?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
i18n: fix misconversion in shell scripts
[git/debian.git] / pretty.c
blob87c44971a1d2070e4d256818e96fccc29472ec4a
1 #include "cache.h"
2 #include "commit.h"
3 #include "utf8.h"
4 #include "diff.h"
5 #include "revision.h"
6 #include "string-list.h"
7 #include "mailmap.h"
8 #include "log-tree.h"
9 #include "notes.h"
10 #include "color.h"
11 #include "reflog-walk.h"
12 #include "gpg-interface.h"
13
14 static char *user_format;
15 static struct cmt_fmt_map {
16 const char *name;
17 enum cmit_fmt format;
18 int is_tformat;
19 int expand_tabs_in_log;
20 int is_alias;
21 const char *user_format;
22 } *commit_formats;
23 static size_t builtin_formats_len;
24 static size_t commit_formats_len;
25 static size_t commit_formats_alloc;
26 static struct cmt_fmt_map *find_commit_format(const char *sought);
27
28 int commit_format_is_empty(enum cmit_fmt fmt)
29 {
30 return fmt == CMIT_FMT_USERFORMAT && !*user_format;
31 }
32
33 static 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
42 static 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
88 static 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 { "fuller", CMIT_FMT_FULLER, 0, 8 },
96 { "full", CMIT_FMT_FULL, 0, 8 },
97 { "oneline", CMIT_FMT_ONELINE, 1, 0 }
98 };
99 commit_formats_len = ARRAY_SIZE(builtin_formats);
100 builtin_formats_len = commit_formats_len;
101 ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
102 memcpy(commit_formats, builtin_formats,
103 sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
104
105 git_config(git_pretty_formats_config, NULL);
106 }
107
108 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
109 const char *original,
110 int num_redirections)
111 {
112 struct cmt_fmt_map *found = NULL;
113 size_t found_match_len = 0;
114 int i;
115
116 if (num_redirections >= commit_formats_len)
117 die("invalid --pretty format: "
118 "'%s' references an alias which points to itself",
119 original);
120
121 for (i = 0; i < commit_formats_len; i++) {
122 size_t match_len;
123
124 if (!starts_with(commit_formats[i].name, sought))
125 continue;
126
127 match_len = strlen(commit_formats[i].name);
128 if (found == NULL || found_match_len > match_len) {
129 found = &commit_formats[i];
130 found_match_len = match_len;
131 }
132 }
133
134 if (found && found->is_alias) {
135 found = find_commit_format_recursive(found->user_format,
136 original,
137 num_redirections+1);
138 }
139
140 return found;
141 }
142
143 static struct cmt_fmt_map *find_commit_format(const char *sought)
144 {
145 if (!commit_formats)
146 setup_commit_formats();
147
148 return find_commit_format_recursive(sought, sought, 0);
149 }
150
151 void get_commit_format(const char *arg, struct rev_info *rev)
152 {
153 struct cmt_fmt_map *commit_format;
154
155 rev->use_terminator = 0;
156 if (!arg) {
157 rev->commit_format = CMIT_FMT_DEFAULT;
158 return;
159 }
160 if (skip_prefix(arg, "format:", &arg)) {
161 save_user_format(rev, arg, 0);
162 return;
163 }
164
165 if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
166 save_user_format(rev, arg, 1);
167 return;
168 }
169
170 commit_format = find_commit_format(arg);
171 if (!commit_format)
172 die("invalid --pretty format: %s", arg);
173
174 rev->commit_format = commit_format->format;
175 rev->use_terminator = commit_format->is_tformat;
176 rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
177 if (commit_format->format == CMIT_FMT_USERFORMAT) {
178 save_user_format(rev, commit_format->user_format,
179 commit_format->is_tformat);
180 }
181 }
182
183 /*
184 * Generic support for pretty-printing the header
185 */
186 static int get_one_line(const char *msg)
187 {
188 int ret = 0;
189
190 for (;;) {
191 char c = *msg++;
192 if (!c)
193 break;
194 ret++;
195 if (c == '\n')
196 break;
197 }
198 return ret;
199 }
200
201 /* High bit set, or ISO-2022-INT */
202 static int non_ascii(int ch)
203 {
204 return !isascii(ch) || ch == '\033';
205 }
206
207 int has_non_ascii(const char *s)
208 {
209 int ch;
210 if (!s)
211 return 0;
212 while ((ch = *s++) != '\0') {
213 if (non_ascii(ch))
214 return 1;
215 }
216 return 0;
217 }
218
219 static int is_rfc822_special(char ch)
220 {
221 switch (ch) {
222 case '(':
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 return 1;
236 default:
237 return 0;
238 }
239 }
240
241 static int needs_rfc822_quoting(const char *s, int len)
242 {
243 int i;
244 for (i = 0; i < len; i++)
245 if (is_rfc822_special(s[i]))
246 return 1;
247 return 0;
248 }
249
250 static int last_line_length(struct strbuf *sb)
251 {
252 int i;
253
254 /* How many bytes are already used on the last line? */
255 for (i = sb->len - 1; i >= 0; i--)
256 if (sb->buf[i] == '\n')
257 break;
258 return sb->len - (i + 1);
259 }
260
261 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
262 {
263 int i;
264
265 /* just a guess, we may have to also backslash-quote */
266 strbuf_grow(out, len + 2);
267
268 strbuf_addch(out, '"');
269 for (i = 0; i < len; i++) {
270 switch (s[i]) {
271 case '"':
272 case '\\':
273 strbuf_addch(out, '\\');
274 /* fall through */
275 default:
276 strbuf_addch(out, s[i]);
277 }
278 }
279 strbuf_addch(out, '"');
280 }
281
282 enum rfc2047_type {
283 RFC2047_SUBJECT,
284 RFC2047_ADDRESS
285 };
286
287 static int is_rfc2047_special(char ch, enum rfc2047_type type)
288 {
289 /*
290 * rfc2047, section 4.2:
291 *
292 * 8-bit values which correspond to printable ASCII characters other
293 * than "=", "?", and "_" (underscore), MAY be represented as those
294 * characters. (But see section 5 for restrictions.) In
295 * particular, SPACE and TAB MUST NOT be represented as themselves
296 * within encoded words.
297 */
298
299 /*
300 * rule out non-ASCII characters and non-printable characters (the
301 * non-ASCII check should be redundant as isprint() is not localized
302 * and only knows about ASCII, but be defensive about that)
303 */
304 if (non_ascii(ch) || !isprint(ch))
305 return 1;
306
307 /*
308 * rule out special printable characters (' ' should be the only
309 * whitespace character considered printable, but be defensive and use
310 * isspace())
311 */
312 if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
313 return 1;
314
315 /*
316 * rfc2047, section 5.3:
317 *
318 * As a replacement for a 'word' entity within a 'phrase', for example,
319 * one that precedes an address in a From, To, or Cc header. The ABNF
320 * definition for 'phrase' from RFC 822 thus becomes:
321 *
322 * phrase = 1*( encoded-word / word )
323 *
324 * In this case the set of characters that may be used in a "Q"-encoded
325 * 'encoded-word' is restricted to: <upper and lower case ASCII
326 * letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
327 * (underscore, ASCII 95.)>. An 'encoded-word' that appears within a
328 * 'phrase' MUST be separated from any adjacent 'word', 'text' or
329 * 'special' by 'linear-white-space'.
330 */
331
332 if (type != RFC2047_ADDRESS)
333 return 0;
334
335 /* '=' and '_' are special cases and have been checked above */
336 return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
337 }
338
339 static int needs_rfc2047_encoding(const char *line, int len,
340 enum rfc2047_type type)
341 {
342 int i;
343
344 for (i = 0; i < len; i++) {
345 int ch = line[i];
346 if (non_ascii(ch) || ch == '\n')
347 return 1;
348 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
349 return 1;
350 }
351
352 return 0;
353 }
354
355 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
356 const char *encoding, enum rfc2047_type type)
357 {
358 static const int max_encoded_length = 76; /* per rfc2047 */
359 int i;
360 int line_len = last_line_length(sb);
361
362 strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
363 strbuf_addf(sb, "=?%s?q?", encoding);
364 line_len += strlen(encoding) + 5; /* 5 for =??q? */
365
366 while (len) {
367 /*
368 * RFC 2047, section 5 (3):
369 *
370 * Each 'encoded-word' MUST represent an integral number of
371 * characters. A multi-octet character may not be split across
372 * adjacent 'encoded- word's.
373 */
374 const unsigned char *p = (const unsigned char *)line;
375 int chrlen = mbs_chrlen(&line, &len, encoding);
376 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
377
378 /* "=%02X" * chrlen, or the byte itself */
379 const char *encoded_fmt = is_special ? "=%02X" : "%c";
380 int encoded_len = is_special ? 3 * chrlen : 1;
381
382 /*
383 * According to RFC 2047, we could encode the special character
384 * ' ' (space) with '_' (underscore) for readability. But many
385 * programs do not understand this and just leave the
386 * underscore in place. Thus, we do nothing special here, which
387 * causes ' ' to be encoded as '=20', avoiding this problem.
388 */
389
390 if (line_len + encoded_len + 2 > max_encoded_length) {
391 /* It won't fit with trailing "?=" --- break the line */
392 strbuf_addf(sb, "?=\n =?%s?q?", encoding);
393 line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
394 }
395
396 for (i = 0; i < chrlen; i++)
397 strbuf_addf(sb, encoded_fmt, p[i]);
398 line_len += encoded_len;
399 }
400 strbuf_addstr(sb, "?=");
401 }
402
403 const char *show_ident_date(const struct ident_split *ident,
404 const struct date_mode *mode)
405 {
406 unsigned long date = 0;
407 long tz = 0;
408
409 if (ident->date_begin && ident->date_end)
410 date = strtoul(ident->date_begin, NULL, 10);
411 if (date_overflows(date))
412 date = 0;
413 else {
414 if (ident->tz_begin && ident->tz_end)
415 tz = strtol(ident->tz_begin, NULL, 10);
416 if (tz >= INT_MAX || tz <= INT_MIN)
417 tz = 0;
418 }
419 return show_date(date, tz, mode);
420 }
421
422 void pp_user_info(struct pretty_print_context *pp,
423 const char *what, struct strbuf *sb,
424 const char *line, const char *encoding)
425 {
426 struct ident_split ident;
427 char *line_end;
428 const char *mailbuf, *namebuf;
429 size_t namelen, maillen;
430 int max_length = 78; /* per rfc2822 */
431
432 if (pp->fmt == CMIT_FMT_ONELINE)
433 return;
434
435 line_end = strchrnul(line, '\n');
436 if (split_ident_line(&ident, line, line_end - line))
437 return;
438
439 mailbuf = ident.mail_begin;
440 maillen = ident.mail_end - ident.mail_begin;
441 namebuf = ident.name_begin;
442 namelen = ident.name_end - ident.name_begin;
443
444 if (pp->mailmap)
445 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
446
447 if (pp->fmt == CMIT_FMT_EMAIL) {
448 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
449 struct strbuf buf = STRBUF_INIT;
450
451 strbuf_addstr(&buf, "From: ");
452 strbuf_add(&buf, namebuf, namelen);
453 strbuf_addstr(&buf, " <");
454 strbuf_add(&buf, mailbuf, maillen);
455 strbuf_addstr(&buf, ">\n");
456 string_list_append(&pp->in_body_headers,
457 strbuf_detach(&buf, NULL));
458
459 mailbuf = pp->from_ident->mail_begin;
460 maillen = pp->from_ident->mail_end - mailbuf;
461 namebuf = pp->from_ident->name_begin;
462 namelen = pp->from_ident->name_end - namebuf;
463 }
464
465 strbuf_addstr(sb, "From: ");
466 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
467 add_rfc2047(sb, namebuf, namelen,
468 encoding, RFC2047_ADDRESS);
469 max_length = 76; /* per rfc2047 */
470 } else if (needs_rfc822_quoting(namebuf, namelen)) {
471 struct strbuf quoted = STRBUF_INIT;
472 add_rfc822_quoted(&quoted, namebuf, namelen);
473 strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
474 -6, 1, max_length);
475 strbuf_release(&quoted);
476 } else {
477 strbuf_add_wrapped_bytes(sb, namebuf, namelen,
478 -6, 1, max_length);
479 }
480
481 if (max_length <
482 last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
483 strbuf_addch(sb, '\n');
484 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
485 } else {
486 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
487 (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, " ",
488 (int)namelen, namebuf, (int)maillen, mailbuf);
489 }
490
491 switch (pp->fmt) {
492 case CMIT_FMT_MEDIUM:
493 strbuf_addf(sb, "Date: %s\n",
494 show_ident_date(&ident, &pp->date_mode));
495 break;
496 case CMIT_FMT_EMAIL:
497 strbuf_addf(sb, "Date: %s\n",
498 show_ident_date(&ident, DATE_MODE(RFC2822)));
499 break;
500 case CMIT_FMT_FULLER:
501 strbuf_addf(sb, "%sDate: %s\n", what,
502 show_ident_date(&ident, &pp->date_mode));
503 break;
504 default:
505 /* notin' */
506 break;
507 }
508 }
509
510 static int is_empty_line(const char *line, int *len_p)
511 {
512 int len = *len_p;
513 while (len && isspace(line[len - 1]))
514 len--;
515 *len_p = len;
516 return !len;
517 }
518
519 static const char *skip_empty_lines(const char *msg)
520 {
521 for (;;) {
522 int linelen = get_one_line(msg);
523 int ll = linelen;
524 if (!linelen)
525 break;
526 if (!is_empty_line(msg, &ll))
527 break;
528 msg += linelen;
529 }
530 return msg;
531 }
532
533 static void add_merge_info(const struct pretty_print_context *pp,
534 struct strbuf *sb, const struct commit *commit)
535 {
536 struct commit_list *parent = commit->parents;
537
538 if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
539 !parent || !parent->next)
540 return;
541
542 strbuf_addstr(sb, "Merge:");
543
544 while (parent) {
545 struct commit *p = parent->item;
546 const char *hex = NULL;
547 if (pp->abbrev)
548 hex = find_unique_abbrev(p->object.oid.hash, pp->abbrev);
549 if (!hex)
550 hex = oid_to_hex(&p->object.oid);
551 parent = parent->next;
552
553 strbuf_addf(sb, " %s", hex);
554 }
555 strbuf_addch(sb, '\n');
556 }
557
558 static 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
565 static 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
596 const 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
664 static 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
675 static 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
730 skip:
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
743 struct chunk {
744 size_t off;
745 size_t len;
746 };
747
748 enum flush_type {
749 no_flush,
750 flush_right,
751 flush_left,
752 flush_left_and_steal,
753 flush_both
754 };
755
756 enum trunc_type {
757 trunc_none,
758 trunc_left,
759 trunc_middle,
760 trunc_right
761 };
762
763 struct 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 struct chunk abbrev_commit_hash;
786 struct chunk abbrev_tree_hash;
787 struct chunk abbrev_parent_hashes;
788 size_t wrap_start;
789 };
790
791 static int add_again(struct strbuf *sb, struct chunk *chunk)
792 {
793 if (chunk->len) {
794 strbuf_adddup(sb, chunk->off, chunk->len);
795 return 1;
796 }
797
798 /*
799 * We haven't seen this chunk before. Our caller is surely
800 * going to add it the hard way now. Remember the most likely
801 * start of the to-be-added chunk: the current end of the
802 * struct strbuf.
803 */
804 chunk->off = sb->len;
805 return 0;
806 }
807
808 static void parse_commit_header(struct format_commit_context *context)
809 {
810 const char *msg = context->message;
811 int i;
812
813 for (i = 0; msg[i]; i++) {
814 const char *name;
815 int eol;
816 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
817 ; /* do nothing */
818
819 if (i == eol) {
820 break;
821 } else if (skip_prefix(msg + i, "author ", &name)) {
822 context->author.off = name - msg;
823 context->author.len = msg + eol - name;
824 } else if (skip_prefix(msg + i, "committer ", &name)) {
825 context->committer.off = name - msg;
826 context->committer.len = msg + eol - name;
827 }
828 i = eol;
829 }
830 context->message_off = i;
831 context->commit_header_parsed = 1;
832 }
833
834 static int istitlechar(char c)
835 {
836 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
837 (c >= '0' && c <= '9') || c == '.' || c == '_';
838 }
839
840 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
841 {
842 size_t trimlen;
843 size_t start_len = sb->len;
844 int space = 2;
845
846 for (; *msg && *msg != '\n'; msg++) {
847 if (istitlechar(*msg)) {
848 if (space == 1)
849 strbuf_addch(sb, '-');
850 space = 0;
851 strbuf_addch(sb, *msg);
852 if (*msg == '.')
853 while (*(msg+1) == '.')
854 msg++;
855 } else
856 space |= 1;
857 }
858
859 /* trim any trailing '.' or '-' characters */
860 trimlen = 0;
861 while (sb->len - trimlen > start_len &&
862 (sb->buf[sb->len - 1 - trimlen] == '.'
863 || sb->buf[sb->len - 1 - trimlen] == '-'))
864 trimlen++;
865 strbuf_remove(sb, sb->len - trimlen, trimlen);
866 }
867
868 const char *format_subject(struct strbuf *sb, const char *msg,
869 const char *line_separator)
870 {
871 int first = 1;
872
873 for (;;) {
874 const char *line = msg;
875 int linelen = get_one_line(line);
876
877 msg += linelen;
878 if (!linelen || is_empty_line(line, &linelen))
879 break;
880
881 if (!sb)
882 continue;
883 strbuf_grow(sb, linelen + 2);
884 if (!first)
885 strbuf_addstr(sb, line_separator);
886 strbuf_add(sb, line, linelen);
887 first = 0;
888 }
889 return msg;
890 }
891
892 static void parse_commit_message(struct format_commit_context *c)
893 {
894 const char *msg = c->message + c->message_off;
895 const char *start = c->message;
896
897 msg = skip_empty_lines(msg);
898 c->subject_off = msg - start;
899
900 msg = format_subject(NULL, msg, NULL);
901 msg = skip_empty_lines(msg);
902 c->body_off = msg - start;
903
904 c->commit_message_parsed = 1;
905 }
906
907 static void strbuf_wrap(struct strbuf *sb, size_t pos,
908 size_t width, size_t indent1, size_t indent2)
909 {
910 struct strbuf tmp = STRBUF_INIT;
911
912 if (pos)
913 strbuf_add(&tmp, sb->buf, pos);
914 strbuf_add_wrapped_text(&tmp, sb->buf + pos,
915 (int) indent1, (int) indent2, (int) width);
916 strbuf_swap(&tmp, sb);
917 strbuf_release(&tmp);
918 }
919
920 static void rewrap_message_tail(struct strbuf *sb,
921 struct format_commit_context *c,
922 size_t new_width, size_t new_indent1,
923 size_t new_indent2)
924 {
925 if (c->width == new_width && c->indent1 == new_indent1 &&
926 c->indent2 == new_indent2)
927 return;
928 if (c->wrap_start < sb->len)
929 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
930 c->wrap_start = sb->len;
931 c->width = new_width;
932 c->indent1 = new_indent1;
933 c->indent2 = new_indent2;
934 }
935
936 static int format_reflog_person(struct strbuf *sb,
937 char part,
938 struct reflog_walk_info *log,
939 const struct date_mode *dmode)
940 {
941 const char *ident;
942
943 if (!log)
944 return 2;
945
946 ident = get_reflog_ident(log);
947 if (!ident)
948 return 2;
949
950 return format_person_part(sb, part, ident, strlen(ident), dmode);
951 }
952
953 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
954 const char *placeholder,
955 struct format_commit_context *c)
956 {
957 const char *rest = placeholder;
958
959 if (placeholder[1] == '(') {
960 const char *begin = placeholder + 2;
961 const char *end = strchr(begin, ')');
962 char color[COLOR_MAXLEN];
963
964 if (!end)
965 return 0;
966 if (skip_prefix(begin, "auto,", &begin)) {
967 if (!want_color(c->pretty_ctx->color))
968 return end - placeholder + 1;
969 }
970 if (color_parse_mem(begin, end - begin, color) < 0)
971 die(_("unable to parse --pretty format"));
972 strbuf_addstr(sb, color);
973 return end - placeholder + 1;
974 }
975 if (skip_prefix(placeholder + 1, "red", &rest))
976 strbuf_addstr(sb, GIT_COLOR_RED);
977 else if (skip_prefix(placeholder + 1, "green", &rest))
978 strbuf_addstr(sb, GIT_COLOR_GREEN);
979 else if (skip_prefix(placeholder + 1, "blue", &rest))
980 strbuf_addstr(sb, GIT_COLOR_BLUE);
981 else if (skip_prefix(placeholder + 1, "reset", &rest))
982 strbuf_addstr(sb, GIT_COLOR_RESET);
983 return rest - placeholder;
984 }
985
986 static size_t parse_padding_placeholder(struct strbuf *sb,
987 const char *placeholder,
988 struct format_commit_context *c)
989 {
990 const char *ch = placeholder;
991 enum flush_type flush_type;
992 int to_column = 0;
993
994 switch (*ch++) {
995 case '<':
996 flush_type = flush_right;
997 break;
998 case '>':
999 if (*ch == '<') {
1000 flush_type = flush_both;
1001 ch++;
1002 } else if (*ch == '>') {
1003 flush_type = flush_left_and_steal;
1004 ch++;
1005 } else
1006 flush_type = flush_left;
1007 break;
1008 default:
1009 return 0;
1010 }
1011
1012 /* the next value means "wide enough to that column" */
1013 if (*ch == '|') {
1014 to_column = 1;
1015 ch++;
1016 }
1017
1018 if (*ch == '(') {
1019 const char *start = ch + 1;
1020 const char *end = start + strcspn(start, ",)");
1021 char *next;
1022 int width;
1023 if (!end || end == start)
1024 return 0;
1025 width = strtoul(start, &next, 10);
1026 if (next == start || width == 0)
1027 return 0;
1028 c->padding = to_column ? -width : width;
1029 c->flush_type = flush_type;
1030
1031 if (*end == ',') {
1032 start = end + 1;
1033 end = strchr(start, ')');
1034 if (!end || end == start)
1035 return 0;
1036 if (starts_with(start, "trunc)"))
1037 c->truncate = trunc_right;
1038 else if (starts_with(start, "ltrunc)"))
1039 c->truncate = trunc_left;
1040 else if (starts_with(start, "mtrunc)"))
1041 c->truncate = trunc_middle;
1042 else
1043 return 0;
1044 } else
1045 c->truncate = trunc_none;
1046
1047 return end - placeholder + 1;
1048 }
1049 return 0;
1050 }
1051
1052 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1053 const char *placeholder,
1054 void *context)
1055 {
1056 struct format_commit_context *c = context;
1057 const struct commit *commit = c->commit;
1058 const char *msg = c->message;
1059 struct commit_list *p;
1060 int h1, h2;
1061
1062 /* these are independent of the commit */
1063 switch (placeholder[0]) {
1064 case 'C':
1065 if (starts_with(placeholder + 1, "(auto)")) {
1066 c->auto_color = 1;
1067 return 7; /* consumed 7 bytes, "C(auto)" */
1068 } else {
1069 int ret = parse_color(sb, placeholder, c);
1070 if (ret)
1071 c->auto_color = 0;
1072 /*
1073 * Otherwise, we decided to treat %C<unknown>
1074 * as a literal string, and the previous
1075 * %C(auto) is still valid.
1076 */
1077 return ret;
1078 }
1079 case 'n': /* newline */
1080 strbuf_addch(sb, '\n');
1081 return 1;
1082 case 'x':
1083 /* %x00 == NUL, %x0a == LF, etc. */
1084 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1085 h1 <= 16 &&
1086 0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1087 h2 <= 16) {
1088 strbuf_addch(sb, (h1<<4)|h2);
1089 return 3;
1090 } else
1091 return 0;
1092 case 'w':
1093 if (placeholder[1] == '(') {
1094 unsigned long width = 0, indent1 = 0, indent2 = 0;
1095 char *next;
1096 const char *start = placeholder + 2;
1097 const char *end = strchr(start, ')');
1098 if (!end)
1099 return 0;
1100 if (end > start) {
1101 width = strtoul(start, &next, 10);
1102 if (*next == ',') {
1103 indent1 = strtoul(next + 1, &next, 10);
1104 if (*next == ',') {
1105 indent2 = strtoul(next + 1,
1106 &next, 10);
1107 }
1108 }
1109 if (*next != ')')
1110 return 0;
1111 }
1112 rewrap_message_tail(sb, c, width, indent1, indent2);
1113 return end - placeholder + 1;
1114 } else
1115 return 0;
1116
1117 case '<':
1118 case '>':
1119 return parse_padding_placeholder(sb, placeholder, c);
1120 }
1121
1122 /* these depend on the commit */
1123 if (!commit->object.parsed)
1124 parse_object(commit->object.oid.hash);
1125
1126 switch (placeholder[0]) {
1127 case 'H': /* commit hash */
1128 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1129 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1130 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1131 return 1;
1132 case 'h': /* abbreviated commit hash */
1133 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1134 if (add_again(sb, &c->abbrev_commit_hash)) {
1135 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1136 return 1;
1137 }
1138 strbuf_addstr(sb, find_unique_abbrev(commit->object.oid.hash,
1139 c->pretty_ctx->abbrev));
1140 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1141 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1142 return 1;
1143 case 'T': /* tree hash */
1144 strbuf_addstr(sb, oid_to_hex(&commit->tree->object.oid));
1145 return 1;
1146 case 't': /* abbreviated tree hash */
1147 if (add_again(sb, &c->abbrev_tree_hash))
1148 return 1;
1149 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.oid.hash,
1150 c->pretty_ctx->abbrev));
1151 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1152 return 1;
1153 case 'P': /* parent hashes */
1154 for (p = commit->parents; p; p = p->next) {
1155 if (p != commit->parents)
1156 strbuf_addch(sb, ' ');
1157 strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1158 }
1159 return 1;
1160 case 'p': /* abbreviated parent hashes */
1161 if (add_again(sb, &c->abbrev_parent_hashes))
1162 return 1;
1163 for (p = commit->parents; p; p = p->next) {
1164 if (p != commit->parents)
1165 strbuf_addch(sb, ' ');
1166 strbuf_addstr(sb, find_unique_abbrev(
1167 p->item->object.oid.hash,
1168 c->pretty_ctx->abbrev));
1169 }
1170 c->abbrev_parent_hashes.len = sb->len -
1171 c->abbrev_parent_hashes.off;
1172 return 1;
1173 case 'm': /* left/right/bottom */
1174 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1175 return 1;
1176 case 'd':
1177 load_ref_decorations(DECORATE_SHORT_REFS);
1178 format_decorations(sb, commit, c->auto_color);
1179 return 1;
1180 case 'D':
1181 load_ref_decorations(DECORATE_SHORT_REFS);
1182 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1183 return 1;
1184 case 'g': /* reflog info */
1185 switch(placeholder[1]) {
1186 case 'd': /* reflog selector */
1187 case 'D':
1188 if (c->pretty_ctx->reflog_info)
1189 get_reflog_selector(sb,
1190 c->pretty_ctx->reflog_info,
1191 &c->pretty_ctx->date_mode,
1192 c->pretty_ctx->date_mode_explicit,
1193 (placeholder[1] == 'd'));
1194 return 2;
1195 case 's': /* reflog message */
1196 if (c->pretty_ctx->reflog_info)
1197 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1198 return 2;
1199 case 'n':
1200 case 'N':
1201 case 'e':
1202 case 'E':
1203 return format_reflog_person(sb,
1204 placeholder[1],
1205 c->pretty_ctx->reflog_info,
1206 &c->pretty_ctx->date_mode);
1207 }
1208 return 0; /* unknown %g placeholder */
1209 case 'N':
1210 if (c->pretty_ctx->notes_message) {
1211 strbuf_addstr(sb, c->pretty_ctx->notes_message);
1212 return 1;
1213 }
1214 return 0;
1215 }
1216
1217 if (placeholder[0] == 'G') {
1218 if (!c->signature_check.result)
1219 check_commit_signature(c->commit, &(c->signature_check));
1220 switch (placeholder[1]) {
1221 case 'G':
1222 if (c->signature_check.gpg_output)
1223 strbuf_addstr(sb, c->signature_check.gpg_output);
1224 break;
1225 case '?':
1226 switch (c->signature_check.result) {
1227 case 'G':
1228 case 'B':
1229 case 'U':
1230 case 'N':
1231 strbuf_addch(sb, c->signature_check.result);
1232 }
1233 break;
1234 case 'S':
1235 if (c->signature_check.signer)
1236 strbuf_addstr(sb, c->signature_check.signer);
1237 break;
1238 case 'K':
1239 if (c->signature_check.key)
1240 strbuf_addstr(sb, c->signature_check.key);
1241 break;
1242 default:
1243 return 0;
1244 }
1245 return 2;
1246 }
1247
1248
1249 /* For the rest we have to parse the commit header. */
1250 if (!c->commit_header_parsed)
1251 parse_commit_header(c);
1252
1253 switch (placeholder[0]) {
1254 case 'a': /* author ... */
1255 return format_person_part(sb, placeholder[1],
1256 msg + c->author.off, c->author.len,
1257 &c->pretty_ctx->date_mode);
1258 case 'c': /* committer ... */
1259 return format_person_part(sb, placeholder[1],
1260 msg + c->committer.off, c->committer.len,
1261 &c->pretty_ctx->date_mode);
1262 case 'e': /* encoding */
1263 if (c->commit_encoding)
1264 strbuf_addstr(sb, c->commit_encoding);
1265 return 1;
1266 case 'B': /* raw body */
1267 /* message_off is always left at the initial newline */
1268 strbuf_addstr(sb, msg + c->message_off + 1);
1269 return 1;
1270 }
1271
1272 /* Now we need to parse the commit message. */
1273 if (!c->commit_message_parsed)
1274 parse_commit_message(c);
1275
1276 switch (placeholder[0]) {
1277 case 's': /* subject */
1278 format_subject(sb, msg + c->subject_off, " ");
1279 return 1;
1280 case 'f': /* sanitized subject */
1281 format_sanitized_subject(sb, msg + c->subject_off);
1282 return 1;
1283 case 'b': /* body */
1284 strbuf_addstr(sb, msg + c->body_off);
1285 return 1;
1286 }
1287 return 0; /* unknown placeholder */
1288 }
1289
1290 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1291 const char *placeholder,
1292 struct format_commit_context *c)
1293 {
1294 struct strbuf local_sb = STRBUF_INIT;
1295 int total_consumed = 0, len, padding = c->padding;
1296 if (padding < 0) {
1297 const char *start = strrchr(sb->buf, '\n');
1298 int occupied;
1299 if (!start)
1300 start = sb->buf;
1301 occupied = utf8_strnwidth(start, -1, 1);
1302 padding = (-padding) - occupied;
1303 }
1304 while (1) {
1305 int modifier = *placeholder == 'C';
1306 int consumed = format_commit_one(&local_sb, placeholder, c);
1307 total_consumed += consumed;
1308
1309 if (!modifier)
1310 break;
1311
1312 placeholder += consumed;
1313 if (*placeholder != '%')
1314 break;
1315 placeholder++;
1316 total_consumed++;
1317 }
1318 len = utf8_strnwidth(local_sb.buf, -1, 1);
1319
1320 if (c->flush_type == flush_left_and_steal) {
1321 const char *ch = sb->buf + sb->len - 1;
1322 while (len > padding && ch > sb->buf) {
1323 const char *p;
1324 if (*ch == ' ') {
1325 ch--;
1326 padding++;
1327 continue;
1328 }
1329 /* check for trailing ansi sequences */
1330 if (*ch != 'm')
1331 break;
1332 p = ch - 1;
1333 while (ch - p < 10 && *p != '\033')
1334 p--;
1335 if (*p != '\033' ||
1336 ch + 1 - p != display_mode_esc_sequence_len(p))
1337 break;
1338 /*
1339 * got a good ansi sequence, put it back to
1340 * local_sb as we're cutting sb
1341 */
1342 strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1343 ch = p - 1;
1344 }
1345 strbuf_setlen(sb, ch + 1 - sb->buf);
1346 c->flush_type = flush_left;
1347 }
1348
1349 if (len > padding) {
1350 switch (c->truncate) {
1351 case trunc_left:
1352 strbuf_utf8_replace(&local_sb,
1353 0, len - (padding - 2),
1354 "..");
1355 break;
1356 case trunc_middle:
1357 strbuf_utf8_replace(&local_sb,
1358 padding / 2 - 1,
1359 len - (padding - 2),
1360 "..");
1361 break;
1362 case trunc_right:
1363 strbuf_utf8_replace(&local_sb,
1364 padding - 2, len - (padding - 2),
1365 "..");
1366 break;
1367 case trunc_none:
1368 break;
1369 }
1370 strbuf_addbuf(sb, &local_sb);
1371 } else {
1372 int sb_len = sb->len, offset = 0;
1373 if (c->flush_type == flush_left)
1374 offset = padding - len;
1375 else if (c->flush_type == flush_both)
1376 offset = (padding - len) / 2;
1377 /*
1378 * we calculate padding in columns, now
1379 * convert it back to chars
1380 */
1381 padding = padding - len + local_sb.len;
1382 strbuf_addchars(sb, ' ', padding);
1383 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1384 local_sb.len);
1385 }
1386 strbuf_release(&local_sb);
1387 c->flush_type = no_flush;
1388 return total_consumed;
1389 }
1390
1391 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1392 const char *placeholder,
1393 void *context)
1394 {
1395 int consumed;
1396 size_t orig_len;
1397 enum {
1398 NO_MAGIC,
1399 ADD_LF_BEFORE_NON_EMPTY,
1400 DEL_LF_BEFORE_EMPTY,
1401 ADD_SP_BEFORE_NON_EMPTY
1402 } magic = NO_MAGIC;
1403
1404 switch (placeholder[0]) {
1405 case '-':
1406 magic = DEL_LF_BEFORE_EMPTY;
1407 break;
1408 case '+':
1409 magic = ADD_LF_BEFORE_NON_EMPTY;
1410 break;
1411 case ' ':
1412 magic = ADD_SP_BEFORE_NON_EMPTY;
1413 break;
1414 default:
1415 break;
1416 }
1417 if (magic != NO_MAGIC)
1418 placeholder++;
1419
1420 orig_len = sb->len;
1421 if (((struct format_commit_context *)context)->flush_type != no_flush)
1422 consumed = format_and_pad_commit(sb, placeholder, context);
1423 else
1424 consumed = format_commit_one(sb, placeholder, context);
1425 if (magic == NO_MAGIC)
1426 return consumed;
1427
1428 if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1429 while (sb->len && sb->buf[sb->len - 1] == '\n')
1430 strbuf_setlen(sb, sb->len - 1);
1431 } else if (orig_len != sb->len) {
1432 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1433 strbuf_insert(sb, orig_len, "\n", 1);
1434 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1435 strbuf_insert(sb, orig_len, " ", 1);
1436 }
1437 return consumed + 1;
1438 }
1439
1440 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1441 void *context)
1442 {
1443 struct userformat_want *w = context;
1444
1445 if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1446 placeholder++;
1447
1448 switch (*placeholder) {
1449 case 'N':
1450 w->notes = 1;
1451 break;
1452 }
1453 return 0;
1454 }
1455
1456 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1457 {
1458 struct strbuf dummy = STRBUF_INIT;
1459
1460 if (!fmt) {
1461 if (!user_format)
1462 return;
1463 fmt = user_format;
1464 }
1465 strbuf_expand(&dummy, fmt, userformat_want_item, w);
1466 strbuf_release(&dummy);
1467 }
1468
1469 void format_commit_message(const struct commit *commit,
1470 const char *format, struct strbuf *sb,
1471 const struct pretty_print_context *pretty_ctx)
1472 {
1473 struct format_commit_context context;
1474 const char *output_enc = pretty_ctx->output_encoding;
1475 const char *utf8 = "UTF-8";
1476
1477 memset(&context, 0, sizeof(context));
1478 context.commit = commit;
1479 context.pretty_ctx = pretty_ctx;
1480 context.wrap_start = sb->len;
1481 /*
1482 * convert a commit message to UTF-8 first
1483 * as far as 'format_commit_item' assumes it in UTF-8
1484 */
1485 context.message = logmsg_reencode(commit,
1486 &context.commit_encoding,
1487 utf8);
1488
1489 strbuf_expand(sb, format, format_commit_item, &context);
1490 rewrap_message_tail(sb, &context, 0, 0, 0);
1491
1492 /* then convert a commit message to an actual output encoding */
1493 if (output_enc) {
1494 if (same_encoding(utf8, output_enc))
1495 output_enc = NULL;
1496 } else {
1497 if (context.commit_encoding &&
1498 !same_encoding(context.commit_encoding, utf8))
1499 output_enc = context.commit_encoding;
1500 }
1501
1502 if (output_enc) {
1503 int outsz;
1504 char *out = reencode_string_len(sb->buf, sb->len,
1505 output_enc, utf8, &outsz);
1506 if (out)
1507 strbuf_attach(sb, out, outsz, outsz + 1);
1508 }
1509
1510 free(context.commit_encoding);
1511 unuse_commit_buffer(commit, context.message);
1512 }
1513
1514 static void pp_header(struct pretty_print_context *pp,
1515 const char *encoding,
1516 const struct commit *commit,
1517 const char **msg_p,
1518 struct strbuf *sb)
1519 {
1520 int parents_shown = 0;
1521
1522 for (;;) {
1523 const char *name, *line = *msg_p;
1524 int linelen = get_one_line(*msg_p);
1525
1526 if (!linelen)
1527 return;
1528 *msg_p += linelen;
1529
1530 if (linelen == 1)
1531 /* End of header */
1532 return;
1533
1534 if (pp->fmt == CMIT_FMT_RAW) {
1535 strbuf_add(sb, line, linelen);
1536 continue;
1537 }
1538
1539 if (starts_with(line, "parent ")) {
1540 if (linelen != 48)
1541 die("bad parent line in commit");
1542 continue;
1543 }
1544
1545 if (!parents_shown) {
1546 unsigned num = commit_list_count(commit->parents);
1547 /* with enough slop */
1548 strbuf_grow(sb, num * 50 + 20);
1549 add_merge_info(pp, sb, commit);
1550 parents_shown = 1;
1551 }
1552
1553 /*
1554 * MEDIUM == DEFAULT shows only author with dates.
1555 * FULL shows both authors but not dates.
1556 * FULLER shows both authors and dates.
1557 */
1558 if (skip_prefix(line, "author ", &name)) {
1559 strbuf_grow(sb, linelen + 80);
1560 pp_user_info(pp, "Author", sb, name, encoding);
1561 }
1562 if (skip_prefix(line, "committer ", &name) &&
1563 (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1564 strbuf_grow(sb, linelen + 80);
1565 pp_user_info(pp, "Commit", sb, name, encoding);
1566 }
1567 }
1568 }
1569
1570 void pp_title_line(struct pretty_print_context *pp,
1571 const char **msg_p,
1572 struct strbuf *sb,
1573 const char *encoding,
1574 int need_8bit_cte)
1575 {
1576 static const int max_length = 78; /* per rfc2047 */
1577 struct strbuf title;
1578
1579 strbuf_init(&title, 80);
1580 *msg_p = format_subject(&title, *msg_p,
1581 pp->preserve_subject ? "\n" : " ");
1582
1583 strbuf_grow(sb, title.len + 1024);
1584 if (pp->subject) {
1585 strbuf_addstr(sb, pp->subject);
1586 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1587 add_rfc2047(sb, title.buf, title.len,
1588 encoding, RFC2047_SUBJECT);
1589 else
1590 strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1591 -last_line_length(sb), 1, max_length);
1592 } else {
1593 strbuf_addbuf(sb, &title);
1594 }
1595 strbuf_addch(sb, '\n');
1596
1597 if (need_8bit_cte == 0) {
1598 int i;
1599 for (i = 0; i < pp->in_body_headers.nr; i++) {
1600 if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1601 need_8bit_cte = 1;
1602 break;
1603 }
1604 }
1605 }
1606
1607 if (need_8bit_cte > 0) {
1608 const char *header_fmt =
1609 "MIME-Version: 1.0\n"
1610 "Content-Type: text/plain; charset=%s\n"
1611 "Content-Transfer-Encoding: 8bit\n";
1612 strbuf_addf(sb, header_fmt, encoding);
1613 }
1614 if (pp->after_subject) {
1615 strbuf_addstr(sb, pp->after_subject);
1616 }
1617 if (pp->fmt == CMIT_FMT_EMAIL) {
1618 strbuf_addch(sb, '\n');
1619 }
1620
1621 if (pp->in_body_headers.nr) {
1622 int i;
1623 for (i = 0; i < pp->in_body_headers.nr; i++) {
1624 strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1625 free(pp->in_body_headers.items[i].string);
1626 }
1627 string_list_clear(&pp->in_body_headers, 0);
1628 strbuf_addch(sb, '\n');
1629 }
1630
1631 strbuf_release(&title);
1632 }
1633
1634 static int pp_utf8_width(const char *start, const char *end)
1635 {
1636 int width = 0;
1637 size_t remain = end - start;
1638
1639 while (remain) {
1640 int n = utf8_width(&start, &remain);
1641 if (n < 0 || !start)
1642 return -1;
1643 width += n;
1644 }
1645 return width;
1646 }
1647
1648 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1649 const char *line, int linelen)
1650 {
1651 const char *tab;
1652
1653 while ((tab = memchr(line, '\t', linelen)) != NULL) {
1654 int width = pp_utf8_width(line, tab);
1655
1656 /*
1657 * If it wasn't well-formed utf8, or it
1658 * had characters with badly defined
1659 * width (control characters etc), just
1660 * give up on trying to align things.
1661 */
1662 if (width < 0)
1663 break;
1664
1665 /* Output the data .. */
1666 strbuf_add(sb, line, tab - line);
1667
1668 /* .. and the de-tabified tab */
1669 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1670
1671 /* Skip over the printed part .. */
1672 linelen -= tab + 1 - line;
1673 line = tab + 1;
1674 }
1675
1676 /*
1677 * Print out everything after the last tab without
1678 * worrying about width - there's nothing more to
1679 * align.
1680 */
1681 strbuf_add(sb, line, linelen);
1682 }
1683
1684 /*
1685 * pp_handle_indent() prints out the intendation, and
1686 * the whole line (without the final newline), after
1687 * de-tabifying.
1688 */
1689 static void pp_handle_indent(struct pretty_print_context *pp,
1690 struct strbuf *sb, int indent,
1691 const char *line, int linelen)
1692 {
1693 strbuf_addchars(sb, ' ', indent);
1694 if (pp->expand_tabs_in_log)
1695 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1696 else
1697 strbuf_add(sb, line, linelen);
1698 }
1699
1700 void pp_remainder(struct pretty_print_context *pp,
1701 const char **msg_p,
1702 struct strbuf *sb,
1703 int indent)
1704 {
1705 int first = 1;
1706 for (;;) {
1707 const char *line = *msg_p;
1708 int linelen = get_one_line(line);
1709 *msg_p += linelen;
1710
1711 if (!linelen)
1712 break;
1713
1714 if (is_empty_line(line, &linelen)) {
1715 if (first)
1716 continue;
1717 if (pp->fmt == CMIT_FMT_SHORT)
1718 break;
1719 }
1720 first = 0;
1721
1722 strbuf_grow(sb, linelen + indent + 20);
1723 if (indent)
1724 pp_handle_indent(pp, sb, indent, line, linelen);
1725 else if (pp->expand_tabs_in_log)
1726 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1727 line, linelen);
1728 else
1729 strbuf_add(sb, line, linelen);
1730 strbuf_addch(sb, '\n');
1731 }
1732 }
1733
1734 void pretty_print_commit(struct pretty_print_context *pp,
1735 const struct commit *commit,
1736 struct strbuf *sb)
1737 {
1738 unsigned long beginning_of_body;
1739 int indent = 4;
1740 const char *msg;
1741 const char *reencoded;
1742 const char *encoding;
1743 int need_8bit_cte = pp->need_8bit_cte;
1744
1745 if (pp->fmt == CMIT_FMT_USERFORMAT) {
1746 format_commit_message(commit, user_format, sb, pp);
1747 return;
1748 }
1749
1750 encoding = get_log_output_encoding();
1751 msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1752
1753 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1754 indent = 0;
1755
1756 /*
1757 * We need to check and emit Content-type: to mark it
1758 * as 8-bit if we haven't done so.
1759 */
1760 if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1761 int i, ch, in_body;
1762
1763 for (in_body = i = 0; (ch = msg[i]); i++) {
1764 if (!in_body) {
1765 /* author could be non 7-bit ASCII but
1766 * the log may be so; skip over the
1767 * header part first.
1768 */
1769 if (ch == '\n' && msg[i+1] == '\n')
1770 in_body = 1;
1771 }
1772 else if (non_ascii(ch)) {
1773 need_8bit_cte = 1;
1774 break;
1775 }
1776 }
1777 }
1778
1779 pp_header(pp, encoding, commit, &msg, sb);
1780 if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1781 strbuf_addch(sb, '\n');
1782 }
1783
1784 /* Skip excess blank lines at the beginning of body, if any... */
1785 msg = skip_empty_lines(msg);
1786
1787 /* These formats treat the title line specially. */
1788 if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1789 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1790
1791 beginning_of_body = sb->len;
1792 if (pp->fmt != CMIT_FMT_ONELINE)
1793 pp_remainder(pp, &msg, sb, indent);
1794 strbuf_rtrim(sb);
1795
1796 /* Make sure there is an EOLN for the non-oneline case */
1797 if (pp->fmt != CMIT_FMT_ONELINE)
1798 strbuf_addch(sb, '\n');
1799
1800 /*
1801 * The caller may append additional body text in e-mail
1802 * format. Make sure we did not strip the blank line
1803 * between the header and the body.
1804 */
1805 if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1806 strbuf_addch(sb, '\n');
1807
1808 unuse_commit_buffer(commit, reencoded);
1809 }
1810
1811 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1812 struct strbuf *sb)
1813 {
1814 struct pretty_print_context pp = {0};
1815 pp.fmt = fmt;
1816 pretty_print_commit(&pp, commit, sb);
1817 }
git for Debian