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 "color.h"
  10
  11 static char *user_format;
  12
  13 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  14 {
  15         free(user_format);
  16         user_format = xstrdup(cp);
  17         if (is_tformat)
  18                 rev->use_terminator = 1;
  19         rev->commit_format = CMIT_FMT_USERFORMAT;
  20 }
  21
  22 void get_commit_format(const char *arg, struct rev_info *rev)
  23 {
  24         int i;
  25         static struct cmt_fmt_map {
  26                 const char *n;
  27                 size_t cmp_len;
  28                 enum cmit_fmt v;
  29         } cmt_fmts[] = {
  30                 { "raw",        1,      CMIT_FMT_RAW },
  31                 { "medium",     1,      CMIT_FMT_MEDIUM },
  32                 { "short",      1,      CMIT_FMT_SHORT },
  33                 { "email",      1,      CMIT_FMT_EMAIL },
  34                 { "full",       5,      CMIT_FMT_FULL },
  35                 { "fuller",     5,      CMIT_FMT_FULLER },
  36                 { "oneline",    1,      CMIT_FMT_ONELINE },
  37         };
  38
  39         rev->use_terminator = 0;
  40         if (!arg || !*arg) {
  41                 rev->commit_format = CMIT_FMT_DEFAULT;
  42                 return;
  43         }
  44         if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
  45                 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
  46                 return;
  47         }
  48         for (i = 0; i < ARRAY_SIZE(cmt_fmts); i++) {
  49                 if (!strncmp(arg, cmt_fmts[i].n, cmt_fmts[i].cmp_len) &&
  50                     !strncmp(arg, cmt_fmts[i].n, strlen(arg))) {
  51                         if (cmt_fmts[i].v == CMIT_FMT_ONELINE)
  52                                 rev->use_terminator = 1;
  53                         rev->commit_format = cmt_fmts[i].v;
  54                         return;
  55                 }
  56         }
  57         if (strchr(arg, '%')) {
  58                 save_user_format(rev, arg, 1);
  59                 return;
  60         }
  61
  62         die("invalid --pretty format: %s", arg);
  63 }
  64
  65 /*
  66  * Generic support for pretty-printing the header
  67  */
  68 static int get_one_line(const char *msg)
  69 {
  70         int ret = 0;
  71
  72         for (;;) {
  73                 char c = *msg++;
  74                 if (!c)
  75                         break;
  76                 ret++;
  77                 if (c == '\n')
  78                         break;
  79         }
  80         return ret;
  81 }
  82
  83 /* High bit set, or ISO-2022-INT */
  84 int non_ascii(int ch)
  85 {
  86         return !isascii(ch) || ch == '\033';
  87 }
  88
  89 int has_non_ascii(const char *s)
  90 {
  91         int ch;
  92         if (!s)
  93                 return 0;
  94         while ((ch = *s++) != '\0') {
  95                 if (non_ascii(ch))
  96                         return 1;
  97         }
  98         return 0;
  99 }
 100
 101 static int is_rfc2047_special(char ch)
 102 {
 103         return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 104 }
 105
 106 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 107                        const char *encoding)
 108 {
 109         int i, last;
 110
 111         for (i = 0; i < len; i++) {
 112                 int ch = line[i];
 113                 if (non_ascii(ch))
 114                         goto needquote;
 115                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 116                         goto needquote;
 117         }
 118         strbuf_add(sb, line, len);
 119         return;
 120
 121 needquote:
 122         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 123         strbuf_addf(sb, "=?%s?q?", encoding);
 124         for (i = last = 0; i < len; i++) {
 125                 unsigned ch = line[i] & 0xFF;
 126                 /*
 127                  * We encode ' ' using '=20' even though rfc2047
 128                  * allows using '_' for readability.  Unfortunately,
 129                  * many programs do not understand this and just
 130                  * leave the underscore in place.
 131                  */
 132                 if (is_rfc2047_special(ch) || ch == ' ') {
 133                         strbuf_add(sb, line + last, i - last);
 134                         strbuf_addf(sb, "=%02X", ch);
 135                         last = i + 1;
 136                 }
 137         }
 138         strbuf_add(sb, line + last, len - last);
 139         strbuf_addstr(sb, "?=");
 140 }
 141
 142 void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 143                   const char *line, enum date_mode dmode,
 144                   const char *encoding)
 145 {
 146         char *date;
 147         int namelen;
 148         unsigned long time;
 149         int tz;
 150
 151         if (fmt == CMIT_FMT_ONELINE)
 152                 return;
 153         date = strchr(line, '>');
 154         if (!date)
 155                 return;
 156         namelen = ++date - line;
 157         time = strtoul(date, &date, 10);
 158         tz = strtol(date, NULL, 10);
 159
 160         if (fmt == CMIT_FMT_EMAIL) {
 161                 char *name_tail = strchr(line, '<');
 162                 int display_name_length;
 163                 if (!name_tail)
 164                         return;
 165                 while (line < name_tail && isspace(name_tail[-1]))
 166                         name_tail--;
 167                 display_name_length = name_tail - line;
 168                 strbuf_addstr(sb, "From: ");
 169                 add_rfc2047(sb, line, display_name_length, encoding);
 170                 strbuf_add(sb, name_tail, namelen - display_name_length);
 171                 strbuf_addch(sb, '\n');
 172         } else {
 173                 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 174                               (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 175                               "    ", namelen, line);
 176         }
 177         switch (fmt) {
 178         case CMIT_FMT_MEDIUM:
 179                 strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 180                 break;
 181         case CMIT_FMT_EMAIL:
 182                 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 183                 break;
 184         case CMIT_FMT_FULLER:
 185                 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 186                 break;
 187         default:
 188                 /* notin' */
 189                 break;
 190         }
 191 }
 192
 193 static int is_empty_line(const char *line, int *len_p)
 194 {
 195         int len = *len_p;
 196         while (len && isspace(line[len-1]))
 197                 len--;
 198         *len_p = len;
 199         return !len;
 200 }
 201
 202 static const char *skip_empty_lines(const char *msg)
 203 {
 204         for (;;) {
 205                 int linelen = get_one_line(msg);
 206                 int ll = linelen;
 207                 if (!linelen)
 208                         break;
 209                 if (!is_empty_line(msg, &ll))
 210                         break;
 211                 msg += linelen;
 212         }
 213         return msg;
 214 }
 215
 216 static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 217                         const struct commit *commit, int abbrev)
 218 {
 219         struct commit_list *parent = commit->parents;
 220
 221         if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 222             !parent || !parent->next)
 223                 return;
 224
 225         strbuf_addstr(sb, "Merge:");
 226
 227         while (parent) {
 228                 struct commit *p = parent->item;
 229                 const char *hex = NULL;
 230                 if (abbrev)
 231                         hex = find_unique_abbrev(p->object.sha1, abbrev);
 232                 if (!hex)
 233                         hex = sha1_to_hex(p->object.sha1);
 234                 parent = parent->next;
 235
 236                 strbuf_addf(sb, " %s", hex);
 237         }
 238         strbuf_addch(sb, '\n');
 239 }
 240
 241 static char *get_header(const struct commit *commit, const char *key)
 242 {
 243         int key_len = strlen(key);
 244         const char *line = commit->buffer;
 245
 246         for (;;) {
 247                 const char *eol = strchr(line, '\n'), *next;
 248
 249                 if (line == eol)
 250                         return NULL;
 251                 if (!eol) {
 252                         eol = line + strlen(line);
 253                         next = NULL;
 254                 } else
 255                         next = eol + 1;
 256                 if (eol - line > key_len &&
 257                     !strncmp(line, key, key_len) &&
 258                     line[key_len] == ' ') {
 259                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 260                 }
 261                 line = next;
 262         }
 263 }
 264
 265 static char *replace_encoding_header(char *buf, const char *encoding)
 266 {
 267         struct strbuf tmp = STRBUF_INIT;
 268         size_t start, len;
 269         char *cp = buf;
 270
 271         /* guess if there is an encoding header before a \n\n */
 272         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 273                 cp = strchr(cp, '\n');
 274                 if (!cp || *++cp == '\n')
 275                         return buf;
 276         }
 277         start = cp - buf;
 278         cp = strchr(cp, '\n');
 279         if (!cp)
 280                 return buf; /* should not happen but be defensive */
 281         len = cp + 1 - (buf + start);
 282
 283         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 284         if (is_encoding_utf8(encoding)) {
 285                 /* we have re-coded to UTF-8; drop the header */
 286                 strbuf_remove(&tmp, start, len);
 287         } else {
 288                 /* just replaces XXXX in 'encoding XXXX\n' */
 289                 strbuf_splice(&tmp, start + strlen("encoding "),
 290                                           len - strlen("encoding \n"),
 291                                           encoding, strlen(encoding));
 292         }
 293         return strbuf_detach(&tmp, NULL);
 294 }
 295
 296 static char *logmsg_reencode(const struct commit *commit,
 297                              const char *output_encoding)
 298 {
 299         static const char *utf8 = "UTF-8";
 300         const char *use_encoding;
 301         char *encoding;
 302         char *out;
 303
 304         if (!*output_encoding)
 305                 return NULL;
 306         encoding = get_header(commit, "encoding");
 307         use_encoding = encoding ? encoding : utf8;
 308         if (!strcmp(use_encoding, output_encoding))
 309                 if (encoding) /* we'll strip encoding header later */
 310                         out = xstrdup(commit->buffer);
 311                 else
 312                         return NULL; /* nothing to do */
 313         else
 314                 out = reencode_string(commit->buffer,
 315                                       output_encoding, use_encoding);
 316         if (out)
 317                 out = replace_encoding_header(out, output_encoding);
 318
 319         free(encoding);
 320         return out;
 321 }
 322
 323 static int mailmap_name(char *email, int email_len, char *name, int name_len)
 324 {
 325         static struct string_list *mail_map;
 326         if (!mail_map) {
 327                 mail_map = xcalloc(1, sizeof(*mail_map));
 328                 read_mailmap(mail_map, NULL);
 329         }
 330         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 331 }
 332
 333 static size_t format_person_part(struct strbuf *sb, char part,
 334                                  const char *msg, int len, enum date_mode dmode)
 335 {
 336         /* currently all placeholders have same length */
 337         const int placeholder_len = 2;
 338         int start, end, tz = 0;
 339         unsigned long date = 0;
 340         char *ep;
 341         const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 342         char person_name[1024];
 343         char person_mail[1024];
 344
 345         /* advance 'end' to point to email start delimiter */
 346         for (end = 0; end < len && msg[end] != '<'; end++)
 347                 ; /* do nothing */
 348
 349         /*
 350          * When end points at the '<' that we found, it should have
 351          * matching '>' later, which means 'end' must be strictly
 352          * below len - 1.
 353          */
 354         if (end >= len - 2)
 355                 goto skip;
 356
 357         /* Seek for both name and email part */
 358         name_start = msg;
 359         name_end = msg+end;
 360         while (name_end > name_start && isspace(*(name_end-1)))
 361                 name_end--;
 362         mail_start = msg+end+1;
 363         mail_end = mail_start;
 364         while (mail_end < msg_end && *mail_end != '>')
 365                 mail_end++;
 366         if (mail_end == msg_end)
 367                 goto skip;
 368         end = mail_end-msg;
 369
 370         if (part == 'N' || part == 'E') { /* mailmap lookup */
 371                 strlcpy(person_name, name_start, name_end-name_start+1);
 372                 strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 373                 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 374                 name_start = person_name;
 375                 name_end = name_start + strlen(person_name);
 376                 mail_start = person_mail;
 377                 mail_end = mail_start +  strlen(person_mail);
 378         }
 379         if (part == 'n' || part == 'N') {       /* name */
 380                 strbuf_add(sb, name_start, name_end-name_start);
 381                 return placeholder_len;
 382         }
 383         if (part == 'e' || part == 'E') {       /* email */
 384                 strbuf_add(sb, mail_start, mail_end-mail_start);
 385                 return placeholder_len;
 386         }
 387
 388         /* advance 'start' to point to date start delimiter */
 389         for (start = end + 1; start < len && isspace(msg[start]); start++)
 390                 ; /* do nothing */
 391         if (start >= len)
 392                 goto skip;
 393         date = strtoul(msg + start, &ep, 10);
 394         if (msg + start == ep)
 395                 goto skip;
 396
 397         if (part == 't') {      /* date, UNIX timestamp */
 398                 strbuf_add(sb, msg + start, ep - (msg + start));
 399                 return placeholder_len;
 400         }
 401
 402         /* parse tz */
 403         for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 404                 ; /* do nothing */
 405         if (start + 1 < len) {
 406                 tz = strtoul(msg + start + 1, NULL, 10);
 407                 if (msg[start] == '-')
 408                         tz = -tz;
 409         }
 410
 411         switch (part) {
 412         case 'd':       /* date */
 413                 strbuf_addstr(sb, show_date(date, tz, dmode));
 414                 return placeholder_len;
 415         case 'D':       /* date, RFC2822 style */
 416                 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 417                 return placeholder_len;
 418         case 'r':       /* date, relative */
 419                 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 420                 return placeholder_len;
 421         case 'i':       /* date, ISO 8601 */
 422                 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 423                 return placeholder_len;
 424         }
 425
 426 skip:
 427         /*
 428          * bogus commit, 'sb' cannot be updated, but we still need to
 429          * compute a valid return value.
 430          */
 431         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 432             || part == 'D' || part == 'r' || part == 'i')
 433                 return placeholder_len;
 434
 435         return 0; /* unknown placeholder */
 436 }
 437
 438 struct chunk {
 439         size_t off;
 440         size_t len;
 441 };
 442
 443 struct format_commit_context {
 444         const struct commit *commit;
 445         enum date_mode dmode;
 446         unsigned commit_header_parsed:1;
 447         unsigned commit_message_parsed:1;
 448         size_t width, indent1, indent2;
 449
 450         /* These offsets are relative to the start of the commit message. */
 451         struct chunk author;
 452         struct chunk committer;
 453         struct chunk encoding;
 454         size_t message_off;
 455         size_t subject_off;
 456         size_t body_off;
 457
 458         /* The following ones are relative to the result struct strbuf. */
 459         struct chunk abbrev_commit_hash;
 460         struct chunk abbrev_tree_hash;
 461         struct chunk abbrev_parent_hashes;
 462         size_t wrap_start;
 463 };
 464
 465 static int add_again(struct strbuf *sb, struct chunk *chunk)
 466 {
 467         if (chunk->len) {
 468                 strbuf_adddup(sb, chunk->off, chunk->len);
 469                 return 1;
 470         }
 471
 472         /*
 473          * We haven't seen this chunk before.  Our caller is surely
 474          * going to add it the hard way now.  Remember the most likely
 475          * start of the to-be-added chunk: the current end of the
 476          * struct strbuf.
 477          */
 478         chunk->off = sb->len;
 479         return 0;
 480 }
 481
 482 static void parse_commit_header(struct format_commit_context *context)
 483 {
 484         const char *msg = context->commit->buffer;
 485         int i;
 486
 487         for (i = 0; msg[i]; i++) {
 488                 int eol;
 489                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 490                         ; /* do nothing */
 491
 492                 if (i == eol) {
 493                         break;
 494                 } else if (!prefixcmp(msg + i, "author ")) {
 495                         context->author.off = i + 7;
 496                         context->author.len = eol - i - 7;
 497                 } else if (!prefixcmp(msg + i, "committer ")) {
 498                         context->committer.off = i + 10;
 499                         context->committer.len = eol - i - 10;
 500                 } else if (!prefixcmp(msg + i, "encoding ")) {
 501                         context->encoding.off = i + 9;
 502                         context->encoding.len = eol - i - 9;
 503                 }
 504                 i = eol;
 505         }
 506         context->message_off = i;
 507         context->commit_header_parsed = 1;
 508 }
 509
 510 static int istitlechar(char c)
 511 {
 512         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 513                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 514 }
 515
 516 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 517 {
 518         size_t trimlen;
 519         size_t start_len = sb->len;
 520         int space = 2;
 521
 522         for (; *msg && *msg != '\n'; msg++) {
 523                 if (istitlechar(*msg)) {
 524                         if (space == 1)
 525                                 strbuf_addch(sb, '-');
 526                         space = 0;
 527                         strbuf_addch(sb, *msg);
 528                         if (*msg == '.')
 529                                 while (*(msg+1) == '.')
 530                                         msg++;
 531                 } else
 532                         space |= 1;
 533         }
 534
 535         /* trim any trailing '.' or '-' characters */
 536         trimlen = 0;
 537         while (sb->len - trimlen > start_len &&
 538                 (sb->buf[sb->len - 1 - trimlen] == '.'
 539                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 540                 trimlen++;
 541         strbuf_remove(sb, sb->len - trimlen, trimlen);
 542 }
 543
 544 const char *format_subject(struct strbuf *sb, const char *msg,
 545                            const char *line_separator)
 546 {
 547         int first = 1;
 548
 549         for (;;) {
 550                 const char *line = msg;
 551                 int linelen = get_one_line(line);
 552
 553                 msg += linelen;
 554                 if (!linelen || is_empty_line(line, &linelen))
 555                         break;
 556
 557                 if (!sb)
 558                         continue;
 559                 strbuf_grow(sb, linelen + 2);
 560                 if (!first)
 561                         strbuf_addstr(sb, line_separator);
 562                 strbuf_add(sb, line, linelen);
 563                 first = 0;
 564         }
 565         return msg;
 566 }
 567
 568 static void parse_commit_message(struct format_commit_context *c)
 569 {
 570         const char *msg = c->commit->buffer + c->message_off;
 571         const char *start = c->commit->buffer;
 572
 573         msg = skip_empty_lines(msg);
 574         c->subject_off = msg - start;
 575
 576         msg = format_subject(NULL, msg, NULL);
 577         msg = skip_empty_lines(msg);
 578         c->body_off = msg - start;
 579
 580         c->commit_message_parsed = 1;
 581 }
 582
 583 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 584 {
 585         struct name_decoration *d;
 586         const char *prefix = " (";
 587
 588         load_ref_decorations(DECORATE_SHORT_REFS);
 589         d = lookup_decoration(&name_decoration, &commit->object);
 590         while (d) {
 591                 strbuf_addstr(sb, prefix);
 592                 prefix = ", ";
 593                 strbuf_addstr(sb, d->name);
 594                 d = d->next;
 595         }
 596         if (prefix[0] == ',')
 597                 strbuf_addch(sb, ')');
 598 }
 599
 600 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 601                         size_t width, size_t indent1, size_t indent2)
 602 {
 603         struct strbuf tmp = STRBUF_INIT;
 604
 605         if (pos)
 606                 strbuf_add(&tmp, sb->buf, pos);
 607         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 608                                 (int) indent1, (int) indent2, (int) width);
 609         strbuf_swap(&tmp, sb);
 610         strbuf_release(&tmp);
 611 }
 612
 613 static void rewrap_message_tail(struct strbuf *sb,
 614                                 struct format_commit_context *c,
 615                                 size_t new_width, size_t new_indent1,
 616                                 size_t new_indent2)
 617 {
 618         if (c->width == new_width && c->indent1 == new_indent1 &&
 619             c->indent2 == new_indent2)
 620                 return;
 621         if (c->wrap_start < sb->len)
 622                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 623         c->wrap_start = sb->len;
 624         c->width = new_width;
 625         c->indent1 = new_indent1;
 626         c->indent2 = new_indent2;
 627 }
 628
 629 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 630                                void *context)
 631 {
 632         struct format_commit_context *c = context;
 633         const struct commit *commit = c->commit;
 634         const char *msg = commit->buffer;
 635         struct commit_list *p;
 636         int h1, h2;
 637
 638         /* these are independent of the commit */
 639         switch (placeholder[0]) {
 640         case 'C':
 641                 if (placeholder[1] == '(') {
 642                         const char *end = strchr(placeholder + 2, ')');
 643                         char color[COLOR_MAXLEN];
 644                         if (!end)
 645                                 return 0;
 646                         color_parse_mem(placeholder + 2,
 647                                         end - (placeholder + 2),
 648                                         "--pretty format", color);
 649                         strbuf_addstr(sb, color);
 650                         return end - placeholder + 1;
 651                 }
 652                 if (!prefixcmp(placeholder + 1, "red")) {
 653                         strbuf_addstr(sb, GIT_COLOR_RED);
 654                         return 4;
 655                 } else if (!prefixcmp(placeholder + 1, "green")) {
 656                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 657                         return 6;
 658                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 659                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 660                         return 5;
 661                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 662                         strbuf_addstr(sb, GIT_COLOR_RESET);
 663                         return 6;
 664                 } else
 665                         return 0;
 666         case 'n':               /* newline */
 667                 strbuf_addch(sb, '\n');
 668                 return 1;
 669         case 'x':
 670                 /* %x00 == NUL, %x0a == LF, etc. */
 671                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 672                     h1 <= 16 &&
 673                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 674                     h2 <= 16) {
 675                         strbuf_addch(sb, (h1<<4)|h2);
 676                         return 3;
 677                 } else
 678                         return 0;
 679         case 'w':
 680                 if (placeholder[1] == '(') {
 681                         unsigned long width = 0, indent1 = 0, indent2 = 0;
 682                         char *next;
 683                         const char *start = placeholder + 2;
 684                         const char *end = strchr(start, ')');
 685                         if (!end)
 686                                 return 0;
 687                         if (end > start) {
 688                                 width = strtoul(start, &next, 10);
 689                                 if (*next == ',') {
 690                                         indent1 = strtoul(next + 1, &next, 10);
 691                                         if (*next == ',') {
 692                                                 indent2 = strtoul(next + 1,
 693                                                                  &next, 10);
 694                                         }
 695                                 }
 696                                 if (*next != ')')
 697                                         return 0;
 698                         }
 699                         rewrap_message_tail(sb, c, width, indent1, indent2);
 700                         return end - placeholder + 1;
 701                 } else
 702                         return 0;
 703         }
 704
 705         /* these depend on the commit */
 706         if (!commit->object.parsed)
 707                 parse_object(commit->object.sha1);
 708
 709         switch (placeholder[0]) {
 710         case 'H':               /* commit hash */
 711                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 712                 return 1;
 713         case 'h':               /* abbreviated commit hash */
 714                 if (add_again(sb, &c->abbrev_commit_hash))
 715                         return 1;
 716                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 717                                                      DEFAULT_ABBREV));
 718                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 719                 return 1;
 720         case 'T':               /* tree hash */
 721                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 722                 return 1;
 723         case 't':               /* abbreviated tree hash */
 724                 if (add_again(sb, &c->abbrev_tree_hash))
 725                         return 1;
 726                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 727                                                      DEFAULT_ABBREV));
 728                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 729                 return 1;
 730         case 'P':               /* parent hashes */
 731                 for (p = commit->parents; p; p = p->next) {
 732                         if (p != commit->parents)
 733                                 strbuf_addch(sb, ' ');
 734                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 735                 }
 736                 return 1;
 737         case 'p':               /* abbreviated parent hashes */
 738                 if (add_again(sb, &c->abbrev_parent_hashes))
 739                         return 1;
 740                 for (p = commit->parents; p; p = p->next) {
 741                         if (p != commit->parents)
 742                                 strbuf_addch(sb, ' ');
 743                         strbuf_addstr(sb, find_unique_abbrev(
 744                                         p->item->object.sha1, DEFAULT_ABBREV));
 745                 }
 746                 c->abbrev_parent_hashes.len = sb->len -
 747                                               c->abbrev_parent_hashes.off;
 748                 return 1;
 749         case 'm':               /* left/right/bottom */
 750                 strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 751                                  ? '-'
 752                                  : (commit->object.flags & SYMMETRIC_LEFT)
 753                                  ? '<'
 754                                  : '>');
 755                 return 1;
 756         case 'd':
 757                 format_decoration(sb, commit);
 758                 return 1;
 759         }
 760
 761         /* For the rest we have to parse the commit header. */
 762         if (!c->commit_header_parsed)
 763                 parse_commit_header(c);
 764
 765         switch (placeholder[0]) {
 766         case 'a':       /* author ... */
 767                 return format_person_part(sb, placeholder[1],
 768                                    msg + c->author.off, c->author.len,
 769                                    c->dmode);
 770         case 'c':       /* committer ... */
 771                 return format_person_part(sb, placeholder[1],
 772                                    msg + c->committer.off, c->committer.len,
 773                                    c->dmode);
 774         case 'e':       /* encoding */
 775                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 776                 return 1;
 777         }
 778
 779         /* Now we need to parse the commit message. */
 780         if (!c->commit_message_parsed)
 781                 parse_commit_message(c);
 782
 783         switch (placeholder[0]) {
 784         case 's':       /* subject */
 785                 format_subject(sb, msg + c->subject_off, " ");
 786                 return 1;
 787         case 'f':       /* sanitized subject */
 788                 format_sanitized_subject(sb, msg + c->subject_off);
 789                 return 1;
 790         case 'b':       /* body */
 791                 strbuf_addstr(sb, msg + c->body_off);
 792                 return 1;
 793         }
 794         return 0;       /* unknown placeholder */
 795 }
 796
 797 void format_commit_message(const struct commit *commit,
 798                            const char *format, struct strbuf *sb,
 799                            enum date_mode dmode)
 800 {
 801         struct format_commit_context context;
 802
 803         memset(&context, 0, sizeof(context));
 804         context.commit = commit;
 805         context.dmode = dmode;
 806         context.wrap_start = sb->len;
 807         strbuf_expand(sb, format, format_commit_item, &context);
 808         rewrap_message_tail(sb, &context, 0, 0, 0);
 809 }
 810
 811 static void pp_header(enum cmit_fmt fmt,
 812                       int abbrev,
 813                       enum date_mode dmode,
 814                       const char *encoding,
 815                       const struct commit *commit,
 816                       const char **msg_p,
 817                       struct strbuf *sb)
 818 {
 819         int parents_shown = 0;
 820
 821         for (;;) {
 822                 const char *line = *msg_p;
 823                 int linelen = get_one_line(*msg_p);
 824
 825                 if (!linelen)
 826                         return;
 827                 *msg_p += linelen;
 828
 829                 if (linelen == 1)
 830                         /* End of header */
 831                         return;
 832
 833                 if (fmt == CMIT_FMT_RAW) {
 834                         strbuf_add(sb, line, linelen);
 835                         continue;
 836                 }
 837
 838                 if (!memcmp(line, "parent ", 7)) {
 839                         if (linelen != 48)
 840                                 die("bad parent line in commit");
 841                         continue;
 842                 }
 843
 844                 if (!parents_shown) {
 845                         struct commit_list *parent;
 846                         int num;
 847                         for (parent = commit->parents, num = 0;
 848                              parent;
 849                              parent = parent->next, num++)
 850                                 ;
 851                         /* with enough slop */
 852                         strbuf_grow(sb, num * 50 + 20);
 853                         add_merge_info(fmt, sb, commit, abbrev);
 854                         parents_shown = 1;
 855                 }
 856
 857                 /*
 858                  * MEDIUM == DEFAULT shows only author with dates.
 859                  * FULL shows both authors but not dates.
 860                  * FULLER shows both authors and dates.
 861                  */
 862                 if (!memcmp(line, "author ", 7)) {
 863                         strbuf_grow(sb, linelen + 80);
 864                         pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
 865                 }
 866                 if (!memcmp(line, "committer ", 10) &&
 867                     (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
 868                         strbuf_grow(sb, linelen + 80);
 869                         pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
 870                 }
 871         }
 872 }
 873
 874 void pp_title_line(enum cmit_fmt fmt,
 875                    const char **msg_p,
 876                    struct strbuf *sb,
 877                    const char *subject,
 878                    const char *after_subject,
 879                    const char *encoding,
 880                    int need_8bit_cte)
 881 {
 882         const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
 883         struct strbuf title;
 884
 885         strbuf_init(&title, 80);
 886         *msg_p = format_subject(&title, *msg_p, line_separator);
 887
 888         strbuf_grow(sb, title.len + 1024);
 889         if (subject) {
 890                 strbuf_addstr(sb, subject);
 891                 add_rfc2047(sb, title.buf, title.len, encoding);
 892         } else {
 893                 strbuf_addbuf(sb, &title);
 894         }
 895         strbuf_addch(sb, '\n');
 896
 897         if (need_8bit_cte > 0) {
 898                 const char *header_fmt =
 899                         "MIME-Version: 1.0\n"
 900                         "Content-Type: text/plain; charset=%s\n"
 901                         "Content-Transfer-Encoding: 8bit\n";
 902                 strbuf_addf(sb, header_fmt, encoding);
 903         }
 904         if (after_subject) {
 905                 strbuf_addstr(sb, after_subject);
 906         }
 907         if (fmt == CMIT_FMT_EMAIL) {
 908                 strbuf_addch(sb, '\n');
 909         }
 910         strbuf_release(&title);
 911 }
 912
 913 void pp_remainder(enum cmit_fmt fmt,
 914                   const char **msg_p,
 915                   struct strbuf *sb,
 916                   int indent)
 917 {
 918         int first = 1;
 919         for (;;) {
 920                 const char *line = *msg_p;
 921                 int linelen = get_one_line(line);
 922                 *msg_p += linelen;
 923
 924                 if (!linelen)
 925                         break;
 926
 927                 if (is_empty_line(line, &linelen)) {
 928                         if (first)
 929                                 continue;
 930                         if (fmt == CMIT_FMT_SHORT)
 931                                 break;
 932                 }
 933                 first = 0;
 934
 935                 strbuf_grow(sb, linelen + indent + 20);
 936                 if (indent) {
 937                         memset(sb->buf + sb->len, ' ', indent);
 938                         strbuf_setlen(sb, sb->len + indent);
 939                 }
 940                 strbuf_add(sb, line, linelen);
 941                 strbuf_addch(sb, '\n');
 942         }
 943 }
 944
 945 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
 946 {
 947         const char *encoding;
 948
 949         encoding = (git_log_output_encoding
 950                     ? git_log_output_encoding
 951                     : git_commit_encoding);
 952         if (!encoding)
 953                 encoding = "UTF-8";
 954         if (encoding_p)
 955                 *encoding_p = encoding;
 956         return logmsg_reencode(commit, encoding);
 957 }
 958
 959 void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
 960                          struct strbuf *sb, int abbrev,
 961                          const char *subject, const char *after_subject,
 962                          enum date_mode dmode, int need_8bit_cte)
 963 {
 964         unsigned long beginning_of_body;
 965         int indent = 4;
 966         const char *msg = commit->buffer;
 967         char *reencoded;
 968         const char *encoding;
 969
 970         if (fmt == CMIT_FMT_USERFORMAT) {
 971                 format_commit_message(commit, user_format, sb, dmode);
 972                 return;
 973         }
 974
 975         reencoded = reencode_commit_message(commit, &encoding);
 976         if (reencoded) {
 977                 msg = reencoded;
 978         }
 979
 980         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
 981                 indent = 0;
 982
 983         /*
 984          * We need to check and emit Content-type: to mark it
 985          * as 8-bit if we haven't done so.
 986          */
 987         if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
 988                 int i, ch, in_body;
 989
 990                 for (in_body = i = 0; (ch = msg[i]); i++) {
 991                         if (!in_body) {
 992                                 /* author could be non 7-bit ASCII but
 993                                  * the log may be so; skip over the
 994                                  * header part first.
 995                                  */
 996                                 if (ch == '\n' && msg[i+1] == '\n')
 997                                         in_body = 1;
 998                         }
 999                         else if (non_ascii(ch)) {
1000                                 need_8bit_cte = 1;
1001                                 break;
1002                         }
1003                 }
1004         }
1005
1006         pp_header(fmt, abbrev, dmode, encoding, commit, &msg, sb);
1007         if (fmt != CMIT_FMT_ONELINE && !subject) {
1008                 strbuf_addch(sb, '\n');
1009         }
1010
1011         /* Skip excess blank lines at the beginning of body, if any... */
1012         msg = skip_empty_lines(msg);
1013
1014         /* These formats treat the title line specially. */
1015         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1016                 pp_title_line(fmt, &msg, sb, subject,
1017                               after_subject, encoding, need_8bit_cte);
1018
1019         beginning_of_body = sb->len;
1020         if (fmt != CMIT_FMT_ONELINE)
1021                 pp_remainder(fmt, &msg, sb, indent);
1022         strbuf_rtrim(sb);
1023
1024         /* Make sure there is an EOLN for the non-oneline case */
1025         if (fmt != CMIT_FMT_ONELINE)
1026                 strbuf_addch(sb, '\n');
1027
1028         /*
1029          * The caller may append additional body text in e-mail
1030          * format.  Make sure we did not strip the blank line
1031          * between the header and the body.
1032          */
1033         if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1034                 strbuf_addch(sb, '\n');
1035         free(reencoded);
1036 }