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         ch = (ch & 0xff);
  87         return ((ch & 0x80) || (ch == 0x1b));
  88 }
  89
  90 static int is_rfc2047_special(char ch)
  91 {
  92         return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
  93 }
  94
  95 static void add_rfc2047(struct strbuf *sb, const char *line, int len,
  96                        const char *encoding)
  97 {
  98         int i, last;
  99
 100         for (i = 0; i < len; i++) {
 101                 int ch = line[i];
 102                 if (non_ascii(ch))
 103                         goto needquote;
 104                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 105                         goto needquote;
 106         }
 107         strbuf_add(sb, line, len);
 108         return;
 109
 110 needquote:
 111         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 112         strbuf_addf(sb, "=?%s?q?", encoding);
 113         for (i = last = 0; i < len; i++) {
 114                 unsigned ch = line[i] & 0xFF;
 115                 /*
 116                  * We encode ' ' using '=20' even though rfc2047
 117                  * allows using '_' for readability.  Unfortunately,
 118                  * many programs do not understand this and just
 119                  * leave the underscore in place.
 120                  */
 121                 if (is_rfc2047_special(ch) || ch == ' ') {
 122                         strbuf_add(sb, line + last, i - last);
 123                         strbuf_addf(sb, "=%02X", ch);
 124                         last = i + 1;
 125                 }
 126         }
 127         strbuf_add(sb, line + last, len - last);
 128         strbuf_addstr(sb, "?=");
 129 }
 130
 131 void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 132                   const char *line, enum date_mode dmode,
 133                   const char *encoding)
 134 {
 135         char *date;
 136         int namelen;
 137         unsigned long time;
 138         int tz;
 139         const char *filler = "    ";
 140
 141         if (fmt == CMIT_FMT_ONELINE)
 142                 return;
 143         date = strchr(line, '>');
 144         if (!date)
 145                 return;
 146         namelen = ++date - line;
 147         time = strtoul(date, &date, 10);
 148         tz = strtol(date, NULL, 10);
 149
 150         if (fmt == CMIT_FMT_EMAIL) {
 151                 char *name_tail = strchr(line, '<');
 152                 int display_name_length;
 153                 if (!name_tail)
 154                         return;
 155                 while (line < name_tail && isspace(name_tail[-1]))
 156                         name_tail--;
 157                 display_name_length = name_tail - line;
 158                 filler = "";
 159                 strbuf_addstr(sb, "From: ");
 160                 add_rfc2047(sb, line, display_name_length, encoding);
 161                 strbuf_add(sb, name_tail, namelen - display_name_length);
 162                 strbuf_addch(sb, '\n');
 163         } else {
 164                 strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 165                               (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 166                               filler, namelen, line);
 167         }
 168         switch (fmt) {
 169         case CMIT_FMT_MEDIUM:
 170                 strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 171                 break;
 172         case CMIT_FMT_EMAIL:
 173                 strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 174                 break;
 175         case CMIT_FMT_FULLER:
 176                 strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 177                 break;
 178         default:
 179                 /* notin' */
 180                 break;
 181         }
 182 }
 183
 184 static int is_empty_line(const char *line, int *len_p)
 185 {
 186         int len = *len_p;
 187         while (len && isspace(line[len-1]))
 188                 len--;
 189         *len_p = len;
 190         return !len;
 191 }
 192
 193 static const char *skip_empty_lines(const char *msg)
 194 {
 195         for (;;) {
 196                 int linelen = get_one_line(msg);
 197                 int ll = linelen;
 198                 if (!linelen)
 199                         break;
 200                 if (!is_empty_line(msg, &ll))
 201                         break;
 202                 msg += linelen;
 203         }
 204         return msg;
 205 }
 206
 207 static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 208                         const struct commit *commit, int abbrev)
 209 {
 210         struct commit_list *parent = commit->parents;
 211
 212         if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 213             !parent || !parent->next)
 214                 return;
 215
 216         strbuf_addstr(sb, "Merge:");
 217
 218         while (parent) {
 219                 struct commit *p = parent->item;
 220                 const char *hex = NULL;
 221                 if (abbrev)
 222                         hex = find_unique_abbrev(p->object.sha1, abbrev);
 223                 if (!hex)
 224                         hex = sha1_to_hex(p->object.sha1);
 225                 parent = parent->next;
 226
 227                 strbuf_addf(sb, " %s", hex);
 228         }
 229         strbuf_addch(sb, '\n');
 230 }
 231
 232 static char *get_header(const struct commit *commit, const char *key)
 233 {
 234         int key_len = strlen(key);
 235         const char *line = commit->buffer;
 236
 237         for (;;) {
 238                 const char *eol = strchr(line, '\n'), *next;
 239
 240                 if (line == eol)
 241                         return NULL;
 242                 if (!eol) {
 243                         eol = line + strlen(line);
 244                         next = NULL;
 245                 } else
 246                         next = eol + 1;
 247                 if (eol - line > key_len &&
 248                     !strncmp(line, key, key_len) &&
 249                     line[key_len] == ' ') {
 250                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 251                 }
 252                 line = next;
 253         }
 254 }
 255
 256 static char *replace_encoding_header(char *buf, const char *encoding)
 257 {
 258         struct strbuf tmp = STRBUF_INIT;
 259         size_t start, len;
 260         char *cp = buf;
 261
 262         /* guess if there is an encoding header before a \n\n */
 263         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 264                 cp = strchr(cp, '\n');
 265                 if (!cp || *++cp == '\n')
 266                         return buf;
 267         }
 268         start = cp - buf;
 269         cp = strchr(cp, '\n');
 270         if (!cp)
 271                 return buf; /* should not happen but be defensive */
 272         len = cp + 1 - (buf + start);
 273
 274         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 275         if (is_encoding_utf8(encoding)) {
 276                 /* we have re-coded to UTF-8; drop the header */
 277                 strbuf_remove(&tmp, start, len);
 278         } else {
 279                 /* just replaces XXXX in 'encoding XXXX\n' */
 280                 strbuf_splice(&tmp, start + strlen("encoding "),
 281                                           len - strlen("encoding \n"),
 282                                           encoding, strlen(encoding));
 283         }
 284         return strbuf_detach(&tmp, NULL);
 285 }
 286
 287 static char *logmsg_reencode(const struct commit *commit,
 288                              const char *output_encoding)
 289 {
 290         static const char *utf8 = "utf-8";
 291         const char *use_encoding;
 292         char *encoding;
 293         char *out;
 294
 295         if (!*output_encoding)
 296                 return NULL;
 297         encoding = get_header(commit, "encoding");
 298         use_encoding = encoding ? encoding : utf8;
 299         if (!strcmp(use_encoding, output_encoding))
 300                 if (encoding) /* we'll strip encoding header later */
 301                         out = xstrdup(commit->buffer);
 302                 else
 303                         return NULL; /* nothing to do */
 304         else
 305                 out = reencode_string(commit->buffer,
 306                                       output_encoding, use_encoding);
 307         if (out)
 308                 out = replace_encoding_header(out, output_encoding);
 309
 310         free(encoding);
 311         return out;
 312 }
 313
 314 static int mailmap_name(char *email, int email_len, char *name, int name_len)
 315 {
 316         static struct string_list *mail_map;
 317         if (!mail_map) {
 318                 mail_map = xcalloc(1, sizeof(*mail_map));
 319                 read_mailmap(mail_map, NULL);
 320         }
 321         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 322 }
 323
 324 static size_t format_person_part(struct strbuf *sb, char part,
 325                                  const char *msg, int len, enum date_mode dmode)
 326 {
 327         /* currently all placeholders have same length */
 328         const int placeholder_len = 2;
 329         int start, end, tz = 0;
 330         unsigned long date = 0;
 331         char *ep;
 332         const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 333         char person_name[1024];
 334         char person_mail[1024];
 335
 336         /* advance 'end' to point to email start delimiter */
 337         for (end = 0; end < len && msg[end] != '<'; end++)
 338                 ; /* do nothing */
 339
 340         /*
 341          * When end points at the '<' that we found, it should have
 342          * matching '>' later, which means 'end' must be strictly
 343          * below len - 1.
 344          */
 345         if (end >= len - 2)
 346                 goto skip;
 347
 348         /* Seek for both name and email part */
 349         name_start = msg;
 350         name_end = msg+end;
 351         while (name_end > name_start && isspace(*(name_end-1)))
 352                 name_end--;
 353         mail_start = msg+end+1;
 354         mail_end = mail_start;
 355         while (mail_end < msg_end && *mail_end != '>')
 356                 mail_end++;
 357         if (mail_end == msg_end)
 358                 goto skip;
 359         end = mail_end-msg;
 360
 361         if (part == 'N' || part == 'E') { /* mailmap lookup */
 362                 strlcpy(person_name, name_start, name_end-name_start+1);
 363                 strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 364                 mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 365                 name_start = person_name;
 366                 name_end = name_start + strlen(person_name);
 367                 mail_start = person_mail;
 368                 mail_end = mail_start +  strlen(person_mail);
 369         }
 370         if (part == 'n' || part == 'N') {       /* name */
 371                 strbuf_add(sb, name_start, name_end-name_start);
 372                 return placeholder_len;
 373         }
 374         if (part == 'e' || part == 'E') {       /* email */
 375                 strbuf_add(sb, mail_start, mail_end-mail_start);
 376                 return placeholder_len;
 377         }
 378
 379         /* advance 'start' to point to date start delimiter */
 380         for (start = end + 1; start < len && isspace(msg[start]); start++)
 381                 ; /* do nothing */
 382         if (start >= len)
 383                 goto skip;
 384         date = strtoul(msg + start, &ep, 10);
 385         if (msg + start == ep)
 386                 goto skip;
 387
 388         if (part == 't') {      /* date, UNIX timestamp */
 389                 strbuf_add(sb, msg + start, ep - (msg + start));
 390                 return placeholder_len;
 391         }
 392
 393         /* parse tz */
 394         for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 395                 ; /* do nothing */
 396         if (start + 1 < len) {
 397                 tz = strtoul(msg + start + 1, NULL, 10);
 398                 if (msg[start] == '-')
 399                         tz = -tz;
 400         }
 401
 402         switch (part) {
 403         case 'd':       /* date */
 404                 strbuf_addstr(sb, show_date(date, tz, dmode));
 405                 return placeholder_len;
 406         case 'D':       /* date, RFC2822 style */
 407                 strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 408                 return placeholder_len;
 409         case 'r':       /* date, relative */
 410                 strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 411                 return placeholder_len;
 412         case 'i':       /* date, ISO 8601 */
 413                 strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 414                 return placeholder_len;
 415         }
 416
 417 skip:
 418         /*
 419          * bogus commit, 'sb' cannot be updated, but we still need to
 420          * compute a valid return value.
 421          */
 422         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 423             || part == 'D' || part == 'r' || part == 'i')
 424                 return placeholder_len;
 425
 426         return 0; /* unknown placeholder */
 427 }
 428
 429 struct chunk {
 430         size_t off;
 431         size_t len;
 432 };
 433
 434 struct format_commit_context {
 435         const struct commit *commit;
 436         enum date_mode dmode;
 437         unsigned commit_header_parsed:1;
 438         unsigned commit_message_parsed:1;
 439
 440         /* These offsets are relative to the start of the commit message. */
 441         struct chunk author;
 442         struct chunk committer;
 443         struct chunk encoding;
 444         size_t message_off;
 445         size_t subject_off;
 446         size_t body_off;
 447
 448         /* The following ones are relative to the result struct strbuf. */
 449         struct chunk abbrev_commit_hash;
 450         struct chunk abbrev_tree_hash;
 451         struct chunk abbrev_parent_hashes;
 452 };
 453
 454 static int add_again(struct strbuf *sb, struct chunk *chunk)
 455 {
 456         if (chunk->len) {
 457                 strbuf_adddup(sb, chunk->off, chunk->len);
 458                 return 1;
 459         }
 460
 461         /*
 462          * We haven't seen this chunk before.  Our caller is surely
 463          * going to add it the hard way now.  Remember the most likely
 464          * start of the to-be-added chunk: the current end of the
 465          * struct strbuf.
 466          */
 467         chunk->off = sb->len;
 468         return 0;
 469 }
 470
 471 static void parse_commit_header(struct format_commit_context *context)
 472 {
 473         const char *msg = context->commit->buffer;
 474         int i;
 475
 476         for (i = 0; msg[i]; i++) {
 477                 int eol;
 478                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 479                         ; /* do nothing */
 480
 481                 if (i == eol) {
 482                         break;
 483                 } else if (!prefixcmp(msg + i, "author ")) {
 484                         context->author.off = i + 7;
 485                         context->author.len = eol - i - 7;
 486                 } else if (!prefixcmp(msg + i, "committer ")) {
 487                         context->committer.off = i + 10;
 488                         context->committer.len = eol - i - 10;
 489                 } else if (!prefixcmp(msg + i, "encoding ")) {
 490                         context->encoding.off = i + 9;
 491                         context->encoding.len = eol - i - 9;
 492                 }
 493                 i = eol;
 494         }
 495         context->message_off = i;
 496         context->commit_header_parsed = 1;
 497 }
 498
 499 const char *format_subject(struct strbuf *sb, const char *msg,
 500                            const char *line_separator)
 501 {
 502         int first = 1;
 503
 504         for (;;) {
 505                 const char *line = msg;
 506                 int linelen = get_one_line(line);
 507
 508                 msg += linelen;
 509                 if (!linelen || is_empty_line(line, &linelen))
 510                         break;
 511
 512                 if (!sb)
 513                         continue;
 514                 strbuf_grow(sb, linelen + 2);
 515                 if (!first)
 516                         strbuf_addstr(sb, line_separator);
 517                 strbuf_add(sb, line, linelen);
 518                 first = 0;
 519         }
 520         return msg;
 521 }
 522
 523 static void parse_commit_message(struct format_commit_context *c)
 524 {
 525         const char *msg = c->commit->buffer + c->message_off;
 526         const char *start = c->commit->buffer;
 527
 528         msg = skip_empty_lines(msg);
 529         c->subject_off = msg - start;
 530
 531         msg = format_subject(NULL, msg, NULL);
 532         msg = skip_empty_lines(msg);
 533         c->body_off = msg - start;
 534
 535         c->commit_message_parsed = 1;
 536 }
 537
 538 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 539 {
 540         struct name_decoration *d;
 541         const char *prefix = " (";
 542
 543         load_ref_decorations();
 544         d = lookup_decoration(&name_decoration, &commit->object);
 545         while (d) {
 546                 strbuf_addstr(sb, prefix);
 547                 prefix = ", ";
 548                 strbuf_addstr(sb, d->name);
 549                 d = d->next;
 550         }
 551         if (prefix[0] == ',')
 552                 strbuf_addch(sb, ')');
 553 }
 554
 555 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 556                                void *context)
 557 {
 558         struct format_commit_context *c = context;
 559         const struct commit *commit = c->commit;
 560         const char *msg = commit->buffer;
 561         struct commit_list *p;
 562         int h1, h2;
 563
 564         /* these are independent of the commit */
 565         switch (placeholder[0]) {
 566         case 'C':
 567                 if (placeholder[1] == '(') {
 568                         const char *end = strchr(placeholder + 2, ')');
 569                         char color[COLOR_MAXLEN];
 570                         if (!end)
 571                                 return 0;
 572                         color_parse_mem(placeholder + 2,
 573                                         end - (placeholder + 2),
 574                                         "--pretty format", color);
 575                         strbuf_addstr(sb, color);
 576                         return end - placeholder + 1;
 577                 }
 578                 if (!prefixcmp(placeholder + 1, "red")) {
 579                         strbuf_addstr(sb, GIT_COLOR_RED);
 580                         return 4;
 581                 } else if (!prefixcmp(placeholder + 1, "green")) {
 582                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 583                         return 6;
 584                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 585                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 586                         return 5;
 587                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 588                         strbuf_addstr(sb, GIT_COLOR_RESET);
 589                         return 6;
 590                 } else
 591                         return 0;
 592         case 'n':               /* newline */
 593                 strbuf_addch(sb, '\n');
 594                 return 1;
 595         case 'x':
 596                 /* %x00 == NUL, %x0a == LF, etc. */
 597                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 598                     h1 <= 16 &&
 599                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 600                     h2 <= 16) {
 601                         strbuf_addch(sb, (h1<<4)|h2);
 602                         return 3;
 603                 } else
 604                         return 0;
 605         }
 606
 607         /* these depend on the commit */
 608         if (!commit->object.parsed)
 609                 parse_object(commit->object.sha1);
 610
 611         switch (placeholder[0]) {
 612         case 'H':               /* commit hash */
 613                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 614                 return 1;
 615         case 'h':               /* abbreviated commit hash */
 616                 if (add_again(sb, &c->abbrev_commit_hash))
 617                         return 1;
 618                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 619                                                      DEFAULT_ABBREV));
 620                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 621                 return 1;
 622         case 'T':               /* tree hash */
 623                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 624                 return 1;
 625         case 't':               /* abbreviated tree hash */
 626                 if (add_again(sb, &c->abbrev_tree_hash))
 627                         return 1;
 628                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 629                                                      DEFAULT_ABBREV));
 630                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 631                 return 1;
 632         case 'P':               /* parent hashes */
 633                 for (p = commit->parents; p; p = p->next) {
 634                         if (p != commit->parents)
 635                                 strbuf_addch(sb, ' ');
 636                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 637                 }
 638                 return 1;
 639         case 'p':               /* abbreviated parent hashes */
 640                 if (add_again(sb, &c->abbrev_parent_hashes))
 641                         return 1;
 642                 for (p = commit->parents; p; p = p->next) {
 643                         if (p != commit->parents)
 644                                 strbuf_addch(sb, ' ');
 645                         strbuf_addstr(sb, find_unique_abbrev(
 646                                         p->item->object.sha1, DEFAULT_ABBREV));
 647                 }
 648                 c->abbrev_parent_hashes.len = sb->len -
 649                                               c->abbrev_parent_hashes.off;
 650                 return 1;
 651         case 'm':               /* left/right/bottom */
 652                 strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 653                                  ? '-'
 654                                  : (commit->object.flags & SYMMETRIC_LEFT)
 655                                  ? '<'
 656                                  : '>');
 657                 return 1;
 658         case 'd':
 659                 format_decoration(sb, commit);
 660                 return 1;
 661         }
 662
 663         /* For the rest we have to parse the commit header. */
 664         if (!c->commit_header_parsed)
 665                 parse_commit_header(c);
 666
 667         switch (placeholder[0]) {
 668         case 'a':       /* author ... */
 669                 return format_person_part(sb, placeholder[1],
 670                                    msg + c->author.off, c->author.len,
 671                                    c->dmode);
 672         case 'c':       /* committer ... */
 673                 return format_person_part(sb, placeholder[1],
 674                                    msg + c->committer.off, c->committer.len,
 675                                    c->dmode);
 676         case 'e':       /* encoding */
 677                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 678                 return 1;
 679         }
 680
 681         /* Now we need to parse the commit message. */
 682         if (!c->commit_message_parsed)
 683                 parse_commit_message(c);
 684
 685         switch (placeholder[0]) {
 686         case 's':       /* subject */
 687                 format_subject(sb, msg + c->subject_off, " ");
 688                 return 1;
 689         case 'b':       /* body */
 690                 strbuf_addstr(sb, msg + c->body_off);
 691                 return 1;
 692         }
 693         return 0;       /* unknown placeholder */
 694 }
 695
 696 void format_commit_message(const struct commit *commit,
 697                            const void *format, struct strbuf *sb,
 698                            enum date_mode dmode)
 699 {
 700         struct format_commit_context context;
 701
 702         memset(&context, 0, sizeof(context));
 703         context.commit = commit;
 704         context.dmode = dmode;
 705         strbuf_expand(sb, format, format_commit_item, &context);
 706 }
 707
 708 static void pp_header(enum cmit_fmt fmt,
 709                       int abbrev,
 710                       enum date_mode dmode,
 711                       const char *encoding,
 712                       const struct commit *commit,
 713                       const char **msg_p,
 714                       struct strbuf *sb)
 715 {
 716         int parents_shown = 0;
 717
 718         for (;;) {
 719                 const char *line = *msg_p;
 720                 int linelen = get_one_line(*msg_p);
 721
 722                 if (!linelen)
 723                         return;
 724                 *msg_p += linelen;
 725
 726                 if (linelen == 1)
 727                         /* End of header */
 728                         return;
 729
 730                 if (fmt == CMIT_FMT_RAW) {
 731                         strbuf_add(sb, line, linelen);
 732                         continue;
 733                 }
 734
 735                 if (!memcmp(line, "parent ", 7)) {
 736                         if (linelen != 48)
 737                                 die("bad parent line in commit");
 738                         continue;
 739                 }
 740
 741                 if (!parents_shown) {
 742                         struct commit_list *parent;
 743                         int num;
 744                         for (parent = commit->parents, num = 0;
 745                              parent;
 746                              parent = parent->next, num++)
 747                                 ;
 748                         /* with enough slop */
 749                         strbuf_grow(sb, num * 50 + 20);
 750                         add_merge_info(fmt, sb, commit, abbrev);
 751                         parents_shown = 1;
 752                 }
 753
 754                 /*
 755                  * MEDIUM == DEFAULT shows only author with dates.
 756                  * FULL shows both authors but not dates.
 757                  * FULLER shows both authors and dates.
 758                  */
 759                 if (!memcmp(line, "author ", 7)) {
 760                         strbuf_grow(sb, linelen + 80);
 761                         pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
 762                 }
 763                 if (!memcmp(line, "committer ", 10) &&
 764                     (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
 765                         strbuf_grow(sb, linelen + 80);
 766                         pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
 767                 }
 768         }
 769 }
 770
 771 void pp_title_line(enum cmit_fmt fmt,
 772                    const char **msg_p,
 773                    struct strbuf *sb,
 774                    const char *subject,
 775                    const char *after_subject,
 776                    const char *encoding,
 777                    int need_8bit_cte)
 778 {
 779         const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
 780         struct strbuf title;
 781
 782         strbuf_init(&title, 80);
 783         *msg_p = format_subject(&title, *msg_p, line_separator);
 784
 785         strbuf_grow(sb, title.len + 1024);
 786         if (subject) {
 787                 strbuf_addstr(sb, subject);
 788                 add_rfc2047(sb, title.buf, title.len, encoding);
 789         } else {
 790                 strbuf_addbuf(sb, &title);
 791         }
 792         strbuf_addch(sb, '\n');
 793
 794         if (need_8bit_cte > 0) {
 795                 const char *header_fmt =
 796                         "MIME-Version: 1.0\n"
 797                         "Content-Type: text/plain; charset=%s\n"
 798                         "Content-Transfer-Encoding: 8bit\n";
 799                 strbuf_addf(sb, header_fmt, encoding);
 800         }
 801         if (after_subject) {
 802                 strbuf_addstr(sb, after_subject);
 803         }
 804         if (fmt == CMIT_FMT_EMAIL) {
 805                 strbuf_addch(sb, '\n');
 806         }
 807         strbuf_release(&title);
 808 }
 809
 810 void pp_remainder(enum cmit_fmt fmt,
 811                   const char **msg_p,
 812                   struct strbuf *sb,
 813                   int indent)
 814 {
 815         int first = 1;
 816         for (;;) {
 817                 const char *line = *msg_p;
 818                 int linelen = get_one_line(line);
 819                 *msg_p += linelen;
 820
 821                 if (!linelen)
 822                         break;
 823
 824                 if (is_empty_line(line, &linelen)) {
 825                         if (first)
 826                                 continue;
 827                         if (fmt == CMIT_FMT_SHORT)
 828                                 break;
 829                 }
 830                 first = 0;
 831
 832                 strbuf_grow(sb, linelen + indent + 20);
 833                 if (indent) {
 834                         memset(sb->buf + sb->len, ' ', indent);
 835                         strbuf_setlen(sb, sb->len + indent);
 836                 }
 837                 strbuf_add(sb, line, linelen);
 838                 strbuf_addch(sb, '\n');
 839         }
 840 }
 841
 842 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
 843 {
 844         const char *encoding;
 845
 846         encoding = (git_log_output_encoding
 847                     ? git_log_output_encoding
 848                     : git_commit_encoding);
 849         if (!encoding)
 850                 encoding = "utf-8";
 851         if (encoding_p)
 852                 *encoding_p = encoding;
 853         return logmsg_reencode(commit, encoding);
 854 }
 855
 856 void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
 857                          struct strbuf *sb, int abbrev,
 858                          const char *subject, const char *after_subject,
 859                          enum date_mode dmode, int need_8bit_cte)
 860 {
 861         unsigned long beginning_of_body;
 862         int indent = 4;
 863         const char *msg = commit->buffer;
 864         char *reencoded;
 865         const char *encoding;
 866
 867         if (fmt == CMIT_FMT_USERFORMAT) {
 868                 format_commit_message(commit, user_format, sb, dmode);
 869                 return;
 870         }
 871
 872         reencoded = reencode_commit_message(commit, &encoding);
 873         if (reencoded) {
 874                 msg = reencoded;
 875         }
 876
 877         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
 878                 indent = 0;
 879
 880         /*
 881          * We need to check and emit Content-type: to mark it
 882          * as 8-bit if we haven't done so.
 883          */
 884         if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
 885                 int i, ch, in_body;
 886
 887                 for (in_body = i = 0; (ch = msg[i]); i++) {
 888                         if (!in_body) {
 889                                 /* author could be non 7-bit ASCII but
 890                                  * the log may be so; skip over the
 891                                  * header part first.
 892                                  */
 893                                 if (ch == '\n' && msg[i+1] == '\n')
 894                                         in_body = 1;
 895                         }
 896                         else if (non_ascii(ch)) {
 897                                 need_8bit_cte = 1;
 898                                 break;
 899                         }
 900                 }
 901         }
 902
 903         pp_header(fmt, abbrev, dmode, encoding, commit, &msg, sb);
 904         if (fmt != CMIT_FMT_ONELINE && !subject) {
 905                 strbuf_addch(sb, '\n');
 906         }
 907
 908         /* Skip excess blank lines at the beginning of body, if any... */
 909         msg = skip_empty_lines(msg);
 910
 911         /* These formats treat the title line specially. */
 912         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
 913                 pp_title_line(fmt, &msg, sb, subject,
 914                               after_subject, encoding, need_8bit_cte);
 915
 916         beginning_of_body = sb->len;
 917         if (fmt != CMIT_FMT_ONELINE)
 918                 pp_remainder(fmt, &msg, sb, indent);
 919         strbuf_rtrim(sb);
 920
 921         /* Make sure there is an EOLN for the non-oneline case */
 922         if (fmt != CMIT_FMT_ONELINE)
 923                 strbuf_addch(sb, '\n');
 924
 925         /*
 926          * The caller may append additional body text in e-mail
 927          * format.  Make sure we did not strip the blank line
 928          * between the header and the body.
 929          */
 930         if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
 931                 strbuf_addch(sb, '\n');
 932         free(reencoded);
 933 }