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 expand_tabs_in_log;
20 int is_alias;
21 const char *user_format;
22} *commit_formats;
23static size_t builtin_formats_len;
24static size_t commit_formats_len;
25static size_t commit_formats_alloc;
26static struct cmt_fmt_map *find_commit_format(const char *sought);
27
28int commit_format_is_empty(enum cmit_fmt fmt)
29{
30 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
31}
32
33static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
34{
35 free(user_format);
36 user_format = xstrdup(cp);
37 if (is_tformat)
38 rev->use_terminator = 1;
39 rev->commit_format = CMIT_FMT_USERFORMAT;
40}
41
42static int git_pretty_formats_config(const char *var, const char *value, void *cb)
43{
44 struct cmt_fmt_map *commit_format = NULL;
45 const char *name;
46 const char *fmt;
47 int i;
48
49 if (!skip_prefix(var, "pretty.", &name))
50 return 0;
51
52 for (i = 0; i < builtin_formats_len; i++) {
53 if (!strcmp(commit_formats[i].name, name))
54 return 0;
55 }
56
57 for (i = builtin_formats_len; i < commit_formats_len; i++) {
58 if (!strcmp(commit_formats[i].name, name)) {
59 commit_format = &commit_formats[i];
60 break;
61 }
62 }
63
64 if (!commit_format) {
65 ALLOC_GROW(commit_formats, commit_formats_len+1,
66 commit_formats_alloc);
67 commit_format = &commit_formats[commit_formats_len];
68 memset(commit_format, 0, sizeof(*commit_format));
69 commit_formats_len++;
70 }
71
72 commit_format->name = xstrdup(name);
73 commit_format->format = CMIT_FMT_USERFORMAT;
74 if (git_config_string(&fmt, var, value))
75 return -1;
76
77 if (skip_prefix(fmt, "format:", &fmt))
78 commit_format->is_tformat = 0;
79 else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
80 commit_format->is_tformat = 1;
81 else
82 commit_format->is_alias = 1;
83 commit_format->user_format = fmt;
84
85 return 0;
86}
87
88static void setup_commit_formats(void)
89{
90 struct cmt_fmt_map builtin_formats[] = {
91 { "raw", CMIT_FMT_RAW, 0, 0 },
92 { "medium", CMIT_FMT_MEDIUM, 0, 8 },
93 { "short", CMIT_FMT_SHORT, 0, 0 },
94 { "email", CMIT_FMT_EMAIL, 0, 0 },
95 { "mboxrd", CMIT_FMT_MBOXRD, 0, 0 },
96 { "fuller", CMIT_FMT_FULLER, 0, 8 },
97 { "full", CMIT_FMT_FULL, 0, 8 },
98 { "oneline", CMIT_FMT_ONELINE, 1, 0 }
99 };
100 commit_formats_len = ARRAY_SIZE(builtin_formats);
101 builtin_formats_len = commit_formats_len;
102 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
103 memcpy(commit_formats, builtin_formats,
104 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
105
106 git_config(git_pretty_formats_config, NULL);
107}
108
109static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
110 const char *original,
111 int num_redirections)
112{
113 struct cmt_fmt_map *found = NULL;
114 size_t found_match_len = 0;
115 int i;
116
117 if (num_redirections >= commit_formats_len)
118 die("invalid --pretty format: "
119 "'%s' references an alias which points to itself",
120 original);
121
122 for (i = 0; i < commit_formats_len; i++) {
123 size_t match_len;
124
125 if (!starts_with(commit_formats[i].name, sought))
126 continue;
127
128 match_len = strlen(commit_formats[i].name);
129 if (found == NULL || found_match_len > match_len) {
130 found = &commit_formats[i];
131 found_match_len = match_len;
132 }
133 }
134
135 if (found && found->is_alias) {
136 found = find_commit_format_recursive(found->user_format,
137 original,
138 num_redirections+1);
139 }
140
141 return found;
142}
143
144static struct cmt_fmt_map *find_commit_format(const char *sought)
145{
146 if (!commit_formats)
147 setup_commit_formats();
148
149 return find_commit_format_recursive(sought, sought, 0);
150}
151
152void get_commit_format(const char *arg, struct rev_info *rev)
153{
154 struct cmt_fmt_map *commit_format;
155
156 rev->use_terminator = 0;
157 if (!arg) {
158 rev->commit_format = CMIT_FMT_DEFAULT;
159 return;
160 }
161 if (skip_prefix(arg, "format:", &arg)) {
162 save_user_format(rev, arg, 0);
163 return;
164 }
165
166 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
167 save_user_format(rev, arg, 1);
168 return;
169 }
170
171 commit_format = find_commit_format(arg);
172 if (!commit_format)
173 die("invalid --pretty format: %s", arg);
174
175 rev->commit_format = commit_format->format;
176 rev->use_terminator = commit_format->is_tformat;
177 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
178 if (commit_format->format == CMIT_FMT_USERFORMAT) {
179 save_user_format(rev, commit_format->user_format,
180 commit_format->is_tformat);
181 }
182}
183
184/*
185 * Generic support for pretty-printing the header
186 */
187static int get_one_line(const char *msg)
188{
189 int ret = 0;
190
191 for (;;) {
192 char c = *msg++;
193 if (!c)
194 break;
195 ret++;
196 if (c == '\n')
197 break;
198 }
199 return ret;
200}
201
202/* High bit set, or ISO-2022-INT */
203static int non_ascii(int ch)
204{
205 return !isascii(ch) || ch == '\033';
206}
207
208int has_non_ascii(const char *s)
209{
210 int ch;
211 if (!s)
212 return 0;
213 while ((ch = *s++) != '\0') {
214 if (non_ascii(ch))
215 return 1;
216 }
217 return 0;
218}
219
220static int is_rfc822_special(char ch)
221{
222 switch (ch) {
223 case '(':
224 case ')':
225 case '<':
226 case '>':
227 case '[':
228 case ']':
229 case ':':
230 case ';':
231 case '@':
232 case ',':
233 case '.':
234 case '"':
235 case '\\':
236 return 1;
237 default:
238 return 0;
239 }
240}
241
242static int needs_rfc822_quoting(const char *s, int len)
243{
244 int i;
245 for (i = 0; i < len; i++)
246 if (is_rfc822_special(s[i]))
247 return 1;
248 return 0;
249}
250
251static int last_line_length(struct strbuf *sb)
252{
253 int i;
254
255 /* How many bytes are already used on the last line? */
256 for (i = sb->len - 1; i >= 0; i--)
257 if (sb->buf[i] == '\n')
258 break;
259 return sb->len - (i + 1);
260}
261
262static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
263{
264 int i;
265
266 /* just a guess, we may have to also backslash-quote */
267 strbuf_grow(out, len + 2);
268
269 strbuf_addch(out, '"');
270 for (i = 0; i < len; i++) {
271 switch (s[i]) {
272 case '"':
273 case '\\':
274 strbuf_addch(out, '\\');
275 /* fall through */
276 default:
277 strbuf_addch(out, s[i]);
278 }
279 }
280 strbuf_addch(out, '"');
281}
282
283enum rfc2047_type {
284 RFC2047_SUBJECT,
285 RFC2047_ADDRESS
286};
287
288static int is_rfc2047_special(char ch, enum rfc2047_type type)
289{
290 /*
291 * rfc2047, section 4.2:
292 *
293 * 8-bit values which correspond to printable ASCII characters other
294 * than "=", "?", and "_" (underscore), MAY be represented as those
295 * characters. (But see section 5 for restrictions.) In
296 * particular, SPACE and TAB MUST NOT be represented as themselves
297 * within encoded words.
298 */
299
300 /*
301 * rule out non-ASCII characters and non-printable characters (the
302 * non-ASCII check should be redundant as isprint() is not localized
303 * and only knows about ASCII, but be defensive about that)
304 */
305 if (non_ascii(ch) || !isprint(ch))
306 return 1;
307
308 /*
309 * rule out special printable characters (' ' should be the only
310 * whitespace character considered printable, but be defensive and use
311 * isspace())
312 */
313 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
314 return 1;
315
316 /*
317 * rfc2047, section 5.3:
318 *
319 * As a replacement for a 'word' entity within a 'phrase', for example,
320 * one that precedes an address in a From, To, or Cc header. The ABNF
321 * definition for 'phrase' from RFC 822 thus becomes:
322 *
323 * phrase = 1*( encoded-word / word )
324 *
325 * In this case the set of characters that may be used in a "Q"-encoded
326 * 'encoded-word' is restricted to: <upper and lower case ASCII
327 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
328 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
329 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
330 * 'special' by 'linear-white-space'.
331 */
332
333 if (type != RFC2047_ADDRESS)
334 return 0;
335
336 /* '=' and '_' are special cases and have been checked above */
337 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
338}
339
340static int needs_rfc2047_encoding(const char *line, int len,
341 enum rfc2047_type type)
342{
343 int i;
344
345 for (i = 0; i < len; i++) {
346 int ch = line[i];
347 if (non_ascii(ch) || ch == '\n')
348 return 1;
349 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
350 return 1;
351 }
352
353 return 0;
354}
355
356static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
357 const char *encoding, enum rfc2047_type type)
358{
359 static const int max_encoded_length = 76; /* per rfc2047 */
360 int i;
361 int line_len = last_line_length(sb);
362
363 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
364 strbuf_addf(sb, "=?%s?q?", encoding);
365 line_len += strlen(encoding) + 5; /* 5 for =??q? */
366
367 while (len) {
368 /*
369 * RFC 2047, section 5 (3):
370 *
371 * Each 'encoded-word' MUST represent an integral number of
372 * characters. A multi-octet character may not be split across
373 * adjacent 'encoded- word's.
374 */
375 const unsigned char *p = (const unsigned char *)line;
376 int chrlen = mbs_chrlen(&line, &len, encoding);
377 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
378
379 /* "=%02X" * chrlen, or the byte itself */
380 const char *encoded_fmt = is_special ? "=%02X" : "%c";
381 int encoded_len = is_special ? 3 * chrlen : 1;
382
383 /*
384 * According to RFC 2047, we could encode the special character
385 * ' ' (space) with '_' (underscore) for readability. But many
386 * programs do not understand this and just leave the
387 * underscore in place. Thus, we do nothing special here, which
388 * causes ' ' to be encoded as '=20', avoiding this problem.
389 */
390
391 if (line_len + encoded_len + 2 > max_encoded_length) {
392 /* It won't fit with trailing "?=" --- break the line */
393 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
394 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
395 }
396
397 for (i = 0; i < chrlen; i++)
398 strbuf_addf(sb, encoded_fmt, p[i]);
399 line_len += encoded_len;
400 }
401 strbuf_addstr(sb, "?=");
402}
403
404const char *show_ident_date(const struct ident_split *ident,
405 const struct date_mode *mode)
406{
407 unsigned long date = 0;
408 long tz = 0;
409
410 if (ident->date_begin && ident->date_end)
411 date = strtoul(ident->date_begin, NULL, 10);
412 if (date_overflows(date))
413 date = 0;
414 else {
415 if (ident->tz_begin && ident->tz_end)
416 tz = strtol(ident->tz_begin, NULL, 10);
417 if (tz >= INT_MAX || tz <= INT_MIN)
418 tz = 0;
419 }
420 return show_date(date, tz, mode);
421}
422
423void pp_user_info(struct pretty_print_context *pp,
424 const char *what, struct strbuf *sb,
425 const char *line, const char *encoding)
426{
427 struct ident_split ident;
428 char *line_end;
429 const char *mailbuf, *namebuf;
430 size_t namelen, maillen;
431 int max_length = 78; /* per rfc2822 */
432
433 if (pp->fmt == CMIT_FMT_ONELINE)
434 return;
435
436 line_end = strchrnul(line, '\n');
437 if (split_ident_line(&ident, line, line_end - line))
438 return;
439
440 mailbuf = ident.mail_begin;
441 maillen = ident.mail_end - ident.mail_begin;
442 namebuf = ident.name_begin;
443 namelen = ident.name_end - ident.name_begin;
444
445 if (pp->mailmap)
446 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
447
448 if (cmit_fmt_is_mail(pp->fmt)) {
449 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
450 struct strbuf buf = STRBUF_INIT;
451
452 strbuf_addstr(&buf, "From: ");
453 strbuf_add(&buf, namebuf, namelen);
454 strbuf_addstr(&buf, " <");
455 strbuf_add(&buf, mailbuf, maillen);
456 strbuf_addstr(&buf, ">\n");
457 string_list_append(&pp->in_body_headers,
458 strbuf_detach(&buf, NULL));
459
460 mailbuf = pp->from_ident->mail_begin;
461 maillen = pp->from_ident->mail_end - mailbuf;
462 namebuf = pp->from_ident->name_begin;
463 namelen = pp->from_ident->name_end - namebuf;
464 }
465
466 strbuf_addstr(sb, "From: ");
467 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
468 add_rfc2047(sb, namebuf, namelen,
469 encoding, RFC2047_ADDRESS);
470 max_length = 76; /* per rfc2047 */
471 } else if (needs_rfc822_quoting(namebuf, namelen)) {
472 struct strbuf quoted = STRBUF_INIT;
473 add_rfc822_quoted("ed, namebuf, namelen);
474 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
475 -6, 1, max_length);
476 strbuf_release("ed);
477 } else {
478 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
479 -6, 1, max_length);
480 }
481
482 if (max_length <
483 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
484 strbuf_addch(sb, '\n');
485 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
486 } else {
487 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
488 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
489 (int)namelen, namebuf, (int)maillen, mailbuf);
490 }
491
492 switch (pp->fmt) {
493 case CMIT_FMT_MEDIUM:
494 strbuf_addf(sb, "Date: %s\n",
495 show_ident_date(&ident, &pp->date_mode));
496 break;
497 case CMIT_FMT_EMAIL:
498 case CMIT_FMT_MBOXRD:
499 strbuf_addf(sb, "Date: %s\n",
500 show_ident_date(&ident, DATE_MODE(RFC2822)));
501 break;
502 case CMIT_FMT_FULLER:
503 strbuf_addf(sb, "%sDate: %s\n", what,
504 show_ident_date(&ident, &pp->date_mode));
505 break;
506 default:
507 /* notin' */
508 break;
509 }
510}
511
512static int is_blank_line(const char *line, int *len_p)
513{
514 int len = *len_p;
515 while (len && isspace(line[len - 1]))
516 len--;
517 *len_p = len;
518 return !len;
519}
520
521const char *skip_blank_lines(const char *msg)
522{
523 for (;;) {
524 int linelen = get_one_line(msg);
525 int ll = linelen;
526 if (!linelen)
527 break;
528 if (!is_blank_line(msg, &ll))
529 break;
530 msg += linelen;
531 }
532 return msg;
533}
534
535static void add_merge_info(const struct pretty_print_context *pp,
536 struct strbuf *sb, const struct commit *commit)
537{
538 struct commit_list *parent = commit->parents;
539
540 if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
541 !parent || !parent->next)
542 return;
543
544 strbuf_addstr(sb, "Merge:");
545
546 while (parent) {
547 struct object_id *oidp = &parent->item->object.oid;
548 strbuf_addch(sb, ' ');
549 if (pp->abbrev)
550 strbuf_add_unique_abbrev(sb, oidp->hash, pp->abbrev);
551 else
552 strbuf_addstr(sb, oid_to_hex(oidp));
553 parent = parent->next;
554 }
555 strbuf_addch(sb, '\n');
556}
557
558static char *get_header(const char *msg, const char *key)
559{
560 size_t len;
561 const char *v = find_commit_header(msg, key, &len);
562 return v ? xmemdupz(v, len) : NULL;
563}
564
565static char *replace_encoding_header(char *buf, const char *encoding)
566{
567 struct strbuf tmp = STRBUF_INIT;
568 size_t start, len;
569 char *cp = buf;
570
571 /* guess if there is an encoding header before a \n\n */
572 while (!starts_with(cp, "encoding ")) {
573 cp = strchr(cp, '\n');
574 if (!cp || *++cp == '\n')
575 return buf;
576 }
577 start = cp - buf;
578 cp = strchr(cp, '\n');
579 if (!cp)
580 return buf; /* should not happen but be defensive */
581 len = cp + 1 - (buf + start);
582
583 strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
584 if (is_encoding_utf8(encoding)) {
585 /* we have re-coded to UTF-8; drop the header */
586 strbuf_remove(&tmp, start, len);
587 } else {
588 /* just replaces XXXX in 'encoding XXXX\n' */
589 strbuf_splice(&tmp, start + strlen("encoding "),
590 len - strlen("encoding \n"),
591 encoding, strlen(encoding));
592 }
593 return strbuf_detach(&tmp, NULL);
594}
595
596const char *logmsg_reencode(const struct commit *commit,
597 char **commit_encoding,
598 const char *output_encoding)
599{
600 static const char *utf8 = "UTF-8";
601 const char *use_encoding;
602 char *encoding;
603 const char *msg = get_commit_buffer(commit, NULL);
604 char *out;
605
606 if (!output_encoding || !*output_encoding) {
607 if (commit_encoding)
608 *commit_encoding = get_header(msg, "encoding");
609 return msg;
610 }
611 encoding = get_header(msg, "encoding");
612 if (commit_encoding)
613 *commit_encoding = encoding;
614 use_encoding = encoding ? encoding : utf8;
615 if (same_encoding(use_encoding, output_encoding)) {
616 /*
617 * No encoding work to be done. If we have no encoding header
618 * at all, then there's nothing to do, and we can return the
619 * message verbatim (whether newly allocated or not).
620 */
621 if (!encoding)
622 return msg;
623
624 /*
625 * Otherwise, we still want to munge the encoding header in the
626 * result, which will be done by modifying the buffer. If we
627 * are using a fresh copy, we can reuse it. But if we are using
628 * the cached copy from get_commit_buffer, we need to duplicate it
629 * to avoid munging the cached copy.
630 */
631 if (msg == get_cached_commit_buffer(commit, NULL))
632 out = xstrdup(msg);
633 else
634 out = (char *)msg;
635 }
636 else {
637 /*
638 * There's actual encoding work to do. Do the reencoding, which
639 * still leaves the header to be replaced in the next step. At
640 * this point, we are done with msg. If we allocated a fresh
641 * copy, we can free it.
642 */
643 out = reencode_string(msg, output_encoding, use_encoding);
644 if (out)
645 unuse_commit_buffer(commit, msg);
646 }
647
648 /*
649 * This replacement actually consumes the buffer we hand it, so we do
650 * not have to worry about freeing the old "out" here.
651 */
652 if (out)
653 out = replace_encoding_header(out, output_encoding);
654
655 if (!commit_encoding)
656 free(encoding);
657 /*
658 * If the re-encoding failed, out might be NULL here; in that
659 * case we just return the commit message verbatim.
660 */
661 return out ? out : msg;
662}
663
664static int mailmap_name(const char **email, size_t *email_len,
665 const char **name, size_t *name_len)
666{
667 static struct string_list *mail_map;
668 if (!mail_map) {
669 mail_map = xcalloc(1, sizeof(*mail_map));
670 read_mailmap(mail_map, NULL);
671 }
672 return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
673}
674
675static size_t format_person_part(struct strbuf *sb, char part,
676 const char *msg, int len,
677 const struct date_mode *dmode)
678{
679 /* currently all placeholders have same length */
680 const int placeholder_len = 2;
681 struct ident_split s;
682 const char *name, *mail;
683 size_t maillen, namelen;
684
685 if (split_ident_line(&s, msg, len) < 0)
686 goto skip;
687
688 name = s.name_begin;
689 namelen = s.name_end - s.name_begin;
690 mail = s.mail_begin;
691 maillen = s.mail_end - s.mail_begin;
692
693 if (part == 'N' || part == 'E') /* mailmap lookup */
694 mailmap_name(&mail, &maillen, &name, &namelen);
695 if (part == 'n' || part == 'N') { /* name */
696 strbuf_add(sb, name, namelen);
697 return placeholder_len;
698 }
699 if (part == 'e' || part == 'E') { /* email */
700 strbuf_add(sb, mail, maillen);
701 return placeholder_len;
702 }
703
704 if (!s.date_begin)
705 goto skip;
706
707 if (part == 't') { /* date, UNIX timestamp */
708 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
709 return placeholder_len;
710 }
711
712 switch (part) {
713 case 'd': /* date */
714 strbuf_addstr(sb, show_ident_date(&s, dmode));
715 return placeholder_len;
716 case 'D': /* date, RFC2822 style */
717 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
718 return placeholder_len;
719 case 'r': /* date, relative */
720 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
721 return placeholder_len;
722 case 'i': /* date, ISO 8601-like */
723 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
724 return placeholder_len;
725 case 'I': /* date, ISO 8601 strict */
726 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
727 return placeholder_len;
728 }
729
730skip:
731 /*
732 * reading from either a bogus commit, or a reflog entry with
733 * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
734 * to compute a valid return value.
735 */
736 if (part == 'n' || part == 'e' || part == 't' || part == 'd'
737 || part == 'D' || part == 'r' || part == 'i')
738 return placeholder_len;
739
740 return 0; /* unknown placeholder */
741}
742
743struct chunk {
744 size_t off;
745 size_t len;
746};
747
748enum flush_type {
749 no_flush,
750 flush_right,
751 flush_left,
752 flush_left_and_steal,
753 flush_both
754};
755
756enum trunc_type {
757 trunc_none,
758 trunc_left,
759 trunc_middle,
760 trunc_right
761};
762
763struct format_commit_context {
764 const struct commit *commit;
765 const struct pretty_print_context *pretty_ctx;
766 unsigned commit_header_parsed:1;
767 unsigned commit_message_parsed:1;
768 struct signature_check signature_check;
769 enum flush_type flush_type;
770 enum trunc_type truncate;
771 const char *message;
772 char *commit_encoding;
773 size_t width, indent1, indent2;
774 int auto_color;
775 int padding;
776
777 /* These offsets are relative to the start of the commit message. */
778 struct chunk author;
779 struct chunk committer;
780 size_t message_off;
781 size_t subject_off;
782 size_t body_off;
783
784 /* The following ones are relative to the result struct strbuf. */
785 size_t wrap_start;
786};
787
788static void parse_commit_header(struct format_commit_context *context)
789{
790 const char *msg = context->message;
791 int i;
792
793 for (i = 0; msg[i]; i++) {
794 const char *name;
795 int eol;
796 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
797 ; /* do nothing */
798
799 if (i == eol) {
800 break;
801 } else if (skip_prefix(msg + i, "author ", &name)) {
802 context->author.off = name - msg;
803 context->author.len = msg + eol - name;
804 } else if (skip_prefix(msg + i, "committer ", &name)) {
805 context->committer.off = name - msg;
806 context->committer.len = msg + eol - name;
807 }
808 i = eol;
809 }
810 context->message_off = i;
811 context->commit_header_parsed = 1;
812}
813
814static int istitlechar(char c)
815{
816 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
817 (c >= '0' && c <= '9') || c == '.' || c == '_';
818}
819
820static void format_sanitized_subject(struct strbuf *sb, const char *msg)
821{
822 size_t trimlen;
823 size_t start_len = sb->len;
824 int space = 2;
825
826 for (; *msg && *msg != '\n'; msg++) {
827 if (istitlechar(*msg)) {
828 if (space == 1)
829 strbuf_addch(sb, '-');
830 space = 0;
831 strbuf_addch(sb, *msg);
832 if (*msg == '.')
833 while (*(msg+1) == '.')
834 msg++;
835 } else
836 space |= 1;
837 }
838
839 /* trim any trailing '.' or '-' characters */
840 trimlen = 0;
841 while (sb->len - trimlen > start_len &&
842 (sb->buf[sb->len - 1 - trimlen] == '.'
843 || sb->buf[sb->len - 1 - trimlen] == '-'))
844 trimlen++;
845 strbuf_remove(sb, sb->len - trimlen, trimlen);
846}
847
848const char *format_subject(struct strbuf *sb, const char *msg,
849 const char *line_separator)
850{
851 int first = 1;
852
853 for (;;) {
854 const char *line = msg;
855 int linelen = get_one_line(line);
856
857 msg += linelen;
858 if (!linelen || is_blank_line(line, &linelen))
859 break;
860
861 if (!sb)
862 continue;
863 strbuf_grow(sb, linelen + 2);
864 if (!first)
865 strbuf_addstr(sb, line_separator);
866 strbuf_add(sb, line, linelen);
867 first = 0;
868 }
869 return msg;
870}
871
872static void parse_commit_message(struct format_commit_context *c)
873{
874 const char *msg = c->message + c->message_off;
875 const char *start = c->message;
876
877 msg = skip_blank_lines(msg);
878 c->subject_off = msg - start;
879
880 msg = format_subject(NULL, msg, NULL);
881 msg = skip_blank_lines(msg);
882 c->body_off = msg - start;
883
884 c->commit_message_parsed = 1;
885}
886
887static void strbuf_wrap(struct strbuf *sb, size_t pos,
888 size_t width, size_t indent1, size_t indent2)
889{
890 struct strbuf tmp = STRBUF_INIT;
891
892 if (pos)
893 strbuf_add(&tmp, sb->buf, pos);
894 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
895 (int) indent1, (int) indent2, (int) width);
896 strbuf_swap(&tmp, sb);
897 strbuf_release(&tmp);
898}
899
900static void rewrap_message_tail(struct strbuf *sb,
901 struct format_commit_context *c,
902 size_t new_width, size_t new_indent1,
903 size_t new_indent2)
904{
905 if (c->width == new_width && c->indent1 == new_indent1 &&
906 c->indent2 == new_indent2)
907 return;
908 if (c->wrap_start < sb->len)
909 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
910 c->wrap_start = sb->len;
911 c->width = new_width;
912 c->indent1 = new_indent1;
913 c->indent2 = new_indent2;
914}
915
916static int format_reflog_person(struct strbuf *sb,
917 char part,
918 struct reflog_walk_info *log,
919 const struct date_mode *dmode)
920{
921 const char *ident;
922
923 if (!log)
924 return 2;
925
926 ident = get_reflog_ident(log);
927 if (!ident)
928 return 2;
929
930 return format_person_part(sb, part, ident, strlen(ident), dmode);
931}
932
933static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
934 const char *placeholder,
935 struct format_commit_context *c)
936{
937 const char *rest = placeholder;
938
939 if (placeholder[1] == '(') {
940 const char *begin = placeholder + 2;
941 const char *end = strchr(begin, ')');
942 char color[COLOR_MAXLEN];
943
944 if (!end)
945 return 0;
946 if (skip_prefix(begin, "auto,", &begin)) {
947 if (!want_color(c->pretty_ctx->color))
948 return end - placeholder + 1;
949 }
950 if (color_parse_mem(begin, end - begin, color) < 0)
951 die(_("unable to parse --pretty format"));
952 strbuf_addstr(sb, color);
953 return end - placeholder + 1;
954 }
955 if (skip_prefix(placeholder + 1, "red", &rest))
956 strbuf_addstr(sb, GIT_COLOR_RED);
957 else if (skip_prefix(placeholder + 1, "green", &rest))
958 strbuf_addstr(sb, GIT_COLOR_GREEN);
959 else if (skip_prefix(placeholder + 1, "blue", &rest))
960 strbuf_addstr(sb, GIT_COLOR_BLUE);
961 else if (skip_prefix(placeholder + 1, "reset", &rest))
962 strbuf_addstr(sb, GIT_COLOR_RESET);
963 return rest - placeholder;
964}
965
966static size_t parse_padding_placeholder(struct strbuf *sb,
967 const char *placeholder,
968 struct format_commit_context *c)
969{
970 const char *ch = placeholder;
971 enum flush_type flush_type;
972 int to_column = 0;
973
974 switch (*ch++) {
975 case '<':
976 flush_type = flush_right;
977 break;
978 case '>':
979 if (*ch == '<') {
980 flush_type = flush_both;
981 ch++;
982 } else if (*ch == '>') {
983 flush_type = flush_left_and_steal;
984 ch++;
985 } else
986 flush_type = flush_left;
987 break;
988 default:
989 return 0;
990 }
991
992 /* the next value means "wide enough to that column" */
993 if (*ch == '|') {
994 to_column = 1;
995 ch++;
996 }
997
998 if (*ch == '(') {
999 const char *start = ch + 1;
1000 const char *end = start + strcspn(start, ",)");
1001 char *next;
1002 int width;
1003 if (!end || end == start)
1004 return 0;
1005 width = strtol(start, &next, 10);
1006 if (next == start || width == 0)
1007 return 0;
1008 if (width < 0) {
1009 if (to_column)
1010 width += term_columns();
1011 if (width < 0)
1012 return 0;
1013 }
1014 c->padding = to_column ? -width : width;
1015 c->flush_type = flush_type;
1016
1017 if (*end == ',') {
1018 start = end + 1;
1019 end = strchr(start, ')');
1020 if (!end || end == start)
1021 return 0;
1022 if (starts_with(start, "trunc)"))
1023 c->truncate = trunc_right;
1024 else if (starts_with(start, "ltrunc)"))
1025 c->truncate = trunc_left;
1026 else if (starts_with(start, "mtrunc)"))
1027 c->truncate = trunc_middle;
1028 else
1029 return 0;
1030 } else
1031 c->truncate = trunc_none;
1032
1033 return end - placeholder + 1;
1034 }
1035 return 0;
1036}
1037
1038static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1039 const char *placeholder,
1040 void *context)
1041{
1042 struct format_commit_context *c = context;
1043 const struct commit *commit = c->commit;
1044 const char *msg = c->message;
1045 struct commit_list *p;
1046 int ch;
1047
1048 /* these are independent of the commit */
1049 switch (placeholder[0]) {
1050 case 'C':
1051 if (starts_with(placeholder + 1, "(auto)")) {
1052 c->auto_color = want_color(c->pretty_ctx->color);
1053 if (c->auto_color && sb->len)
1054 strbuf_addstr(sb, GIT_COLOR_RESET);
1055 return 7; /* consumed 7 bytes, "C(auto)" */
1056 } else {
1057 int ret = parse_color(sb, placeholder, c);
1058 if (ret)
1059 c->auto_color = 0;
1060 /*
1061 * Otherwise, we decided to treat %C<unknown>
1062 * as a literal string, and the previous
1063 * %C(auto) is still valid.
1064 */
1065 return ret;
1066 }
1067 case 'n': /* newline */
1068 strbuf_addch(sb, '\n');
1069 return 1;
1070 case 'x':
1071 /* %x00 == NUL, %x0a == LF, etc. */
1072 ch = hex2chr(placeholder + 1);
1073 if (ch < 0)
1074 return 0;
1075 strbuf_addch(sb, ch);
1076 return 3;
1077 case 'w':
1078 if (placeholder[1] == '(') {
1079 unsigned long width = 0, indent1 = 0, indent2 = 0;
1080 char *next;
1081 const char *start = placeholder + 2;
1082 const char *end = strchr(start, ')');
1083 if (!end)
1084 return 0;
1085 if (end > start) {
1086 width = strtoul(start, &next, 10);
1087 if (*next == ',') {
1088 indent1 = strtoul(next + 1, &next, 10);
1089 if (*next == ',') {
1090 indent2 = strtoul(next + 1,
1091 &next, 10);
1092 }
1093 }
1094 if (*next != ')')
1095 return 0;
1096 }
1097 rewrap_message_tail(sb, c, width, indent1, indent2);
1098 return end - placeholder + 1;
1099 } else
1100 return 0;
1101
1102 case '<':
1103 case '>':
1104 return parse_padding_placeholder(sb, placeholder, c);
1105 }
1106
1107 /* these depend on the commit */
1108 if (!commit->object.parsed)
1109 parse_object(commit->object.oid.hash);
1110
1111 switch (placeholder[0]) {
1112 case 'H': /* commit hash */
1113 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1114 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1115 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1116 return 1;
1117 case 'h': /* abbreviated commit hash */
1118 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1119 strbuf_add_unique_abbrev(sb, commit->object.oid.hash,
1120 c->pretty_ctx->abbrev);
1121 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1122 return 1;
1123 case 'T': /* tree hash */
1124 strbuf_addstr(sb, oid_to_hex(&commit->tree->object.oid));
1125 return 1;
1126 case 't': /* abbreviated tree hash */
1127 strbuf_add_unique_abbrev(sb, commit->tree->object.oid.hash,
1128 c->pretty_ctx->abbrev);
1129 return 1;
1130 case 'P': /* parent hashes */
1131 for (p = commit->parents; p; p = p->next) {
1132 if (p != commit->parents)
1133 strbuf_addch(sb, ' ');
1134 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1135 }
1136 return 1;
1137 case 'p': /* abbreviated parent hashes */
1138 for (p = commit->parents; p; p = p->next) {
1139 if (p != commit->parents)
1140 strbuf_addch(sb, ' ');
1141 strbuf_add_unique_abbrev(sb, p->item->object.oid.hash,
1142 c->pretty_ctx->abbrev);
1143 }
1144 return 1;
1145 case 'm': /* left/right/bottom */
1146 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1147 return 1;
1148 case 'd':
1149 load_ref_decorations(DECORATE_SHORT_REFS);
1150 format_decorations(sb, commit, c->auto_color);
1151 return 1;
1152 case 'D':
1153 load_ref_decorations(DECORATE_SHORT_REFS);
1154 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1155 return 1;
1156 case 'g': /* reflog info */
1157 switch(placeholder[1]) {
1158 case 'd': /* reflog selector */
1159 case 'D':
1160 if (c->pretty_ctx->reflog_info)
1161 get_reflog_selector(sb,
1162 c->pretty_ctx->reflog_info,
1163 &c->pretty_ctx->date_mode,
1164 c->pretty_ctx->date_mode_explicit,
1165 (placeholder[1] == 'd'));
1166 return 2;
1167 case 's': /* reflog message */
1168 if (c->pretty_ctx->reflog_info)
1169 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1170 return 2;
1171 case 'n':
1172 case 'N':
1173 case 'e':
1174 case 'E':
1175 return format_reflog_person(sb,
1176 placeholder[1],
1177 c->pretty_ctx->reflog_info,
1178 &c->pretty_ctx->date_mode);
1179 }
1180 return 0; /* unknown %g placeholder */
1181 case 'N':
1182 if (c->pretty_ctx->notes_message) {
1183 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1184 return 1;
1185 }
1186 return 0;
1187 }
1188
1189 if (placeholder[0] == 'G') {
1190 if (!c->signature_check.result)
1191 check_commit_signature(c->commit, &(c->signature_check));
1192 switch (placeholder[1]) {
1193 case 'G':
1194 if (c->signature_check.gpg_output)
1195 strbuf_addstr(sb, c->signature_check.gpg_output);
1196 break;
1197 case '?':
1198 switch (c->signature_check.result) {
1199 case 'G':
1200 case 'B':
1201 case 'U':
1202 case 'N':
1203 strbuf_addch(sb, c->signature_check.result);
1204 }
1205 break;
1206 case 'S':
1207 if (c->signature_check.signer)
1208 strbuf_addstr(sb, c->signature_check.signer);
1209 break;
1210 case 'K':
1211 if (c->signature_check.key)
1212 strbuf_addstr(sb, c->signature_check.key);
1213 break;
1214 default:
1215 return 0;
1216 }
1217 return 2;
1218 }
1219
1220
1221 /* For the rest we have to parse the commit header. */
1222 if (!c->commit_header_parsed)
1223 parse_commit_header(c);
1224
1225 switch (placeholder[0]) {
1226 case 'a': /* author ... */
1227 return format_person_part(sb, placeholder[1],
1228 msg + c->author.off, c->author.len,
1229 &c->pretty_ctx->date_mode);
1230 case 'c': /* committer ... */
1231 return format_person_part(sb, placeholder[1],
1232 msg + c->committer.off, c->committer.len,
1233 &c->pretty_ctx->date_mode);
1234 case 'e': /* encoding */
1235 if (c->commit_encoding)
1236 strbuf_addstr(sb, c->commit_encoding);
1237 return 1;
1238 case 'B': /* raw body */
1239 /* message_off is always left at the initial newline */
1240 strbuf_addstr(sb, msg + c->message_off + 1);
1241 return 1;
1242 }
1243
1244 /* Now we need to parse the commit message. */
1245 if (!c->commit_message_parsed)
1246 parse_commit_message(c);
1247
1248 switch (placeholder[0]) {
1249 case 's': /* subject */
1250 format_subject(sb, msg + c->subject_off, " ");
1251 return 1;
1252 case 'f': /* sanitized subject */
1253 format_sanitized_subject(sb, msg + c->subject_off);
1254 return 1;
1255 case 'b': /* body */
1256 strbuf_addstr(sb, msg + c->body_off);
1257 return 1;
1258 }
1259 return 0; /* unknown placeholder */
1260}
1261
1262static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1263 const char *placeholder,
1264 struct format_commit_context *c)
1265{
1266 struct strbuf local_sb = STRBUF_INIT;
1267 int total_consumed = 0, len, padding = c->padding;
1268 if (padding < 0) {
1269 const char *start = strrchr(sb->buf, '\n');
1270 int occupied;
1271 if (!start)
1272 start = sb->buf;
1273 occupied = utf8_strnwidth(start, -1, 1);
1274 occupied += c->pretty_ctx->graph_width;
1275 padding = (-padding) - occupied;
1276 }
1277 while (1) {
1278 int modifier = *placeholder == 'C';
1279 int consumed = format_commit_one(&local_sb, placeholder, c);
1280 total_consumed += consumed;
1281
1282 if (!modifier)
1283 break;
1284
1285 placeholder += consumed;
1286 if (*placeholder != '%')
1287 break;
1288 placeholder++;
1289 total_consumed++;
1290 }
1291 len = utf8_strnwidth(local_sb.buf, -1, 1);
1292
1293 if (c->flush_type == flush_left_and_steal) {
1294 const char *ch = sb->buf + sb->len - 1;
1295 while (len > padding && ch > sb->buf) {
1296 const char *p;
1297 if (*ch == ' ') {
1298 ch--;
1299 padding++;
1300 continue;
1301 }
1302 /* check for trailing ansi sequences */
1303 if (*ch != 'm')
1304 break;
1305 p = ch - 1;
1306 while (ch - p < 10 && *p != '\033')
1307 p--;
1308 if (*p != '\033' ||
1309 ch + 1 - p != display_mode_esc_sequence_len(p))
1310 break;
1311 /*
1312 * got a good ansi sequence, put it back to
1313 * local_sb as we're cutting sb
1314 */
1315 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1316 ch = p - 1;
1317 }
1318 strbuf_setlen(sb, ch + 1 - sb->buf);
1319 c->flush_type = flush_left;
1320 }
1321
1322 if (len > padding) {
1323 switch (c->truncate) {
1324 case trunc_left:
1325 strbuf_utf8_replace(&local_sb,
1326 0, len - (padding - 2),
1327 "..");
1328 break;
1329 case trunc_middle:
1330 strbuf_utf8_replace(&local_sb,
1331 padding / 2 - 1,
1332 len - (padding - 2),
1333 "..");
1334 break;
1335 case trunc_right:
1336 strbuf_utf8_replace(&local_sb,
1337 padding - 2, len - (padding - 2),
1338 "..");
1339 break;
1340 case trunc_none:
1341 break;
1342 }
1343 strbuf_addbuf(sb, &local_sb);
1344 } else {
1345 int sb_len = sb->len, offset = 0;
1346 if (c->flush_type == flush_left)
1347 offset = padding - len;
1348 else if (c->flush_type == flush_both)
1349 offset = (padding - len) / 2;
1350 /*
1351 * we calculate padding in columns, now
1352 * convert it back to chars
1353 */
1354 padding = padding - len + local_sb.len;
1355 strbuf_addchars(sb, ' ', padding);
1356 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1357 local_sb.len);
1358 }
1359 strbuf_release(&local_sb);
1360 c->flush_type = no_flush;
1361 return total_consumed;
1362}
1363
1364static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1365 const char *placeholder,
1366 void *context)
1367{
1368 int consumed;
1369 size_t orig_len;
1370 enum {
1371 NO_MAGIC,
1372 ADD_LF_BEFORE_NON_EMPTY,
1373 DEL_LF_BEFORE_EMPTY,
1374 ADD_SP_BEFORE_NON_EMPTY
1375 } magic = NO_MAGIC;
1376
1377 switch (placeholder[0]) {
1378 case '-':
1379 magic = DEL_LF_BEFORE_EMPTY;
1380 break;
1381 case '+':
1382 magic = ADD_LF_BEFORE_NON_EMPTY;
1383 break;
1384 case ' ':
1385 magic = ADD_SP_BEFORE_NON_EMPTY;
1386 break;
1387 default:
1388 break;
1389 }
1390 if (magic != NO_MAGIC)
1391 placeholder++;
1392
1393 orig_len = sb->len;
1394 if (((struct format_commit_context *)context)->flush_type != no_flush)
1395 consumed = format_and_pad_commit(sb, placeholder, context);
1396 else
1397 consumed = format_commit_one(sb, placeholder, context);
1398 if (magic == NO_MAGIC)
1399 return consumed;
1400
1401 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1402 while (sb->len && sb->buf[sb->len - 1] == '\n')
1403 strbuf_setlen(sb, sb->len - 1);
1404 } else if (orig_len != sb->len) {
1405 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1406 strbuf_insert(sb, orig_len, "\n", 1);
1407 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1408 strbuf_insert(sb, orig_len, " ", 1);
1409 }
1410 return consumed + 1;
1411}
1412
1413static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1414 void *context)
1415{
1416 struct userformat_want *w = context;
1417
1418 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1419 placeholder++;
1420
1421 switch (*placeholder) {
1422 case 'N':
1423 w->notes = 1;
1424 break;
1425 }
1426 return 0;
1427}
1428
1429void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1430{
1431 struct strbuf dummy = STRBUF_INIT;
1432
1433 if (!fmt) {
1434 if (!user_format)
1435 return;
1436 fmt = user_format;
1437 }
1438 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1439 strbuf_release(&dummy);
1440}
1441
1442void format_commit_message(const struct commit *commit,
1443 const char *format, struct strbuf *sb,
1444 const struct pretty_print_context *pretty_ctx)
1445{
1446 struct format_commit_context context;
1447 const char *output_enc = pretty_ctx->output_encoding;
1448 const char *utf8 = "UTF-8";
1449
1450 memset(&context, 0, sizeof(context));
1451 context.commit = commit;
1452 context.pretty_ctx = pretty_ctx;
1453 context.wrap_start = sb->len;
1454 /*
1455 * convert a commit message to UTF-8 first
1456 * as far as 'format_commit_item' assumes it in UTF-8
1457 */
1458 context.message = logmsg_reencode(commit,
1459 &context.commit_encoding,
1460 utf8);
1461
1462 strbuf_expand(sb, format, format_commit_item, &context);
1463 rewrap_message_tail(sb, &context, 0, 0, 0);
1464
1465 /* then convert a commit message to an actual output encoding */
1466 if (output_enc) {
1467 if (same_encoding(utf8, output_enc))
1468 output_enc = NULL;
1469 } else {
1470 if (context.commit_encoding &&
1471 !same_encoding(context.commit_encoding, utf8))
1472 output_enc = context.commit_encoding;
1473 }
1474
1475 if (output_enc) {
1476 int outsz;
1477 char *out = reencode_string_len(sb->buf, sb->len,
1478 output_enc, utf8, &outsz);
1479 if (out)
1480 strbuf_attach(sb, out, outsz, outsz + 1);
1481 }
1482
1483 free(context.commit_encoding);
1484 unuse_commit_buffer(commit, context.message);
1485}
1486
1487static void pp_header(struct pretty_print_context *pp,
1488 const char *encoding,
1489 const struct commit *commit,
1490 const char **msg_p,
1491 struct strbuf *sb)
1492{
1493 int parents_shown = 0;
1494
1495 for (;;) {
1496 const char *name, *line = *msg_p;
1497 int linelen = get_one_line(*msg_p);
1498
1499 if (!linelen)
1500 return;
1501 *msg_p += linelen;
1502
1503 if (linelen == 1)
1504 /* End of header */
1505 return;
1506
1507 if (pp->fmt == CMIT_FMT_RAW) {
1508 strbuf_add(sb, line, linelen);
1509 continue;
1510 }
1511
1512 if (starts_with(line, "parent ")) {
1513 if (linelen != 48)
1514 die("bad parent line in commit");
1515 continue;
1516 }
1517
1518 if (!parents_shown) {
1519 unsigned num = commit_list_count(commit->parents);
1520 /* with enough slop */
1521 strbuf_grow(sb, num * 50 + 20);
1522 add_merge_info(pp, sb, commit);
1523 parents_shown = 1;
1524 }
1525
1526 /*
1527 * MEDIUM == DEFAULT shows only author with dates.
1528 * FULL shows both authors but not dates.
1529 * FULLER shows both authors and dates.
1530 */
1531 if (skip_prefix(line, "author ", &name)) {
1532 strbuf_grow(sb, linelen + 80);
1533 pp_user_info(pp, "Author", sb, name, encoding);
1534 }
1535 if (skip_prefix(line, "committer ", &name) &&
1536 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1537 strbuf_grow(sb, linelen + 80);
1538 pp_user_info(pp, "Commit", sb, name, encoding);
1539 }
1540 }
1541}
1542
1543void pp_title_line(struct pretty_print_context *pp,
1544 const char **msg_p,
1545 struct strbuf *sb,
1546 const char *encoding,
1547 int need_8bit_cte)
1548{
1549 static const int max_length = 78; /* per rfc2047 */
1550 struct strbuf title;
1551
1552 strbuf_init(&title, 80);
1553 *msg_p = format_subject(&title, *msg_p,
1554 pp->preserve_subject ? "\n" : " ");
1555
1556 strbuf_grow(sb, title.len + 1024);
1557 if (pp->subject) {
1558 strbuf_addstr(sb, pp->subject);
1559 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1560 add_rfc2047(sb, title.buf, title.len,
1561 encoding, RFC2047_SUBJECT);
1562 else
1563 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1564 -last_line_length(sb), 1, max_length);
1565 } else {
1566 strbuf_addbuf(sb, &title);
1567 }
1568 strbuf_addch(sb, '\n');
1569
1570 if (need_8bit_cte == 0) {
1571 int i;
1572 for (i = 0; i < pp->in_body_headers.nr; i++) {
1573 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1574 need_8bit_cte = 1;
1575 break;
1576 }
1577 }
1578 }
1579
1580 if (need_8bit_cte > 0) {
1581 const char *header_fmt =
1582 "MIME-Version: 1.0\n"
1583 "Content-Type: text/plain; charset=%s\n"
1584 "Content-Transfer-Encoding: 8bit\n";
1585 strbuf_addf(sb, header_fmt, encoding);
1586 }
1587 if (pp->after_subject) {
1588 strbuf_addstr(sb, pp->after_subject);
1589 }
1590 if (cmit_fmt_is_mail(pp->fmt)) {
1591 strbuf_addch(sb, '\n');
1592 }
1593
1594 if (pp->in_body_headers.nr) {
1595 int i;
1596 for (i = 0; i < pp->in_body_headers.nr; i++) {
1597 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1598 free(pp->in_body_headers.items[i].string);
1599 }
1600 string_list_clear(&pp->in_body_headers, 0);
1601 strbuf_addch(sb, '\n');
1602 }
1603
1604 strbuf_release(&title);
1605}
1606
1607static int pp_utf8_width(const char *start, const char *end)
1608{
1609 int width = 0;
1610 size_t remain = end - start;
1611
1612 while (remain) {
1613 int n = utf8_width(&start, &remain);
1614 if (n < 0 || !start)
1615 return -1;
1616 width += n;
1617 }
1618 return width;
1619}
1620
1621static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1622 const char *line, int linelen)
1623{
1624 const char *tab;
1625
1626 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1627 int width = pp_utf8_width(line, tab);
1628
1629 /*
1630 * If it wasn't well-formed utf8, or it
1631 * had characters with badly defined
1632 * width (control characters etc), just
1633 * give up on trying to align things.
1634 */
1635 if (width < 0)
1636 break;
1637
1638 /* Output the data .. */
1639 strbuf_add(sb, line, tab - line);
1640
1641 /* .. and the de-tabified tab */
1642 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1643
1644 /* Skip over the printed part .. */
1645 linelen -= tab + 1 - line;
1646 line = tab + 1;
1647 }
1648
1649 /*
1650 * Print out everything after the last tab without
1651 * worrying about width - there's nothing more to
1652 * align.
1653 */
1654 strbuf_add(sb, line, linelen);
1655}
1656
1657/*
1658 * pp_handle_indent() prints out the intendation, and
1659 * the whole line (without the final newline), after
1660 * de-tabifying.
1661 */
1662static void pp_handle_indent(struct pretty_print_context *pp,
1663 struct strbuf *sb, int indent,
1664 const char *line, int linelen)
1665{
1666 strbuf_addchars(sb, ' ', indent);
1667 if (pp->expand_tabs_in_log)
1668 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1669 else
1670 strbuf_add(sb, line, linelen);
1671}
1672
1673static int is_mboxrd_from(const char *line, int len)
1674{
1675 /*
1676 * a line matching /^From $/ here would only have len == 4
1677 * at this point because is_empty_line would've trimmed all
1678 * trailing space
1679 */
1680 return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1681}
1682
1683void pp_remainder(struct pretty_print_context *pp,
1684 const char **msg_p,
1685 struct strbuf *sb,
1686 int indent)
1687{
1688 int first = 1;
1689 for (;;) {
1690 const char *line = *msg_p;
1691 int linelen = get_one_line(line);
1692 *msg_p += linelen;
1693
1694 if (!linelen)
1695 break;
1696
1697 if (is_blank_line(line, &linelen)) {
1698 if (first)
1699 continue;
1700 if (pp->fmt == CMIT_FMT_SHORT)
1701 break;
1702 }
1703 first = 0;
1704
1705 strbuf_grow(sb, linelen + indent + 20);
1706 if (indent)
1707 pp_handle_indent(pp, sb, indent, line, linelen);
1708 else if (pp->expand_tabs_in_log)
1709 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1710 line, linelen);
1711 else {
1712 if (pp->fmt == CMIT_FMT_MBOXRD &&
1713 is_mboxrd_from(line, linelen))
1714 strbuf_addch(sb, '>');
1715
1716 strbuf_add(sb, line, linelen);
1717 }
1718 strbuf_addch(sb, '\n');
1719 }
1720}
1721
1722void pretty_print_commit(struct pretty_print_context *pp,
1723 const struct commit *commit,
1724 struct strbuf *sb)
1725{
1726 unsigned long beginning_of_body;
1727 int indent = 4;
1728 const char *msg;
1729 const char *reencoded;
1730 const char *encoding;
1731 int need_8bit_cte = pp->need_8bit_cte;
1732
1733 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1734 format_commit_message(commit, user_format, sb, pp);
1735 return;
1736 }
1737
1738 encoding = get_log_output_encoding();
1739 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1740
1741 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1742 indent = 0;
1743
1744 /*
1745 * We need to check and emit Content-type: to mark it
1746 * as 8-bit if we haven't done so.
1747 */
1748 if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1749 int i, ch, in_body;
1750
1751 for (in_body = i = 0; (ch = msg[i]); i++) {
1752 if (!in_body) {
1753 /* author could be non 7-bit ASCII but
1754 * the log may be so; skip over the
1755 * header part first.
1756 */
1757 if (ch == '\n' && msg[i+1] == '\n')
1758 in_body = 1;
1759 }
1760 else if (non_ascii(ch)) {
1761 need_8bit_cte = 1;
1762 break;
1763 }
1764 }
1765 }
1766
1767 pp_header(pp, encoding, commit, &msg, sb);
1768 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1769 strbuf_addch(sb, '\n');
1770 }
1771
1772 /* Skip excess blank lines at the beginning of body, if any... */
1773 msg = skip_blank_lines(msg);
1774
1775 /* These formats treat the title line specially. */
1776 if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1777 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1778
1779 beginning_of_body = sb->len;
1780 if (pp->fmt != CMIT_FMT_ONELINE)
1781 pp_remainder(pp, &msg, sb, indent);
1782 strbuf_rtrim(sb);
1783
1784 /* Make sure there is an EOLN for the non-oneline case */
1785 if (pp->fmt != CMIT_FMT_ONELINE)
1786 strbuf_addch(sb, '\n');
1787
1788 /*
1789 * The caller may append additional body text in e-mail
1790 * format. Make sure we did not strip the blank line
1791 * between the header and the body.
1792 */
1793 if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1794 strbuf_addch(sb, '\n');
1795
1796 unuse_commit_buffer(commit, reencoded);
1797}
1798
1799void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1800 struct strbuf *sb)
1801{
1802 struct pretty_print_context pp = {0};
1803 pp.fmt = fmt;
1804 pretty_print_commit(&pp, commit, sb);
1805}