumed;
1325 if (*placeholder != '%')
1326 break;
1327 placeholder++;
1328 total_consumed++;
1329 }
1330 len = utf8_strnwidth(local_sb.buf, -1, 1);
1331
1332 if (c->flush_type == flush_left_and_steal) {
1333 const char *ch = sb->buf + sb->len - 1;
1334 while (len > padding && ch > sb->buf) {
1335 const char *p;
1336 if (*ch == ' ') {
1337 ch--;
1338 padding++;
1339 continue;
1340 }
1341 /* check for trailing ansi sequences */
1342 if (*ch != 'm')
1343 break;
1344 p = ch - 1;
1345 while (ch - p < 10 && *p != '\033')
1346 p--;
1347 if (*p != '\033' ||
1348 ch + 1 - p != display_mode_esc_sequence_len(p))
1349 break;
1350 /*
1351 * got a good ansi sequence, put it back to
1352 * local_sb as we're cutting sb
1353 */
1354 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1355 ch = p - 1;
1356 }
1357 strbuf_setlen(sb, ch + 1 - sb->buf);
1358 c->flush_type = flush_left;
1359 }
1360
1361 if (len > padding) {
1362 switch (c->truncate) {
1363 case trunc_left:
1364 strbuf_utf8_replace(&local_sb,
1365 0, len - (padding - 2),
1366 "..");
1367 break;
1368 case trunc_middle:
1369 strbuf_utf8_replace(&local_sb,
1370 padding / 2 - 1,
1371 len - (padding - 2),
1372 "..");
1373 break;
1374 case trunc_right:
1375 strbuf_utf8_replace(&local_sb,
1376 padding - 2, len - (padding - 2),
1377 "..");
1378 break;
1379 case trunc_none:
1380 break;
1381 }
1382 strbuf_addbuf(sb, &local_sb);
1383 } else {
1384 int sb_len = sb->len, offset = 0;
1385 if (c->flush_type == flush_left)
1386 offset = padding - len;
1387 else if (c->flush_type == flush_both)
1388 offset = (padding - len) / 2;
1389 /*
1390 * we calculate padding in columns, now
1391 * convert it back to chars
1392 */
1393 padding = padding - len + local_sb.len;
1394 strbuf_grow(sb, padding);
1395 strbuf_setlen(sb, sb_len + padding);
1396 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1397 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1398 local_sb.len);
1399 }
1400 strbuf_release(&local_sb);
1401 c->flush_type = no_flush;
1402 return total_consumed;
1403}
1404
1405static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1406 const char *placeholder,
1407 void *context)
1408{
1409 int consumed;
1410 size_t orig_len;
1411 enum {
1412 NO_MAGIC,
1413 ADD_LF_BEFORE_NON_EMPTY,
1414 DEL_LF_BEFORE_EMPTY,
1415 ADD_SP_BEFORE_NON_EMPTY
1416 } magic = NO_MAGIC;
1417
1418 switch (placeholder[0]) {
1419 case '-':
1420 magic = DEL_LF_BEFORE_EMPTY;
1421 break;
1422 case '+':
1423 magic = ADD_LF_BEFORE_NON_EMPTY;
1424 break;
1425 case ' ':
1426 magic = ADD_SP_BEFORE_NON_EMPTY;
1427 break;
1428 default:
1429 break;
1430 }
1431 if (magic != NO_MAGIC)
1432 placeholder++;
1433
1434 orig_len = sb->len;
1435 if (((struct format_commit_context *)context)->flush_type != no_flush)
1436 consumed = format_and_pad_commit(sb, placeholder, context);
1437 else
1438 consumed = format_commit_one(sb, placeholder, context);
1439 if (magic == NO_MAGIC)
1440 return consumed;
1441
1442 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1443 while (sb->len && sb->buf[sb->len - 1] == '\n')
1444 strbuf_setlen(sb, sb->len - 1);
1445 } else if (orig_len != sb->len) {
1446 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1447 strbuf_insert(sb, orig_len, "\n", 1);
1448 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1449 strbuf_insert(sb, orig_len, " ", 1);
1450 }
1451 return consumed + 1;
1452}
1453
1454static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1455 void *context)
1456{
1457 struct userformat_want *w = context;
1458
1459 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1460 placeholder++;
1461
1462 switch (*placeholder) {
1463 case 'N':
1464 w->notes = 1;
1465 break;
1466 }
1467 return 0;
1468}
1469
1470void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1471{
1472 struct strbuf dummy = STRBUF_INIT;
1473
1474 if (!fmt) {
1475 if (!user_format)
1476 return;
1477 fmt = user_format;
1478 }
1479 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1480 strbuf_release(&dummy);
1481}
1482
1483void format_commit_message(const struct commit *commit,
1484 const char *format, struct strbuf *sb,
1485 const struct pretty_print_context *pretty_ctx)
1486{
1487 struct format_commit_context context;
1488 const char *output_enc = pretty_ctx->output_encoding;
1489 const char *utf8 = "UTF-8";
1490
1491 memset(&context, 0, sizeof(context));
1492 context.commit = commit;
1493 context.pretty_ctx = pretty_ctx;
1494 context.wrap_start = sb->len;
1495 /*
1496 * convert a commit message to UTF-8 first
1497 * as far as 'format_commit_item' assumes it in UTF-8
1498 */
1499 context.message = logmsg_reencode(commit,
1500 &context.commit_encoding,
1501 utf8);
1502
1503 strbuf_expand(sb, format, format_commit_item, &context);
1504 rewrap_message_tail(sb, &context, 0, 0, 0);
1505
1506 /* then convert a commit message to an actual output encoding */
1507 if (output_enc) {
1508 if (same_encoding(utf8, output_enc))
1509 output_enc = NULL;
1510 } else {
1511 if (context.commit_encoding &&
1512 !same_encoding(context.commit_encoding, utf8))
1513 output_enc = context.commit_encoding;
1514 }
1515
1516 if (output_enc) {
1517 int outsz;
1518 char *out = reencode_string_len(sb->buf, sb->len,
1519 output_enc, utf8, &outsz);
1520 if (out)
1521 strbuf_attach(sb, out, outsz, outsz + 1);
1522 }
1523
1524 free(context.commit_encoding);
1525 unuse_commit_buffer(commit, context.message);
1526}
1527
1528static void pp_header(struct pretty_print_context *pp,
1529 const char *encoding,
1530 const struct commit *commit,
1531 const char **msg_p,
1532 struct strbuf *sb)
1533{
1534 int parents_shown = 0;
1535
1536 for (;;) {
1537 const char *line = *msg_p;
1538 int linelen = get_one_line(*msg_p);
1539
1540 if (!linelen)
1541 return;
1542 *msg_p += linelen;
1543
1544 if (linelen == 1)
1545 /* End of header */
1546 return;
1547
1548 if (pp->fmt == CMIT_FMT_RAW) {
1549 strbuf_add(sb, line, linelen);
1550 continue;
1551 }
1552
1553 if (starts_with(line, "parent ")) {
1554 if (linelen != 48)
1555 die("bad parent line in commit");
1556 continue;
1557 }
1558
1559 if (!parents_shown) {
1560 unsigned num = commit_list_count(commit->parents);
1561 /* with enough slop */
1562 strbuf_grow(sb, num * 50 + 20);
1563 add_merge_info(pp, sb, commit);
1564 parents_shown = 1;
1565 }
1566
1567 /*
1568 * MEDIUM == DEFAULT shows only author with dates.
1569 * FULL shows both authors but not dates.
1570 * FULLER shows both authors and dates.
1571 */
1572 if (starts_with(line, "author ")) {
1573 strbuf_grow(sb, linelen + 80);
1574 pp_user_info(pp, "Author", sb, line + 7, encoding);
1575 }
1576 if (starts_with(line, "committer ") &&
1577 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1578 strbuf_grow(sb, linelen + 80);
1579 pp_user_info(pp, "Commit", sb, line + 10, encoding);
1580 }
1581 }
1582}
1583
1584void pp_title_line(struct pretty_print_context *pp,
1585 const char **msg_p,
1586 struct strbuf *sb,
1587 const char *encoding,
1588 int need_8bit_cte)
1589{
1590 static const int max_length = 78; /* per rfc2047 */
1591 struct strbuf title;
1592
1593 strbuf_init(&title, 80);
1594 *msg_p = format_subject(&title, *msg_p,
1595 pp->preserve_subject ? "\n" : " ");
1596
1597 strbuf_grow(sb, title.len + 1024);
1598 if (pp->subject) {
1599 strbuf_addstr(sb, pp->subject);
1600 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1601 add_rfc2047(sb, title.buf, title.len,
1602 encoding, RFC2047_SUBJECT);
1603 else
1604 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1605 -last_line_length(sb), 1, max_length);
1606 } else {
1607 strbuf_addbuf(sb, &title);
1608 }
1609 strbuf_addch(sb, '\n');
1610
1611 if (need_8bit_cte == 0) {
1612 int i;
1613 for (i = 0; i < pp->in_body_headers.nr; i++) {
1614 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1615 need_8bit_cte = 1;
1616 break;
1617 }
1618 }
1619 }
1620
1621 if (need_8bit_cte > 0) {
1622 const char *header_fmt =
1623 "MIME-Version: 1.0\n"
1624 "Content-Type: text/plain; charset=%s\n"
1625 "Content-Transfer-Encoding: 8bit\n";
1626 strbuf_addf(sb, header_fmt, encoding);
1627 }
1628 if (pp->after_subject) {
1629 strbuf_addstr(sb, pp->after_subject);
1630 }
1631 if (pp->fmt == CMIT_FMT_EMAIL) {
1632 strbuf_addch(sb, '\n');
1633 }
1634
1635 if (pp->in_body_headers.nr) {
1636 int i;
1637 for (i = 0; i < pp->in_body_headers.nr; i++) {
1638 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1639 free(pp->in_body_headers.items[i].string);
1640 }
1641 string_list_clear(&pp->in_body_headers, 0);
1642 strbuf_addch(sb, '\n');
1643 }
1644
1645 strbuf_release(&title);
1646}
1647
1648void pp_remainder(struct pretty_print_context *pp,
1649 const char **msg_p,
1650 struct strbuf *sb,
1651 int indent)
1652{
1653 int first = 1;
1654 for (;;) {
1655 const char *line = *msg_p;
1656 int linelen = get_one_line(line);
1657 *msg_p += linelen;
1658
1659 if (!linelen)
1660 break;
1661
1662 if (is_empty_line(line, &linelen)) {
1663 if (first)
1664 continue;
1665 if (pp->fmt == CMIT_FMT_SHORT)
1666 break;
1667 }
1668 first = 0;
1669
1670 strbuf_grow(sb, linelen + indent + 20);
1671 if (indent) {
1672 memset(sb->buf + sb->len, ' ', indent);
1673 strbuf_setlen(sb, sb->len + indent);
1674 }
1675 strbuf_add(sb, line, linelen);
1676 strbuf_addch(sb, '\n');
1677 }
1678}
1679
1680void pretty_print_commit(struct pretty_print_context *pp,
1681 const struct commit *commit,
1682 struct strbuf *sb)
1683{
1684 unsigned long beginning_of_body;
1685 int indent = 4;
1686 const char *msg;
1687 const char *reencoded;
1688 const char *encoding;
1689 int need_8bit_cte = pp->need_8bit_cte;
1690
1691 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1692 format_commit_message(commit, user_format, sb, pp);
1693 return;
1694 }
1695
1696 encoding = get_log_output_encoding();
1697 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1698
1699 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1700 indent = 0;
1701
1702 /*
1703 * We need to check and emit Content-type: to mark it
1704 * as 8-bit if we haven't done so.
1705 */
1706 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1707 int i, ch, in_body;
1708
1709 for (in_body = i = 0; (ch = msg[i]); i++) {
1710 if (!in_body) {
1711 /* author could be non 7-bit ASCII but
1712 * the log may be so; skip over the
1713 * header part first.
1714 */
1715 if (ch == '\n' && msg[i+1] == '\n')
1716 in_body = 1;
1717 }
1718 else if (non_ascii(ch)) {
1719 need_8bit_cte = 1;
1720 break;
1721 }
1722 }
1723 }
1724
1725 pp_header(pp, encoding, commit, &msg, sb);
1726 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1727 strbuf_addch(sb, '\n');
1728 }
1729
1730 /* Skip excess blank lines at the beginning of body, if any... */
1731 msg = skip_empty_lines(msg);
1732
1733 /* These formats treat the title line specially. */
1734 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1735 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1736
1737 beginning_of_body = sb->len;
1738 if (pp->fmt != CMIT_FMT_ONELINE)
1739 pp_remainder(pp, &msg, sb, indent);
1740 strbuf_rtrim(sb);
1741
1742 /* Make sure there is an EOLN for the non-oneline case */
1743 if (pp->fmt != CMIT_FMT_ONELINE)
1744 strbuf_addch(sb, '\n');
1745
1746 /*
1747 * The caller may append additional body text in e-mail
1748 * format. Make sure we did not strip the blank line
1749 * between the header and the body.
1750 */
1751 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1752 strbuf_addch(sb, '\n');
1753
1754 unuse_commit_buffer(commit, reencoded);
1755}
1756
1757void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1758 struct strbuf *sb)
1759{
1760 struct pretty_print_context pp = {0};
1761 pp.fmt = fmt;
1762 pretty_print_commit(&pp, commit, sb);
1763}