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
 449         /* These offsets are relative to the start of the commit message. */
 450         struct chunk author;
 451         struct chunk committer;
 452         struct chunk encoding;
 453         size_t message_off;
 454         size_t subject_off;
 455         size_t body_off;
 456
 457         /* The following ones are relative to the result struct strbuf. */
 458         struct chunk abbrev_commit_hash;
 459         struct chunk abbrev_tree_hash;
 460         struct chunk abbrev_parent_hashes;
 461 };
 462
 463 static int add_again(struct strbuf *sb, struct chunk *chunk)
 464 {
 465         if (chunk->len) {
 466                 strbuf_adddup(sb, chunk->off, chunk->len);
 467                 return 1;
 468         }
 469
 470         /*
 471          * We haven't seen this chunk before.  Our caller is surely
 472          * going to add it the hard way now.  Remember the most likely
 473          * start of the to-be-added chunk: the current end of the
 474          * struct strbuf.
 475          */
 476         chunk->off = sb->len;
 477         return 0;
 478 }
 479
 480 static void parse_commit_header(struct format_commit_context *context)
 481 {
 482         const char *msg = context->commit->buffer;
 483         int i;
 484
 485         for (i = 0; msg[i]; i++) {
 486                 int eol;
 487                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 488                         ; /* do nothing */
 489
 490                 if (i == eol) {
 491                         break;
 492                 } else if (!prefixcmp(msg + i, "author ")) {
 493                         context->author.off = i + 7;
 494                         context->author.len = eol - i - 7;
 495                 } else if (!prefixcmp(msg + i, "committer ")) {
 496                         context->committer.off = i + 10;
 497                         context->committer.len = eol - i - 10;
 498                 } else if (!prefixcmp(msg + i, "encoding ")) {
 499                         context->encoding.off = i + 9;
 500                         context->encoding.len = eol - i - 9;
 501                 }
 502                 i = eol;
 503         }
 504         context->message_off = i;
 505         context->commit_header_parsed = 1;
 506 }
 507
 508 static int istitlechar(char c)
 509 {
 510         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 511                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 512 }
 513
 514 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 515 {
 516         size_t trimlen;
 517         size_t start_len = sb->len;
 518         int space = 2;
 519
 520         for (; *msg && *msg != '\n'; msg++) {
 521                 if (istitlechar(*msg)) {
 522                         if (space == 1)
 523                                 strbuf_addch(sb, '-');
 524                         space = 0;
 525                         strbuf_addch(sb, *msg);
 526                         if (*msg == '.')
 527                                 while (*(msg+1) == '.')
 528                                         msg++;
 529                 } else
 530                         space |= 1;
 531         }
 532
 533         /* trim any trailing '.' or '-' characters */
 534         trimlen = 0;
 535         while (sb->len - trimlen > start_len &&
 536                 (sb->buf[sb->len - 1 - trimlen] == '.'
 537                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 538                 trimlen++;
 539         strbuf_remove(sb, sb->len - trimlen, trimlen);
 540 }
 541
 542 const char *format_subject(struct strbuf *sb, const char *msg,
 543                            const char *line_separator)
 544 {
 545         int first = 1;
 546
 547         for (;;) {
 548                 const char *line = msg;
 549                 int linelen = get_one_line(line);
 550
 551                 msg += linelen;
 552                 if (!linelen || is_empty_line(line, &linelen))
 553                         break;
 554
 555                 if (!sb)
 556                         continue;
 557                 strbuf_grow(sb, linelen + 2);
 558                 if (!first)
 559                         strbuf_addstr(sb, line_separator);
 560                 strbuf_add(sb, line, linelen);
 561                 first = 0;
 562         }
 563         return msg;
 564 }
 565
 566 static void parse_commit_message(struct format_commit_context *c)
 567 {
 568         const char *msg = c->commit->buffer + c->message_off;
 569         const char *start = c->commit->buffer;
 570
 571         msg = skip_empty_lines(msg);
 572         c->subject_off = msg - start;
 573
 574         msg = format_subject(NULL, msg, NULL);
 575         msg = skip_empty_lines(msg);
 576         c->body_off = msg - start;
 577
 578         c->commit_message_parsed = 1;
 579 }
 580
 581 static void format_decoration(struct strbuf *sb, const struct commit *commit)
 582 {
 583         struct name_decoration *d;
 584         const char *prefix = " (";
 585
 586         load_ref_decorations(DECORATE_SHORT_REFS);
 587         d = lookup_decoration(&name_decoration, &commit->object);
 588         while (d) {
 589                 strbuf_addstr(sb, prefix);
 590                 prefix = ", ";
 591                 strbuf_addstr(sb, d->name);
 592                 d = d->next;
 593         }
 594         if (prefix[0] == ',')
 595                 strbuf_addch(sb, ')');
 596 }
 597
 598 static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 599                                 void *context)
 600 {
 601         struct format_commit_context *c = context;
 602         const struct commit *commit = c->commit;
 603         const char *msg = commit->buffer;
 604         struct commit_list *p;
 605         int h1, h2;
 606
 607         /* these are independent of the commit */
 608         switch (placeholder[0]) {
 609         case 'C':
 610                 if (placeholder[1] == '(') {
 611                         const char *end = strchr(placeholder + 2, ')');
 612                         char color[COLOR_MAXLEN];
 613                         if (!end)
 614                                 return 0;
 615                         color_parse_mem(placeholder + 2,
 616                                         end - (placeholder + 2),
 617                                         "--pretty format", color);
 618                         strbuf_addstr(sb, color);
 619                         return end - placeholder + 1;
 620                 }
 621                 if (!prefixcmp(placeholder + 1, "red")) {
 622                         strbuf_addstr(sb, GIT_COLOR_RED);
 623                         return 4;
 624                 } else if (!prefixcmp(placeholder + 1, "green")) {
 625                         strbuf_addstr(sb, GIT_COLOR_GREEN);
 626                         return 6;
 627                 } else if (!prefixcmp(placeholder + 1, "blue")) {
 628                         strbuf_addstr(sb, GIT_COLOR_BLUE);
 629                         return 5;
 630                 } else if (!prefixcmp(placeholder + 1, "reset")) {
 631                         strbuf_addstr(sb, GIT_COLOR_RESET);
 632                         return 6;
 633                 } else
 634                         return 0;
 635         case 'n':               /* newline */
 636                 strbuf_addch(sb, '\n');
 637                 return 1;
 638         case 'x':
 639                 /* %x00 == NUL, %x0a == LF, etc. */
 640                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 641                     h1 <= 16 &&
 642                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 643                     h2 <= 16) {
 644                         strbuf_addch(sb, (h1<<4)|h2);
 645                         return 3;
 646                 } else
 647                         return 0;
 648         }
 649
 650         /* these depend on the commit */
 651         if (!commit->object.parsed)
 652                 parse_object(commit->object.sha1);
 653
 654         switch (placeholder[0]) {
 655         case 'H':               /* commit hash */
 656                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 657                 return 1;
 658         case 'h':               /* abbreviated commit hash */
 659                 if (add_again(sb, &c->abbrev_commit_hash))
 660                         return 1;
 661                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 662                                                      DEFAULT_ABBREV));
 663                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 664                 return 1;
 665         case 'T':               /* tree hash */
 666                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 667                 return 1;
 668         case 't':               /* abbreviated tree hash */
 669                 if (add_again(sb, &c->abbrev_tree_hash))
 670                         return 1;
 671                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 672                                                      DEFAULT_ABBREV));
 673                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 674                 return 1;
 675         case 'P':               /* parent hashes */
 676                 for (p = commit->parents; p; p = p->next) {
 677                         if (p != commit->parents)
 678                                 strbuf_addch(sb, ' ');
 679                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 680                 }
 681                 return 1;
 682         case 'p':               /* abbreviated parent hashes */
 683                 if (add_again(sb, &c->abbrev_parent_hashes))
 684                         return 1;
 685                 for (p = commit->parents; p; p = p->next) {
 686                         if (p != commit->parents)
 687                                 strbuf_addch(sb, ' ');
 688                         strbuf_addstr(sb, find_unique_abbrev(
 689                                         p->item->object.sha1, DEFAULT_ABBREV));
 690                 }
 691                 c->abbrev_parent_hashes.len = sb->len -
 692                                               c->abbrev_parent_hashes.off;
 693                 return 1;
 694         case 'm':               /* left/right/bottom */
 695                 strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 696                                  ? '-'
 697                                  : (commit->object.flags & SYMMETRIC_LEFT)
 698                                  ? '<'
 699                                  : '>');
 700                 return 1;
 701         case 'd':
 702                 format_decoration(sb, commit);
 703                 return 1;
 704         }
 705
 706         /* For the rest we have to parse the commit header. */
 707         if (!c->commit_header_parsed)
 708                 parse_commit_header(c);
 709
 710         switch (placeholder[0]) {
 711         case 'a':       /* author ... */
 712                 return format_person_part(sb, placeholder[1],
 713                                    msg + c->author.off, c->author.len,
 714                                    c->dmode);
 715         case 'c':       /* committer ... */
 716                 return format_person_part(sb, placeholder[1],
 717                                    msg + c->committer.off, c->committer.len,
 718                                    c->dmode);
 719         case 'e':       /* encoding */
 720                 strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 721                 return 1;
 722         }
 723
 724         /* Now we need to parse the commit message. */
 725         if (!c->commit_message_parsed)
 726                 parse_commit_message(c);
 727
 728         switch (placeholder[0]) {
 729         case 's':       /* subject */
 730                 format_subject(sb, msg + c->subject_off, " ");
 731                 return 1;
 732         case 'f':       /* sanitized subject */
 733                 format_sanitized_subject(sb, msg + c->subject_off);
 734                 return 1;
 735         case 'b':       /* body */
 736                 strbuf_addstr(sb, msg + c->body_off);
 737                 return 1;
 738         }
 739         return 0;       /* unknown placeholder */
 740 }
 741
 742 static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 743                                  void *context)
 744 {
 745         int consumed;
 746         size_t orig_len;
 747         enum {
 748                 NO_MAGIC,
 749                 ADD_LF_BEFORE_NON_EMPTY,
 750                 DEL_LF_BEFORE_EMPTY,
 751         } magic = NO_MAGIC;
 752
 753         switch (placeholder[0]) {
 754         case '-':
 755                 magic = DEL_LF_BEFORE_EMPTY;
 756                 break;
 757         case '+':
 758                 magic = ADD_LF_BEFORE_NON_EMPTY;
 759                 break;
 760         default:
 761                 break;
 762         }
 763         if (magic != NO_MAGIC)
 764                 placeholder++;
 765
 766         orig_len = sb->len;
 767         consumed = format_commit_one(sb, placeholder, context);
 768         if (magic == NO_MAGIC)
 769                 return consumed;
 770
 771         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
 772                 while (sb->len && sb->buf[sb->len - 1] == '\n')
 773                         strbuf_setlen(sb, sb->len - 1);
 774         } else if ((orig_len != sb->len) && magic == ADD_LF_BEFORE_NON_EMPTY) {
 775                 strbuf_insert(sb, orig_len, "\n", 1);
 776         }
 777         return consumed + 1;
 778 }
 779
 780 void format_commit_message(const struct commit *commit,
 781                            const void *format, struct strbuf *sb,
 782                            enum date_mode dmode)
 783 {
 784         struct format_commit_context context;
 785
 786         memset(&context, 0, sizeof(context));
 787         context.commit = commit;
 788         context.dmode = dmode;
 789         strbuf_expand(sb, format, format_commit_item, &context);
 790 }
 791
 792 static void pp_header(enum cmit_fmt fmt,
 793                       int abbrev,
 794                       enum date_mode dmode,
 795                       const char *encoding,
 796                       const struct commit *commit,
 797                       const char **msg_p,
 798                       struct strbuf *sb)
 799 {
 800         int parents_shown = 0;
 801
 802         for (;;) {
 803                 const char *line = *msg_p;
 804                 int linelen = get_one_line(*msg_p);
 805
 806                 if (!linelen)
 807                         return;
 808                 *msg_p += linelen;
 809
 810                 if (linelen == 1)
 811                         /* End of header */
 812                         return;
 813
 814                 if (fmt == CMIT_FMT_RAW) {
 815                         strbuf_add(sb, line, linelen);
 816                         continue;
 817                 }
 818
 819                 if (!memcmp(line, "parent ", 7)) {
 820                         if (linelen != 48)
 821                                 die("bad parent line in commit");
 822                         continue;
 823                 }
 824
 825                 if (!parents_shown) {
 826                         struct commit_list *parent;
 827                         int num;
 828                         for (parent = commit->parents, num = 0;
 829                              parent;
 830                              parent = parent->next, num++)
 831                                 ;
 832                         /* with enough slop */
 833                         strbuf_grow(sb, num * 50 + 20);
 834                         add_merge_info(fmt, sb, commit, abbrev);
 835                         parents_shown = 1;
 836                 }
 837
 838                 /*
 839                  * MEDIUM == DEFAULT shows only author with dates.
 840                  * FULL shows both authors but not dates.
 841                  * FULLER shows both authors and dates.
 842                  */
 843                 if (!memcmp(line, "author ", 7)) {
 844                         strbuf_grow(sb, linelen + 80);
 845                         pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
 846                 }
 847                 if (!memcmp(line, "committer ", 10) &&
 848                     (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
 849                         strbuf_grow(sb, linelen + 80);
 850                         pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
 851                 }
 852         }
 853 }
 854
 855 void pp_title_line(enum cmit_fmt fmt,
 856                    const char **msg_p,
 857                    struct strbuf *sb,
 858                    const char *subject,
 859                    const char *after_subject,
 860                    const char *encoding,
 861                    int need_8bit_cte)
 862 {
 863         const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
 864         struct strbuf title;
 865
 866         strbuf_init(&title, 80);
 867         *msg_p = format_subject(&title, *msg_p, line_separator);
 868
 869         strbuf_grow(sb, title.len + 1024);
 870         if (subject) {
 871                 strbuf_addstr(sb, subject);
 872                 add_rfc2047(sb, title.buf, title.len, encoding);
 873         } else {
 874                 strbuf_addbuf(sb, &title);
 875         }
 876         strbuf_addch(sb, '\n');
 877
 878         if (need_8bit_cte > 0) {
 879                 const char *header_fmt =
 880                         "MIME-Version: 1.0\n"
 881                         "Content-Type: text/plain; charset=%s\n"
 882                         "Content-Transfer-Encoding: 8bit\n";
 883                 strbuf_addf(sb, header_fmt, encoding);
 884         }
 885         if (after_subject) {
 886                 strbuf_addstr(sb, after_subject);
 887         }
 888         if (fmt == CMIT_FMT_EMAIL) {
 889                 strbuf_addch(sb, '\n');
 890         }
 891         strbuf_release(&title);
 892 }
 893
 894 void pp_remainder(enum cmit_fmt fmt,
 895                   const char **msg_p,
 896                   struct strbuf *sb,
 897                   int indent)
 898 {
 899         int first = 1;
 900         for (;;) {
 901                 const char *line = *msg_p;
 902                 int linelen = get_one_line(line);
 903                 *msg_p += linelen;
 904
 905                 if (!linelen)
 906                         break;
 907
 908                 if (is_empty_line(line, &linelen)) {
 909                         if (first)
 910                                 continue;
 911                         if (fmt == CMIT_FMT_SHORT)
 912                                 break;
 913                 }
 914                 first = 0;
 915
 916                 strbuf_grow(sb, linelen + indent + 20);
 917                 if (indent) {
 918                         memset(sb->buf + sb->len, ' ', indent);
 919                         strbuf_setlen(sb, sb->len + indent);
 920                 }
 921                 strbuf_add(sb, line, linelen);
 922                 strbuf_addch(sb, '\n');
 923         }
 924 }
 925
 926 char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
 927 {
 928         const char *encoding;
 929
 930         encoding = (git_log_output_encoding
 931                     ? git_log_output_encoding
 932                     : git_commit_encoding);
 933         if (!encoding)
 934                 encoding = "UTF-8";
 935         if (encoding_p)
 936                 *encoding_p = encoding;
 937         return logmsg_reencode(commit, encoding);
 938 }
 939
 940 void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
 941                          struct strbuf *sb, int abbrev,
 942                          const char *subject, const char *after_subject,
 943                          enum date_mode dmode, int need_8bit_cte)
 944 {
 945         unsigned long beginning_of_body;
 946         int indent = 4;
 947         const char *msg = commit->buffer;
 948         char *reencoded;
 949         const char *encoding;
 950
 951         if (fmt == CMIT_FMT_USERFORMAT) {
 952                 format_commit_message(commit, user_format, sb, dmode);
 953                 return;
 954         }
 955
 956         reencoded = reencode_commit_message(commit, &encoding);
 957         if (reencoded) {
 958                 msg = reencoded;
 959         }
 960
 961         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
 962                 indent = 0;
 963
 964         /*
 965          * We need to check and emit Content-type: to mark it
 966          * as 8-bit if we haven't done so.
 967          */
 968         if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
 969                 int i, ch, in_body;
 970
 971                 for (in_body = i = 0; (ch = msg[i]); i++) {
 972                         if (!in_body) {
 973                                 /* author could be non 7-bit ASCII but
 974                                  * the log may be so; skip over the
 975                                  * header part first.
 976                                  */
 977                                 if (ch == '\n' && msg[i+1] == '\n')
 978                                         in_body = 1;
 979                         }
 980                         else if (non_ascii(ch)) {
 981                                 need_8bit_cte = 1;
 982                                 break;
 983                         }
 984                 }
 985         }
 986
 987         pp_header(fmt, abbrev, dmode, encoding, commit, &msg, sb);
 988         if (fmt != CMIT_FMT_ONELINE && !subject) {
 989                 strbuf_addch(sb, '\n');
 990         }
 991
 992         /* Skip excess blank lines at the beginning of body, if any... */
 993         msg = skip_empty_lines(msg);
 994
 995         /* These formats treat the title line specially. */
 996         if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
 997                 pp_title_line(fmt, &msg, sb, subject,
 998                               after_subject, encoding, need_8bit_cte);
 999
1000         beginning_of_body = sb->len;
1001         if (fmt != CMIT_FMT_ONELINE)
1002                 pp_remainder(fmt, &msg, sb, indent);
1003         strbuf_rtrim(sb);
1004
1005         /* Make sure there is an EOLN for the non-oneline case */
1006         if (fmt != CMIT_FMT_ONELINE)
1007                 strbuf_addch(sb, '\n');
1008
1009         /*
1010          * The caller may append additional body text in e-mail
1011          * format.  Make sure we did not strip the blank line
1012          * between the header and the body.
1013          */
1014         if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1015                 strbuf_addch(sb, '\n');
1016         free(reencoded);
1017 }