pretty.c

   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
  13 static char *user_format;
  14 static struct cmt_fmt_map {
  15         const char *name;
  16         enum cmit_fmt format;
  17         int is_tformat;
  18         int is_alias;
  19         const char *user_format;
  20 } *commit_formats;
  21 static size_t builtin_formats_len;
  22 static size_t commit_formats_len;
  23 static size_t commit_formats_alloc;
  24 static struct cmt_fmt_map *find_commit_format(const char *sought);
  25
  26 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  27 {
  28         free(user_format);
  29         user_format = xstrdup(cp);
  30         if (is_tformat)
  31                 rev->use_terminator = 1;
  32         rev->commit_format = CMIT_FMT_USERFORMAT;
  33 }
  34
  35 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  36 {
  37         struct cmt_fmt_map *commit_format = NULL;
  38         const char *name;
  39         const char *fmt;
  40         int i;
  41
  42         if (prefixcmp(var, "pretty."))
  43                 return 0;
  44
  45         name = var + strlen("pretty.");
  46         for (i = 0; i < builtin_formats_len; i++) {
  47                 if (!strcmp(commit_formats[i].name, name))
  48                         return 0;
  49         }
  50
  51         for (i = builtin_formats_len; i < commit_formats_len; i++) {
  52                 if (!strcmp(commit_formats[i].name, name)) {
  53                         commit_format = &commit_formats[i];
  54                         break;
  55                 }
  56         }
  57
  58         if (!commit_format) {
  59                 ALLOC_GROW(commit_formats, commit_formats_len+1,
  60                            commit_formats_alloc);
  61                 commit_format = &commit_formats[commit_formats_len];
  62                 memset(commit_format, 0, sizeof(*commit_format));
  63                 commit_formats_len++;
  64         }
  65
  66         commit_format->name = xstrdup(name);
  67         commit_format->format = CMIT_FMT_USERFORMAT;
  68         git_config_string(&fmt, var, value);
  69         if (!prefixcmp(fmt, "format:") || !prefixcmp(fmt, "tformat:")) {
  70                 commit_format->is_tformat = fmt[0] == 't';
  71                 fmt = strchr(fmt, ':') + 1;
  72         } else if (strchr(fmt, '%'))
  73                 commit_format->is_tformat = 1;
  74         else
  75                 commit_format->is_alias = 1;
  76         commit_format->user_format = fmt;
  77
  78         return 0;
  79 }
  80
  81 static void setup_commit_formats(void)
  82 {
  83         struct cmt_fmt_map builtin_formats[] = {
  84                 { "raw",        CMIT_FMT_RAW,           0 },
  85                 { "medium",     CMIT_FMT_MEDIUM,        0 },
  86                 { "short",      CMIT_FMT_SHORT,         0 },
  87                 { "email",      CMIT_FMT_EMAIL,         0 },
  88                 { "fuller",     CMIT_FMT_FULLER,        0 },
  89                 { "full",       CMIT_FMT_FULL,          0 },
  90                 { "oneline",    CMIT_FMT_ONELINE,       1 }
  91         };
  92         commit_formats_len = ARRAY_SIZE(builtin_formats);
  93         builtin_formats_len = commit_formats_len;
  94         ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
  95         memcpy(commit_formats, builtin_formats,
  96                sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  97
  98         git_config(git_pretty_formats_config, NULL);
  99 }
 100
 101 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 102                                                         const char *original,
 103                                                         int num_redirections)
 104 {
 105         struct cmt_fmt_map *found = NULL;
 106         size_t found_match_len = 0;
 107         int i;
 108
 109         if (num_redirections >= commit_formats_len)
 110                 die("invalid --pretty format: "
 111                     "'%s' references an alias which points to itself",
 112                     original);
 113
 114         for (i = 0; i < commit_formats_len; i++) {
 115                 size_t match_len;
 116
 117                 if (prefixcmp(commit_formats[i].name, sought))
 118                         continue;
 119
 120                 match_len = strlen(commit_formats[i].name);
 121                 if (found == NULL || found_match_len > match_len) {
 122                         found = &commit_formats[i];
 123                         found_match_len = match_len;
 124                 }
 125         }
 126
 127         if (found && found->is_alias) {
 128                 found = find_commit_format_recursive(found->user_format,
 129                                                      original,
 130                                                      num_redirections+1);
 131         }
 132
 133         return found;
 134 }
 135
 136 static struct cmt_fmt_map *find_commit_format(const char *sought)
 137 {
 138         if (!commit_formats)
 139                 setup_commit_formats();
 140
 141         return find_commit_format_recursive(sought, sought, 0);
 142 }
 143
 144 void get_commit_format(const char *arg, struct rev_info *rev)
 145 {
 146         struct cmt_fmt_map *commit_format;
 147
 148         rev->use_terminator = 0;
 149         if (!arg || !*arg) {
 150                 rev->commit_format = CMIT_FMT_DEFAULT;
 151                 return;
 152         }
 153         if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
 154                 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 155                 return;
 156         }
 157
 158         if (strchr(arg, '%')) {
 159                 save_user_format(rev, arg, 1);
 160                 return;
 161         }
 162
 163         commit_format = find_commit_format(arg);
 164         if (!commit_format)
 165                 die("invalid --pretty format: %s", arg);
 166
 167         rev->commit_format = commit_format->format;
 168         rev->use_terminator = commit_format->is_tformat;
 169         if (commit_format->format == CMIT_FMT_USERFORMAT) {
 170                 save_user_format(rev, commit_format->user_format,
 171                                  commit_format->is_tformat);
 172         }
 173 }
 174
 175 /*
 176  * Generic support for pretty-printing the header
 177  */
 178 static int get_one_line(const char *msg)
 179 {
 180         int ret = 0;
 181
 182         for (;;) {
 183                 char c = *msg++;
 184                 if (!c)
 185                         break;
 186                 ret++;
 187                 if (c == '\n')
 188                         break;
 189         }
 190         return ret;
 191 }
 192
 193 /* High bit set, or ISO-2022-INT */
 194 static int non_ascii(int ch)
 195 {
 196         return !isascii(ch) || ch == '\033';
 197 }
 198
 199 int has_non_ascii(const char *s)
 200 {
 201         int ch;
 202         if (!s)
 203                 return 0;
 204         while ((ch = *s++) != '\0') {
 205                 if (non_ascii(ch))
 206                         return 1;
 207         }
 208         return 0;
 209 }
 210
 211 static int is_rfc2047_special(char ch)
 212 {
 213         return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 214 }
 215
 216 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 217                        const char *encoding)
 218 {
 219         static const int max_length = 78; /* per rfc2822 */
 220         int i;
 221         int line_len;
 222
 223         /* How many bytes are already used on the current line? */
 224         for (i = sb->len - 1; i >= 0; i--)
 225                 if (sb->buf[i] == '\n')
 226                         break;
 227         line_len = sb->len - (i+1);
 228
 229         for (i = 0; i < len; i++) {
 230                 int ch = line[i];
 231                 if (non_ascii(ch) || ch == '\n')
 232                         goto needquote;
 233                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 234                         goto needquote;
 235         }
 236         strbuf_add_wrapped_bytes(sb, line, len, 0, 1, max_length - line_len);
 237         return;
 238
 239 needquote:
 240         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 241         strbuf_addf(sb, "=?%s?q?", encoding);
 242         line_len += strlen(encoding) + 5; /* 5 for =??q? */
 243         for (i = 0; i < len; i++) {
 244                 unsigned ch = line[i] & 0xFF;
 245
 246                 if (line_len >= max_length - 2) {
 247                         strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 248                         line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 249                 }
 250
 251                 /*
 252                  * We encode ' ' using '=20' even though rfc2047
 253                  * allows using '_' for readability.  Unfortunately,
 254                  * many programs do not understand this and just
 255                  * leave the underscore in place.
 256                  */
 257                 if (is_rfc2047_special(ch) || ch == ' ' || ch == '\n') {
 258                         strbuf_addf(sb, "=%02X", ch);
 259                         line_len += 3;
 260                 }
 261                 else {
 262                         strbuf_addch(sb, ch);
 263                         line_len++;
 264                 }
 265         }
 266         strbuf_addstr(sb, "?=");
 267 }
 268
 269 void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 270                   const char *line, enum date_mode dmode,
 271                   const char *encoding)
 272 {
 273         char *date;
 274         int namelen;
 275         unsigned long time;
 276         int tz;
 277
 278         if (fmt == CMIT_FMT_ONELINE)
 279                 return;
 280         date = strchr(line, '>');
 281         if (!date)
 282                 return;
 283         namelen = ++date - line;
 284         time = strtoul(date, &date, 10);
 285         tz = strtol(date, NULL, 10);
 286
 287         if (fmt == CMIT_FMT_EMAIL) {
 288                 char *name_tail = strchr(line, '<');
 289                 int display_name_length;
 290                 if (!name_tail)
 291                         return;
 292                 while (line < name_tail && isspace(name_tail[-1]))
 293                         name_tail--;
 294                 display_name_length = name_tail - line;
 295                 strbuf_addstr(sb, "From: ");
 296                 add_rfc2047(sb, line, display_name_length, encoding);
 297                 strbuf_add(sb, name_tail, namelen - display_name_length);
 298                 strbuf_addch(sb, '\n');
 299         } else {
 300                 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 301                               (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 302                               "    ", namelen, line);
 303         }
 304         switch (fmt) {
 305         case CMIT_FMT_MEDIUM:
 306                 strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 307                 break;
 308         case CMIT_FMT_EMAIL:
 309                 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 310                 break;
 311         case CMIT_FMT_FULLER:
 312                 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 313                 break;
 314         default:
 315                 /* notin' */
 316                 break;
 317         }
 318 }
 319
 320 static int is_empty_line(const char *line, int *len_p)
 321 {
 322         int len = *len_p;
 323         while (len && isspace(line[len-1]))
 324                 len--;
 325         *len_p = len;
 326         return !len;
 327 }
 328
 329 static const char *skip_empty_lines(const char *msg)
 330 {
 331         for (;;) {
 332                 int linelen = get_one_line(msg);
 333                 int ll = linelen;
 334                 if (!linelen)
 335                         break;
 336                 if (!is_empty_line(msg, &ll))
 337                         break;
 338                 msg += linelen;
 339         }
 340         return msg;
 341 }
 342
 343 static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 344                         const struct commit *commit, int abbrev)
 345 {
 346         struct commit_list *parent = commit->parents;
 347
 348         if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 349             !parent || !parent->next)
 350                 return;
 351
 352         strbuf_addstr(sb, "Merge:");
 353
 354         while (parent) {
 355                 struct commit *p = parent->item;
 356                 const char *hex = NULL;
 357                 if (abbrev)
 358                         hex = find_unique_abbrev(p->object.sha1, abbrev);
 359                 if (!hex)
 360                         hex = sha1_to_hex(p->object.sha1);
 361                 parent = parent->next;
 362
 363                 strbuf_addf(sb, " %s", hex);
 364         }
 365         strbuf_addch(sb, '\n');
 366 }
 367
 368 static char *get_header(const struct commit *commit, const char *key)
 369 {
 370         int key_len = strlen(key);
 371         const char *line = commit->buffer;
 372
 373         for (;;) {
 374                 const char *eol = strchr(line, '\n'), *next;
 375
 376                 if (line == eol)
 377                         return NULL;
 378                 if (!eol) {
 379                         eol = line + strlen(line);
 380                         next = NULL;
 381                 } else
 382                         next = eol + 1;
 383                 if (eol - line > key_len &&
 384                     !strncmp(line, key, key_len) &&
 385                     line[key_len] == ' ') {
 386                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 387                 }
 388                 line = next;
 389         }
 390 }
 391
 392 static char *replace_encoding_header(char *buf, const char *encoding)
 393 {
 394         struct strbuf tmp = STRBUF_INIT;
 395         size_t start, len;
 396         char *cp = buf;
 397
 398         /* guess if there is an encoding header before a \n\n */
 399         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 400                 cp = strchr(cp, '\n');
 401                 if (!cp || *++cp == '\n')
 402                         return buf;
 403         }
 404         start = cp - buf;
 405         cp = strchr(cp, '\n');
 406         if (!cp)
 407                 return buf; /* should not happen but be defensive */
 408         len = cp + 1 - (buf + start);
 409
 410         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 411         if (is_encoding_utf8(encoding)) {
 412                 /* we have re-coded to UTF-8; drop the header */
 413                 strbuf_remove(&tmp, start, len);
 414         } else {
 415                 /* just replaces XXXX in 'encoding XXXX\n' */
 416                 strbuf_splice(&tmp, start + strlen("encoding "),
 417                                           len - strlen("encoding \n"),
 418                                           encoding, strlen(encoding));
 419         }
 420         return strbuf_detach(&tmp, NULL);
 421 }
 422
 423 char *logmsg_reencode(const struct commit *commit,
 424                       const char *output_encoding)
 425 {
 426         static const char *utf8 = "UTF-8";
 427         const char *use_encoding;
 428         char *encoding;
 429         char *out;
 430
 431         if (!*output_encoding)
 432                 return NULL;
 433         encoding = get_header(commit, "encoding");
 434         use_encoding = encoding ? encoding : utf8;
 435         if (!strcmp(use_encoding, output_encoding))
 436                 if (encoding) /* we'll strip encoding header later */
 437                         out = xstrdup(commit->buffer);
 438                 else
 439                         return NULL; /* nothing to do */
 440         else
 441                 out = reencode_string(commit->buffer,
 442                                       output_encoding, use_encoding);
 443         if (out)
 444                 out = replace_encoding_header(out, output_encoding);
 445
 446         free(encoding);
 447         return out;
 448 }
 449
 450 static int mailmap_name(char *email, int email_len, char *name, int name_len)
 451 {
 452         static struct string_list *mail_map;
 453         if (!mail_map) {
 454                 mail_map = xcalloc(1, sizeof(*mail_map));
 455                 read_mailmap(mail_map, NULL);
 456         }
 457         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 458 }
 459
 460 static size_t format_person_part(struct strbuf *sb, char part,
 461                                  const char *msg, int len, enum date_mode dmode)
 462 {
 463         /* currently all placeholders have same length */
 464         const int placeholder_len = 2;
 465         int start, end, tz = 0;
 466         unsigned long date = 0;
 467         char *ep;
 468         const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 469         char person_name[1024];
 470         char person_mail[1024];
 471
 472         /* advance 'end' to point to email start delimiter */
 473         for (end = 0; end < len && msg[end] != '<'; end++)
 474                 ; /* do nothing */
 475
 476         /*
 477          * When end points at the '<' that we found, it should have
 478          * matching '>' later, which means 'end' must be strictly
 479          * below len - 1.
 480          */
 481         if (end >= len - 2)
 482                 goto skip;
 483
 484         /* Seek for both name and email part */
 485         name_start = msg;
 486         name_end = msg+end;
 487         while (name_end > name_start && isspace(*(name_end-1)))
 488                 name_end--;
 489         mail_start = msg+end+1;
 490         mail_end = mail_start;
 491         while (mail_end < msg_end && *mail_end != '>')
 492                 mail_end++;
 493         if (mail_end == msg_end)
 494                 goto skip;
 495         end = mail_end-msg;
 496
 497         if (part == 'N' || part == 'E') { /* mailmap lookup */
 498                 strlcpy(person_name, name_start, name_end-name_start+1);
 499                 strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 500                 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 501                 name_start = person_name;
 502                 name_end = name_start + strlen(person_name);
 503                 mail_start = person_mail;
 504                 mail_end = mail_start +  strlen(person_mail);
 505         }
 506         if (part == 'n' || part == 'N') {       /* name */
 507                 strbuf_add(sb, name_start, name_end-name_start);
 508                 return placeholder_len;
 509         }
 510         if (part == 'e' || part == 'E') {       /* email */
 511                 strbuf_add(sb, mail_start, mail_end-mail_start);
 512                 return placeholder_len;
 513         }
 514
 515         /* advance 'start' to point to date start delimiter */
 516         for (start = end + 1; start < len && isspace(msg[start]); start++)
 517                 ; /* do nothing */
 518         if (start >= len)
 519                 goto skip;
 520         date = strtoul(msg + start, &ep, 10);
 521         if (msg + start == ep)
 522                 goto skip;
 523
 524         if (part == 't') {      /* date, UNIX timestamp */
 525                 strbuf_add(sb, msg + start, ep - (msg + start));
 526                 return placeholder_len;
 527         }
 528
 529         /* parse tz */
 530         for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 531                 ; /* do nothing */
 532         if (start + 1 < len) {
 533                 tz = strtoul(msg + start + 1, NULL, 10);
 534                 if (msg[start] == '-')
 535                         tz = -tz;
 536         }
 537
 538         switch (part) {
 539         case 'd':       /* date */
 540                 strbuf_addstr(sb, show_date(date, tz, dmode));
 541                 return placeholder_len;
 542         case 'D':       /* date, RFC2822 style */
 543                 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 544                 return placeholder_len;
 545         case 'r':       /* date, relative */
 546                 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 547                 return placeholder_len;
 548         case 'i':       /* date, ISO 8601 */
 549                 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 550                 return placeholder_len;
 551         }
 552
 553 skip:
 554         /*
 555          * bogus commit, 'sb' cannot be updated, but we still need to
 556          * compute a valid return value.
 557          */
 558         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 559             || part == 'D' || part == 'r' || part == 'i')
 560                 return placeholder_len;
 561
 562         return 0; /* unknown placeholder */
 563 }
 564
 565 struct chunk {
 566         size_t off;
 567         size_t len;
 568 };
 569
 570 struct format_commit_context {
 571         const struct commit *commit;
 572         const struct pretty_print_context *pretty_ctx;
 573         unsigned commit_header_parsed:1;
 574         unsigned commit_message_parsed:1;
 575         char *message;
 576         size_t width, indent1, indent2;
 577
 578         /* These offsets are relative to the start of the commit message. */
 579         struct chunk author;
 580         struct chunk committer;
 581         struct chunk encoding;
 582         size_t message_off;
 583         size_t subject_off;
 584         size_t body_off;
 585
 586         /* The following ones are relative to the result struct strbuf. */
 587         struct chunk abbrev_commit_hash;
 588         struct chunk abbrev_tree_hash;
 589         struct chunk abbrev_parent_hashes;
 590         size_t wrap_start;
 591 };
 592
 593 static int add_again(struct strbuf *sb, struct chunk *chunk)
 594 {
 595         if (chunk->len) {
 596                 strbuf_adddup(sb, chunk->off, chunk->len);
 597                 return 1;
 598         }
 599
 600         /*
 601          * We haven't seen this chunk before.  Our caller is surely
 602          * going to add it the hard way now.  Remember the most likely
 603          * start of the to-be-added chunk: the current end of the
 604          * struct strbuf.
 605          */
 606         chunk->off = sb->len;
 607         return 0;
 608 }
 609
 610 static void parse_commit_header(struct format_commit_context *context)
 611 {
 612         const char *msg = context->message;
 613         int i;
 614
 615         for (i = 0; msg[i]; i++) {
 616                 int eol;
 617                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 618                         ; /* do nothing */
 619
 620                 if (i == eol) {
 621                         break;
 622                 } else if (!prefixcmp(msg + i, "author ")) {
 623                         context->author.off = i + 7;
 624                         context->author.len = eol - i - 7;
 625                 } else if (!prefixcmp(msg + i, "committer ")) {
 626                         context->committer.off = i + 10;
 627                         context->committer.len = eol - i - 10;
 628                 } else if (!prefixcmp(msg + i, "encoding ")) {
 629                         context->encoding.off = i + 9;
 630                         context->encoding.len = eol - i - 9;
 631                 }
 632                 i = eol;
 633         }
 634         context->message_off = i;
 635         context->commit_header_parsed = 1;
 636 }
 637
 638 static int istitlechar(char c)
 639 {
 640         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 641                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 642 }
 643
 644 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 645 {
 646         size_t trimlen;
 647         size_t start_len = sb->len;
 648         int space = 2;
 649
 650         for (; *msg && *msg != '\n'; msg++) {
 651                 if (istitlechar(*msg)) {
 652                         if (space == 1)
 653                                 strbuf_addch(sb, '-');
 654                         space = 0;
 655                         strbuf_addch(sb, *msg);
 656                         if (*msg == '.')
 657                                 while (*(msg+1) == '.')
 658                                         msg++;
 659                 } else
 660                         space |= 1;
 661         }
 662
 663         /* trim any trailing '.' or '-' characters */
 664         trimlen = 0;
 665         while (sb->len - trimlen > start_len &&
 666                 (sb->buf[sb->len - 1 - trimlen] == '.'
 667                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 668                 trimlen++;
 669         strbuf_remove(sb, sb->len - trimlen, trimlen);
 670 }
 671
 672 const char *format_subject(struct strbuf *sb, const char *msg,
 673                            const char *line_separator)
 674 {
 675         int first = 1;
 676
 677         for (;;) {
 678                 const char *line = msg;
 679                 int linelen = get_one_line(line);
 680
 681                 msg += linelen;
 682                 if (!linelen || is_empty_line(line, &linelen))
 683                         break;
 684
 685                 if (!sb)
 686                         continue;
 687                 strbuf_grow(sb, linelen + 2);
 688                 if (!first)
 689                         strbuf_addstr(sb, line_separator);
 690                 strbuf_add(sb, line, linelen);
 691                 first = 0;
 692         }
 693         return msg;
 694 }
 695
 696 static void parse_commit_message(struct format_commit_context *c)
 697 {
 698         const char *msg = c->message + c->message_off;
 699         const char *start = c->message;
 700
 701         msg = skip_empty_lines(msg);
 702         c->subject_off = msg - start;
 703
 704         msg = format_subject(NULL, msg, NULL);
 705         msg = skip_empty_lines(msg);
 706         c->body_off = msg - start;
 707
 708         c->commit_message_parsed = 1;
 709 }
 710
 711 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 712 {
 713         struct name_decoration *d;
 714         const char *prefix = " (";
 715
 716         load_ref_decorations(DECORATE_SHORT_REFS);
 717         d = lookup_decoration(&name_decoration, &commit->object);
 718         while (d) {
 719                 strbuf_addstr(sb, prefix);
 720                 prefix = ", ";
 721                 strbuf_addstr(sb, d->name);
 722                 d = d->next;
 723         }
 724         if (prefix[0] == ',')
 725                 strbuf_addch(sb, ')');
 726 }
 727
 728 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 729                         size_t width, size_t indent1, size_t indent2)
 730 {
 731         struct strbuf tmp = STRBUF_INIT;
 732
 733         if (pos)
 734                 strbuf_add(&tmp, sb->buf, pos);
 735         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 736                                 (int) indent1, (int) indent2, (int) width);
 737         strbuf_swap(&tmp, sb);
 738         strbuf_release(&tmp);
 739 }
 740
 741 static void rewrap_message_tail(struct strbuf *sb,
 742                                 struct format_commit_context *c,
 743                                 size_t new_width, size_t new_indent1,
 744                                 size_t new_indent2)
 745 {
 746         if (c->width == new_width && c->indent1 == new_indent1 &&
 747             c->indent2 == new_indent2)
 748                 return;
 749         if (c->wrap_start < sb->len)
 750                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 751         c->wrap_start = sb->len;
 752         c->width = new_width;
 753         c->indent1 = new_indent1;
 754         c->indent2 = new_indent2;
 755 }
 756
 757 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 758                                 void *context)
 759 {
 760         struct format_commit_context *c = context;
 761         const struct commit *commit = c->commit;
 762         const char *msg = c->message;
 763         struct commit_list *p;
 764         int h1, h2;
 765
 766         /* these are independent of the commit */
 767         switch (placeholder[0]) {
 768         case 'C':
 769                 if (placeholder[1] == '(') {
 770                         const char *end = strchr(placeholder + 2, ')');
 771                         char color[COLOR_MAXLEN];
 772                         if (!end)
 773                                 return 0;
 774                         color_parse_mem(placeholder + 2,
 775                                         end - (placeholder + 2),
 776                                         "--pretty format", color);
 777                         strbuf_addstr(sb, color);
 778                         return end - placeholder + 1;
 779                 }
 780                 if (!prefixcmp(placeholder + 1, "red")) {
 781                         strbuf_addstr(sb, GIT_COLOR_RED);
 782                         return 4;
 783                 } else if (!prefixcmp(placeholder + 1, "green")) {
 784                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 785                         return 6;
 786                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 787                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 788                         return 5;
 789                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 790                         strbuf_addstr(sb, GIT_COLOR_RESET);
 791                         return 6;
 792                 } else
 793                         return 0;
 794         case 'n':               /* newline */
 795                 strbuf_addch(sb, '\n');
 796                 return 1;
 797         case 'x':
 798                 /* %x00 == NUL, %x0a == LF, etc. */
 799                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 800                     h1 <= 16 &&
 801                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 802                     h2 <= 16) {
 803                         strbuf_addch(sb, (h1<<4)|h2);
 804                         return 3;
 805                 } else
 806                         return 0;
 807         case 'w':
 808                 if (placeholder[1] == '(') {
 809                         unsigned long width = 0, indent1 = 0, indent2 = 0;
 810                         char *next;
 811                         const char *start = placeholder + 2;
 812                         const char *end = strchr(start, ')');
 813                         if (!end)
 814                                 return 0;
 815                         if (end > start) {
 816                                 width = strtoul(start, &next, 10);
 817                                 if (*next == ',') {
 818                                         indent1 = strtoul(next + 1, &next, 10);
 819                                         if (*next == ',') {
 820                                                 indent2 = strtoul(next + 1,
 821                                                                  &next, 10);
 822                                         }
 823                                 }
 824                                 if (*next != ')')
 825                                         return 0;
 826                         }
 827                         rewrap_message_tail(sb, c, width, indent1, indent2);
 828                         return end - placeholder + 1;
 829                 } else
 830                         return 0;
 831         }
 832
 833         /* these depend on the commit */
 834         if (!commit->object.parsed)
 835                 parse_object(commit->object.sha1);
 836
 837         switch (placeholder[0]) {
 838         case 'H':               /* commit hash */
 839                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 840                 return 1;
 841         case 'h':               /* abbreviated commit hash */
 842                 if (add_again(sb, &c->abbrev_commit_hash))
 843                         return 1;
 844                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 845                                                      c->pretty_ctx->abbrev));
 846                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 847                 return 1;
 848         case 'T':               /* tree hash */
 849                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 850                 return 1;
 851         case 't':               /* abbreviated tree hash */
 852                 if (add_again(sb, &c->abbrev_tree_hash))
 853                         return 1;
 854                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 855                                                      c->pretty_ctx->abbrev));
 856                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 857                 return 1;
 858         case 'P':               /* parent hashes */
 859                 for (p = commit->parents; p; p = p->next) {
 860                         if (p != commit->parents)
 861                                 strbuf_addch(sb, ' ');
 862                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 863                 }
 864                 return 1;
 865         case 'p':               /* abbreviated parent hashes */
 866                 if (add_again(sb, &c->abbrev_parent_hashes))
 867                         return 1;
 868                 for (p = commit->parents; p; p = p->next) {
 869                         if (p != commit->parents)
 870                                 strbuf_addch(sb, ' ');
 871                         strbuf_addstr(sb, find_unique_abbrev(
 872                                         p->item->object.sha1,
 873                                         c->pretty_ctx->abbrev));
 874                 }
 875                 c->abbrev_parent_hashes.len = sb->len -
 876                                               c->abbrev_parent_hashes.off;
 877                 return 1;
 878         case 'm':               /* left/right/bottom */
 879                 strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 880                                  ? '-'
 881                                  : (commit->object.flags & SYMMETRIC_LEFT)
 882                                  ? '<'
 883                                  : '>');
 884                 return 1;
 885         case 'd':
 886                 format_decoration(sb, commit);
 887                 return 1;
 888         case 'g':               /* reflog info */
 889                 switch(placeholder[1]) {
 890                 case 'd':       /* reflog selector */
 891                 case 'D':
 892                         if (c->pretty_ctx->reflog_info)
 893                                 get_reflog_selector(sb,
 894                                                     c->pretty_ctx->reflog_info,
 895                                                     c->pretty_ctx->date_mode,
 896                                                     (placeholder[1] == 'd'));
 897                         return 2;
 898                 case 's':       /* reflog message */
 899                         if (c->pretty_ctx->reflog_info)
 900                                 get_reflog_message(sb, c->pretty_ctx->reflog_info);
 901                         return 2;
 902                 }
 903                 return 0;       /* unknown %g placeholder */
 904         case 'N':
 905                 if (c->pretty_ctx->show_notes) {
 906                         format_display_notes(commit->object.sha1, sb,
 907                                     get_log_output_encoding(), 0);
 908                         return 1;
 909                 }
 910                 return 0;
 911         }
 912
 913         /* For the rest we have to parse the commit header. */
 914         if (!c->commit_header_parsed)
 915                 parse_commit_header(c);
 916
 917         switch (placeholder[0]) {
 918         case 'a':       /* author ... */
 919                 return format_person_part(sb, placeholder[1],
 920                                    msg + c->author.off, c->author.len,
 921                                    c->pretty_ctx->date_mode);
 922         case 'c':       /* committer ... */
 923                 return format_person_part(sb, placeholder[1],
 924                                    msg + c->committer.off, c->committer.len,
 925                                    c->pretty_ctx->date_mode);
 926         case 'e':       /* encoding */
 927                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 928                 return 1;
 929         case 'B':       /* raw body */
 930                 /* message_off is always left at the initial newline */
 931                 strbuf_addstr(sb, msg + c->message_off + 1);
 932                 return 1;
 933         }
 934
 935         /* Now we need to parse the commit message. */
 936         if (!c->commit_message_parsed)
 937                 parse_commit_message(c);
 938
 939         switch (placeholder[0]) {
 940         case 's':       /* subject */
 941                 format_subject(sb, msg + c->subject_off, " ");
 942                 return 1;
 943         case 'f':       /* sanitized subject */
 944                 format_sanitized_subject(sb, msg + c->subject_off);
 945                 return 1;
 946         case 'b':       /* body */
 947                 strbuf_addstr(sb, msg + c->body_off);
 948                 return 1;
 949         }
 950         return 0;       /* unknown placeholder */
 951 }
 952
 953 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 954                                  void *context)
 955 {
 956         int consumed;
 957         size_t orig_len;
 958         enum {
 959                 NO_MAGIC,
 960                 ADD_LF_BEFORE_NON_EMPTY,
 961                 DEL_LF_BEFORE_EMPTY,
 962                 ADD_SP_BEFORE_NON_EMPTY
 963         } magic = NO_MAGIC;
 964
 965         switch (placeholder[0]) {
 966         case '-':
 967                 magic = DEL_LF_BEFORE_EMPTY;
 968                 break;
 969         case '+':
 970                 magic = ADD_LF_BEFORE_NON_EMPTY;
 971                 break;
 972         case ' ':
 973                 magic = ADD_SP_BEFORE_NON_EMPTY;
 974                 break;
 975         default:
 976                 break;
 977         }
 978         if (magic != NO_MAGIC)
 979                 placeholder++;
 980
 981         orig_len = sb->len;
 982         consumed = format_commit_one(sb, placeholder, context);
 983         if (magic == NO_MAGIC)
 984                 return consumed;
 985
 986         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
 987                 while (sb->len && sb->buf[sb->len - 1] == '\n')
 988                         strbuf_setlen(sb, sb->len - 1);
 989         } else if (orig_len != sb->len) {
 990                 if (magic == ADD_LF_BEFORE_NON_EMPTY)
 991                         strbuf_insert(sb, orig_len, "\n", 1);
 992                 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
 993                         strbuf_insert(sb, orig_len, " ", 1);
 994         }
 995         return consumed + 1;
 996 }
 997
 998 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
 999                                    void *context)
1000 {
1001         struct userformat_want *w = context;
1002
1003         if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1004                 placeholder++;
1005
1006         switch (*placeholder) {
1007         case 'N':
1008                 w->notes = 1;
1009                 break;
1010         }
1011         return 0;
1012 }
1013
1014 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1015 {
1016         struct strbuf dummy = STRBUF_INIT;
1017
1018         if (!fmt) {
1019                 if (!user_format)
1020                         return;
1021                 fmt = user_format;
1022         }
1023         strbuf_expand(&dummy, user_format, userformat_want_item, w);
1024         strbuf_release(&dummy);
1025 }
1026
1027 void format_commit_message(const struct commit *commit,
1028                            const char *format, struct strbuf *sb,
1029                            const struct pretty_print_context *pretty_ctx)
1030 {
1031         struct format_commit_context context;
1032         static const char utf8[] = "UTF-8";
1033         const char *enc;
1034         const char *output_enc = pretty_ctx->output_encoding;
1035
1036         memset(&context, 0, sizeof(context));
1037         context.commit = commit;
1038         context.pretty_ctx = pretty_ctx;
1039         context.wrap_start = sb->len;
1040         context.message = commit->buffer;
1041         if (output_enc) {
1042                 enc = get_header(commit, "encoding");
1043                 enc = enc ? enc : utf8;
1044                 if (strcmp(enc, output_enc))
1045                         context.message = logmsg_reencode(commit, output_enc);
1046         }
1047
1048         strbuf_expand(sb, format, format_commit_item, &context);
1049         rewrap_message_tail(sb, &context, 0, 0, 0);
1050
1051         if (context.message != commit->buffer)
1052                 free(context.message);
1053 }
1054
1055 static void pp_header(enum cmit_fmt fmt,
1056                       int abbrev,
1057                       enum date_mode dmode,
1058                       const char *encoding,
1059                       const struct commit *commit,
1060                       const char **msg_p,
1061                       struct strbuf *sb)
1062 {
1063         int parents_shown = 0;
1064
1065         for (;;) {
1066                 const char *line = *msg_p;
1067                 int linelen = get_one_line(*msg_p);
1068
1069                 if (!linelen)
1070                         return;
1071                 *msg_p += linelen;
1072
1073                 if (linelen == 1)
1074                         /* End of header */
1075                         return;
1076
1077                 if (fmt == CMIT_FMT_RAW) {
1078                         strbuf_add(sb, line, linelen);
1079                         continue;
1080                 }
1081
1082                 if (!memcmp(line, "parent ", 7)) {
1083                         if (linelen != 48)
1084                                 die("bad parent line in commit");
1085                         continue;
1086                 }
1087
1088                 if (!parents_shown) {
1089                         struct commit_list *parent;
1090                         int num;
1091                         for (parent = commit->parents, num = 0;
1092                              parent;
1093                              parent = parent->next, num++)
1094                                 ;
1095                         /* with enough slop */
1096                         strbuf_grow(sb, num * 50 + 20);
1097                         add_merge_info(fmt, sb, commit, abbrev);
1098                         parents_shown = 1;
1099                 }
1100
1101                 /*
1102                  * MEDIUM == DEFAULT shows only author with dates.
1103                  * FULL shows both authors but not dates.
1104                  * FULLER shows both authors and dates.
1105                  */
1106                 if (!memcmp(line, "author ", 7)) {
1107                         strbuf_grow(sb, linelen + 80);
1108                         pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
1109                 }
1110                 if (!memcmp(line, "committer ", 10) &&
1111                     (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
1112                         strbuf_grow(sb, linelen + 80);
1113                         pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
1114                 }
1115         }
1116 }
1117
1118 void pp_title_line(enum cmit_fmt fmt,
1119                    const char **msg_p,
1120                    struct strbuf *sb,
1121                    const char *subject,
1122                    const char *after_subject,
1123                    const char *encoding,
1124                    int need_8bit_cte)
1125 {
1126         struct strbuf title;
1127
1128         strbuf_init(&title, 80);
1129         *msg_p = format_subject(&title, *msg_p, " ");
1130
1131         strbuf_grow(sb, title.len + 1024);
1132         if (subject) {
1133                 strbuf_addstr(sb, subject);
1134                 add_rfc2047(sb, title.buf, title.len, encoding);
1135         } else {
1136                 strbuf_addbuf(sb, &title);
1137         }
1138         strbuf_addch(sb, '\n');
1139
1140         if (need_8bit_cte > 0) {
1141                 const char *header_fmt =
1142                         "MIME-Version: 1.0\n"
1143                         "Content-Type: text/plain; charset=%s\n"
1144                         "Content-Transfer-Encoding: 8bit\n";
1145                 strbuf_addf(sb, header_fmt, encoding);
1146         }
1147         if (after_subject) {
1148                 strbuf_addstr(sb, after_subject);
1149         }
1150         if (fmt == CMIT_FMT_EMAIL) {
1151                 strbuf_addch(sb, '\n');
1152         }
1153         strbuf_release(&title);
1154 }
1155
1156 void pp_remainder(enum cmit_fmt fmt,
1157                   const char **msg_p,
1158                   struct strbuf *sb,
1159                   int indent)
1160 {
1161         int first = 1;
1162         for (;;) {
1163                 const char *line = *msg_p;
1164                 int linelen = get_one_line(line);
1165                 *msg_p += linelen;
1166
1167                 if (!linelen)
1168                         break;
1169
1170                 if (is_empty_line(line, &linelen)) {
1171                         if (first)
1172                                 continue;
1173                         if (fmt == CMIT_FMT_SHORT)
1174                                 break;
1175                 }
1176                 first = 0;
1177
1178                 strbuf_grow(sb, linelen + indent + 20);
1179                 if (indent) {
1180                         memset(sb->buf + sb->len, ' ', indent);
1181                         strbuf_setlen(sb, sb->len + indent);
1182                 }
1183                 strbuf_add(sb, line, linelen);
1184                 strbuf_addch(sb, '\n');
1185         }
1186 }
1187
1188 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1189 {
1190         const char *encoding;
1191
1192         encoding = get_log_output_encoding();
1193         if (encoding_p)
1194                 *encoding_p = encoding;
1195         return logmsg_reencode(commit, encoding);
1196 }
1197
1198 void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
1199                          struct strbuf *sb,
1200                          const struct pretty_print_context *context)
1201 {
1202         unsigned long beginning_of_body;
1203         int indent = 4;
1204         const char *msg = commit->buffer;
1205         char *reencoded;
1206         const char *encoding;
1207         int need_8bit_cte = context->need_8bit_cte;
1208
1209         if (fmt == CMIT_FMT_USERFORMAT) {
1210                 format_commit_message(commit, user_format, sb, context);
1211                 return;
1212         }
1213
1214         reencoded = reencode_commit_message(commit, &encoding);
1215         if (reencoded) {
1216                 msg = reencoded;
1217         }
1218
1219         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1220                 indent = 0;
1221
1222         /*
1223          * We need to check and emit Content-type: to mark it
1224          * as 8-bit if we haven't done so.
1225          */
1226         if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1227                 int i, ch, in_body;
1228
1229                 for (in_body = i = 0; (ch = msg[i]); i++) {
1230                         if (!in_body) {
1231                                 /* author could be non 7-bit ASCII but
1232                                  * the log may be so; skip over the
1233                                  * header part first.
1234                                  */
1235                                 if (ch == '\n' && msg[i+1] == '\n')
1236                                         in_body = 1;
1237                         }
1238                         else if (non_ascii(ch)) {
1239                                 need_8bit_cte = 1;
1240                                 break;
1241                         }
1242                 }
1243         }
1244
1245         pp_header(fmt, context->abbrev, context->date_mode, encoding,
1246                   commit, &msg, sb);
1247         if (fmt != CMIT_FMT_ONELINE && !context->subject) {
1248                 strbuf_addch(sb, '\n');
1249         }
1250
1251         /* Skip excess blank lines at the beginning of body, if any... */
1252         msg = skip_empty_lines(msg);
1253
1254         /* These formats treat the title line specially. */
1255         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1256                 pp_title_line(fmt, &msg, sb, context->subject,
1257                               context->after_subject, encoding, need_8bit_cte);
1258
1259         beginning_of_body = sb->len;
1260         if (fmt != CMIT_FMT_ONELINE)
1261                 pp_remainder(fmt, &msg, sb, indent);
1262         strbuf_rtrim(sb);
1263
1264         /* Make sure there is an EOLN for the non-oneline case */
1265         if (fmt != CMIT_FMT_ONELINE)
1266                 strbuf_addch(sb, '\n');
1267
1268         /*
1269          * The caller may append additional body text in e-mail
1270          * format.  Make sure we did not strip the blank line
1271          * between the header and the body.
1272          */
1273         if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1274                 strbuf_addch(sb, '\n');
1275
1276         if (context->show_notes)
1277                 format_display_notes(commit->object.sha1, sb, encoding,
1278                                      NOTES_SHOW_HEADER | NOTES_INDENT);
1279
1280         free(reencoded);
1281 }