pretty.c

   1 #include "cache.h"
   2 #include "commit.h"
   3 #include "utf8.h"
   4 #include "diff.h"
   5 #include "revision.h"
   6 #include "string-list.h"
   7 #include "mailmap.h"
   8 #include "log-tree.h"
   9 #include "notes.h"
  10 #include "color.h"
  11 #include "reflog-walk.h"
  12 #include "gpg-interface.h"
  13
  14 static char *user_format;
  15 static struct cmt_fmt_map {
  16         const char *name;
  17         enum cmit_fmt format;
  18         int is_tformat;
  19         int is_alias;
  20         const char *user_format;
  21 } *commit_formats;
  22 static size_t builtin_formats_len;
  23 static size_t commit_formats_len;
  24 static size_t commit_formats_alloc;
  25 static struct cmt_fmt_map *find_commit_format(const char *sought);
  26
  27 int commit_format_is_empty(enum cmit_fmt fmt)
  28 {
  29         return fmt == CMIT_FMT_USERFORMAT && !*user_format;
  30 }
  31
  32 static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  33 {
  34         free(user_format);
  35         user_format = xstrdup(cp);
  36         if (is_tformat)
  37                 rev->use_terminator = 1;
  38         rev->commit_format = CMIT_FMT_USERFORMAT;
  39 }
  40
  41 static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  42 {
  43         struct cmt_fmt_map *commit_format = NULL;
  44         const char *name;
  45         const char *fmt;
  46         int i;
  47
  48         if (!skip_prefix(var, "pretty.", &name))
  49                 return 0;
  50
  51         for (i = 0; i < builtin_formats_len; i++) {
  52                 if (!strcmp(commit_formats[i].name, name))
  53                         return 0;
  54         }
  55
  56         for (i = builtin_formats_len; i < commit_formats_len; i++) {
  57                 if (!strcmp(commit_formats[i].name, name)) {
  58                         commit_format = &commit_formats[i];
  59                         break;
  60                 }
  61         }
  62
  63         if (!commit_format) {
  64                 ALLOC_GROW(commit_formats, commit_formats_len+1,
  65                            commit_formats_alloc);
  66                 commit_format = &commit_formats[commit_formats_len];
  67                 memset(commit_format, 0, sizeof(*commit_format));
  68                 commit_formats_len++;
  69         }
  70
  71         commit_format->name = xstrdup(name);
  72         commit_format->format = CMIT_FMT_USERFORMAT;
  73         git_config_string(&fmt, var, value);
  74         if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
  75                 commit_format->is_tformat = fmt[0] == 't';
  76                 fmt = strchr(fmt, ':') + 1;
  77         } else if (strchr(fmt, '%'))
  78                 commit_format->is_tformat = 1;
  79         else
  80                 commit_format->is_alias = 1;
  81         commit_format->user_format = fmt;
  82
  83         return 0;
  84 }
  85
  86 static void setup_commit_formats(void)
  87 {
  88         struct cmt_fmt_map builtin_formats[] = {
  89                 { "raw",        CMIT_FMT_RAW,           0 },
  90                 { "medium",     CMIT_FMT_MEDIUM,        0 },
  91                 { "short",      CMIT_FMT_SHORT,         0 },
  92                 { "email",      CMIT_FMT_EMAIL,         0 },
  93                 { "fuller",     CMIT_FMT_FULLER,        0 },
  94                 { "full",       CMIT_FMT_FULL,          0 },
  95                 { "oneline",    CMIT_FMT_ONELINE,       1 }
  96         };
  97         commit_formats_len = ARRAY_SIZE(builtin_formats);
  98         builtin_formats_len = commit_formats_len;
  99         ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
 100         memcpy(commit_formats, builtin_formats,
 101                sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
 102
 103         git_config(git_pretty_formats_config, NULL);
 104 }
 105
 106 static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 107                                                         const char *original,
 108                                                         int num_redirections)
 109 {
 110         struct cmt_fmt_map *found = NULL;
 111         size_t found_match_len = 0;
 112         int i;
 113
 114         if (num_redirections >= commit_formats_len)
 115                 die("invalid --pretty format: "
 116                     "'%s' references an alias which points to itself",
 117                     original);
 118
 119         for (i = 0; i < commit_formats_len; i++) {
 120                 size_t match_len;
 121
 122                 if (!starts_with(commit_formats[i].name, sought))
 123                         continue;
 124
 125                 match_len = strlen(commit_formats[i].name);
 126                 if (found == NULL || found_match_len > match_len) {
 127                         found = &commit_formats[i];
 128                         found_match_len = match_len;
 129                 }
 130         }
 131
 132         if (found && found->is_alias) {
 133                 found = find_commit_format_recursive(found->user_format,
 134                                                      original,
 135                                                      num_redirections+1);
 136         }
 137
 138         return found;
 139 }
 140
 141 static struct cmt_fmt_map *find_commit_format(const char *sought)
 142 {
 143         if (!commit_formats)
 144                 setup_commit_formats();
 145
 146         return find_commit_format_recursive(sought, sought, 0);
 147 }
 148
 149 void get_commit_format(const char *arg, struct rev_info *rev)
 150 {
 151         struct cmt_fmt_map *commit_format;
 152
 153         rev->use_terminator = 0;
 154         if (!arg) {
 155                 rev->commit_format = CMIT_FMT_DEFAULT;
 156                 return;
 157         }
 158         if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
 159                 save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 160                 return;
 161         }
 162
 163         if (!*arg || strchr(arg, '%')) {
 164                 save_user_format(rev, arg, 1);
 165                 return;
 166         }
 167
 168         commit_format = find_commit_format(arg);
 169         if (!commit_format)
 170                 die("invalid --pretty format: %s", arg);
 171
 172         rev->commit_format = commit_format->format;
 173         rev->use_terminator = commit_format->is_tformat;
 174         if (commit_format->format == CMIT_FMT_USERFORMAT) {
 175                 save_user_format(rev, commit_format->user_format,
 176                                  commit_format->is_tformat);
 177         }
 178 }
 179
 180 /*
 181  * Generic support for pretty-printing the header
 182  */
 183 static int get_one_line(const char *msg)
 184 {
 185         int ret = 0;
 186
 187         for (;;) {
 188                 char c = *msg++;
 189                 if (!c)
 190                         break;
 191                 ret++;
 192                 if (c == '\n')
 193                         break;
 194         }
 195         return ret;
 196 }
 197
 198 /* High bit set, or ISO-2022-INT */
 199 static int non_ascii(int ch)
 200 {
 201         return !isascii(ch) || ch == '\033';
 202 }
 203
 204 int has_non_ascii(const char *s)
 205 {
 206         int ch;
 207         if (!s)
 208                 return 0;
 209         while ((ch = *s++) != '\0') {
 210                 if (non_ascii(ch))
 211                         return 1;
 212         }
 213         return 0;
 214 }
 215
 216 static int is_rfc822_special(char ch)
 217 {
 218         switch (ch) {
 219         case '(':
 220         case ')':
 221         case '<':
 222         case '>':
 223         case '[':
 224         case ']':
 225         case ':':
 226         case ';':
 227         case '@':
 228         case ',':
 229         case '.':
 230         case '"':
 231         case '\\':
 232                 return 1;
 233         default:
 234                 return 0;
 235         }
 236 }
 237
 238 static int needs_rfc822_quoting(const char *s, int len)
 239 {
 240         int i;
 241         for (i = 0; i < len; i++)
 242                 if (is_rfc822_special(s[i]))
 243                         return 1;
 244         return 0;
 245 }
 246
 247 static int last_line_length(struct strbuf *sb)
 248 {
 249         int i;
 250
 251         /* How many bytes are already used on the last line? */
 252         for (i = sb->len - 1; i >= 0; i--)
 253                 if (sb->buf[i] == '\n')
 254                         break;
 255         return sb->len - (i + 1);
 256 }
 257
 258 static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 259 {
 260         int i;
 261
 262         /* just a guess, we may have to also backslash-quote */
 263         strbuf_grow(out, len + 2);
 264
 265         strbuf_addch(out, '"');
 266         for (i = 0; i < len; i++) {
 267                 switch (s[i]) {
 268                 case '"':
 269                 case '\\':
 270                         strbuf_addch(out, '\\');
 271                         /* fall through */
 272                 default:
 273                         strbuf_addch(out, s[i]);
 274                 }
 275         }
 276         strbuf_addch(out, '"');
 277 }
 278
 279 enum rfc2047_type {
 280         RFC2047_SUBJECT,
 281         RFC2047_ADDRESS
 282 };
 283
 284 static int is_rfc2047_special(char ch, enum rfc2047_type type)
 285 {
 286         /*
 287          * rfc2047, section 4.2:
 288          *
 289          *    8-bit values which correspond to printable ASCII characters other
 290          *    than "=", "?", and "_" (underscore), MAY be represented as those
 291          *    characters.  (But see section 5 for restrictions.)  In
 292          *    particular, SPACE and TAB MUST NOT be represented as themselves
 293          *    within encoded words.
 294          */
 295
 296         /*
 297          * rule out non-ASCII characters and non-printable characters (the
 298          * non-ASCII check should be redundant as isprint() is not localized
 299          * and only knows about ASCII, but be defensive about that)
 300          */
 301         if (non_ascii(ch) || !isprint(ch))
 302                 return 1;
 303
 304         /*
 305          * rule out special printable characters (' ' should be the only
 306          * whitespace character considered printable, but be defensive and use
 307          * isspace())
 308          */
 309         if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 310                 return 1;
 311
 312         /*
 313          * rfc2047, section 5.3:
 314          *
 315          *    As a replacement for a 'word' entity within a 'phrase', for example,
 316          *    one that precedes an address in a From, To, or Cc header.  The ABNF
 317          *    definition for 'phrase' from RFC 822 thus becomes:
 318          *
 319          *    phrase = 1*( encoded-word / word )
 320          *
 321          *    In this case the set of characters that may be used in a "Q"-encoded
 322          *    'encoded-word' is restricted to: <upper and lower case ASCII
 323          *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 324          *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 325          *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 326          *    'special' by 'linear-white-space'.
 327          */
 328
 329         if (type != RFC2047_ADDRESS)
 330                 return 0;
 331
 332         /* '=' and '_' are special cases and have been checked above */
 333         return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 334 }
 335
 336 static int needs_rfc2047_encoding(const char *line, int len,
 337                                   enum rfc2047_type type)
 338 {
 339         int i;
 340
 341         for (i = 0; i < len; i++) {
 342                 int ch = line[i];
 343                 if (non_ascii(ch) || ch == '\n')
 344                         return 1;
 345                 if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 346                         return 1;
 347         }
 348
 349         return 0;
 350 }
 351
 352 static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 353                        const char *encoding, enum rfc2047_type type)
 354 {
 355         static const int max_encoded_length = 76; /* per rfc2047 */
 356         int i;
 357         int line_len = last_line_length(sb);
 358
 359         strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 360         strbuf_addf(sb, "=?%s?q?", encoding);
 361         line_len += strlen(encoding) + 5; /* 5 for =??q? */
 362
 363         while (len) {
 364                 /*
 365                  * RFC 2047, section 5 (3):
 366                  *
 367                  * Each 'encoded-word' MUST represent an integral number of
 368                  * characters.  A multi-octet character may not be split across
 369                  * adjacent 'encoded- word's.
 370                  */
 371                 const unsigned char *p = (const unsigned char *)line;
 372                 int chrlen = mbs_chrlen(&line, &len, encoding);
 373                 int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 374
 375                 /* "=%02X" * chrlen, or the byte itself */
 376                 const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 377                 int         encoded_len = is_special ? 3 * chrlen : 1;
 378
 379                 /*
 380                  * According to RFC 2047, we could encode the special character
 381                  * ' ' (space) with '_' (underscore) for readability. But many
 382                  * programs do not understand this and just leave the
 383                  * underscore in place. Thus, we do nothing special here, which
 384                  * causes ' ' to be encoded as '=20', avoiding this problem.
 385                  */
 386
 387                 if (line_len + encoded_len + 2 > max_encoded_length) {
 388                         /* It won't fit with trailing "?=" --- break the line */
 389                         strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 390                         line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 391                 }
 392
 393                 for (i = 0; i < chrlen; i++)
 394                         strbuf_addf(sb, encoded_fmt, p[i]);
 395                 line_len += encoded_len;
 396         }
 397         strbuf_addstr(sb, "?=");
 398 }
 399
 400 const char *show_ident_date(const struct ident_split *ident,
 401                             enum date_mode mode)
 402 {
 403         unsigned long date = 0;
 404         long tz = 0;
 405
 406         if (ident->date_begin && ident->date_end)
 407                 date = strtoul(ident->date_begin, NULL, 10);
 408         if (date_overflows(date))
 409                 date = 0;
 410         else {
 411                 if (ident->tz_begin && ident->tz_end)
 412                         tz = strtol(ident->tz_begin, NULL, 10);
 413                 if (tz >= INT_MAX || tz <= INT_MIN)
 414                         tz = 0;
 415         }
 416         return show_date(date, tz, mode);
 417 }
 418
 419 void pp_user_info(struct pretty_print_context *pp,
 420                   const char *what, struct strbuf *sb,
 421                   const char *line, const char *encoding)
 422 {
 423         struct ident_split ident;
 424         char *line_end;
 425         const char *mailbuf, *namebuf;
 426         size_t namelen, maillen;
 427         int max_length = 78; /* per rfc2822 */
 428
 429         if (pp->fmt == CMIT_FMT_ONELINE)
 430                 return;
 431
 432         line_end = strchrnul(line, '\n');
 433         if (split_ident_line(&ident, line, line_end - line))
 434                 return;
 435
 436         mailbuf = ident.mail_begin;
 437         maillen = ident.mail_end - ident.mail_begin;
 438         namebuf = ident.name_begin;
 439         namelen = ident.name_end - ident.name_begin;
 440
 441         if (pp->mailmap)
 442                 map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 443
 444         if (pp->fmt == CMIT_FMT_EMAIL) {
 445                 if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 446                         struct strbuf buf = STRBUF_INIT;
 447
 448                         strbuf_addstr(&buf, "From: ");
 449                         strbuf_add(&buf, namebuf, namelen);
 450                         strbuf_addstr(&buf, " <");
 451                         strbuf_add(&buf, mailbuf, maillen);
 452                         strbuf_addstr(&buf, ">\n");
 453                         string_list_append(&pp->in_body_headers,
 454                                            strbuf_detach(&buf, NULL));
 455
 456                         mailbuf = pp->from_ident->mail_begin;
 457                         maillen = pp->from_ident->mail_end - mailbuf;
 458                         namebuf = pp->from_ident->name_begin;
 459                         namelen = pp->from_ident->name_end - namebuf;
 460                 }
 461
 462                 strbuf_addstr(sb, "From: ");
 463                 if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 464                         add_rfc2047(sb, namebuf, namelen,
 465                                     encoding, RFC2047_ADDRESS);
 466                         max_length = 76; /* per rfc2047 */
 467                 } else if (needs_rfc822_quoting(namebuf, namelen)) {
 468                         struct strbuf quoted = STRBUF_INIT;
 469                         add_rfc822_quoted(&quoted, namebuf, namelen);
 470                         strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 471                                                         -6, 1, max_length);
 472                         strbuf_release(&quoted);
 473                 } else {
 474                         strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 475                                                  -6, 1, max_length);
 476                 }
 477
 478                 if (max_length <
 479                     last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 480                         strbuf_addch(sb, '\n');
 481                 strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 482         } else {
 483                 strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 484                             (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 485                             (int)namelen, namebuf, (int)maillen, mailbuf);
 486         }
 487
 488         switch (pp->fmt) {
 489         case CMIT_FMT_MEDIUM:
 490                 strbuf_addf(sb, "Date:   %s\n",
 491                             show_ident_date(&ident, pp->date_mode));
 492                 break;
 493         case CMIT_FMT_EMAIL:
 494                 strbuf_addf(sb, "Date: %s\n",
 495                             show_ident_date(&ident, DATE_RFC2822));
 496                 break;
 497         case CMIT_FMT_FULLER:
 498                 strbuf_addf(sb, "%sDate: %s\n", what,
 499                             show_ident_date(&ident, pp->date_mode));
 500                 break;
 501         default:
 502                 /* notin' */
 503                 break;
 504         }
 505 }
 506
 507 static int is_empty_line(const char *line, int *len_p)
 508 {
 509         int len = *len_p;
 510         while (len && isspace(line[len - 1]))
 511                 len--;
 512         *len_p = len;
 513         return !len;
 514 }
 515
 516 static const char *skip_empty_lines(const char *msg)
 517 {
 518         for (;;) {
 519                 int linelen = get_one_line(msg);
 520                 int ll = linelen;
 521                 if (!linelen)
 522                         break;
 523                 if (!is_empty_line(msg, &ll))
 524                         break;
 525                 msg += linelen;
 526         }
 527         return msg;
 528 }
 529
 530 static void add_merge_info(const struct pretty_print_context *pp,
 531                            struct strbuf *sb, const struct commit *commit)
 532 {
 533         struct commit_list *parent = commit->parents;
 534
 535         if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 536             !parent || !parent->next)
 537                 return;
 538
 539         strbuf_addstr(sb, "Merge:");
 540
 541         while (parent) {
 542                 struct commit *p = parent->item;
 543                 const char *hex = NULL;
 544                 if (pp->abbrev)
 545                         hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
 546                 if (!hex)
 547                         hex = sha1_to_hex(p->object.sha1);
 548                 parent = parent->next;
 549
 550                 strbuf_addf(sb, " %s", hex);
 551         }
 552         strbuf_addch(sb, '\n');
 553 }
 554
 555 static char *get_header(const struct commit *commit, const char *msg,
 556                         const char *key)
 557 {
 558         int key_len = strlen(key);
 559         const char *line = msg;
 560
 561         while (line) {
 562                 const char *eol = strchrnul(line, '\n'), *next;
 563
 564                 if (line == eol)
 565                         return NULL;
 566                 if (!*eol) {
 567                         warning("malformed commit (header is missing newline): %s",
 568                                 sha1_to_hex(commit->object.sha1));
 569                         next = NULL;
 570                 } else
 571                         next = eol + 1;
 572                 if (eol - line > key_len &&
 573                     !strncmp(line, key, key_len) &&
 574                     line[key_len] == ' ') {
 575                         return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 576                 }
 577                 line = next;
 578         }
 579         return NULL;
 580 }
 581
 582 static char *replace_encoding_header(char *buf, const char *encoding)
 583 {
 584         struct strbuf tmp = STRBUF_INIT;
 585         size_t start, len;
 586         char *cp = buf;
 587
 588         /* guess if there is an encoding header before a \n\n */
 589         while (strncmp(cp, "encoding ", strlen("encoding "))) {
 590                 cp = strchr(cp, '\n');
 591                 if (!cp || *++cp == '\n')
 592                         return buf;
 593         }
 594         start = cp - buf;
 595         cp = strchr(cp, '\n');
 596         if (!cp)
 597                 return buf; /* should not happen but be defensive */
 598         len = cp + 1 - (buf + start);
 599
 600         strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 601         if (is_encoding_utf8(encoding)) {
 602                 /* we have re-coded to UTF-8; drop the header */
 603                 strbuf_remove(&tmp, start, len);
 604         } else {
 605                 /* just replaces XXXX in 'encoding XXXX\n' */
 606                 strbuf_splice(&tmp, start + strlen("encoding "),
 607                                           len - strlen("encoding \n"),
 608                                           encoding, strlen(encoding));
 609         }
 610         return strbuf_detach(&tmp, NULL);
 611 }
 612
 613 const char *logmsg_reencode(const struct commit *commit,
 614                             char **commit_encoding,
 615                             const char *output_encoding)
 616 {
 617         static const char *utf8 = "UTF-8";
 618         const char *use_encoding;
 619         char *encoding;
 620         const char *msg = get_commit_buffer(commit, NULL);
 621         char *out;
 622
 623         if (!output_encoding || !*output_encoding) {
 624                 if (commit_encoding)
 625                         *commit_encoding =
 626                                 get_header(commit, msg, "encoding");
 627                 return msg;
 628         }
 629         encoding = get_header(commit, msg, "encoding");
 630         if (commit_encoding)
 631                 *commit_encoding = encoding;
 632         use_encoding = encoding ? encoding : utf8;
 633         if (same_encoding(use_encoding, output_encoding)) {
 634                 /*
 635                  * No encoding work to be done. If we have no encoding header
 636                  * at all, then there's nothing to do, and we can return the
 637                  * message verbatim (whether newly allocated or not).
 638                  */
 639                 if (!encoding)
 640                         return msg;
 641
 642                 /*
 643                  * Otherwise, we still want to munge the encoding header in the
 644                  * result, which will be done by modifying the buffer. If we
 645                  * are using a fresh copy, we can reuse it. But if we are using
 646                  * the cached copy from get_commit_buffer, we need to duplicate it
 647                  * to avoid munging the cached copy.
 648                  */
 649                 if (msg == get_cached_commit_buffer(commit, NULL))
 650                         out = xstrdup(msg);
 651                 else
 652                         out = (char *)msg;
 653         }
 654         else {
 655                 /*
 656                  * There's actual encoding work to do. Do the reencoding, which
 657                  * still leaves the header to be replaced in the next step. At
 658                  * this point, we are done with msg. If we allocated a fresh
 659                  * copy, we can free it.
 660                  */
 661                 out = reencode_string(msg, output_encoding, use_encoding);
 662                 if (out)
 663                         unuse_commit_buffer(commit, msg);
 664         }
 665
 666         /*
 667          * This replacement actually consumes the buffer we hand it, so we do
 668          * not have to worry about freeing the old "out" here.
 669          */
 670         if (out)
 671                 out = replace_encoding_header(out, output_encoding);
 672
 673         if (!commit_encoding)
 674                 free(encoding);
 675         /*
 676          * If the re-encoding failed, out might be NULL here; in that
 677          * case we just return the commit message verbatim.
 678          */
 679         return out ? out : msg;
 680 }
 681
 682 static int mailmap_name(const char **email, size_t *email_len,
 683                         const char **name, size_t *name_len)
 684 {
 685         static struct string_list *mail_map;
 686         if (!mail_map) {
 687                 mail_map = xcalloc(1, sizeof(*mail_map));
 688                 read_mailmap(mail_map, NULL);
 689         }
 690         return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 691 }
 692
 693 static size_t format_person_part(struct strbuf *sb, char part,
 694                                  const char *msg, int len, enum date_mode dmode)
 695 {
 696         /* currently all placeholders have same length */
 697         const int placeholder_len = 2;
 698         struct ident_split s;
 699         const char *name, *mail;
 700         size_t maillen, namelen;
 701
 702         if (split_ident_line(&s, msg, len) < 0)
 703                 goto skip;
 704
 705         name = s.name_begin;
 706         namelen = s.name_end - s.name_begin;
 707         mail = s.mail_begin;
 708         maillen = s.mail_end - s.mail_begin;
 709
 710         if (part == 'N' || part == 'E') /* mailmap lookup */
 711                 mailmap_name(&mail, &maillen, &name, &namelen);
 712         if (part == 'n' || part == 'N') {       /* name */
 713                 strbuf_add(sb, name, namelen);
 714                 return placeholder_len;
 715         }
 716         if (part == 'e' || part == 'E') {       /* email */
 717                 strbuf_add(sb, mail, maillen);
 718                 return placeholder_len;
 719         }
 720
 721         if (!s.date_begin)
 722                 goto skip;
 723
 724         if (part == 't') {      /* date, UNIX timestamp */
 725                 strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 726                 return placeholder_len;
 727         }
 728
 729         switch (part) {
 730         case 'd':       /* date */
 731                 strbuf_addstr(sb, show_ident_date(&s, dmode));
 732                 return placeholder_len;
 733         case 'D':       /* date, RFC2822 style */
 734                 strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
 735                 return placeholder_len;
 736         case 'r':       /* date, relative */
 737                 strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
 738                 return placeholder_len;
 739         case 'i':       /* date, ISO 8601 */
 740                 strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
 741                 return placeholder_len;
 742         }
 743
 744 skip:
 745         /*
 746          * reading from either a bogus commit, or a reflog entry with
 747          * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 748          * to compute a valid return value.
 749          */
 750         if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 751             || part == 'D' || part == 'r' || part == 'i')
 752                 return placeholder_len;
 753
 754         return 0; /* unknown placeholder */
 755 }
 756
 757 struct chunk {
 758         size_t off;
 759         size_t len;
 760 };
 761
 762 enum flush_type {
 763         no_flush,
 764         flush_right,
 765         flush_left,
 766         flush_left_and_steal,
 767         flush_both
 768 };
 769
 770 enum trunc_type {
 771         trunc_none,
 772         trunc_left,
 773         trunc_middle,
 774         trunc_right
 775 };
 776
 777 struct format_commit_context {
 778         const struct commit *commit;
 779         const struct pretty_print_context *pretty_ctx;
 780         unsigned commit_header_parsed:1;
 781         unsigned commit_message_parsed:1;
 782         struct signature_check signature_check;
 783         enum flush_type flush_type;
 784         enum trunc_type truncate;
 785         const char *message;
 786         char *commit_encoding;
 787         size_t width, indent1, indent2;
 788         int auto_color;
 789         int padding;
 790
 791         /* These offsets are relative to the start of the commit message. */
 792         struct chunk author;
 793         struct chunk committer;
 794         size_t message_off;
 795         size_t subject_off;
 796         size_t body_off;
 797
 798         /* The following ones are relative to the result struct strbuf. */
 799         struct chunk abbrev_commit_hash;
 800         struct chunk abbrev_tree_hash;
 801         struct chunk abbrev_parent_hashes;
 802         size_t wrap_start;
 803 };
 804
 805 static int add_again(struct strbuf *sb, struct chunk *chunk)
 806 {
 807         if (chunk->len) {
 808                 strbuf_adddup(sb, chunk->off, chunk->len);
 809                 return 1;
 810         }
 811
 812         /*
 813          * We haven't seen this chunk before.  Our caller is surely
 814          * going to add it the hard way now.  Remember the most likely
 815          * start of the to-be-added chunk: the current end of the
 816          * struct strbuf.
 817          */
 818         chunk->off = sb->len;
 819         return 0;
 820 }
 821
 822 static void parse_commit_header(struct format_commit_context *context)
 823 {
 824         const char *msg = context->message;
 825         int i;
 826
 827         for (i = 0; msg[i]; i++) {
 828                 int eol;
 829                 for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 830                         ; /* do nothing */
 831
 832                 if (i == eol) {
 833                         break;
 834                 } else if (starts_with(msg + i, "author ")) {
 835                         context->author.off = i + 7;
 836                         context->author.len = eol - i - 7;
 837                 } else if (starts_with(msg + i, "committer ")) {
 838                         context->committer.off = i + 10;
 839                         context->committer.len = eol - i - 10;
 840                 }
 841                 i = eol;
 842         }
 843         context->message_off = i;
 844         context->commit_header_parsed = 1;
 845 }
 846
 847 static int istitlechar(char c)
 848 {
 849         return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 850                 (c >= '0' && c <= '9') || c == '.' || c == '_';
 851 }
 852
 853 static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 854 {
 855         size_t trimlen;
 856         size_t start_len = sb->len;
 857         int space = 2;
 858
 859         for (; *msg && *msg != '\n'; msg++) {
 860                 if (istitlechar(*msg)) {
 861                         if (space == 1)
 862                                 strbuf_addch(sb, '-');
 863                         space = 0;
 864                         strbuf_addch(sb, *msg);
 865                         if (*msg == '.')
 866                                 while (*(msg+1) == '.')
 867                                         msg++;
 868                 } else
 869                         space |= 1;
 870         }
 871
 872         /* trim any trailing '.' or '-' characters */
 873         trimlen = 0;
 874         while (sb->len - trimlen > start_len &&
 875                 (sb->buf[sb->len - 1 - trimlen] == '.'
 876                 || sb->buf[sb->len - 1 - trimlen] == '-'))
 877                 trimlen++;
 878         strbuf_remove(sb, sb->len - trimlen, trimlen);
 879 }
 880
 881 const char *format_subject(struct strbuf *sb, const char *msg,
 882                            const char *line_separator)
 883 {
 884         int first = 1;
 885
 886         for (;;) {
 887                 const char *line = msg;
 888                 int linelen = get_one_line(line);
 889
 890                 msg += linelen;
 891                 if (!linelen || is_empty_line(line, &linelen))
 892                         break;
 893
 894                 if (!sb)
 895                         continue;
 896                 strbuf_grow(sb, linelen + 2);
 897                 if (!first)
 898                         strbuf_addstr(sb, line_separator);
 899                 strbuf_add(sb, line, linelen);
 900                 first = 0;
 901         }
 902         return msg;
 903 }
 904
 905 static void parse_commit_message(struct format_commit_context *c)
 906 {
 907         const char *msg = c->message + c->message_off;
 908         const char *start = c->message;
 909
 910         msg = skip_empty_lines(msg);
 911         c->subject_off = msg - start;
 912
 913         msg = format_subject(NULL, msg, NULL);
 914         msg = skip_empty_lines(msg);
 915         c->body_off = msg - start;
 916
 917         c->commit_message_parsed = 1;
 918 }
 919
 920 static void strbuf_wrap(struct strbuf *sb, size_t pos,
 921                         size_t width, size_t indent1, size_t indent2)
 922 {
 923         struct strbuf tmp = STRBUF_INIT;
 924
 925         if (pos)
 926                 strbuf_add(&tmp, sb->buf, pos);
 927         strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 928                                 (int) indent1, (int) indent2, (int) width);
 929         strbuf_swap(&tmp, sb);
 930         strbuf_release(&tmp);
 931 }
 932
 933 static void rewrap_message_tail(struct strbuf *sb,
 934                                 struct format_commit_context *c,
 935                                 size_t new_width, size_t new_indent1,
 936                                 size_t new_indent2)
 937 {
 938         if (c->width == new_width && c->indent1 == new_indent1 &&
 939             c->indent2 == new_indent2)
 940                 return;
 941         if (c->wrap_start < sb->len)
 942                 strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 943         c->wrap_start = sb->len;
 944         c->width = new_width;
 945         c->indent1 = new_indent1;
 946         c->indent2 = new_indent2;
 947 }
 948
 949 static int format_reflog_person(struct strbuf *sb,
 950                                 char part,
 951                                 struct reflog_walk_info *log,
 952                                 enum date_mode dmode)
 953 {
 954         const char *ident;
 955
 956         if (!log)
 957                 return 2;
 958
 959         ident = get_reflog_ident(log);
 960         if (!ident)
 961                 return 2;
 962
 963         return format_person_part(sb, part, ident, strlen(ident), dmode);
 964 }
 965
 966 static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 967                           const char *placeholder,
 968                           struct format_commit_context *c)
 969 {
 970         if (placeholder[1] == '(') {
 971                 const char *begin = placeholder + 2;
 972                 const char *end = strchr(begin, ')');
 973                 char color[COLOR_MAXLEN];
 974
 975                 if (!end)
 976                         return 0;
 977                 if (starts_with(begin, "auto,")) {
 978                         if (!want_color(c->pretty_ctx->color))
 979                                 return end - placeholder + 1;
 980                         begin += 5;
 981                 }
 982                 color_parse_mem(begin,
 983                                 end - begin,
 984                                 "--pretty format", color);
 985                 strbuf_addstr(sb, color);
 986                 return end - placeholder + 1;
 987         }
 988         if (starts_with(placeholder + 1, "red")) {
 989                 strbuf_addstr(sb, GIT_COLOR_RED);
 990                 return 4;
 991         } else if (starts_with(placeholder + 1, "green")) {
 992                 strbuf_addstr(sb, GIT_COLOR_GREEN);
 993                 return 6;
 994         } else if (starts_with(placeholder + 1, "blue")) {
 995                 strbuf_addstr(sb, GIT_COLOR_BLUE);
 996                 return 5;
 997         } else if (starts_with(placeholder + 1, "reset")) {
 998                 strbuf_addstr(sb, GIT_COLOR_RESET);
 999                 return 6;
1000         } else
1001                 return 0;
1002 }
1003
1004 static size_t parse_padding_placeholder(struct strbuf *sb,
1005                                         const char *placeholder,
1006                                         struct format_commit_context *c)
1007 {
1008         const char *ch = placeholder;
1009         enum flush_type flush_type;
1010         int to_column = 0;
1011
1012         switch (*ch++) {
1013         case '<':
1014                 flush_type = flush_right;
1015                 break;
1016         case '>':
1017                 if (*ch == '<') {
1018                         flush_type = flush_both;
1019                         ch++;
1020                 } else if (*ch == '>') {
1021                         flush_type = flush_left_and_steal;
1022                         ch++;
1023                 } else
1024                         flush_type = flush_left;
1025                 break;
1026         default:
1027                 return 0;
1028         }
1029
1030         /* the next value means "wide enough to that column" */
1031         if (*ch == '|') {
1032                 to_column = 1;
1033                 ch++;
1034         }
1035
1036         if (*ch == '(') {
1037                 const char *start = ch + 1;
1038                 const char *end = start + strcspn(start, ",)");
1039                 char *next;
1040                 int width;
1041                 if (!end || end == start)
1042                         return 0;
1043                 width = strtoul(start, &next, 10);
1044                 if (next == start || width == 0)
1045                         return 0;
1046                 c->padding = to_column ? -width : width;
1047                 c->flush_type = flush_type;
1048
1049                 if (*end == ',') {
1050                         start = end + 1;
1051                         end = strchr(start, ')');
1052                         if (!end || end == start)
1053                                 return 0;
1054                         if (starts_with(start, "trunc)"))
1055                                 c->truncate = trunc_right;
1056                         else if (starts_with(start, "ltrunc)"))
1057                                 c->truncate = trunc_left;
1058                         else if (starts_with(start, "mtrunc)"))
1059                                 c->truncate = trunc_middle;
1060                         else
1061                                 return 0;
1062                 } else
1063                         c->truncate = trunc_none;
1064
1065                 return end - placeholder + 1;
1066         }
1067         return 0;
1068 }
1069
1070 static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1071                                 const char *placeholder,
1072                                 void *context)
1073 {
1074         struct format_commit_context *c = context;
1075         const struct commit *commit = c->commit;
1076         const char *msg = c->message;
1077         struct commit_list *p;
1078         int h1, h2;
1079
1080         /* these are independent of the commit */
1081         switch (placeholder[0]) {
1082         case 'C':
1083                 if (starts_with(placeholder + 1, "(auto)")) {
1084                         c->auto_color = 1;
1085                         return 7; /* consumed 7 bytes, "C(auto)" */
1086                 } else {
1087                         int ret = parse_color(sb, placeholder, c);
1088                         if (ret)
1089                                 c->auto_color = 0;
1090                         /*
1091                          * Otherwise, we decided to treat %C<unknown>
1092                          * as a literal string, and the previous
1093                          * %C(auto) is still valid.
1094                          */
1095                         return ret;
1096                 }
1097         case 'n':               /* newline */
1098                 strbuf_addch(sb, '\n');
1099                 return 1;
1100         case 'x':
1101                 /* %x00 == NUL, %x0a == LF, etc. */
1102                 if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1103                     h1 <= 16 &&
1104                     0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1105                     h2 <= 16) {
1106                         strbuf_addch(sb, (h1<<4)|h2);
1107                         return 3;
1108                 } else
1109                         return 0;
1110         case 'w':
1111                 if (placeholder[1] == '(') {
1112                         unsigned long width = 0, indent1 = 0, indent2 = 0;
1113                         char *next;
1114                         const char *start = placeholder + 2;
1115                         const char *end = strchr(start, ')');
1116                         if (!end)
1117                                 return 0;
1118                         if (end > start) {
1119                                 width = strtoul(start, &next, 10);
1120                                 if (*next == ',') {
1121                                         indent1 = strtoul(next + 1, &next, 10);
1122                                         if (*next == ',') {
1123                                                 indent2 = strtoul(next + 1,
1124                                                                  &next, 10);
1125                                         }
1126                                 }
1127                                 if (*next != ')')
1128                                         return 0;
1129                         }
1130                         rewrap_message_tail(sb, c, width, indent1, indent2);
1131                         return end - placeholder + 1;
1132                 } else
1133                         return 0;
1134
1135         case '<':
1136         case '>':
1137                 return parse_padding_placeholder(sb, placeholder, c);
1138         }
1139
1140         /* these depend on the commit */
1141         if (!commit->object.parsed)
1142                 parse_object(commit->object.sha1);
1143
1144         switch (placeholder[0]) {
1145         case 'H':               /* commit hash */
1146                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1147                 strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1148                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1149                 return 1;
1150         case 'h':               /* abbreviated commit hash */
1151                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1152                 if (add_again(sb, &c->abbrev_commit_hash)) {
1153                         strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1154                         return 1;
1155                 }
1156                 strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1157                                                      c->pretty_ctx->abbrev));
1158                 strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1159                 c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1160                 return 1;
1161         case 'T':               /* tree hash */
1162                 strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1163                 return 1;
1164         case 't':               /* abbreviated tree hash */
1165                 if (add_again(sb, &c->abbrev_tree_hash))
1166                         return 1;
1167                 strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1168                                                      c->pretty_ctx->abbrev));
1169                 c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1170                 return 1;
1171         case 'P':               /* parent hashes */
1172                 for (p = commit->parents; p; p = p->next) {
1173                         if (p != commit->parents)
1174                                 strbuf_addch(sb, ' ');
1175                         strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1176                 }
1177                 return 1;
1178         case 'p':               /* abbreviated parent hashes */
1179                 if (add_again(sb, &c->abbrev_parent_hashes))
1180                         return 1;
1181                 for (p = commit->parents; p; p = p->next) {
1182                         if (p != commit->parents)
1183                                 strbuf_addch(sb, ' ');
1184                         strbuf_addstr(sb, find_unique_abbrev(
1185                                         p->item->object.sha1,
1186                                         c->pretty_ctx->abbrev));
1187                 }
1188                 c->abbrev_parent_hashes.len = sb->len -
1189                                               c->abbrev_parent_hashes.off;
1190                 return 1;
1191         case 'm':               /* left/right/bottom */
1192                 strbuf_addstr(sb, get_revision_mark(NULL, commit));
1193                 return 1;
1194         case 'd':
1195                 load_ref_decorations(DECORATE_SHORT_REFS);
1196                 format_decorations(sb, commit, c->auto_color);
1197                 return 1;
1198         case 'g':               /* reflog info */
1199                 switch(placeholder[1]) {
1200                 case 'd':       /* reflog selector */
1201                 case 'D':
1202                         if (c->pretty_ctx->reflog_info)
1203                                 get_reflog_selector(sb,
1204                                                     c->pretty_ctx->reflog_info,
1205                                                     c->pretty_ctx->date_mode,
1206                                                     c->pretty_ctx->date_mode_explicit,
1207                                                     (placeholder[1] == 'd'));
1208                         return 2;
1209                 case 's':       /* reflog message */
1210                         if (c->pretty_ctx->reflog_info)
1211                                 get_reflog_message(sb, c->pretty_ctx->reflog_info);
1212                         return 2;
1213                 case 'n':
1214                 case 'N':
1215                 case 'e':
1216                 case 'E':
1217                         return format_reflog_person(sb,
1218                                                     placeholder[1],
1219                                                     c->pretty_ctx->reflog_info,
1220                                                     c->pretty_ctx->date_mode);
1221                 }
1222                 return 0;       /* unknown %g placeholder */
1223         case 'N':
1224                 if (c->pretty_ctx->notes_message) {
1225                         strbuf_addstr(sb, c->pretty_ctx->notes_message);
1226                         return 1;
1227                 }
1228                 return 0;
1229         }
1230
1231         if (placeholder[0] == 'G') {
1232                 if (!c->signature_check.result)
1233                         check_commit_signature(c->commit, &(c->signature_check));
1234                 switch (placeholder[1]) {
1235                 case 'G':
1236                         if (c->signature_check.gpg_output)
1237                                 strbuf_addstr(sb, c->signature_check.gpg_output);
1238                         break;
1239                 case '?':
1240                         switch (c->signature_check.result) {
1241                         case 'G':
1242                         case 'B':
1243                         case 'U':
1244                         case 'N':
1245                                 strbuf_addch(sb, c->signature_check.result);
1246                         }
1247                         break;
1248                 case 'S':
1249                         if (c->signature_check.signer)
1250                                 strbuf_addstr(sb, c->signature_check.signer);
1251                         break;
1252                 case 'K':
1253                         if (c->signature_check.key)
1254                                 strbuf_addstr(sb, c->signature_check.key);
1255                         break;
1256                 default:
1257                         return 0;
1258                 }
1259                 return 2;
1260         }
1261
1262
1263         /* For the rest we have to parse the commit header. */
1264         if (!c->commit_header_parsed)
1265                 parse_commit_header(c);
1266
1267         switch (placeholder[0]) {
1268         case 'a':       /* author ... */
1269                 return format_person_part(sb, placeholder[1],
1270                                    msg + c->author.off, c->author.len,
1271                                    c->pretty_ctx->date_mode);
1272         case 'c':       /* committer ... */
1273                 return format_person_part(sb, placeholder[1],
1274                                    msg + c->committer.off, c->committer.len,
1275                                    c->pretty_ctx->date_mode);
1276         case 'e':       /* encoding */
1277                 if (c->commit_encoding)
1278                         strbuf_addstr(sb, c->commit_encoding);
1279                 return 1;
1280         case 'B':       /* raw body */
1281                 /* message_off is always left at the initial newline */
1282                 strbuf_addstr(sb, msg + c->message_off + 1);
1283                 return 1;
1284         }
1285
1286         /* Now we need to parse the commit message. */
1287         if (!c->commit_message_parsed)
1288                 parse_commit_message(c);
1289
1290         switch (placeholder[0]) {
1291         case 's':       /* subject */
1292                 format_subject(sb, msg + c->subject_off, " ");
1293                 return 1;
1294         case 'f':       /* sanitized subject */
1295                 format_sanitized_subject(sb, msg + c->subject_off);
1296                 return 1;
1297         case 'b':       /* body */
1298                 strbuf_addstr(sb, msg + c->body_off);
1299                 return 1;
1300         }
1301         return 0;       /* unknown placeholder */
1302 }
1303
1304 static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1305                                     const char *placeholder,
1306                                     struct format_commit_context *c)
1307 {
1308         struct strbuf local_sb = STRBUF_INIT;
1309         int total_consumed = 0, len, padding = c->padding;
1310         if (padding < 0) {
1311                 const char *start = strrchr(sb->buf, '\n');
1312                 int occupied;
1313                 if (!start)
1314                         start = sb->buf;
1315                 occupied = utf8_strnwidth(start, -1, 1);
1316                 padding = (-padding) - occupied;
1317         }
1318         while (1) {
1319                 int modifier = *placeholder == 'C';
1320                 int consumed = format_commit_one(&local_sb, placeholder, c);
1321                 total_consumed += consumed;
1322
1323                 if (!modifier)
1324                         break;
1325
1326                 placeholder += consumed;
1327                 if (*placeholder != '%')
1328                         break;
1329                 placeholder++;
1330                 total_consumed++;
1331         }
1332         len = utf8_strnwidth(local_sb.buf, -1, 1);
1333
1334         if (c->flush_type == flush_left_and_steal) {
1335                 const char *ch = sb->buf + sb->len - 1;
1336                 while (len > padding && ch > sb->buf) {
1337                         const char *p;
1338                         if (*ch == ' ') {
1339                                 ch--;
1340                                 padding++;
1341                                 continue;
1342                         }
1343                         /* check for trailing ansi sequences */
1344                         if (*ch != 'm')
1345                                 break;
1346                         p = ch - 1;
1347                         while (ch - p < 10 && *p != '\033')
1348                                 p--;
1349                         if (*p != '\033' ||
1350                             ch + 1 - p != display_mode_esc_sequence_len(p))
1351                                 break;
1352                         /*
1353                          * got a good ansi sequence, put it back to
1354                          * local_sb as we're cutting sb
1355                          */
1356                         strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1357                         ch = p - 1;
1358                 }
1359                 strbuf_setlen(sb, ch + 1 - sb->buf);
1360                 c->flush_type = flush_left;
1361         }
1362
1363         if (len > padding) {
1364                 switch (c->truncate) {
1365                 case trunc_left:
1366                         strbuf_utf8_replace(&local_sb,
1367                                             0, len - (padding - 2),
1368                                             "..");
1369                         break;
1370                 case trunc_middle:
1371                         strbuf_utf8_replace(&local_sb,
1372                                             padding / 2 - 1,
1373                                             len - (padding - 2),
1374                                             "..");
1375                         break;
1376                 case trunc_right:
1377                         strbuf_utf8_replace(&local_sb,
1378                                             padding - 2, len - (padding - 2),
1379                                             "..");
1380                         break;
1381                 case trunc_none:
1382                         break;
1383                 }
1384                 strbuf_addbuf(sb, &local_sb);
1385         } else {
1386                 int sb_len = sb->len, offset = 0;
1387                 if (c->flush_type == flush_left)
1388                         offset = padding - len;
1389                 else if (c->flush_type == flush_both)
1390                         offset = (padding - len) / 2;
1391                 /*
1392                  * we calculate padding in columns, now
1393                  * convert it back to chars
1394                  */
1395                 padding = padding - len + local_sb.len;
1396                 strbuf_grow(sb, padding);
1397                 strbuf_setlen(sb, sb_len + padding);
1398                 memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1399                 memcpy(sb->buf + sb_len + offset, local_sb.buf,
1400                        local_sb.len);
1401         }
1402         strbuf_release(&local_sb);
1403         c->flush_type = no_flush;
1404         return total_consumed;
1405 }
1406
1407 static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1408                                  const char *placeholder,
1409                                  void *context)
1410 {
1411         int consumed;
1412         size_t orig_len;
1413         enum {
1414                 NO_MAGIC,
1415                 ADD_LF_BEFORE_NON_EMPTY,
1416                 DEL_LF_BEFORE_EMPTY,
1417                 ADD_SP_BEFORE_NON_EMPTY
1418         } magic = NO_MAGIC;
1419
1420         switch (placeholder[0]) {
1421         case '-':
1422                 magic = DEL_LF_BEFORE_EMPTY;
1423                 break;
1424         case '+':
1425                 magic = ADD_LF_BEFORE_NON_EMPTY;
1426                 break;
1427         case ' ':
1428                 magic = ADD_SP_BEFORE_NON_EMPTY;
1429                 break;
1430         default:
1431                 break;
1432         }
1433         if (magic != NO_MAGIC)
1434                 placeholder++;
1435
1436         orig_len = sb->len;
1437         if (((struct format_commit_context *)context)->flush_type != no_flush)
1438                 consumed = format_and_pad_commit(sb, placeholder, context);
1439         else
1440                 consumed = format_commit_one(sb, placeholder, context);
1441         if (magic == NO_MAGIC)
1442                 return consumed;
1443
1444         if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1445                 while (sb->len && sb->buf[sb->len - 1] == '\n')
1446                         strbuf_setlen(sb, sb->len - 1);
1447         } else if (orig_len != sb->len) {
1448                 if (magic == ADD_LF_BEFORE_NON_EMPTY)
1449                         strbuf_insert(sb, orig_len, "\n", 1);
1450                 else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1451                         strbuf_insert(sb, orig_len, " ", 1);
1452         }
1453         return consumed + 1;
1454 }
1455
1456 static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1457                                    void *context)
1458 {
1459         struct userformat_want *w = context;
1460
1461         if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1462                 placeholder++;
1463
1464         switch (*placeholder) {
1465         case 'N':
1466                 w->notes = 1;
1467                 break;
1468         }
1469         return 0;
1470 }
1471
1472 void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1473 {
1474         struct strbuf dummy = STRBUF_INIT;
1475
1476         if (!fmt) {
1477                 if (!user_format)
1478                         return;
1479                 fmt = user_format;
1480         }
1481         strbuf_expand(&dummy, fmt, userformat_want_item, w);
1482         strbuf_release(&dummy);
1483 }
1484
1485 void format_commit_message(const struct commit *commit,
1486                            const char *format, struct strbuf *sb,
1487                            const struct pretty_print_context *pretty_ctx)
1488 {
1489         struct format_commit_context context;
1490         const char *output_enc = pretty_ctx->output_encoding;
1491         const char *utf8 = "UTF-8";
1492
1493         memset(&context, 0, sizeof(context));
1494         context.commit = commit;
1495         context.pretty_ctx = pretty_ctx;
1496         context.wrap_start = sb->len;
1497         /*
1498          * convert a commit message to UTF-8 first
1499          * as far as 'format_commit_item' assumes it in UTF-8
1500          */
1501         context.message = logmsg_reencode(commit,
1502                                           &context.commit_encoding,
1503                                           utf8);
1504
1505         strbuf_expand(sb, format, format_commit_item, &context);
1506         rewrap_message_tail(sb, &context, 0, 0, 0);
1507
1508         /* then convert a commit message to an actual output encoding */
1509         if (output_enc) {
1510                 if (same_encoding(utf8, output_enc))
1511                         output_enc = NULL;
1512         } else {
1513                 if (context.commit_encoding &&
1514                     !same_encoding(context.commit_encoding, utf8))
1515                         output_enc = context.commit_encoding;
1516         }
1517
1518         if (output_enc) {
1519                 int outsz;
1520                 char *out = reencode_string_len(sb->buf, sb->len,
1521                                                 output_enc, utf8, &outsz);
1522                 if (out)
1523                         strbuf_attach(sb, out, outsz, outsz + 1);
1524         }
1525
1526         free(context.commit_encoding);
1527         unuse_commit_buffer(commit, context.message);
1528 }
1529
1530 static void pp_header(struct pretty_print_context *pp,
1531                       const char *encoding,
1532                       const struct commit *commit,
1533                       const char **msg_p,
1534                       struct strbuf *sb)
1535 {
1536         int parents_shown = 0;
1537
1538         for (;;) {
1539                 const char *line = *msg_p;
1540                 int linelen = get_one_line(*msg_p);
1541
1542                 if (!linelen)
1543                         return;
1544                 *msg_p += linelen;
1545
1546                 if (linelen == 1)
1547                         /* End of header */
1548                         return;
1549
1550                 if (pp->fmt == CMIT_FMT_RAW) {
1551                         strbuf_add(sb, line, linelen);
1552                         continue;
1553                 }
1554
1555                 if (starts_with(line, "parent ")) {
1556                         if (linelen != 48)
1557                                 die("bad parent line in commit");
1558                         continue;
1559                 }
1560
1561                 if (!parents_shown) {
1562                         unsigned num = commit_list_count(commit->parents);
1563                         /* with enough slop */
1564                         strbuf_grow(sb, num * 50 + 20);
1565                         add_merge_info(pp, sb, commit);
1566                         parents_shown = 1;
1567                 }
1568
1569                 /*
1570                  * MEDIUM == DEFAULT shows only author with dates.
1571                  * FULL shows both authors but not dates.
1572                  * FULLER shows both authors and dates.
1573                  */
1574                 if (starts_with(line, "author ")) {
1575                         strbuf_grow(sb, linelen + 80);
1576                         pp_user_info(pp, "Author", sb, line + 7, encoding);
1577                 }
1578                 if (starts_with(line, "committer ") &&
1579                     (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1580                         strbuf_grow(sb, linelen + 80);
1581                         pp_user_info(pp, "Commit", sb, line + 10, encoding);
1582                 }
1583         }
1584 }
1585
1586 void pp_title_line(struct pretty_print_context *pp,
1587                    const char **msg_p,
1588                    struct strbuf *sb,
1589                    const char *encoding,
1590                    int need_8bit_cte)
1591 {
1592         static const int max_length = 78; /* per rfc2047 */
1593         struct strbuf title;
1594
1595         strbuf_init(&title, 80);
1596         *msg_p = format_subject(&title, *msg_p,
1597                                 pp->preserve_subject ? "\n" : " ");
1598
1599         strbuf_grow(sb, title.len + 1024);
1600         if (pp->subject) {
1601                 strbuf_addstr(sb, pp->subject);
1602                 if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1603                         add_rfc2047(sb, title.buf, title.len,
1604                                                 encoding, RFC2047_SUBJECT);
1605                 else
1606                         strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1607                                          -last_line_length(sb), 1, max_length);
1608         } else {
1609                 strbuf_addbuf(sb, &title);
1610         }
1611         strbuf_addch(sb, '\n');
1612
1613         if (need_8bit_cte == 0) {
1614                 int i;
1615                 for (i = 0; i < pp->in_body_headers.nr; i++) {
1616                         if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1617                                 need_8bit_cte = 1;
1618                                 break;
1619                         }
1620                 }
1621         }
1622
1623         if (need_8bit_cte > 0) {
1624                 const char *header_fmt =
1625                         "MIME-Version: 1.0\n"
1626                         "Content-Type: text/plain; charset=%s\n"
1627                         "Content-Transfer-Encoding: 8bit\n";
1628                 strbuf_addf(sb, header_fmt, encoding);
1629         }
1630         if (pp->after_subject) {
1631                 strbuf_addstr(sb, pp->after_subject);
1632         }
1633         if (pp->fmt == CMIT_FMT_EMAIL) {
1634                 strbuf_addch(sb, '\n');
1635         }
1636
1637         if (pp->in_body_headers.nr) {
1638                 int i;
1639                 for (i = 0; i < pp->in_body_headers.nr; i++) {
1640                         strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1641                         free(pp->in_body_headers.items[i].string);
1642                 }
1643                 string_list_clear(&pp->in_body_headers, 0);
1644                 strbuf_addch(sb, '\n');
1645         }
1646
1647         strbuf_release(&title);
1648 }
1649
1650 void pp_remainder(struct pretty_print_context *pp,
1651                   const char **msg_p,
1652                   struct strbuf *sb,
1653                   int indent)
1654 {
1655         int first = 1;
1656         for (;;) {
1657                 const char *line = *msg_p;
1658                 int linelen = get_one_line(line);
1659                 *msg_p += linelen;
1660
1661                 if (!linelen)
1662                         break;
1663
1664                 if (is_empty_line(line, &linelen)) {
1665                         if (first)
1666                                 continue;
1667                         if (pp->fmt == CMIT_FMT_SHORT)
1668                                 break;
1669                 }
1670                 first = 0;
1671
1672                 strbuf_grow(sb, linelen + indent + 20);
1673                 if (indent) {
1674                         memset(sb->buf + sb->len, ' ', indent);
1675                         strbuf_setlen(sb, sb->len + indent);
1676                 }
1677                 strbuf_add(sb, line, linelen);
1678                 strbuf_addch(sb, '\n');
1679         }
1680 }
1681
1682 void pretty_print_commit(struct pretty_print_context *pp,
1683                          const struct commit *commit,
1684                          struct strbuf *sb)
1685 {
1686         unsigned long beginning_of_body;
1687         int indent = 4;
1688         const char *msg;
1689         const char *reencoded;
1690         const char *encoding;
1691         int need_8bit_cte = pp->need_8bit_cte;
1692
1693         if (pp->fmt == CMIT_FMT_USERFORMAT) {
1694                 format_commit_message(commit, user_format, sb, pp);
1695                 return;
1696         }
1697
1698         encoding = get_log_output_encoding();
1699         msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1700
1701         if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1702                 indent = 0;
1703
1704         /*
1705          * We need to check and emit Content-type: to mark it
1706          * as 8-bit if we haven't done so.
1707          */
1708         if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1709                 int i, ch, in_body;
1710
1711                 for (in_body = i = 0; (ch = msg[i]); i++) {
1712                         if (!in_body) {
1713                                 /* author could be non 7-bit ASCII but
1714                                  * the log may be so; skip over the
1715                                  * header part first.
1716                                  */
1717                                 if (ch == '\n' && msg[i+1] == '\n')
1718                                         in_body = 1;
1719                         }
1720                         else if (non_ascii(ch)) {
1721                                 need_8bit_cte = 1;
1722                                 break;
1723                         }
1724                 }
1725         }
1726
1727         pp_header(pp, encoding, commit, &msg, sb);
1728         if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1729                 strbuf_addch(sb, '\n');
1730         }
1731
1732         /* Skip excess blank lines at the beginning of body, if any... */
1733         msg = skip_empty_lines(msg);
1734
1735         /* These formats treat the title line specially. */
1736         if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1737                 pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1738
1739         beginning_of_body = sb->len;
1740         if (pp->fmt != CMIT_FMT_ONELINE)
1741                 pp_remainder(pp, &msg, sb, indent);
1742         strbuf_rtrim(sb);
1743
1744         /* Make sure there is an EOLN for the non-oneline case */
1745         if (pp->fmt != CMIT_FMT_ONELINE)
1746                 strbuf_addch(sb, '\n');
1747
1748         /*
1749          * The caller may append additional body text in e-mail
1750          * format.  Make sure we did not strip the blank line
1751          * between the header and the body.
1752          */
1753         if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1754                 strbuf_addch(sb, '\n');
1755
1756         unuse_commit_buffer(commit, reencoded);
1757 }
1758
1759 void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1760                     struct strbuf *sb)
1761 {
1762         struct pretty_print_context pp = {0};
1763         pp.fmt = fmt;
1764         pretty_print_commit(&pp, commit, sb);
1765 }