git.oblomov.eu Git - git/blob - pretty.c

   1 #include "cache.h"
   2 #include "config.h"
   3 #include "commit.h"
   4 #include "utf8.h"
   5 #include "diff.h"
   6 #include "revision.h"
   7 #include "string-list.h"
   8 #include "mailmap.h"
   9 #include "log-tree.h"
  10 #include "notes.h"
  11 #include "color.h"
  12 #include "reflog-walk.h"
  13 #include "gpg-interface.h"
  14 #include "trailer.h"
  15
  16 static char *user_format;
  17 static struct cmt_fmt_map {
  18         const char *name;
  19         enum cmit_fmt format;
  20         int is_tformat;
  21         int expand_tabs_in_log;
  22         int is_alias;
  23         enum date_mode_type default_date_mode_type;
  24         const char *user_format;
  25 } *commit_formats;
  26 static size_t builtin_formats_len;
  27 static size_t commit_formats_len;
  28 static size_t commit_formats_alloc;
  29 static struct cmt_fmt_map *find_commit_format(const char *sought);
  30
  31 int commit_format_is_empty(enum cmit_fmt fmt)
  32 {
  33         return fmt == CMIT_FMT_USERFORMAT && !*user_format;
  34 }
  35
  36 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  37 {
  38         free(user_format);
  39         user_format = xstrdup(cp);
  40         if (is_tformat)
  41                 rev->use_terminator = 1;
  42         rev->commit_format = CMIT_FMT_USERFORMAT;
  43 }
  44
  45 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  46 {
  47         struct cmt_fmt_map *commit_format = NULL;
  48         const char *name;
  49         const char *fmt;
  50         int i;
  51
  52         if (!skip_prefix(var, "pretty.", &name))
  53                 return 0;
  54
  55         for (i = 0; i < builtin_formats_len; i++) {
  56                 if (!strcmp(commit_formats[i].name, name))
  57                         return 0;
  58         }
  59
  60         for (i = builtin_formats_len; i < commit_formats_len; i++) {
  61                 if (!strcmp(commit_formats[i].name, name)) {
  62                         commit_format = &commit_formats[i];
  63                         break;
  64                 }
  65         }
  66
  67         if (!commit_format) {
  68                 ALLOC_GROW(commit_formats, commit_formats_len+1,
  69                            commit_formats_alloc);
  70                 commit_format = &commit_formats[commit_formats_len];
  71                 memset(commit_format, 0, sizeof(*commit_format));
  72                 commit_formats_len++;
  73         }
  74
  75         commit_format->name = xstrdup(name);
  76         commit_format->format = CMIT_FMT_USERFORMAT;
  77         if (git_config_string(&fmt, var, value))
  78                 return -1;
  79
  80         if (skip_prefix(fmt, "format:", &fmt))
  81                 commit_format->is_tformat = 0;
  82         else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
  83                 commit_format->is_tformat = 1;
  84         else
  85                 commit_format->is_alias = 1;
  86         commit_format->user_format = fmt;
  87
  88         return 0;
  89 }
  90
  91 static void setup_commit_formats(void)
  92 {
  93         struct cmt_fmt_map builtin_formats[] = {
  94                 { "raw",        CMIT_FMT_RAW,           0,      0 },
  95                 { "medium",     CMIT_FMT_MEDIUM,        0,      8 },
  96                 { "short",      CMIT_FMT_SHORT,         0,      0 },
  97                 { "email",      CMIT_FMT_EMAIL,         0,      0 },
  98                 { "mboxrd",     CMIT_FMT_MBOXRD,        0,      0 },
  99                 { "fuller",     CMIT_FMT_FULLER,        0,      8 },
 100                 { "full",       CMIT_FMT_FULL,          0,      8 },
 101                 { "oneline",    CMIT_FMT_ONELINE,       1,      0 },
 102                 { "reference",  CMIT_FMT_USERFORMAT,    1,      0,
 103                         0, DATE_SHORT, "%C(auto)%h (%s, %ad)" },
 104                 /*
 105                  * Please update $__git_log_pretty_formats in
 106                  * git-completion.bash when you add new formats.
 107                  */
 108         };
 109         commit_formats_len = ARRAY_SIZE(builtin_formats);
 110         builtin_formats_len = commit_formats_len;
 111         ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
 112         COPY_ARRAY(commit_formats, builtin_formats,
 113                    ARRAY_SIZE(builtin_formats));
 114
 115         git_config(git_pretty_formats_config, NULL);
 116 }
 117
 118 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 119                                                         const char *original,
 120                                                         int num_redirections)
 121 {
 122         struct cmt_fmt_map *found = NULL;
 123         size_t found_match_len = 0;
 124         int i;
 125
 126         if (num_redirections >= commit_formats_len)
 127                 die("invalid --pretty format: "
 128                     "'%s' references an alias which points to itself",
 129                     original);
 130
 131         for (i = 0; i < commit_formats_len; i++) {
 132                 size_t match_len;
 133
 134                 if (!starts_with(commit_formats[i].name, sought))
 135                         continue;
 136
 137                 match_len = strlen(commit_formats[i].name);
 138                 if (found == NULL || found_match_len > match_len) {
 139                         found = &commit_formats[i];
 140                         found_match_len = match_len;
 141                 }
 142         }
 143
 144         if (found && found->is_alias) {
 145                 found = find_commit_format_recursive(found->user_format,
 146                                                      original,
 147                                                      num_redirections+1);
 148         }
 149
 150         return found;
 151 }
 152
 153 static struct cmt_fmt_map *find_commit_format(const char *sought)
 154 {
 155         if (!commit_formats)
 156                 setup_commit_formats();
 157
 158         return find_commit_format_recursive(sought, sought, 0);
 159 }
 160
 161 void get_commit_format(const char *arg, struct rev_info *rev)
 162 {
 163         struct cmt_fmt_map *commit_format;
 164
 165         rev->use_terminator = 0;
 166         if (!arg) {
 167                 rev->commit_format = CMIT_FMT_DEFAULT;
 168                 return;
 169         }
 170         if (skip_prefix(arg, "format:", &arg)) {
 171                 save_user_format(rev, arg, 0);
 172                 return;
 173         }
 174
 175         if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
 176                 save_user_format(rev, arg, 1);
 177                 return;
 178         }
 179
 180         commit_format = find_commit_format(arg);
 181         if (!commit_format)
 182                 die("invalid --pretty format: %s", arg);
 183
 184         rev->commit_format = commit_format->format;
 185         rev->use_terminator = commit_format->is_tformat;
 186         rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
 187         if (!rev->date_mode_explicit && commit_format->default_date_mode_type)
 188                 rev->date_mode.type = commit_format->default_date_mode_type;
 189         if (commit_format->format == CMIT_FMT_USERFORMAT) {
 190                 save_user_format(rev, commit_format->user_format,
 191                                  commit_format->is_tformat);
 192         }
 193 }
 194
 195 /*
 196  * Generic support for pretty-printing the header
 197  */
 198 static int get_one_line(const char *msg)
 199 {
 200         int ret = 0;
 201
 202         for (;;) {
 203                 char c = *msg++;
 204                 if (!c)
 205                         break;
 206                 ret++;
 207                 if (c == '\n')
 208                         break;
 209         }
 210         return ret;
 211 }
 212
 213 /* High bit set, or ISO-2022-INT */
 214 static int non_ascii(int ch)
 215 {
 216         return !isascii(ch) || ch == '\033';
 217 }
 218
 219 int has_non_ascii(const char *s)
 220 {
 221         int ch;
 222         if (!s)
 223                 return 0;
 224         while ((ch = *s++) != '\0') {
 225                 if (non_ascii(ch))
 226                         return 1;
 227         }
 228         return 0;
 229 }
 230
 231 static int is_rfc822_special(char ch)
 232 {
 233         switch (ch) {
 234         case '(':
 235         case ')':
 236         case '<':
 237         case '>':
 238         case '[':
 239         case ']':
 240         case ':':
 241         case ';':
 242         case '@':
 243         case ',':
 244         case '.':
 245         case '"':
 246         case '\\':
 247                 return 1;
 248         default:
 249                 return 0;
 250         }
 251 }
 252
 253 static int needs_rfc822_quoting(const char *s, int len)
 254 {
 255         int i;
 256         for (i = 0; i < len; i++)
 257                 if (is_rfc822_special(s[i]))
 258                         return 1;
 259         return 0;
 260 }
 261
 262 static int last_line_length(struct strbuf *sb)
 263 {
 264         int i;
 265
 266         /* How many bytes are already used on the last line? */
 267         for (i = sb->len - 1; i >= 0; i--)
 268                 if (sb->buf[i] == '\n')
 269                         break;
 270         return sb->len - (i + 1);
 271 }
 272
 273 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 274 {
 275         int i;
 276
 277         /* just a guess, we may have to also backslash-quote */
 278         strbuf_grow(out, len + 2);
 279
 280         strbuf_addch(out, '"');
 281         for (i = 0; i < len; i++) {
 282                 switch (s[i]) {
 283                 case '"':
 284                 case '\\':
 285                         strbuf_addch(out, '\\');
 286                         /* fall through */
 287                 default:
 288                         strbuf_addch(out, s[i]);
 289                 }
 290         }
 291         strbuf_addch(out, '"');
 292 }
 293
 294 enum rfc2047_type {
 295         RFC2047_SUBJECT,
 296         RFC2047_ADDRESS
 297 };
 298
 299 static int is_rfc2047_special(char ch, enum rfc2047_type type)
 300 {
 301         /*
 302          * rfc2047, section 4.2:
 303          *
 304          *    8-bit values which correspond to printable ASCII characters other
 305          *    than "=", "?", and "_" (underscore), MAY be represented as those
 306          *    characters.  (But see section 5 for restrictions.)  In
 307          *    particular, SPACE and TAB MUST NOT be represented as themselves
 308          *    within encoded words.
 309          */
 310
 311         /*
 312          * rule out non-ASCII characters and non-printable characters (the
 313          * non-ASCII check should be redundant as isprint() is not localized
 314          * and only knows about ASCII, but be defensive about that)
 315          */
 316         if (non_ascii(ch) || !isprint(ch))
 317                 return 1;
 318
 319         /*
 320          * rule out special printable characters (' ' should be the only
 321          * whitespace character considered printable, but be defensive and use
 322          * isspace())
 323          */
 324         if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 325                 return 1;
 326
 327         /*
 328          * rfc2047, section 5.3:
 329          *
 330          *    As a replacement for a 'word' entity within a 'phrase', for example,
 331          *    one that precedes an address in a From, To, or Cc header.  The ABNF
 332          *    definition for 'phrase' from RFC 822 thus becomes:
 333          *
 334          *    phrase = 1*( encoded-word / word )
 335          *
 336          *    In this case the set of characters that may be used in a "Q"-encoded
 337          *    'encoded-word' is restricted to: <upper and lower case ASCII
 338          *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 339          *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 340          *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 341          *    'special' by 'linear-white-space'.
 342          */
 343
 344         if (type != RFC2047_ADDRESS)
 345                 return 0;
 346
 347         /* '=' and '_' are special cases and have been checked above */
 348         return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 349 }
 350
 351 static int needs_rfc2047_encoding(const char *line, int len)
 352 {
 353         int i;
 354
 355         for (i = 0; i < len; i++) {
 356                 int ch = line[i];
 357                 if (non_ascii(ch) || ch == '\n')
 358                         return 1;
 359                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 360                         return 1;
 361         }
 362
 363         return 0;
 364 }
 365
 366 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 367                        const char *encoding, enum rfc2047_type type)
 368 {
 369         static const int max_encoded_length = 76; /* per rfc2047 */
 370         int i;
 371         int line_len = last_line_length(sb);
 372
 373         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 374         strbuf_addf(sb, "=?%s?q?", encoding);
 375         line_len += strlen(encoding) + 5; /* 5 for =??q? */
 376
 377         while (len) {
 378                 /*
 379                  * RFC 2047, section 5 (3):
 380                  *
 381                  * Each 'encoded-word' MUST represent an integral number of
 382                  * characters.  A multi-octet character may not be split across
 383                  * adjacent 'encoded- word's.
 384                  */
 385                 const unsigned char *p = (const unsigned char *)line;
 386                 int chrlen = mbs_chrlen(&line, &len, encoding);
 387                 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 388
 389                 /* "=%02X" * chrlen, or the byte itself */
 390                 const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 391                 int         encoded_len = is_special ? 3 * chrlen : 1;
 392
 393                 /*
 394                  * According to RFC 2047, we could encode the special character
 395                  * ' ' (space) with '_' (underscore) for readability. But many
 396                  * programs do not understand this and just leave the
 397                  * underscore in place. Thus, we do nothing special here, which
 398                  * causes ' ' to be encoded as '=20', avoiding this problem.
 399                  */
 400
 401                 if (line_len + encoded_len + 2 > max_encoded_length) {
 402                         /* It won't fit with trailing "?=" --- break the line */
 403                         strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 404                         line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 405                 }
 406
 407                 for (i = 0; i < chrlen; i++)
 408                         strbuf_addf(sb, encoded_fmt, p[i]);
 409                 line_len += encoded_len;
 410         }
 411         strbuf_addstr(sb, "?=");
 412 }
 413
 414 const char *show_ident_date(const struct ident_split *ident,
 415                             const struct date_mode *mode)
 416 {
 417         timestamp_t date = 0;
 418         long tz = 0;
 419
 420         if (ident->date_begin && ident->date_end)
 421                 date = parse_timestamp(ident->date_begin, NULL, 10);
 422         if (date_overflows(date))
 423                 date = 0;
 424         else {
 425                 if (ident->tz_begin && ident->tz_end)
 426                         tz = strtol(ident->tz_begin, NULL, 10);
 427                 if (tz >= INT_MAX || tz <= INT_MIN)
 428                         tz = 0;
 429         }
 430         return show_date(date, tz, mode);
 431 }
 432
 433 void pp_user_info(struct pretty_print_context *pp,
 434                   const char *what, struct strbuf *sb,
 435                   const char *line, const char *encoding)
 436 {
 437         struct ident_split ident;
 438         char *line_end;
 439         const char *mailbuf, *namebuf;
 440         size_t namelen, maillen;
 441         int max_length = 78; /* per rfc2822 */
 442
 443         if (pp->fmt == CMIT_FMT_ONELINE)
 444                 return;
 445
 446         line_end = strchrnul(line, '\n');
 447         if (split_ident_line(&ident, line, line_end - line))
 448                 return;
 449
 450         mailbuf = ident.mail_begin;
 451         maillen = ident.mail_end - ident.mail_begin;
 452         namebuf = ident.name_begin;
 453         namelen = ident.name_end - ident.name_begin;
 454
 455         if (pp->mailmap)
 456                 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 457
 458         if (cmit_fmt_is_mail(pp->fmt)) {
 459                 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 460                         struct strbuf buf = STRBUF_INIT;
 461
 462                         strbuf_addstr(&buf, "From: ");
 463                         strbuf_add(&buf, namebuf, namelen);
 464                         strbuf_addstr(&buf, " <");
 465                         strbuf_add(&buf, mailbuf, maillen);
 466                         strbuf_addstr(&buf, ">\n");
 467                         string_list_append(&pp->in_body_headers,
 468                                            strbuf_detach(&buf, NULL));
 469
 470                         mailbuf = pp->from_ident->mail_begin;
 471                         maillen = pp->from_ident->mail_end - mailbuf;
 472                         namebuf = pp->from_ident->name_begin;
 473                         namelen = pp->from_ident->name_end - namebuf;
 474                 }
 475
 476                 strbuf_addstr(sb, "From: ");
 477                 if (pp->encode_email_headers &&
 478                     needs_rfc2047_encoding(namebuf, namelen)) {
 479                         add_rfc2047(sb, namebuf, namelen,
 480                                     encoding, RFC2047_ADDRESS);
 481                         max_length = 76; /* per rfc2047 */
 482                 } else if (needs_rfc822_quoting(namebuf, namelen)) {
 483                         struct strbuf quoted = STRBUF_INIT;
 484                         add_rfc822_quoted(&quoted, namebuf, namelen);
 485                         strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 486                                                         -6, 1, max_length);
 487                         strbuf_release(&quoted);
 488                 } else {
 489                         strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 490                                                  -6, 1, max_length);
 491                 }
 492
 493                 if (max_length <
 494                     last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 495                         strbuf_addch(sb, '\n');
 496                 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 497         } else {
 498                 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 499                             (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 500                             (int)namelen, namebuf, (int)maillen, mailbuf);
 501         }
 502
 503         switch (pp->fmt) {
 504         case CMIT_FMT_MEDIUM:
 505                 strbuf_addf(sb, "Date:   %s\n",
 506                             show_ident_date(&ident, &pp->date_mode));
 507                 break;
 508         case CMIT_FMT_EMAIL:
 509         case CMIT_FMT_MBOXRD:
 510                 strbuf_addf(sb, "Date: %s\n",
 511                             show_ident_date(&ident, DATE_MODE(RFC2822)));
 512                 break;
 513         case CMIT_FMT_FULLER:
 514                 strbuf_addf(sb, "%sDate: %s\n", what,
 515                             show_ident_date(&ident, &pp->date_mode));
 516                 break;
 517         default:
 518                 /* notin' */
 519                 break;
 520         }
 521 }
 522
 523 static int is_blank_line(const char *line, int *len_p)
 524 {
 525         int len = *len_p;
 526         while (len && isspace(line[len - 1]))
 527                 len--;
 528         *len_p = len;
 529         return !len;
 530 }
 531
 532 const char *skip_blank_lines(const char *msg)
 533 {
 534         for (;;) {
 535                 int linelen = get_one_line(msg);
 536                 int ll = linelen;
 537                 if (!linelen)
 538                         break;
 539                 if (!is_blank_line(msg, &ll))
 540                         break;
 541                 msg += linelen;
 542         }
 543         return msg;
 544 }
 545
 546 static void add_merge_info(const struct pretty_print_context *pp,
 547                            struct strbuf *sb, const struct commit *commit)
 548 {
 549         struct commit_list *parent = commit->parents;
 550
 551         if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
 552             !parent || !parent->next)
 553                 return;
 554
 555         strbuf_addstr(sb, "Merge:");
 556
 557         while (parent) {
 558                 struct object_id *oidp = &parent->item->object.oid;
 559                 strbuf_addch(sb, ' ');
 560                 if (pp->abbrev)
 561                         strbuf_add_unique_abbrev(sb, oidp, pp->abbrev);
 562                 else
 563                         strbuf_addstr(sb, oid_to_hex(oidp));
 564                 parent = parent->next;
 565         }
 566         strbuf_addch(sb, '\n');
 567 }
 568
 569 static char *get_header(const char *msg, const char *key)
 570 {
 571         size_t len;
 572         const char *v = find_commit_header(msg, key, &len);
 573         return v ? xmemdupz(v, len) : NULL;
 574 }
 575
 576 static char *replace_encoding_header(char *buf, const char *encoding)
 577 {
 578         struct strbuf tmp = STRBUF_INIT;
 579         size_t start, len;
 580         char *cp = buf;
 581
 582         /* guess if there is an encoding header before a \n\n */
 583         while (!starts_with(cp, "encoding ")) {
 584                 cp = strchr(cp, '\n');
 585                 if (!cp || *++cp == '\n')
 586                         return buf;
 587         }
 588         start = cp - buf;
 589         cp = strchr(cp, '\n');
 590         if (!cp)
 591                 return buf; /* should not happen but be defensive */
 592         len = cp + 1 - (buf + start);
 593
 594         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 595         if (is_encoding_utf8(encoding)) {
 596                 /* we have re-coded to UTF-8; drop the header */
 597                 strbuf_remove(&tmp, start, len);
 598         } else {
 599                 /* just replaces XXXX in 'encoding XXXX\n' */
 600                 strbuf_splice(&tmp, start + strlen("encoding "),
 601                                           len - strlen("encoding \n"),
 602                                           encoding, strlen(encoding));
 603         }
 604         return strbuf_detach(&tmp, NULL);
 605 }
 606
 607 const char *repo_logmsg_reencode(struct repository *r,
 608                                  const struct commit *commit,
 609                                  char **commit_encoding,
 610                                  const char *output_encoding)
 611 {
 612         static const char *utf8 = "UTF-8";
 613         const char *use_encoding;
 614         char *encoding;
 615         const char *msg = repo_get_commit_buffer(r, commit, NULL);
 616         char *out;
 617
 618         if (!output_encoding || !*output_encoding) {
 619                 if (commit_encoding)
 620                         *commit_encoding = get_header(msg, "encoding");
 621                 return msg;
 622         }
 623         encoding = get_header(msg, "encoding");
 624         if (commit_encoding)
 625                 *commit_encoding = encoding;
 626         use_encoding = encoding ? encoding : utf8;
 627         if (same_encoding(use_encoding, output_encoding)) {
 628                 /*
 629                  * No encoding work to be done. If we have no encoding header
 630                  * at all, then there's nothing to do, and we can return the
 631                  * message verbatim (whether newly allocated or not).
 632                  */
 633                 if (!encoding)
 634                         return msg;
 635
 636                 /*
 637                  * Otherwise, we still want to munge the encoding header in the
 638                  * result, which will be done by modifying the buffer. If we
 639                  * are using a fresh copy, we can reuse it. But if we are using
 640                  * the cached copy from get_commit_buffer, we need to duplicate it
 641                  * to avoid munging the cached copy.
 642                  */
 643                 if (msg == get_cached_commit_buffer(r, commit, NULL))
 644                         out = xstrdup(msg);
 645                 else
 646                         out = (char *)msg;
 647         }
 648         else {
 649                 /*
 650                  * There's actual encoding work to do. Do the reencoding, which
 651                  * still leaves the header to be replaced in the next step. At
 652                  * this point, we are done with msg. If we allocated a fresh
 653                  * copy, we can free it.
 654                  */
 655                 out = reencode_string(msg, output_encoding, use_encoding);
 656                 if (out)
 657                         repo_unuse_commit_buffer(r, commit, msg);
 658         }
 659
 660         /*
 661          * This replacement actually consumes the buffer we hand it, so we do
 662          * not have to worry about freeing the old "out" here.
 663          */
 664         if (out)
 665                 out = replace_encoding_header(out, output_encoding);
 666
 667         if (!commit_encoding)
 668                 free(encoding);
 669         /*
 670          * If the re-encoding failed, out might be NULL here; in that
 671          * case we just return the commit message verbatim.
 672          */
 673         return out ? out : msg;
 674 }
 675
 676 static int mailmap_name(const char **email, size_t *email_len,
 677                         const char **name, size_t *name_len)
 678 {
 679         static struct string_list *mail_map;
 680         if (!mail_map) {
 681                 mail_map = xcalloc(1, sizeof(*mail_map));
 682                 read_mailmap(mail_map);
 683         }
 684         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 685 }
 686
 687 static size_t format_person_part(struct strbuf *sb, char part,
 688                                  const char *msg, int len,
 689                                  const struct date_mode *dmode)
 690 {
 691         /* currently all placeholders have same length */
 692         const int placeholder_len = 2;
 693         struct ident_split s;
 694         const char *name, *mail;
 695         size_t maillen, namelen;
 696
 697         if (split_ident_line(&s, msg, len) < 0)
 698                 goto skip;
 699
 700         name = s.name_begin;
 701         namelen = s.name_end - s.name_begin;
 702         mail = s.mail_begin;
 703         maillen = s.mail_end - s.mail_begin;
 704
 705         if (part == 'N' || part == 'E' || part == 'L') /* mailmap lookup */
 706                 mailmap_name(&mail, &maillen, &name, &namelen);
 707         if (part == 'n' || part == 'N') {       /* name */
 708                 strbuf_add(sb, name, namelen);
 709                 return placeholder_len;
 710         }
 711         if (part == 'e' || part == 'E') {       /* email */
 712                 strbuf_add(sb, mail, maillen);
 713                 return placeholder_len;
 714         }
 715         if (part == 'l' || part == 'L') {       /* local-part */
 716                 const char *at = memchr(mail, '@', maillen);
 717                 if (at)
 718                         maillen = at - mail;
 719                 strbuf_add(sb, mail, maillen);
 720                 return placeholder_len;
 721         }
 722
 723         if (!s.date_begin)
 724                 goto skip;
 725
 726         if (part == 't') {      /* date, UNIX timestamp */
 727                 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 728                 return placeholder_len;
 729         }
 730
 731         switch (part) {
 732         case 'd':       /* date */
 733                 strbuf_addstr(sb, show_ident_date(&s, dmode));
 734                 return placeholder_len;
 735         case 'D':       /* date, RFC2822 style */
 736                 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
 737                 return placeholder_len;
 738         case 'r':       /* date, relative */
 739                 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
 740                 return placeholder_len;
 741         case 'i':       /* date, ISO 8601-like */
 742                 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
 743                 return placeholder_len;
 744         case 'I':       /* date, ISO 8601 strict */
 745                 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
 746                 return placeholder_len;
 747         case 's':
 748                 strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(SHORT)));
 749                 return placeholder_len;
 750         }
 751
 752 skip:
 753         /*
 754          * reading from either a bogus commit, or a reflog entry with
 755          * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 756          * to compute a valid return value.
 757          */
 758         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 759             || part == 'D' || part == 'r' || part == 'i')
 760                 return placeholder_len;
 761
 762         return 0; /* unknown placeholder */
 763 }
 764
 765 struct chunk {
 766         size_t off;
 767         size_t len;
 768 };
 769
 770 enum flush_type {
 771         no_flush,
 772         flush_right,
 773         flush_left,
 774         flush_left_and_steal,
 775         flush_both
 776 };
 777
 778 enum trunc_type {
 779         trunc_none,
 780         trunc_left,
 781         trunc_middle,
 782         trunc_right
 783 };
 784
 785 struct format_commit_context {
 786         const struct commit *commit;
 787         const struct pretty_print_context *pretty_ctx;
 788         unsigned commit_header_parsed:1;
 789         unsigned commit_message_parsed:1;
 790         struct signature_check signature_check;
 791         enum flush_type flush_type;
 792         enum trunc_type truncate;
 793         const char *message;
 794         char *commit_encoding;
 795         size_t width, indent1, indent2;
 796         int auto_color;
 797         int padding;
 798
 799         /* These offsets are relative to the start of the commit message. */
 800         struct chunk author;
 801         struct chunk committer;
 802         size_t message_off;
 803         size_t subject_off;
 804         size_t body_off;
 805
 806         /* The following ones are relative to the result struct strbuf. */
 807         size_t wrap_start;
 808 };
 809
 810 static void parse_commit_header(struct format_commit_context *context)
 811 {
 812         const char *msg = context->message;
 813         int i;
 814
 815         for (i = 0; msg[i]; i++) {
 816                 const char *name;
 817                 int eol;
 818                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 819                         ; /* do nothing */
 820
 821                 if (i == eol) {
 822                         break;
 823                 } else if (skip_prefix(msg + i, "author ", &name)) {
 824                         context->author.off = name - msg;
 825                         context->author.len = msg + eol - name;
 826                 } else if (skip_prefix(msg + i, "committer ", &name)) {
 827                         context->committer.off = name - msg;
 828                         context->committer.len = msg + eol - name;
 829                 }
 830                 i = eol;
 831         }
 832         context->message_off = i;
 833         context->commit_header_parsed = 1;
 834 }
 835
 836 static int istitlechar(char c)
 837 {
 838         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 839                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 840 }
 841
 842 void format_sanitized_subject(struct strbuf *sb, const char *msg, size_t len)
 843 {
 844         size_t trimlen;
 845         size_t start_len = sb->len;
 846         int space = 2;
 847         int i;
 848
 849         for (i = 0; i < len; i++) {
 850                 if (istitlechar(msg[i])) {
 851                         if (space == 1)
 852                                 strbuf_addch(sb, '-');
 853                         space = 0;
 854                         strbuf_addch(sb, msg[i]);
 855                         if (msg[i] == '.')
 856                                 while (msg[i+1] == '.')
 857                                         i++;
 858                 } else
 859                         space |= 1;
 860         }
 861
 862         /* trim any trailing '.' or '-' characters */
 863         trimlen = 0;
 864         while (sb->len - trimlen > start_len &&
 865                 (sb->buf[sb->len - 1 - trimlen] == '.'
 866                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 867                 trimlen++;
 868         strbuf_remove(sb, sb->len - trimlen, trimlen);
 869 }
 870
 871 const char *format_subject(struct strbuf *sb, const char *msg,
 872                            const char *line_separator)
 873 {
 874         int first = 1;
 875
 876         for (;;) {
 877                 const char *line = msg;
 878                 int linelen = get_one_line(line);
 879
 880                 msg += linelen;
 881                 if (!linelen || is_blank_line(line, &linelen))
 882                         break;
 883
 884                 if (!sb)
 885                         continue;
 886                 strbuf_grow(sb, linelen + 2);
 887                 if (!first)
 888                         strbuf_addstr(sb, line_separator);
 889                 strbuf_add(sb, line, linelen);
 890                 first = 0;
 891         }
 892         return msg;
 893 }
 894
 895 static void parse_commit_message(struct format_commit_context *c)
 896 {
 897         const char *msg = c->message + c->message_off;
 898         const char *start = c->message;
 899
 900         msg = skip_blank_lines(msg);
 901         c->subject_off = msg - start;
 902
 903         msg = format_subject(NULL, msg, NULL);
 904         msg = skip_blank_lines(msg);
 905         c->body_off = msg - start;
 906
 907         c->commit_message_parsed = 1;
 908 }
 909
 910 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 911                         size_t width, size_t indent1, size_t indent2)
 912 {
 913         struct strbuf tmp = STRBUF_INIT;
 914
 915         if (pos)
 916                 strbuf_add(&tmp, sb->buf, pos);
 917         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 918                                 (int) indent1, (int) indent2, (int) width);
 919         strbuf_swap(&tmp, sb);
 920         strbuf_release(&tmp);
 921 }
 922
 923 static void rewrap_message_tail(struct strbuf *sb,
 924                                 struct format_commit_context *c,
 925                                 size_t new_width, size_t new_indent1,
 926                                 size_t new_indent2)
 927 {
 928         if (c->width == new_width && c->indent1 == new_indent1 &&
 929             c->indent2 == new_indent2)
 930                 return;
 931         if (c->wrap_start < sb->len)
 932                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 933         c->wrap_start = sb->len;
 934         c->width = new_width;
 935         c->indent1 = new_indent1;
 936         c->indent2 = new_indent2;
 937 }
 938
 939 static int format_reflog_person(struct strbuf *sb,
 940                                 char part,
 941                                 struct reflog_walk_info *log,
 942                                 const struct date_mode *dmode)
 943 {
 944         const char *ident;
 945
 946         if (!log)
 947                 return 2;
 948
 949         ident = get_reflog_ident(log);
 950         if (!ident)
 951                 return 2;
 952
 953         return format_person_part(sb, part, ident, strlen(ident), dmode);
 954 }
 955
 956 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 957                           const char *placeholder,
 958                           struct format_commit_context *c)
 959 {
 960         const char *rest = placeholder;
 961         const char *basic_color = NULL;
 962
 963         if (placeholder[1] == '(') {
 964                 const char *begin = placeholder + 2;
 965                 const char *end = strchr(begin, ')');
 966                 char color[COLOR_MAXLEN];
 967
 968                 if (!end)
 969                         return 0;
 970
 971                 if (skip_prefix(begin, "auto,", &begin)) {
 972                         if (!want_color(c->pretty_ctx->color))
 973                                 return end - placeholder + 1;
 974                 } else if (skip_prefix(begin, "always,", &begin)) {
 975                         /* nothing to do; we do not respect want_color at all */
 976                 } else {
 977                         /* the default is the same as "auto" */
 978                         if (!want_color(c->pretty_ctx->color))
 979                                 return end - placeholder + 1;
 980                 }
 981
 982                 if (color_parse_mem(begin, end - begin, color) < 0)
 983                         die(_("unable to parse --pretty format"));
 984                 strbuf_addstr(sb, color);
 985                 return end - placeholder + 1;
 986         }
 987
 988         /*
 989          * We handle things like "%C(red)" above; for historical reasons, there
 990          * are a few colors that can be specified without parentheses (and
 991          * they cannot support things like "auto" or "always" at all).
 992          */
 993         if (skip_prefix(placeholder + 1, "red", &rest))
 994                 basic_color = GIT_COLOR_RED;
 995         else if (skip_prefix(placeholder + 1, "green", &rest))
 996                 basic_color = GIT_COLOR_GREEN;
 997         else if (skip_prefix(placeholder + 1, "blue", &rest))
 998                 basic_color = GIT_COLOR_BLUE;
 999         else if (skip_prefix(placeholder + 1, "reset", &rest))
1000                 basic_color = GIT_COLOR_RESET;
1001
1002         if (basic_color && want_color(c->pretty_ctx->color))
1003                 strbuf_addstr(sb, basic_color);
1004
1005         return rest - placeholder;
1006 }
1007
1008 static size_t parse_padding_placeholder(const char *placeholder,
1009                                         struct format_commit_context *c)
1010 {
1011         const char *ch = placeholder;
1012         enum flush_type flush_type;
1013         int to_column = 0;
1014
1015         switch (*ch++) {
1016         case '<':
1017                 flush_type = flush_right;
1018                 break;
1019         case '>':
1020                 if (*ch == '<') {
1021                         flush_type = flush_both;
1022                         ch++;
1023                 } else if (*ch == '>') {
1024                         flush_type = flush_left_and_steal;
1025                         ch++;
1026                 } else
1027                         flush_type = flush_left;
1028                 break;
1029         default:
1030                 return 0;
1031         }
1032
1033         /* the next value means "wide enough to that column" */
1034         if (*ch == '|') {
1035                 to_column = 1;
1036                 ch++;
1037         }
1038
1039         if (*ch == '(') {
1040                 const char *start = ch + 1;
1041                 const char *end = start + strcspn(start, ",)");
1042                 char *next;
1043                 int width;
1044                 if (!end || end == start)
1045                         return 0;
1046                 width = strtol(start, &next, 10);
1047                 if (next == start || width == 0)
1048                         return 0;
1049                 if (width < 0) {
1050                         if (to_column)
1051                                 width += term_columns();
1052                         if (width < 0)
1053                                 return 0;
1054                 }
1055                 c->padding = to_column ? -width : width;
1056                 c->flush_type = flush_type;
1057
1058                 if (*end == ',') {
1059                         start = end + 1;
1060                         end = strchr(start, ')');
1061                         if (!end || end == start)
1062                                 return 0;
1063                         if (starts_with(start, "trunc)"))
1064                                 c->truncate = trunc_right;
1065                         else if (starts_with(start, "ltrunc)"))
1066                                 c->truncate = trunc_left;
1067                         else if (starts_with(start, "mtrunc)"))
1068                                 c->truncate = trunc_middle;
1069                         else
1070                                 return 0;
1071                 } else
1072                         c->truncate = trunc_none;
1073
1074                 return end - placeholder + 1;
1075         }
1076         return 0;
1077 }
1078
1079 static int match_placeholder_arg_value(const char *to_parse, const char *candidate,
1080                                        const char **end, const char **valuestart,
1081                                        size_t *valuelen)
1082 {
1083         const char *p;
1084
1085         if (!(skip_prefix(to_parse, candidate, &p)))
1086                 return 0;
1087         if (valuestart) {
1088                 if (*p == '=') {
1089                         *valuestart = p + 1;
1090                         *valuelen = strcspn(*valuestart, ",)");
1091                         p = *valuestart + *valuelen;
1092                 } else {
1093                         if (*p != ',' && *p != ')')
1094                                 return 0;
1095                         *valuestart = NULL;
1096                         *valuelen = 0;
1097                 }
1098         }
1099         if (*p == ',') {
1100                 *end = p + 1;
1101                 return 1;
1102         }
1103         if (*p == ')') {
1104                 *end = p;
1105                 return 1;
1106         }
1107         return 0;
1108 }
1109
1110 static int match_placeholder_bool_arg(const char *to_parse, const char *candidate,
1111                                       const char **end, int *val)
1112 {
1113         const char *argval;
1114         char *strval;
1115         size_t arglen;
1116         int v;
1117
1118         if (!match_placeholder_arg_value(to_parse, candidate, end, &argval, &arglen))
1119                 return 0;
1120
1121         if (!argval) {
1122                 *val = 1;
1123                 return 1;
1124         }
1125
1126         strval = xstrndup(argval, arglen);
1127         v = git_parse_maybe_bool(strval);
1128         free(strval);
1129
1130         if (v == -1)
1131                 return 0;
1132
1133         *val = v;
1134
1135         return 1;
1136 }
1137
1138 static int format_trailer_match_cb(const struct strbuf *key, void *ud)
1139 {
1140         const struct string_list *list = ud;
1141         const struct string_list_item *item;
1142
1143         for_each_string_list_item (item, list) {
1144                 if (key->len == (uintptr_t)item->util &&
1145                     !strncasecmp(item->string, key->buf, key->len))
1146                         return 1;
1147         }
1148         return 0;
1149 }
1150
1151 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1152                                 const char *placeholder,
1153                                 void *context)
1154 {
1155         struct format_commit_context *c = context;
1156         const struct commit *commit = c->commit;
1157         const char *msg = c->message;
1158         struct commit_list *p;
1159         const char *arg, *eol;
1160         size_t res;
1161         char **slot;
1162
1163         /* these are independent of the commit */
1164         res = strbuf_expand_literal_cb(sb, placeholder, NULL);
1165         if (res)
1166                 return res;
1167
1168         switch (placeholder[0]) {
1169         case 'C':
1170                 if (starts_with(placeholder + 1, "(auto)")) {
1171                         c->auto_color = want_color(c->pretty_ctx->color);
1172                         if (c->auto_color && sb->len)
1173                                 strbuf_addstr(sb, GIT_COLOR_RESET);
1174                         return 7; /* consumed 7 bytes, "C(auto)" */
1175                 } else {
1176                         int ret = parse_color(sb, placeholder, c);
1177                         if (ret)
1178                                 c->auto_color = 0;
1179                         /*
1180                          * Otherwise, we decided to treat %C<unknown>
1181                          * as a literal string, and the previous
1182                          * %C(auto) is still valid.
1183                          */
1184                         return ret;
1185                 }
1186         case 'w':
1187                 if (placeholder[1] == '(') {
1188                         unsigned long width = 0, indent1 = 0, indent2 = 0;
1189                         char *next;
1190                         const char *start = placeholder + 2;
1191                         const char *end = strchr(start, ')');
1192                         if (!end)
1193                                 return 0;
1194                         if (end > start) {
1195                                 width = strtoul(start, &next, 10);
1196                                 if (*next == ',') {
1197                                         indent1 = strtoul(next + 1, &next, 10);
1198                                         if (*next == ',') {
1199                                                 indent2 = strtoul(next + 1,
1200                                                                  &next, 10);
1201                                         }
1202                                 }
1203                                 if (*next != ')')
1204                                         return 0;
1205                         }
1206                         rewrap_message_tail(sb, c, width, indent1, indent2);
1207                         return end - placeholder + 1;
1208                 } else
1209                         return 0;
1210
1211         case '<':
1212         case '>':
1213                 return parse_padding_placeholder(placeholder, c);
1214         }
1215
1216         /* these depend on the commit */
1217         if (!commit->object.parsed)
1218                 parse_object(the_repository, &commit->object.oid);
1219
1220         switch (placeholder[0]) {
1221         case 'H':               /* commit hash */
1222                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1223                 strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1224                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1225                 return 1;
1226         case 'h':               /* abbreviated commit hash */
1227                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1228                 strbuf_add_unique_abbrev(sb, &commit->object.oid,
1229                                          c->pretty_ctx->abbrev);
1230                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1231                 return 1;
1232         case 'T':               /* tree hash */
1233                 strbuf_addstr(sb, oid_to_hex(get_commit_tree_oid(commit)));
1234                 return 1;
1235         case 't':               /* abbreviated tree hash */
1236                 strbuf_add_unique_abbrev(sb,
1237                                          get_commit_tree_oid(commit),
1238                                          c->pretty_ctx->abbrev);
1239                 return 1;
1240         case 'P':               /* parent hashes */
1241                 for (p = commit->parents; p; p = p->next) {
1242                         if (p != commit->parents)
1243                                 strbuf_addch(sb, ' ');
1244                         strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1245                 }
1246                 return 1;
1247         case 'p':               /* abbreviated parent hashes */
1248                 for (p = commit->parents; p; p = p->next) {
1249                         if (p != commit->parents)
1250                                 strbuf_addch(sb, ' ');
1251                         strbuf_add_unique_abbrev(sb, &p->item->object.oid,
1252                                                  c->pretty_ctx->abbrev);
1253                 }
1254                 return 1;
1255         case 'm':               /* left/right/bottom */
1256                 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1257                 return 1;
1258         case 'd':
1259                 format_decorations(sb, commit, c->auto_color);
1260                 return 1;
1261         case 'D':
1262                 format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1263                 return 1;
1264         case 'S':               /* tag/branch like --source */
1265                 if (!(c->pretty_ctx->rev && c->pretty_ctx->rev->sources))
1266                         return 0;
1267                 slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
1268                 if (!(slot && *slot))
1269                         return 0;
1270                 strbuf_addstr(sb, *slot);
1271                 return 1;
1272         case 'g':               /* reflog info */
1273                 switch(placeholder[1]) {
1274                 case 'd':       /* reflog selector */
1275                 case 'D':
1276                         if (c->pretty_ctx->reflog_info)
1277                                 get_reflog_selector(sb,
1278                                                     c->pretty_ctx->reflog_info,
1279                                                     &c->pretty_ctx->date_mode,
1280                                                     c->pretty_ctx->date_mode_explicit,
1281                                                     (placeholder[1] == 'd'));
1282                         return 2;
1283                 case 's':       /* reflog message */
1284                         if (c->pretty_ctx->reflog_info)
1285                                 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1286                         return 2;
1287                 case 'n':
1288                 case 'N':
1289                 case 'e':
1290                 case 'E':
1291                         return format_reflog_person(sb,
1292                                                     placeholder[1],
1293                                                     c->pretty_ctx->reflog_info,
1294                                                     &c->pretty_ctx->date_mode);
1295                 }
1296                 return 0;       /* unknown %g placeholder */
1297         case 'N':
1298                 if (c->pretty_ctx->notes_message) {
1299                         strbuf_addstr(sb, c->pretty_ctx->notes_message);
1300                         return 1;
1301                 }
1302                 return 0;
1303         }
1304
1305         if (placeholder[0] == 'G') {
1306                 if (!c->signature_check.result)
1307                         check_commit_signature(c->commit, &(c->signature_check));
1308                 switch (placeholder[1]) {
1309                 case 'G':
1310                         if (c->signature_check.gpg_output)
1311                                 strbuf_addstr(sb, c->signature_check.gpg_output);
1312                         break;
1313                 case '?':
1314                         switch (c->signature_check.result) {
1315                         case 'G':
1316                                 switch (c->signature_check.trust_level) {
1317                                 case TRUST_UNDEFINED:
1318                                 case TRUST_NEVER:
1319                                         strbuf_addch(sb, 'U');
1320                                         break;
1321                                 default:
1322                                         strbuf_addch(sb, 'G');
1323                                         break;
1324                                 }
1325                                 break;
1326                         case 'B':
1327                         case 'E':
1328                         case 'N':
1329                         case 'X':
1330                         case 'Y':
1331                         case 'R':
1332                                 strbuf_addch(sb, c->signature_check.result);
1333                         }
1334                         break;
1335                 case 'S':
1336                         if (c->signature_check.signer)
1337                                 strbuf_addstr(sb, c->signature_check.signer);
1338                         break;
1339                 case 'K':
1340                         if (c->signature_check.key)
1341                                 strbuf_addstr(sb, c->signature_check.key);
1342                         break;
1343                 case 'F':
1344                         if (c->signature_check.fingerprint)
1345                                 strbuf_addstr(sb, c->signature_check.fingerprint);
1346                         break;
1347                 case 'P':
1348                         if (c->signature_check.primary_key_fingerprint)
1349                                 strbuf_addstr(sb, c->signature_check.primary_key_fingerprint);
1350                         break;
1351                 case 'T':
1352                         switch (c->signature_check.trust_level) {
1353                         case TRUST_UNDEFINED:
1354                                 strbuf_addstr(sb, "undefined");
1355                                 break;
1356                         case TRUST_NEVER:
1357                                 strbuf_addstr(sb, "never");
1358                                 break;
1359                         case TRUST_MARGINAL:
1360                                 strbuf_addstr(sb, "marginal");
1361                                 break;
1362                         case TRUST_FULLY:
1363                                 strbuf_addstr(sb, "fully");
1364                                 break;
1365                         case TRUST_ULTIMATE:
1366                                 strbuf_addstr(sb, "ultimate");
1367                                 break;
1368                         }
1369                         break;
1370                 default:
1371                         return 0;
1372                 }
1373                 return 2;
1374         }
1375
1376
1377         /* For the rest we have to parse the commit header. */
1378         if (!c->commit_header_parsed)
1379                 parse_commit_header(c);
1380
1381         switch (placeholder[0]) {
1382         case 'a':       /* author ... */
1383                 return format_person_part(sb, placeholder[1],
1384                                    msg + c->author.off, c->author.len,
1385                                    &c->pretty_ctx->date_mode);
1386         case 'c':       /* committer ... */
1387                 return format_person_part(sb, placeholder[1],
1388                                    msg + c->committer.off, c->committer.len,
1389                                    &c->pretty_ctx->date_mode);
1390         case 'e':       /* encoding */
1391                 if (c->commit_encoding)
1392                         strbuf_addstr(sb, c->commit_encoding);
1393                 return 1;
1394         case 'B':       /* raw body */
1395                 /* message_off is always left at the initial newline */
1396                 strbuf_addstr(sb, msg + c->message_off + 1);
1397                 return 1;
1398         }
1399
1400         /* Now we need to parse the commit message. */
1401         if (!c->commit_message_parsed)
1402                 parse_commit_message(c);
1403
1404         switch (placeholder[0]) {
1405         case 's':       /* subject */
1406                 format_subject(sb, msg + c->subject_off, " ");
1407                 return 1;
1408         case 'f':       /* sanitized subject */
1409                 eol = strchrnul(msg + c->subject_off, '\n');
1410                 format_sanitized_subject(sb, msg + c->subject_off, eol - (msg + c->subject_off));
1411                 return 1;
1412         case 'b':       /* body */
1413                 strbuf_addstr(sb, msg + c->body_off);
1414                 return 1;
1415         }
1416
1417         if (skip_prefix(placeholder, "(trailers", &arg)) {
1418                 struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1419                 struct string_list filter_list = STRING_LIST_INIT_NODUP;
1420                 struct strbuf sepbuf = STRBUF_INIT;
1421                 struct strbuf kvsepbuf = STRBUF_INIT;
1422                 size_t ret = 0;
1423
1424                 opts.no_divider = 1;
1425
1426                 if (*arg == ':') {
1427                         arg++;
1428                         for (;;) {
1429                                 const char *argval;
1430                                 size_t arglen;
1431
1432                                 if (match_placeholder_arg_value(arg, "key", &arg, &argval, &arglen)) {
1433                                         uintptr_t len = arglen;
1434
1435                                         if (!argval)
1436                                                 goto trailer_out;
1437
1438                                         if (len && argval[len - 1] == ':')
1439                                                 len--;
1440                                         string_list_append(&filter_list, argval)->util = (char *)len;
1441
1442                                         opts.filter = format_trailer_match_cb;
1443                                         opts.filter_data = &filter_list;
1444                                         opts.only_trailers = 1;
1445                                 } else if (match_placeholder_arg_value(arg, "separator", &arg, &argval, &arglen)) {
1446                                         char *fmt;
1447
1448                                         strbuf_reset(&sepbuf);
1449                                         fmt = xstrndup(argval, arglen);
1450                                         strbuf_expand(&sepbuf, fmt, strbuf_expand_literal_cb, NULL);
1451                                         free(fmt);
1452                                         opts.separator = &sepbuf;
1453                                 } else if (match_placeholder_arg_value(arg, "key_value_separator", &arg, &argval, &arglen)) {
1454                                         char *fmt;
1455
1456                                         strbuf_reset(&kvsepbuf);
1457                                         fmt = xstrndup(argval, arglen);
1458                                         strbuf_expand(&kvsepbuf, fmt, strbuf_expand_literal_cb, NULL);
1459                                         free(fmt);
1460                                         opts.key_value_separator = &kvsepbuf;
1461                                 } else if (!match_placeholder_bool_arg(arg, "only", &arg, &opts.only_trailers) &&
1462                                            !match_placeholder_bool_arg(arg, "unfold", &arg, &opts.unfold) &&
1463                                            !match_placeholder_bool_arg(arg, "keyonly", &arg, &opts.key_only) &&
1464                                            !match_placeholder_bool_arg(arg, "valueonly", &arg, &opts.value_only))
1465                                         break;
1466                         }
1467                 }
1468                 if (*arg == ')') {
1469                         format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1470                         ret = arg - placeholder + 1;
1471                 }
1472         trailer_out:
1473                 string_list_clear(&filter_list, 0);
1474                 strbuf_release(&sepbuf);
1475                 return ret;
1476         }
1477
1478         return 0;       /* unknown placeholder */
1479 }
1480
1481 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1482                                     const char *placeholder,
1483                                     struct format_commit_context *c)
1484 {
1485         struct strbuf local_sb = STRBUF_INIT;
1486         int total_consumed = 0, len, padding = c->padding;
1487         if (padding < 0) {
1488                 const char *start = strrchr(sb->buf, '\n');
1489                 int occupied;
1490                 if (!start)
1491                         start = sb->buf;
1492                 occupied = utf8_strnwidth(start, -1, 1);
1493                 occupied += c->pretty_ctx->graph_width;
1494                 padding = (-padding) - occupied;
1495         }
1496         while (1) {
1497                 int modifier = *placeholder == 'C';
1498                 int consumed = format_commit_one(&local_sb, placeholder, c);
1499                 total_consumed += consumed;
1500
1501                 if (!modifier)
1502                         break;
1503
1504                 placeholder += consumed;
1505                 if (*placeholder != '%')
1506                         break;
1507                 placeholder++;
1508                 total_consumed++;
1509         }
1510         len = utf8_strnwidth(local_sb.buf, -1, 1);
1511
1512         if (c->flush_type == flush_left_and_steal) {
1513                 const char *ch = sb->buf + sb->len - 1;
1514                 while (len > padding && ch > sb->buf) {
1515                         const char *p;
1516                         if (*ch == ' ') {
1517                                 ch--;
1518                                 padding++;
1519                                 continue;
1520                         }
1521                         /* check for trailing ansi sequences */
1522                         if (*ch != 'm')
1523                                 break;
1524                         p = ch - 1;
1525                         while (ch - p < 10 && *p != '\033')
1526                                 p--;
1527                         if (*p != '\033' ||
1528                             ch + 1 - p != display_mode_esc_sequence_len(p))
1529                                 break;
1530                         /*
1531                          * got a good ansi sequence, put it back to
1532                          * local_sb as we're cutting sb
1533                          */
1534                         strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1535                         ch = p - 1;
1536                 }
1537                 strbuf_setlen(sb, ch + 1 - sb->buf);
1538                 c->flush_type = flush_left;
1539         }
1540
1541         if (len > padding) {
1542                 switch (c->truncate) {
1543                 case trunc_left:
1544                         strbuf_utf8_replace(&local_sb,
1545                                             0, len - (padding - 2),
1546                                             "..");
1547                         break;
1548                 case trunc_middle:
1549                         strbuf_utf8_replace(&local_sb,
1550                                             padding / 2 - 1,
1551                                             len - (padding - 2),
1552                                             "..");
1553                         break;
1554                 case trunc_right:
1555                         strbuf_utf8_replace(&local_sb,
1556                                             padding - 2, len - (padding - 2),
1557                                             "..");
1558                         break;
1559                 case trunc_none:
1560                         break;
1561                 }
1562                 strbuf_addbuf(sb, &local_sb);
1563         } else {
1564                 int sb_len = sb->len, offset = 0;
1565                 if (c->flush_type == flush_left)
1566                         offset = padding - len;
1567                 else if (c->flush_type == flush_both)
1568                         offset = (padding - len) / 2;
1569                 /*
1570                  * we calculate padding in columns, now
1571                  * convert it back to chars
1572                  */
1573                 padding = padding - len + local_sb.len;
1574                 strbuf_addchars(sb, ' ', padding);
1575                 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1576                        local_sb.len);
1577         }
1578         strbuf_release(&local_sb);
1579         c->flush_type = no_flush;
1580         return total_consumed;
1581 }
1582
1583 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1584                                  const char *placeholder,
1585                                  void *context)
1586 {
1587         int consumed;
1588         size_t orig_len;
1589         enum {
1590                 NO_MAGIC,
1591                 ADD_LF_BEFORE_NON_EMPTY,
1592                 DEL_LF_BEFORE_EMPTY,
1593                 ADD_SP_BEFORE_NON_EMPTY
1594         } magic = NO_MAGIC;
1595
1596         switch (placeholder[0]) {
1597         case '-':
1598                 magic = DEL_LF_BEFORE_EMPTY;
1599                 break;
1600         case '+':
1601                 magic = ADD_LF_BEFORE_NON_EMPTY;
1602                 break;
1603         case ' ':
1604                 magic = ADD_SP_BEFORE_NON_EMPTY;
1605                 break;
1606         default:
1607                 break;
1608         }
1609         if (magic != NO_MAGIC)
1610                 placeholder++;
1611
1612         orig_len = sb->len;
1613         if (((struct format_commit_context *)context)->flush_type != no_flush)
1614                 consumed = format_and_pad_commit(sb, placeholder, context);
1615         else
1616                 consumed = format_commit_one(sb, placeholder, context);
1617         if (magic == NO_MAGIC)
1618                 return consumed;
1619
1620         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1621                 while (sb->len && sb->buf[sb->len - 1] == '\n')
1622                         strbuf_setlen(sb, sb->len - 1);
1623         } else if (orig_len != sb->len) {
1624                 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1625                         strbuf_insertstr(sb, orig_len, "\n");
1626                 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1627                         strbuf_insertstr(sb, orig_len, " ");
1628         }
1629         return consumed + 1;
1630 }
1631
1632 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1633                                    void *context)
1634 {
1635         struct userformat_want *w = context;
1636
1637         if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1638                 placeholder++;
1639
1640         switch (*placeholder) {
1641         case 'N':
1642                 w->notes = 1;
1643                 break;
1644         case 'S':
1645                 w->source = 1;
1646                 break;
1647         }
1648         return 0;
1649 }
1650
1651 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1652 {
1653         struct strbuf dummy = STRBUF_INIT;
1654
1655         if (!fmt) {
1656                 if (!user_format)
1657                         return;
1658                 fmt = user_format;
1659         }
1660         strbuf_expand(&dummy, fmt, userformat_want_item, w);
1661         strbuf_release(&dummy);
1662 }
1663
1664 void repo_format_commit_message(struct repository *r,
1665                                 const struct commit *commit,
1666                                 const char *format, struct strbuf *sb,
1667                                 const struct pretty_print_context *pretty_ctx)
1668 {
1669         struct format_commit_context context = {
1670                 .commit = commit,
1671                 .pretty_ctx = pretty_ctx,
1672                 .wrap_start = sb->len
1673         };
1674         const char *output_enc = pretty_ctx->output_encoding;
1675         const char *utf8 = "UTF-8";
1676
1677         /*
1678          * convert a commit message to UTF-8 first
1679          * as far as 'format_commit_item' assumes it in UTF-8
1680          */
1681         context.message = repo_logmsg_reencode(r, commit,
1682                                                &context.commit_encoding,
1683                                                utf8);
1684
1685         strbuf_expand(sb, format, format_commit_item, &context);
1686         rewrap_message_tail(sb, &context, 0, 0, 0);
1687
1688         /* then convert a commit message to an actual output encoding */
1689         if (output_enc) {
1690                 if (same_encoding(utf8, output_enc))
1691                         output_enc = NULL;
1692         } else {
1693                 if (context.commit_encoding &&
1694                     !same_encoding(context.commit_encoding, utf8))
1695                         output_enc = context.commit_encoding;
1696         }
1697
1698         if (output_enc) {
1699                 size_t outsz;
1700                 char *out = reencode_string_len(sb->buf, sb->len,
1701                                                 output_enc, utf8, &outsz);
1702                 if (out)
1703                         strbuf_attach(sb, out, outsz, outsz + 1);
1704         }
1705
1706         free(context.commit_encoding);
1707         repo_unuse_commit_buffer(r, commit, context.message);
1708 }
1709
1710 static void pp_header(struct pretty_print_context *pp,
1711                       const char *encoding,
1712                       const struct commit *commit,
1713                       const char **msg_p,
1714                       struct strbuf *sb)
1715 {
1716         int parents_shown = 0;
1717
1718         for (;;) {
1719                 const char *name, *line = *msg_p;
1720                 int linelen = get_one_line(*msg_p);
1721
1722                 if (!linelen)
1723                         return;
1724                 *msg_p += linelen;
1725
1726                 if (linelen == 1)
1727                         /* End of header */
1728                         return;
1729
1730                 if (pp->fmt == CMIT_FMT_RAW) {
1731                         strbuf_add(sb, line, linelen);
1732                         continue;
1733                 }
1734
1735                 if (starts_with(line, "parent ")) {
1736                         if (linelen != the_hash_algo->hexsz + 8)
1737                                 die("bad parent line in commit");
1738                         continue;
1739                 }
1740
1741                 if (!parents_shown) {
1742                         unsigned num = commit_list_count(commit->parents);
1743                         /* with enough slop */
1744                         strbuf_grow(sb, num * (GIT_MAX_HEXSZ + 10) + 20);
1745                         add_merge_info(pp, sb, commit);
1746                         parents_shown = 1;
1747                 }
1748
1749                 /*
1750                  * MEDIUM == DEFAULT shows only author with dates.
1751                  * FULL shows both authors but not dates.
1752                  * FULLER shows both authors and dates.
1753                  */
1754                 if (skip_prefix(line, "author ", &name)) {
1755                         strbuf_grow(sb, linelen + 80);
1756                         pp_user_info(pp, "Author", sb, name, encoding);
1757                 }
1758                 if (skip_prefix(line, "committer ", &name) &&
1759                     (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1760                         strbuf_grow(sb, linelen + 80);
1761                         pp_user_info(pp, "Commit", sb, name, encoding);
1762                 }
1763         }
1764 }
1765
1766 void pp_title_line(struct pretty_print_context *pp,
1767                    const char **msg_p,
1768                    struct strbuf *sb,
1769                    const char *encoding,
1770                    int need_8bit_cte)
1771 {
1772         static const int max_length = 78; /* per rfc2047 */
1773         struct strbuf title;
1774
1775         strbuf_init(&title, 80);
1776         *msg_p = format_subject(&title, *msg_p,
1777                                 pp->preserve_subject ? "\n" : " ");
1778
1779         strbuf_grow(sb, title.len + 1024);
1780         if (pp->print_email_subject) {
1781                 if (pp->rev)
1782                         fmt_output_email_subject(sb, pp->rev);
1783                 if (pp->encode_email_headers &&
1784                     needs_rfc2047_encoding(title.buf, title.len))
1785                         add_rfc2047(sb, title.buf, title.len,
1786                                                 encoding, RFC2047_SUBJECT);
1787                 else
1788                         strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1789                                          -last_line_length(sb), 1, max_length);
1790         } else {
1791                 strbuf_addbuf(sb, &title);
1792         }
1793         strbuf_addch(sb, '\n');
1794
1795         if (need_8bit_cte == 0) {
1796                 int i;
1797                 for (i = 0; i < pp->in_body_headers.nr; i++) {
1798                         if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1799                                 need_8bit_cte = 1;
1800                                 break;
1801                         }
1802                 }
1803         }
1804
1805         if (need_8bit_cte > 0) {
1806                 const char *header_fmt =
1807                         "MIME-Version: 1.0\n"
1808                         "Content-Type: text/plain; charset=%s\n"
1809                         "Content-Transfer-Encoding: 8bit\n";
1810                 strbuf_addf(sb, header_fmt, encoding);
1811         }
1812         if (pp->after_subject) {
1813                 strbuf_addstr(sb, pp->after_subject);
1814         }
1815         if (cmit_fmt_is_mail(pp->fmt)) {
1816                 strbuf_addch(sb, '\n');
1817         }
1818
1819         if (pp->in_body_headers.nr) {
1820                 int i;
1821                 for (i = 0; i < pp->in_body_headers.nr; i++) {
1822                         strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1823                         free(pp->in_body_headers.items[i].string);
1824                 }
1825                 string_list_clear(&pp->in_body_headers, 0);
1826                 strbuf_addch(sb, '\n');
1827         }
1828
1829         strbuf_release(&title);
1830 }
1831
1832 static int pp_utf8_width(const char *start, const char *end)
1833 {
1834         int width = 0;
1835         size_t remain = end - start;
1836
1837         while (remain) {
1838                 int n = utf8_width(&start, &remain);
1839                 if (n < 0 || !start)
1840                         return -1;
1841                 width += n;
1842         }
1843         return width;
1844 }
1845
1846 static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1847                                  const char *line, int linelen)
1848 {
1849         const char *tab;
1850
1851         while ((tab = memchr(line, '\t', linelen)) != NULL) {
1852                 int width = pp_utf8_width(line, tab);
1853
1854                 /*
1855                  * If it wasn't well-formed utf8, or it
1856                  * had characters with badly defined
1857                  * width (control characters etc), just
1858                  * give up on trying to align things.
1859                  */
1860                 if (width < 0)
1861                         break;
1862
1863                 /* Output the data .. */
1864                 strbuf_add(sb, line, tab - line);
1865
1866                 /* .. and the de-tabified tab */
1867                 strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1868
1869                 /* Skip over the printed part .. */
1870                 linelen -= tab + 1 - line;
1871                 line = tab + 1;
1872         }
1873
1874         /*
1875          * Print out everything after the last tab without
1876          * worrying about width - there's nothing more to
1877          * align.
1878          */
1879         strbuf_add(sb, line, linelen);
1880 }
1881
1882 /*
1883  * pp_handle_indent() prints out the intendation, and
1884  * the whole line (without the final newline), after
1885  * de-tabifying.
1886  */
1887 static void pp_handle_indent(struct pretty_print_context *pp,
1888                              struct strbuf *sb, int indent,
1889                              const char *line, int linelen)
1890 {
1891         strbuf_addchars(sb, ' ', indent);
1892         if (pp->expand_tabs_in_log)
1893                 strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1894         else
1895                 strbuf_add(sb, line, linelen);
1896 }
1897
1898 static int is_mboxrd_from(const char *line, int len)
1899 {
1900         /*
1901          * a line matching /^From $/ here would only have len == 4
1902          * at this point because is_empty_line would've trimmed all
1903          * trailing space
1904          */
1905         return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1906 }
1907
1908 void pp_remainder(struct pretty_print_context *pp,
1909                   const char **msg_p,
1910                   struct strbuf *sb,
1911                   int indent)
1912 {
1913         int first = 1;
1914         for (;;) {
1915                 const char *line = *msg_p;
1916                 int linelen = get_one_line(line);
1917                 *msg_p += linelen;
1918
1919                 if (!linelen)
1920                         break;
1921
1922                 if (is_blank_line(line, &linelen)) {
1923                         if (first)
1924                                 continue;
1925                         if (pp->fmt == CMIT_FMT_SHORT)
1926                                 break;
1927                 }
1928                 first = 0;
1929
1930                 strbuf_grow(sb, linelen + indent + 20);
1931                 if (indent)
1932                         pp_handle_indent(pp, sb, indent, line, linelen);
1933                 else if (pp->expand_tabs_in_log)
1934                         strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1935                                              line, linelen);
1936                 else {
1937                         if (pp->fmt == CMIT_FMT_MBOXRD &&
1938                                         is_mboxrd_from(line, linelen))
1939                                 strbuf_addch(sb, '>');
1940
1941                         strbuf_add(sb, line, linelen);
1942                 }
1943                 strbuf_addch(sb, '\n');
1944         }
1945 }
1946
1947 void pretty_print_commit(struct pretty_print_context *pp,
1948                          const struct commit *commit,
1949                          struct strbuf *sb)
1950 {
1951         unsigned long beginning_of_body;
1952         int indent = 4;
1953         const char *msg;
1954         const char *reencoded;
1955         const char *encoding;
1956         int need_8bit_cte = pp->need_8bit_cte;
1957
1958         if (pp->fmt == CMIT_FMT_USERFORMAT) {
1959                 format_commit_message(commit, user_format, sb, pp);
1960                 return;
1961         }
1962
1963         encoding = get_log_output_encoding();
1964         msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1965
1966         if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1967                 indent = 0;
1968
1969         /*
1970          * We need to check and emit Content-type: to mark it
1971          * as 8-bit if we haven't done so.
1972          */
1973         if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1974                 int i, ch, in_body;
1975
1976                 for (in_body = i = 0; (ch = msg[i]); i++) {
1977                         if (!in_body) {
1978                                 /* author could be non 7-bit ASCII but
1979                                  * the log may be so; skip over the
1980                                  * header part first.
1981                                  */
1982                                 if (ch == '\n' && msg[i+1] == '\n')
1983                                         in_body = 1;
1984                         }
1985                         else if (non_ascii(ch)) {
1986                                 need_8bit_cte = 1;
1987                                 break;
1988                         }
1989                 }
1990         }
1991
1992         pp_header(pp, encoding, commit, &msg, sb);
1993         if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
1994                 strbuf_addch(sb, '\n');
1995         }
1996
1997         /* Skip excess blank lines at the beginning of body, if any... */
1998         msg = skip_blank_lines(msg);
1999
2000         /* These formats treat the title line specially. */
2001         if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
2002                 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
2003
2004         beginning_of_body = sb->len;
2005         if (pp->fmt != CMIT_FMT_ONELINE)
2006                 pp_remainder(pp, &msg, sb, indent);
2007         strbuf_rtrim(sb);
2008
2009         /* Make sure there is an EOLN for the non-oneline case */
2010         if (pp->fmt != CMIT_FMT_ONELINE)
2011                 strbuf_addch(sb, '\n');
2012
2013         /*
2014          * The caller may append additional body text in e-mail
2015          * format.  Make sure we did not strip the blank line
2016          * between the header and the body.
2017          */
2018         if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
2019                 strbuf_addch(sb, '\n');
2020
2021         unuse_commit_buffer(commit, reencoded);
2022 }
2023
2024 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
2025                     struct strbuf *sb)
2026 {
2027         struct pretty_print_context pp = {0};
2028         pp.fmt = fmt;
2029         pretty_print_commit(&pp, commit, sb);
2030 }