pretty.con commit Merge branch 'ta/parse-commit-with-skip-prefix' (27ac2b1)
   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
  14static char *user_format;
  15static 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;
  22static size_t builtin_formats_len;
  23static size_t commit_formats_len;
  24static size_t commit_formats_alloc;
  25static struct cmt_fmt_map *find_commit_format(const char *sought);
  26
  27static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  28{
  29        free(user_format);
  30        user_format = xstrdup(cp);
  31        if (is_tformat)
  32                rev->use_terminator = 1;
  33        rev->commit_format = CMIT_FMT_USERFORMAT;
  34}
  35
  36static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  37{
  38        struct cmt_fmt_map *commit_format = NULL;
  39        const char *name;
  40        const char *fmt;
  41        int i;
  42
  43        if (!starts_with(var, "pretty."))
  44                return 0;
  45
  46        name = var + strlen("pretty.");
  47        for (i = 0; i < builtin_formats_len; i++) {
  48                if (!strcmp(commit_formats[i].name, name))
  49                        return 0;
  50        }
  51
  52        for (i = builtin_formats_len; i < commit_formats_len; i++) {
  53                if (!strcmp(commit_formats[i].name, name)) {
  54                        commit_format = &commit_formats[i];
  55                        break;
  56                }
  57        }
  58
  59        if (!commit_format) {
  60                ALLOC_GROW(commit_formats, commit_formats_len+1,
  61                           commit_formats_alloc);
  62                commit_format = &commit_formats[commit_formats_len];
  63                memset(commit_format, 0, sizeof(*commit_format));
  64                commit_formats_len++;
  65        }
  66
  67        commit_format->name = xstrdup(name);
  68        commit_format->format = CMIT_FMT_USERFORMAT;
  69        git_config_string(&fmt, var, value);
  70        if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
  71                commit_format->is_tformat = fmt[0] == 't';
  72                fmt = strchr(fmt, ':') + 1;
  73        } else if (strchr(fmt, '%'))
  74                commit_format->is_tformat = 1;
  75        else
  76                commit_format->is_alias = 1;
  77        commit_format->user_format = fmt;
  78
  79        return 0;
  80}
  81
  82static void setup_commit_formats(void)
  83{
  84        struct cmt_fmt_map builtin_formats[] = {
  85                { "raw",        CMIT_FMT_RAW,           0 },
  86                { "medium",     CMIT_FMT_MEDIUM,        0 },
  87                { "short",      CMIT_FMT_SHORT,         0 },
  88                { "email",      CMIT_FMT_EMAIL,         0 },
  89                { "fuller",     CMIT_FMT_FULLER,        0 },
  90                { "full",       CMIT_FMT_FULL,          0 },
  91                { "oneline",    CMIT_FMT_ONELINE,       1 }
  92        };
  93        commit_formats_len = ARRAY_SIZE(builtin_formats);
  94        builtin_formats_len = commit_formats_len;
  95        ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
  96        memcpy(commit_formats, builtin_formats,
  97               sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  98
  99        git_config(git_pretty_formats_config, NULL);
 100}
 101
 102static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 103                                                        const char *original,
 104                                                        int num_redirections)
 105{
 106        struct cmt_fmt_map *found = NULL;
 107        size_t found_match_len = 0;
 108        int i;
 109
 110        if (num_redirections >= commit_formats_len)
 111                die("invalid --pretty format: "
 112                    "'%s' references an alias which points to itself",
 113                    original);
 114
 115        for (i = 0; i < commit_formats_len; i++) {
 116                size_t match_len;
 117
 118                if (!starts_with(commit_formats[i].name, sought))
 119                        continue;
 120
 121                match_len = strlen(commit_formats[i].name);
 122                if (found == NULL || found_match_len > match_len) {
 123                        found = &commit_formats[i];
 124                        found_match_len = match_len;
 125                }
 126        }
 127
 128        if (found && found->is_alias) {
 129                found = find_commit_format_recursive(found->user_format,
 130                                                     original,
 131                                                     num_redirections+1);
 132        }
 133
 134        return found;
 135}
 136
 137static struct cmt_fmt_map *find_commit_format(const char *sought)
 138{
 139        if (!commit_formats)
 140                setup_commit_formats();
 141
 142        return find_commit_format_recursive(sought, sought, 0);
 143}
 144
 145void get_commit_format(const char *arg, struct rev_info *rev)
 146{
 147        struct cmt_fmt_map *commit_format;
 148
 149        rev->use_terminator = 0;
 150        if (!arg || !*arg) {
 151                rev->commit_format = CMIT_FMT_DEFAULT;
 152                return;
 153        }
 154        if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
 155                save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 156                return;
 157        }
 158
 159        if (strchr(arg, '%')) {
 160                save_user_format(rev, arg, 1);
 161                return;
 162        }
 163
 164        commit_format = find_commit_format(arg);
 165        if (!commit_format)
 166                die("invalid --pretty format: %s", arg);
 167
 168        rev->commit_format = commit_format->format;
 169        rev->use_terminator = commit_format->is_tformat;
 170        if (commit_format->format == CMIT_FMT_USERFORMAT) {
 171                save_user_format(rev, commit_format->user_format,
 172                                 commit_format->is_tformat);
 173        }
 174}
 175
 176/*
 177 * Generic support for pretty-printing the header
 178 */
 179static int get_one_line(const char *msg)
 180{
 181        int ret = 0;
 182
 183        for (;;) {
 184                char c = *msg++;
 185                if (!c)
 186                        break;
 187                ret++;
 188                if (c == '\n')
 189                        break;
 190        }
 191        return ret;
 192}
 193
 194/* High bit set, or ISO-2022-INT */
 195static int non_ascii(int ch)
 196{
 197        return !isascii(ch) || ch == '\033';
 198}
 199
 200int has_non_ascii(const char *s)
 201{
 202        int ch;
 203        if (!s)
 204                return 0;
 205        while ((ch = *s++) != '\0') {
 206                if (non_ascii(ch))
 207                        return 1;
 208        }
 209        return 0;
 210}
 211
 212static int is_rfc822_special(char ch)
 213{
 214        switch (ch) {
 215        case '(':
 216        case ')':
 217        case '<':
 218        case '>':
 219        case '[':
 220        case ']':
 221        case ':':
 222        case ';':
 223        case '@':
 224        case ',':
 225        case '.':
 226        case '"':
 227        case '\\':
 228                return 1;
 229        default:
 230                return 0;
 231        }
 232}
 233
 234static int needs_rfc822_quoting(const char *s, int len)
 235{
 236        int i;
 237        for (i = 0; i < len; i++)
 238                if (is_rfc822_special(s[i]))
 239                        return 1;
 240        return 0;
 241}
 242
 243static int last_line_length(struct strbuf *sb)
 244{
 245        int i;
 246
 247        /* How many bytes are already used on the last line? */
 248        for (i = sb->len - 1; i >= 0; i--)
 249                if (sb->buf[i] == '\n')
 250                        break;
 251        return sb->len - (i + 1);
 252}
 253
 254static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 255{
 256        int i;
 257
 258        /* just a guess, we may have to also backslash-quote */
 259        strbuf_grow(out, len + 2);
 260
 261        strbuf_addch(out, '"');
 262        for (i = 0; i < len; i++) {
 263                switch (s[i]) {
 264                case '"':
 265                case '\\':
 266                        strbuf_addch(out, '\\');
 267                        /* fall through */
 268                default:
 269                        strbuf_addch(out, s[i]);
 270                }
 271        }
 272        strbuf_addch(out, '"');
 273}
 274
 275enum rfc2047_type {
 276        RFC2047_SUBJECT,
 277        RFC2047_ADDRESS,
 278};
 279
 280static int is_rfc2047_special(char ch, enum rfc2047_type type)
 281{
 282        /*
 283         * rfc2047, section 4.2:
 284         *
 285         *    8-bit values which correspond to printable ASCII characters other
 286         *    than "=", "?", and "_" (underscore), MAY be represented as those
 287         *    characters.  (But see section 5 for restrictions.)  In
 288         *    particular, SPACE and TAB MUST NOT be represented as themselves
 289         *    within encoded words.
 290         */
 291
 292        /*
 293         * rule out non-ASCII characters and non-printable characters (the
 294         * non-ASCII check should be redundant as isprint() is not localized
 295         * and only knows about ASCII, but be defensive about that)
 296         */
 297        if (non_ascii(ch) || !isprint(ch))
 298                return 1;
 299
 300        /*
 301         * rule out special printable characters (' ' should be the only
 302         * whitespace character considered printable, but be defensive and use
 303         * isspace())
 304         */
 305        if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 306                return 1;
 307
 308        /*
 309         * rfc2047, section 5.3:
 310         *
 311         *    As a replacement for a 'word' entity within a 'phrase', for example,
 312         *    one that precedes an address in a From, To, or Cc header.  The ABNF
 313         *    definition for 'phrase' from RFC 822 thus becomes:
 314         *
 315         *    phrase = 1*( encoded-word / word )
 316         *
 317         *    In this case the set of characters that may be used in a "Q"-encoded
 318         *    'encoded-word' is restricted to: <upper and lower case ASCII
 319         *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 320         *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 321         *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 322         *    'special' by 'linear-white-space'.
 323         */
 324
 325        if (type != RFC2047_ADDRESS)
 326                return 0;
 327
 328        /* '=' and '_' are special cases and have been checked above */
 329        return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 330}
 331
 332static int needs_rfc2047_encoding(const char *line, int len,
 333                                  enum rfc2047_type type)
 334{
 335        int i;
 336
 337        for (i = 0; i < len; i++) {
 338                int ch = line[i];
 339                if (non_ascii(ch) || ch == '\n')
 340                        return 1;
 341                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 342                        return 1;
 343        }
 344
 345        return 0;
 346}
 347
 348static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 349                       const char *encoding, enum rfc2047_type type)
 350{
 351        static const int max_encoded_length = 76; /* per rfc2047 */
 352        int i;
 353        int line_len = last_line_length(sb);
 354
 355        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 356        strbuf_addf(sb, "=?%s?q?", encoding);
 357        line_len += strlen(encoding) + 5; /* 5 for =??q? */
 358
 359        while (len) {
 360                /*
 361                 * RFC 2047, section 5 (3):
 362                 *
 363                 * Each 'encoded-word' MUST represent an integral number of
 364                 * characters.  A multi-octet character may not be split across
 365                 * adjacent 'encoded- word's.
 366                 */
 367                const unsigned char *p = (const unsigned char *)line;
 368                int chrlen = mbs_chrlen(&line, &len, encoding);
 369                int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 370
 371                /* "=%02X" * chrlen, or the byte itself */
 372                const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 373                int         encoded_len = is_special ? 3 * chrlen : 1;
 374
 375                /*
 376                 * According to RFC 2047, we could encode the special character
 377                 * ' ' (space) with '_' (underscore) for readability. But many
 378                 * programs do not understand this and just leave the
 379                 * underscore in place. Thus, we do nothing special here, which
 380                 * causes ' ' to be encoded as '=20', avoiding this problem.
 381                 */
 382
 383                if (line_len + encoded_len + 2 > max_encoded_length) {
 384                        /* It won't fit with trailing "?=" --- break the line */
 385                        strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 386                        line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 387                }
 388
 389                for (i = 0; i < chrlen; i++)
 390                        strbuf_addf(sb, encoded_fmt, p[i]);
 391                line_len += encoded_len;
 392        }
 393        strbuf_addstr(sb, "?=");
 394}
 395
 396static const char *show_ident_date(const struct ident_split *ident,
 397                                   enum date_mode mode)
 398{
 399        unsigned long date = 0;
 400        long tz = 0;
 401
 402        if (ident->date_begin && ident->date_end)
 403                date = strtoul(ident->date_begin, NULL, 10);
 404        if (date_overflows(date))
 405                date = 0;
 406        else {
 407                if (ident->tz_begin && ident->tz_end)
 408                        tz = strtol(ident->tz_begin, NULL, 10);
 409                if (tz >= INT_MAX || tz <= INT_MIN)
 410                        tz = 0;
 411        }
 412        return show_date(date, tz, mode);
 413}
 414
 415void pp_user_info(struct pretty_print_context *pp,
 416                  const char *what, struct strbuf *sb,
 417                  const char *line, const char *encoding)
 418{
 419        struct ident_split ident;
 420        char *line_end;
 421        const char *mailbuf, *namebuf;
 422        size_t namelen, maillen;
 423        int max_length = 78; /* per rfc2822 */
 424
 425        if (pp->fmt == CMIT_FMT_ONELINE)
 426                return;
 427
 428        line_end = strchrnul(line, '\n');
 429        if (split_ident_line(&ident, line, line_end - line))
 430                return;
 431
 432        mailbuf = ident.mail_begin;
 433        maillen = ident.mail_end - ident.mail_begin;
 434        namebuf = ident.name_begin;
 435        namelen = ident.name_end - ident.name_begin;
 436
 437        if (pp->mailmap)
 438                map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 439
 440        if (pp->fmt == CMIT_FMT_EMAIL) {
 441                if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 442                        struct strbuf buf = STRBUF_INIT;
 443
 444                        strbuf_addstr(&buf, "From: ");
 445                        strbuf_add(&buf, namebuf, namelen);
 446                        strbuf_addstr(&buf, " <");
 447                        strbuf_add(&buf, mailbuf, maillen);
 448                        strbuf_addstr(&buf, ">\n");
 449                        string_list_append(&pp->in_body_headers,
 450                                           strbuf_detach(&buf, NULL));
 451
 452                        mailbuf = pp->from_ident->mail_begin;
 453                        maillen = pp->from_ident->mail_end - mailbuf;
 454                        namebuf = pp->from_ident->name_begin;
 455                        namelen = pp->from_ident->name_end - namebuf;
 456                }
 457
 458                strbuf_addstr(sb, "From: ");
 459                if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 460                        add_rfc2047(sb, namebuf, namelen,
 461                                    encoding, RFC2047_ADDRESS);
 462                        max_length = 76; /* per rfc2047 */
 463                } else if (needs_rfc822_quoting(namebuf, namelen)) {
 464                        struct strbuf quoted = STRBUF_INIT;
 465                        add_rfc822_quoted(&quoted, namebuf, namelen);
 466                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 467                                                        -6, 1, max_length);
 468                        strbuf_release(&quoted);
 469                } else {
 470                        strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 471                                                 -6, 1, max_length);
 472                }
 473
 474                if (max_length <
 475                    last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 476                        strbuf_addch(sb, '\n');
 477                strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 478        } else {
 479                strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 480                            (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 481                            (int)namelen, namebuf, (int)maillen, mailbuf);
 482        }
 483
 484        switch (pp->fmt) {
 485        case CMIT_FMT_MEDIUM:
 486                strbuf_addf(sb, "Date:   %s\n",
 487                            show_ident_date(&ident, pp->date_mode));
 488                break;
 489        case CMIT_FMT_EMAIL:
 490                strbuf_addf(sb, "Date: %s\n",
 491                            show_ident_date(&ident, DATE_RFC2822));
 492                break;
 493        case CMIT_FMT_FULLER:
 494                strbuf_addf(sb, "%sDate: %s\n", what,
 495                            show_ident_date(&ident, pp->date_mode));
 496                break;
 497        default:
 498                /* notin' */
 499                break;
 500        }
 501}
 502
 503static int is_empty_line(const char *line, int *len_p)
 504{
 505        int len = *len_p;
 506        while (len && isspace(line[len - 1]))
 507                len--;
 508        *len_p = len;
 509        return !len;
 510}
 511
 512static const char *skip_empty_lines(const char *msg)
 513{
 514        for (;;) {
 515                int linelen = get_one_line(msg);
 516                int ll = linelen;
 517                if (!linelen)
 518                        break;
 519                if (!is_empty_line(msg, &ll))
 520                        break;
 521                msg += linelen;
 522        }
 523        return msg;
 524}
 525
 526static void add_merge_info(const struct pretty_print_context *pp,
 527                           struct strbuf *sb, const struct commit *commit)
 528{
 529        struct commit_list *parent = commit->parents;
 530
 531        if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 532            !parent || !parent->next)
 533                return;
 534
 535        strbuf_addstr(sb, "Merge:");
 536
 537        while (parent) {
 538                struct commit *p = parent->item;
 539                const char *hex = NULL;
 540                if (pp->abbrev)
 541                        hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
 542                if (!hex)
 543                        hex = sha1_to_hex(p->object.sha1);
 544                parent = parent->next;
 545
 546                strbuf_addf(sb, " %s", hex);
 547        }
 548        strbuf_addch(sb, '\n');
 549}
 550
 551static char *get_header(const struct commit *commit, const char *msg,
 552                        const char *key)
 553{
 554        int key_len = strlen(key);
 555        const char *line = msg;
 556
 557        while (line) {
 558                const char *eol = strchr(line, '\n'), *next;
 559
 560                if (line == eol)
 561                        return NULL;
 562                if (!eol) {
 563                        warning("malformed commit (header is missing newline): %s",
 564                                sha1_to_hex(commit->object.sha1));
 565                        eol = line + strlen(line);
 566                        next = NULL;
 567                } else
 568                        next = eol + 1;
 569                if (eol - line > key_len &&
 570                    !strncmp(line, key, key_len) &&
 571                    line[key_len] == ' ') {
 572                        return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 573                }
 574                line = next;
 575        }
 576        return NULL;
 577}
 578
 579static char *replace_encoding_header(char *buf, const char *encoding)
 580{
 581        struct strbuf tmp = STRBUF_INIT;
 582        size_t start, len;
 583        char *cp = buf;
 584
 585        /* guess if there is an encoding header before a \n\n */
 586        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 587                cp = strchr(cp, '\n');
 588                if (!cp || *++cp == '\n')
 589                        return buf;
 590        }
 591        start = cp - buf;
 592        cp = strchr(cp, '\n');
 593        if (!cp)
 594                return buf; /* should not happen but be defensive */
 595        len = cp + 1 - (buf + start);
 596
 597        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 598        if (is_encoding_utf8(encoding)) {
 599                /* we have re-coded to UTF-8; drop the header */
 600                strbuf_remove(&tmp, start, len);
 601        } else {
 602                /* just replaces XXXX in 'encoding XXXX\n' */
 603                strbuf_splice(&tmp, start + strlen("encoding "),
 604                                          len - strlen("encoding \n"),
 605                                          encoding, strlen(encoding));
 606        }
 607        return strbuf_detach(&tmp, NULL);
 608}
 609
 610char *logmsg_reencode(const struct commit *commit,
 611                      char **commit_encoding,
 612                      const char *output_encoding)
 613{
 614        static const char *utf8 = "UTF-8";
 615        const char *use_encoding;
 616        char *encoding;
 617        char *msg = commit->buffer;
 618        char *out;
 619
 620        if (!msg) {
 621                enum object_type type;
 622                unsigned long size;
 623
 624                msg = read_sha1_file(commit->object.sha1, &type, &size);
 625                if (!msg)
 626                        die("Cannot read commit object %s",
 627                            sha1_to_hex(commit->object.sha1));
 628                if (type != OBJ_COMMIT)
 629                        die("Expected commit for '%s', got %s",
 630                            sha1_to_hex(commit->object.sha1), typename(type));
 631        }
 632
 633        if (!output_encoding || !*output_encoding) {
 634                if (commit_encoding)
 635                        *commit_encoding =
 636                                get_header(commit, msg, "encoding");
 637                return msg;
 638        }
 639        encoding = get_header(commit, msg, "encoding");
 640        if (commit_encoding)
 641                *commit_encoding = encoding;
 642        use_encoding = encoding ? encoding : utf8;
 643        if (same_encoding(use_encoding, output_encoding)) {
 644                /*
 645                 * No encoding work to be done. If we have no encoding header
 646                 * at all, then there's nothing to do, and we can return the
 647                 * message verbatim (whether newly allocated or not).
 648                 */
 649                if (!encoding)
 650                        return msg;
 651
 652                /*
 653                 * Otherwise, we still want to munge the encoding header in the
 654                 * result, which will be done by modifying the buffer. If we
 655                 * are using a fresh copy, we can reuse it. But if we are using
 656                 * the cached copy from commit->buffer, we need to duplicate it
 657                 * to avoid munging commit->buffer.
 658                 */
 659                out = msg;
 660                if (out == commit->buffer)
 661                        out = xstrdup(out);
 662        }
 663        else {
 664                /*
 665                 * There's actual encoding work to do. Do the reencoding, which
 666                 * still leaves the header to be replaced in the next step. At
 667                 * this point, we are done with msg. If we allocated a fresh
 668                 * copy, we can free it.
 669                 */
 670                out = reencode_string(msg, output_encoding, use_encoding);
 671                if (out && msg != commit->buffer)
 672                        free(msg);
 673        }
 674
 675        /*
 676         * This replacement actually consumes the buffer we hand it, so we do
 677         * not have to worry about freeing the old "out" here.
 678         */
 679        if (out)
 680                out = replace_encoding_header(out, output_encoding);
 681
 682        if (!commit_encoding)
 683                free(encoding);
 684        /*
 685         * If the re-encoding failed, out might be NULL here; in that
 686         * case we just return the commit message verbatim.
 687         */
 688        return out ? out : msg;
 689}
 690
 691void logmsg_free(char *msg, const struct commit *commit)
 692{
 693        if (msg != commit->buffer)
 694                free(msg);
 695}
 696
 697static int mailmap_name(const char **email, size_t *email_len,
 698                        const char **name, size_t *name_len)
 699{
 700        static struct string_list *mail_map;
 701        if (!mail_map) {
 702                mail_map = xcalloc(1, sizeof(*mail_map));
 703                read_mailmap(mail_map, NULL);
 704        }
 705        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 706}
 707
 708static size_t format_person_part(struct strbuf *sb, char part,
 709                                 const char *msg, int len, enum date_mode dmode)
 710{
 711        /* currently all placeholders have same length */
 712        const int placeholder_len = 2;
 713        struct ident_split s;
 714        const char *name, *mail;
 715        size_t maillen, namelen;
 716
 717        if (split_ident_line(&s, msg, len) < 0)
 718                goto skip;
 719
 720        name = s.name_begin;
 721        namelen = s.name_end - s.name_begin;
 722        mail = s.mail_begin;
 723        maillen = s.mail_end - s.mail_begin;
 724
 725        if (part == 'N' || part == 'E') /* mailmap lookup */
 726                mailmap_name(&mail, &maillen, &name, &namelen);
 727        if (part == 'n' || part == 'N') {       /* name */
 728                strbuf_add(sb, name, namelen);
 729                return placeholder_len;
 730        }
 731        if (part == 'e' || part == 'E') {       /* email */
 732                strbuf_add(sb, mail, maillen);
 733                return placeholder_len;
 734        }
 735
 736        if (!s.date_begin)
 737                goto skip;
 738
 739        if (part == 't') {      /* date, UNIX timestamp */
 740                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 741                return placeholder_len;
 742        }
 743
 744        switch (part) {
 745        case 'd':       /* date */
 746                strbuf_addstr(sb, show_ident_date(&s, dmode));
 747                return placeholder_len;
 748        case 'D':       /* date, RFC2822 style */
 749                strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
 750                return placeholder_len;
 751        case 'r':       /* date, relative */
 752                strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
 753                return placeholder_len;
 754        case 'i':       /* date, ISO 8601 */
 755                strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
 756                return placeholder_len;
 757        }
 758
 759skip:
 760        /*
 761         * reading from either a bogus commit, or a reflog entry with
 762         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 763         * to compute a valid return value.
 764         */
 765        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 766            || part == 'D' || part == 'r' || part == 'i')
 767                return placeholder_len;
 768
 769        return 0; /* unknown placeholder */
 770}
 771
 772struct chunk {
 773        size_t off;
 774        size_t len;
 775};
 776
 777enum flush_type {
 778        no_flush,
 779        flush_right,
 780        flush_left,
 781        flush_left_and_steal,
 782        flush_both
 783};
 784
 785enum trunc_type {
 786        trunc_none,
 787        trunc_left,
 788        trunc_middle,
 789        trunc_right
 790};
 791
 792struct format_commit_context {
 793        const struct commit *commit;
 794        const struct pretty_print_context *pretty_ctx;
 795        unsigned commit_header_parsed:1;
 796        unsigned commit_message_parsed:1;
 797        struct signature_check signature_check;
 798        enum flush_type flush_type;
 799        enum trunc_type truncate;
 800        char *message;
 801        char *commit_encoding;
 802        size_t width, indent1, indent2;
 803        int auto_color;
 804        int padding;
 805
 806        /* These offsets are relative to the start of the commit message. */
 807        struct chunk author;
 808        struct chunk committer;
 809        size_t message_off;
 810        size_t subject_off;
 811        size_t body_off;
 812
 813        /* The following ones are relative to the result struct strbuf. */
 814        struct chunk abbrev_commit_hash;
 815        struct chunk abbrev_tree_hash;
 816        struct chunk abbrev_parent_hashes;
 817        size_t wrap_start;
 818};
 819
 820static int add_again(struct strbuf *sb, struct chunk *chunk)
 821{
 822        if (chunk->len) {
 823                strbuf_adddup(sb, chunk->off, chunk->len);
 824                return 1;
 825        }
 826
 827        /*
 828         * We haven't seen this chunk before.  Our caller is surely
 829         * going to add it the hard way now.  Remember the most likely
 830         * start of the to-be-added chunk: the current end of the
 831         * struct strbuf.
 832         */
 833        chunk->off = sb->len;
 834        return 0;
 835}
 836
 837static void parse_commit_header(struct format_commit_context *context)
 838{
 839        const char *msg = context->message;
 840        int i;
 841
 842        for (i = 0; msg[i]; i++) {
 843                int eol;
 844                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 845                        ; /* do nothing */
 846
 847                if (i == eol) {
 848                        break;
 849                } else if (starts_with(msg + i, "author ")) {
 850                        context->author.off = i + 7;
 851                        context->author.len = eol - i - 7;
 852                } else if (starts_with(msg + i, "committer ")) {
 853                        context->committer.off = i + 10;
 854                        context->committer.len = eol - i - 10;
 855                }
 856                i = eol;
 857        }
 858        context->message_off = i;
 859        context->commit_header_parsed = 1;
 860}
 861
 862static int istitlechar(char c)
 863{
 864        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 865                (c >= '0' && c <= '9') || c == '.' || c == '_';
 866}
 867
 868static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 869{
 870        size_t trimlen;
 871        size_t start_len = sb->len;
 872        int space = 2;
 873
 874        for (; *msg && *msg != '\n'; msg++) {
 875                if (istitlechar(*msg)) {
 876                        if (space == 1)
 877                                strbuf_addch(sb, '-');
 878                        space = 0;
 879                        strbuf_addch(sb, *msg);
 880                        if (*msg == '.')
 881                                while (*(msg+1) == '.')
 882                                        msg++;
 883                } else
 884                        space |= 1;
 885        }
 886
 887        /* trim any trailing '.' or '-' characters */
 888        trimlen = 0;
 889        while (sb->len - trimlen > start_len &&
 890                (sb->buf[sb->len - 1 - trimlen] == '.'
 891                || sb->buf[sb->len - 1 - trimlen] == '-'))
 892                trimlen++;
 893        strbuf_remove(sb, sb->len - trimlen, trimlen);
 894}
 895
 896const char *format_subject(struct strbuf *sb, const char *msg,
 897                           const char *line_separator)
 898{
 899        int first = 1;
 900
 901        for (;;) {
 902                const char *line = msg;
 903                int linelen = get_one_line(line);
 904
 905                msg += linelen;
 906                if (!linelen || is_empty_line(line, &linelen))
 907                        break;
 908
 909                if (!sb)
 910                        continue;
 911                strbuf_grow(sb, linelen + 2);
 912                if (!first)
 913                        strbuf_addstr(sb, line_separator);
 914                strbuf_add(sb, line, linelen);
 915                first = 0;
 916        }
 917        return msg;
 918}
 919
 920static void parse_commit_message(struct format_commit_context *c)
 921{
 922        const char *msg = c->message + c->message_off;
 923        const char *start = c->message;
 924
 925        msg = skip_empty_lines(msg);
 926        c->subject_off = msg - start;
 927
 928        msg = format_subject(NULL, msg, NULL);
 929        msg = skip_empty_lines(msg);
 930        c->body_off = msg - start;
 931
 932        c->commit_message_parsed = 1;
 933}
 934
 935static void strbuf_wrap(struct strbuf *sb, size_t pos,
 936                        size_t width, size_t indent1, size_t indent2)
 937{
 938        struct strbuf tmp = STRBUF_INIT;
 939
 940        if (pos)
 941                strbuf_add(&tmp, sb->buf, pos);
 942        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 943                                (int) indent1, (int) indent2, (int) width);
 944        strbuf_swap(&tmp, sb);
 945        strbuf_release(&tmp);
 946}
 947
 948static void rewrap_message_tail(struct strbuf *sb,
 949                                struct format_commit_context *c,
 950                                size_t new_width, size_t new_indent1,
 951                                size_t new_indent2)
 952{
 953        if (c->width == new_width && c->indent1 == new_indent1 &&
 954            c->indent2 == new_indent2)
 955                return;
 956        if (c->wrap_start < sb->len)
 957                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 958        c->wrap_start = sb->len;
 959        c->width = new_width;
 960        c->indent1 = new_indent1;
 961        c->indent2 = new_indent2;
 962}
 963
 964static int format_reflog_person(struct strbuf *sb,
 965                                char part,
 966                                struct reflog_walk_info *log,
 967                                enum date_mode dmode)
 968{
 969        const char *ident;
 970
 971        if (!log)
 972                return 2;
 973
 974        ident = get_reflog_ident(log);
 975        if (!ident)
 976                return 2;
 977
 978        return format_person_part(sb, part, ident, strlen(ident), dmode);
 979}
 980
 981static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 982                          const char *placeholder,
 983                          struct format_commit_context *c)
 984{
 985        if (placeholder[1] == '(') {
 986                const char *begin = placeholder + 2;
 987                const char *end = strchr(begin, ')');
 988                char color[COLOR_MAXLEN];
 989
 990                if (!end)
 991                        return 0;
 992                if (starts_with(begin, "auto,")) {
 993                        if (!want_color(c->pretty_ctx->color))
 994                                return end - placeholder + 1;
 995                        begin += 5;
 996                }
 997                color_parse_mem(begin,
 998                                end - begin,
 999                                "--pretty format", color);
1000                strbuf_addstr(sb, color);
1001                return end - placeholder + 1;
1002        }
1003        if (starts_with(placeholder + 1, "red")) {
1004                strbuf_addstr(sb, GIT_COLOR_RED);
1005                return 4;
1006        } else if (starts_with(placeholder + 1, "green")) {
1007                strbuf_addstr(sb, GIT_COLOR_GREEN);
1008                return 6;
1009        } else if (starts_with(placeholder + 1, "blue")) {
1010                strbuf_addstr(sb, GIT_COLOR_BLUE);
1011                return 5;
1012        } else if (starts_with(placeholder + 1, "reset")) {
1013                strbuf_addstr(sb, GIT_COLOR_RESET);
1014                return 6;
1015        } else
1016                return 0;
1017}
1018
1019static size_t parse_padding_placeholder(struct strbuf *sb,
1020                                        const char *placeholder,
1021                                        struct format_commit_context *c)
1022{
1023        const char *ch = placeholder;
1024        enum flush_type flush_type;
1025        int to_column = 0;
1026
1027        switch (*ch++) {
1028        case '<':
1029                flush_type = flush_right;
1030                break;
1031        case '>':
1032                if (*ch == '<') {
1033                        flush_type = flush_both;
1034                        ch++;
1035                } else if (*ch == '>') {
1036                        flush_type = flush_left_and_steal;
1037                        ch++;
1038                } else
1039                        flush_type = flush_left;
1040                break;
1041        default:
1042                return 0;
1043        }
1044
1045        /* the next value means "wide enough to that column" */
1046        if (*ch == '|') {
1047                to_column = 1;
1048                ch++;
1049        }
1050
1051        if (*ch == '(') {
1052                const char *start = ch + 1;
1053                const char *end = start + strcspn(start, ",)");
1054                char *next;
1055                int width;
1056                if (!end || end == start)
1057                        return 0;
1058                width = strtoul(start, &next, 10);
1059                if (next == start || width == 0)
1060                        return 0;
1061                c->padding = to_column ? -width : width;
1062                c->flush_type = flush_type;
1063
1064                if (*end == ',') {
1065                        start = end + 1;
1066                        end = strchr(start, ')');
1067                        if (!end || end == start)
1068                                return 0;
1069                        if (starts_with(start, "trunc)"))
1070                                c->truncate = trunc_right;
1071                        else if (starts_with(start, "ltrunc)"))
1072                                c->truncate = trunc_left;
1073                        else if (starts_with(start, "mtrunc)"))
1074                                c->truncate = trunc_middle;
1075                        else
1076                                return 0;
1077                } else
1078                        c->truncate = trunc_none;
1079
1080                return end - placeholder + 1;
1081        }
1082        return 0;
1083}
1084
1085static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1086                                const char *placeholder,
1087                                void *context)
1088{
1089        struct format_commit_context *c = context;
1090        const struct commit *commit = c->commit;
1091        const char *msg = c->message;
1092        struct commit_list *p;
1093        int h1, h2;
1094
1095        /* these are independent of the commit */
1096        switch (placeholder[0]) {
1097        case 'C':
1098                if (starts_with(placeholder + 1, "(auto)")) {
1099                        c->auto_color = 1;
1100                        return 7; /* consumed 7 bytes, "C(auto)" */
1101                } else {
1102                        int ret = parse_color(sb, placeholder, c);
1103                        if (ret)
1104                                c->auto_color = 0;
1105                        /*
1106                         * Otherwise, we decided to treat %C<unknown>
1107                         * as a literal string, and the previous
1108                         * %C(auto) is still valid.
1109                         */
1110                        return ret;
1111                }
1112        case 'n':               /* newline */
1113                strbuf_addch(sb, '\n');
1114                return 1;
1115        case 'x':
1116                /* %x00 == NUL, %x0a == LF, etc. */
1117                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1118                    h1 <= 16 &&
1119                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1120                    h2 <= 16) {
1121                        strbuf_addch(sb, (h1<<4)|h2);
1122                        return 3;
1123                } else
1124                        return 0;
1125        case 'w':
1126                if (placeholder[1] == '(') {
1127                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1128                        char *next;
1129                        const char *start = placeholder + 2;
1130                        const char *end = strchr(start, ')');
1131                        if (!end)
1132                                return 0;
1133                        if (end > start) {
1134                                width = strtoul(start, &next, 10);
1135                                if (*next == ',') {
1136                                        indent1 = strtoul(next + 1, &next, 10);
1137                                        if (*next == ',') {
1138                                                indent2 = strtoul(next + 1,
1139                                                                 &next, 10);
1140                                        }
1141                                }
1142                                if (*next != ')')
1143                                        return 0;
1144                        }
1145                        rewrap_message_tail(sb, c, width, indent1, indent2);
1146                        return end - placeholder + 1;
1147                } else
1148                        return 0;
1149
1150        case '<':
1151        case '>':
1152                return parse_padding_placeholder(sb, placeholder, c);
1153        }
1154
1155        /* these depend on the commit */
1156        if (!commit->object.parsed)
1157                parse_object(commit->object.sha1);
1158
1159        switch (placeholder[0]) {
1160        case 'H':               /* commit hash */
1161                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1162                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1163                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1164                return 1;
1165        case 'h':               /* abbreviated commit hash */
1166                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1167                if (add_again(sb, &c->abbrev_commit_hash)) {
1168                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1169                        return 1;
1170                }
1171                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1172                                                     c->pretty_ctx->abbrev));
1173                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1174                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1175                return 1;
1176        case 'T':               /* tree hash */
1177                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1178                return 1;
1179        case 't':               /* abbreviated tree hash */
1180                if (add_again(sb, &c->abbrev_tree_hash))
1181                        return 1;
1182                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1183                                                     c->pretty_ctx->abbrev));
1184                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1185                return 1;
1186        case 'P':               /* parent hashes */
1187                for (p = commit->parents; p; p = p->next) {
1188                        if (p != commit->parents)
1189                                strbuf_addch(sb, ' ');
1190                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1191                }
1192                return 1;
1193        case 'p':               /* abbreviated parent hashes */
1194                if (add_again(sb, &c->abbrev_parent_hashes))
1195                        return 1;
1196                for (p = commit->parents; p; p = p->next) {
1197                        if (p != commit->parents)
1198                                strbuf_addch(sb, ' ');
1199                        strbuf_addstr(sb, find_unique_abbrev(
1200                                        p->item->object.sha1,
1201                                        c->pretty_ctx->abbrev));
1202                }
1203                c->abbrev_parent_hashes.len = sb->len -
1204                                              c->abbrev_parent_hashes.off;
1205                return 1;
1206        case 'm':               /* left/right/bottom */
1207                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1208                return 1;
1209        case 'd':
1210                load_ref_decorations(DECORATE_SHORT_REFS);
1211                format_decorations(sb, commit, c->auto_color);
1212                return 1;
1213        case 'g':               /* reflog info */
1214                switch(placeholder[1]) {
1215                case 'd':       /* reflog selector */
1216                case 'D':
1217                        if (c->pretty_ctx->reflog_info)
1218                                get_reflog_selector(sb,
1219                                                    c->pretty_ctx->reflog_info,
1220                                                    c->pretty_ctx->date_mode,
1221                                                    c->pretty_ctx->date_mode_explicit,
1222                                                    (placeholder[1] == 'd'));
1223                        return 2;
1224                case 's':       /* reflog message */
1225                        if (c->pretty_ctx->reflog_info)
1226                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1227                        return 2;
1228                case 'n':
1229                case 'N':
1230                case 'e':
1231                case 'E':
1232                        return format_reflog_person(sb,
1233                                                    placeholder[1],
1234                                                    c->pretty_ctx->reflog_info,
1235                                                    c->pretty_ctx->date_mode);
1236                }
1237                return 0;       /* unknown %g placeholder */
1238        case 'N':
1239                if (c->pretty_ctx->notes_message) {
1240                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1241                        return 1;
1242                }
1243                return 0;
1244        }
1245
1246        if (placeholder[0] == 'G') {
1247                if (!c->signature_check.result)
1248                        check_commit_signature(c->commit, &(c->signature_check));
1249                switch (placeholder[1]) {
1250                case 'G':
1251                        if (c->signature_check.gpg_output)
1252                                strbuf_addstr(sb, c->signature_check.gpg_output);
1253                        break;
1254                case '?':
1255                        switch (c->signature_check.result) {
1256                        case 'G':
1257                        case 'B':
1258                        case 'U':
1259                        case 'N':
1260                                strbuf_addch(sb, c->signature_check.result);
1261                        }
1262                        break;
1263                case 'S':
1264                        if (c->signature_check.signer)
1265                                strbuf_addstr(sb, c->signature_check.signer);
1266                        break;
1267                case 'K':
1268                        if (c->signature_check.key)
1269                                strbuf_addstr(sb, c->signature_check.key);
1270                        break;
1271                }
1272                return 2;
1273        }
1274
1275
1276        /* For the rest we have to parse the commit header. */
1277        if (!c->commit_header_parsed)
1278                parse_commit_header(c);
1279
1280        switch (placeholder[0]) {
1281        case 'a':       /* author ... */
1282                return format_person_part(sb, placeholder[1],
1283                                   msg + c->author.off, c->author.len,
1284                                   c->pretty_ctx->date_mode);
1285        case 'c':       /* committer ... */
1286                return format_person_part(sb, placeholder[1],
1287                                   msg + c->committer.off, c->committer.len,
1288                                   c->pretty_ctx->date_mode);
1289        case 'e':       /* encoding */
1290                if (c->commit_encoding)
1291                        strbuf_addstr(sb, c->commit_encoding);
1292                return 1;
1293        case 'B':       /* raw body */
1294                /* message_off is always left at the initial newline */
1295                strbuf_addstr(sb, msg + c->message_off + 1);
1296                return 1;
1297        }
1298
1299        /* Now we need to parse the commit message. */
1300        if (!c->commit_message_parsed)
1301                parse_commit_message(c);
1302
1303        switch (placeholder[0]) {
1304        case 's':       /* subject */
1305                format_subject(sb, msg + c->subject_off, " ");
1306                return 1;
1307        case 'f':       /* sanitized subject */
1308                format_sanitized_subject(sb, msg + c->subject_off);
1309                return 1;
1310        case 'b':       /* body */
1311                strbuf_addstr(sb, msg + c->body_off);
1312                return 1;
1313        }
1314        return 0;       /* unknown placeholder */
1315}
1316
1317static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1318                                    const char *placeholder,
1319                                    struct format_commit_context *c)
1320{
1321        struct strbuf local_sb = STRBUF_INIT;
1322        int total_consumed = 0, len, padding = c->padding;
1323        if (padding < 0) {
1324                const char *start = strrchr(sb->buf, '\n');
1325                int occupied;
1326                if (!start)
1327                        start = sb->buf;
1328                occupied = utf8_strnwidth(start, -1, 1);
1329                padding = (-padding) - occupied;
1330        }
1331        while (1) {
1332                int modifier = *placeholder == 'C';
1333                int consumed = format_commit_one(&local_sb, placeholder, c);
1334                total_consumed += consumed;
1335
1336                if (!modifier)
1337                        break;
1338
1339                placeholder += consumed;
1340                if (*placeholder != '%')
1341                        break;
1342                placeholder++;
1343                total_consumed++;
1344        }
1345        len = utf8_strnwidth(local_sb.buf, -1, 1);
1346
1347        if (c->flush_type == flush_left_and_steal) {
1348                const char *ch = sb->buf + sb->len - 1;
1349                while (len > padding && ch > sb->buf) {
1350                        const char *p;
1351                        if (*ch == ' ') {
1352                                ch--;
1353                                padding++;
1354                                continue;
1355                        }
1356                        /* check for trailing ansi sequences */
1357                        if (*ch != 'm')
1358                                break;
1359                        p = ch - 1;
1360                        while (ch - p < 10 && *p != '\033')
1361                                p--;
1362                        if (*p != '\033' ||
1363                            ch + 1 - p != display_mode_esc_sequence_len(p))
1364                                break;
1365                        /*
1366                         * got a good ansi sequence, put it back to
1367                         * local_sb as we're cutting sb
1368                         */
1369                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1370                        ch = p - 1;
1371                }
1372                strbuf_setlen(sb, ch + 1 - sb->buf);
1373                c->flush_type = flush_left;
1374        }
1375
1376        if (len > padding) {
1377                switch (c->truncate) {
1378                case trunc_left:
1379                        strbuf_utf8_replace(&local_sb,
1380                                            0, len - (padding - 2),
1381                                            "..");
1382                        break;
1383                case trunc_middle:
1384                        strbuf_utf8_replace(&local_sb,
1385                                            padding / 2 - 1,
1386                                            len - (padding - 2),
1387                                            "..");
1388                        break;
1389                case trunc_right:
1390                        strbuf_utf8_replace(&local_sb,
1391                                            padding - 2, len - (padding - 2),
1392                                            "..");
1393                        break;
1394                case trunc_none:
1395                        break;
1396                }
1397                strbuf_addstr(sb, local_sb.buf);
1398        } else {
1399                int sb_len = sb->len, offset = 0;
1400                if (c->flush_type == flush_left)
1401                        offset = padding - len;
1402                else if (c->flush_type == flush_both)
1403                        offset = (padding - len) / 2;
1404                /*
1405                 * we calculate padding in columns, now
1406                 * convert it back to chars
1407                 */
1408                padding = padding - len + local_sb.len;
1409                strbuf_grow(sb, padding);
1410                strbuf_setlen(sb, sb_len + padding);
1411                memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1412                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1413                       local_sb.len);
1414        }
1415        strbuf_release(&local_sb);
1416        c->flush_type = no_flush;
1417        return total_consumed;
1418}
1419
1420static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1421                                 const char *placeholder,
1422                                 void *context)
1423{
1424        int consumed;
1425        size_t orig_len;
1426        enum {
1427                NO_MAGIC,
1428                ADD_LF_BEFORE_NON_EMPTY,
1429                DEL_LF_BEFORE_EMPTY,
1430                ADD_SP_BEFORE_NON_EMPTY
1431        } magic = NO_MAGIC;
1432
1433        switch (placeholder[0]) {
1434        case '-':
1435                magic = DEL_LF_BEFORE_EMPTY;
1436                break;
1437        case '+':
1438                magic = ADD_LF_BEFORE_NON_EMPTY;
1439                break;
1440        case ' ':
1441                magic = ADD_SP_BEFORE_NON_EMPTY;
1442                break;
1443        default:
1444                break;
1445        }
1446        if (magic != NO_MAGIC)
1447                placeholder++;
1448
1449        orig_len = sb->len;
1450        if (((struct format_commit_context *)context)->flush_type != no_flush)
1451                consumed = format_and_pad_commit(sb, placeholder, context);
1452        else
1453                consumed = format_commit_one(sb, placeholder, context);
1454        if (magic == NO_MAGIC)
1455                return consumed;
1456
1457        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1458                while (sb->len && sb->buf[sb->len - 1] == '\n')
1459                        strbuf_setlen(sb, sb->len - 1);
1460        } else if (orig_len != sb->len) {
1461                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1462                        strbuf_insert(sb, orig_len, "\n", 1);
1463                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1464                        strbuf_insert(sb, orig_len, " ", 1);
1465        }
1466        return consumed + 1;
1467}
1468
1469static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1470                                   void *context)
1471{
1472        struct userformat_want *w = context;
1473
1474        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1475                placeholder++;
1476
1477        switch (*placeholder) {
1478        case 'N':
1479                w->notes = 1;
1480                break;
1481        }
1482        return 0;
1483}
1484
1485void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1486{
1487        struct strbuf dummy = STRBUF_INIT;
1488
1489        if (!fmt) {
1490                if (!user_format)
1491                        return;
1492                fmt = user_format;
1493        }
1494        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1495        strbuf_release(&dummy);
1496}
1497
1498void format_commit_message(const struct commit *commit,
1499                           const char *format, struct strbuf *sb,
1500                           const struct pretty_print_context *pretty_ctx)
1501{
1502        struct format_commit_context context;
1503        const char *output_enc = pretty_ctx->output_encoding;
1504        const char *utf8 = "UTF-8";
1505
1506        memset(&context, 0, sizeof(context));
1507        context.commit = commit;
1508        context.pretty_ctx = pretty_ctx;
1509        context.wrap_start = sb->len;
1510        context.message = logmsg_reencode(commit,
1511                                          &context.commit_encoding,
1512                                          output_enc);
1513
1514        strbuf_expand(sb, format, format_commit_item, &context);
1515        rewrap_message_tail(sb, &context, 0, 0, 0);
1516
1517        if (output_enc) {
1518                if (same_encoding(utf8, output_enc))
1519                        output_enc = NULL;
1520        } else {
1521                if (context.commit_encoding &&
1522                    !same_encoding(context.commit_encoding, utf8))
1523                        output_enc = context.commit_encoding;
1524        }
1525
1526        if (output_enc) {
1527                int outsz;
1528                char *out = reencode_string_len(sb->buf, sb->len,
1529                                                output_enc, utf8, &outsz);
1530                if (out)
1531                        strbuf_attach(sb, out, outsz, outsz + 1);
1532        }
1533
1534        free(context.commit_encoding);
1535        logmsg_free(context.message, commit);
1536        free(context.signature_check.gpg_output);
1537        free(context.signature_check.signer);
1538}
1539
1540static void pp_header(struct pretty_print_context *pp,
1541                      const char *encoding,
1542                      const struct commit *commit,
1543                      const char **msg_p,
1544                      struct strbuf *sb)
1545{
1546        int parents_shown = 0;
1547
1548        for (;;) {
1549                const char *line = *msg_p;
1550                int linelen = get_one_line(*msg_p);
1551
1552                if (!linelen)
1553                        return;
1554                *msg_p += linelen;
1555
1556                if (linelen == 1)
1557                        /* End of header */
1558                        return;
1559
1560                if (pp->fmt == CMIT_FMT_RAW) {
1561                        strbuf_add(sb, line, linelen);
1562                        continue;
1563                }
1564
1565                if (starts_with(line, "parent ")) {
1566                        if (linelen != 48)
1567                                die("bad parent line in commit");
1568                        continue;
1569                }
1570
1571                if (!parents_shown) {
1572                        struct commit_list *parent;
1573                        int num;
1574                        for (parent = commit->parents, num = 0;
1575                             parent;
1576                             parent = parent->next, num++)
1577                                ;
1578                        /* with enough slop */
1579                        strbuf_grow(sb, num * 50 + 20);
1580                        add_merge_info(pp, sb, commit);
1581                        parents_shown = 1;
1582                }
1583
1584                /*
1585                 * MEDIUM == DEFAULT shows only author with dates.
1586                 * FULL shows both authors but not dates.
1587                 * FULLER shows both authors and dates.
1588                 */
1589                if (starts_with(line, "author ")) {
1590                        strbuf_grow(sb, linelen + 80);
1591                        pp_user_info(pp, "Author", sb, line + 7, encoding);
1592                }
1593                if (starts_with(line, "committer ") &&
1594                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1595                        strbuf_grow(sb, linelen + 80);
1596                        pp_user_info(pp, "Commit", sb, line + 10, encoding);
1597                }
1598        }
1599}
1600
1601void pp_title_line(struct pretty_print_context *pp,
1602                   const char **msg_p,
1603                   struct strbuf *sb,
1604                   const char *encoding,
1605                   int need_8bit_cte)
1606{
1607        static const int max_length = 78; /* per rfc2047 */
1608        struct strbuf title;
1609
1610        strbuf_init(&title, 80);
1611        *msg_p = format_subject(&title, *msg_p,
1612                                pp->preserve_subject ? "\n" : " ");
1613
1614        strbuf_grow(sb, title.len + 1024);
1615        if (pp->subject) {
1616                strbuf_addstr(sb, pp->subject);
1617                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1618                        add_rfc2047(sb, title.buf, title.len,
1619                                                encoding, RFC2047_SUBJECT);
1620                else
1621                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1622                                         -last_line_length(sb), 1, max_length);
1623        } else {
1624                strbuf_addbuf(sb, &title);
1625        }
1626        strbuf_addch(sb, '\n');
1627
1628        if (need_8bit_cte == 0) {
1629                int i;
1630                for (i = 0; i < pp->in_body_headers.nr; i++) {
1631                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1632                                need_8bit_cte = 1;
1633                                break;
1634                        }
1635                }
1636        }
1637
1638        if (need_8bit_cte > 0) {
1639                const char *header_fmt =
1640                        "MIME-Version: 1.0\n"
1641                        "Content-Type: text/plain; charset=%s\n"
1642                        "Content-Transfer-Encoding: 8bit\n";
1643                strbuf_addf(sb, header_fmt, encoding);
1644        }
1645        if (pp->after_subject) {
1646                strbuf_addstr(sb, pp->after_subject);
1647        }
1648        if (pp->fmt == CMIT_FMT_EMAIL) {
1649                strbuf_addch(sb, '\n');
1650        }
1651
1652        if (pp->in_body_headers.nr) {
1653                int i;
1654                for (i = 0; i < pp->in_body_headers.nr; i++) {
1655                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1656                        free(pp->in_body_headers.items[i].string);
1657                }
1658                string_list_clear(&pp->in_body_headers, 0);
1659                strbuf_addch(sb, '\n');
1660        }
1661
1662        strbuf_release(&title);
1663}
1664
1665void pp_remainder(struct pretty_print_context *pp,
1666                  const char **msg_p,
1667                  struct strbuf *sb,
1668                  int indent)
1669{
1670        int first = 1;
1671        for (;;) {
1672                const char *line = *msg_p;
1673                int linelen = get_one_line(line);
1674                *msg_p += linelen;
1675
1676                if (!linelen)
1677                        break;
1678
1679                if (is_empty_line(line, &linelen)) {
1680                        if (first)
1681                                continue;
1682                        if (pp->fmt == CMIT_FMT_SHORT)
1683                                break;
1684                }
1685                first = 0;
1686
1687                strbuf_grow(sb, linelen + indent + 20);
1688                if (indent) {
1689                        memset(sb->buf + sb->len, ' ', indent);
1690                        strbuf_setlen(sb, sb->len + indent);
1691                }
1692                strbuf_add(sb, line, linelen);
1693                strbuf_addch(sb, '\n');
1694        }
1695}
1696
1697void pretty_print_commit(struct pretty_print_context *pp,
1698                         const struct commit *commit,
1699                         struct strbuf *sb)
1700{
1701        unsigned long beginning_of_body;
1702        int indent = 4;
1703        const char *msg;
1704        char *reencoded;
1705        const char *encoding;
1706        int need_8bit_cte = pp->need_8bit_cte;
1707
1708        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1709                format_commit_message(commit, user_format, sb, pp);
1710                return;
1711        }
1712
1713        encoding = get_log_output_encoding();
1714        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1715
1716        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1717                indent = 0;
1718
1719        /*
1720         * We need to check and emit Content-type: to mark it
1721         * as 8-bit if we haven't done so.
1722         */
1723        if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1724                int i, ch, in_body;
1725
1726                for (in_body = i = 0; (ch = msg[i]); i++) {
1727                        if (!in_body) {
1728                                /* author could be non 7-bit ASCII but
1729                                 * the log may be so; skip over the
1730                                 * header part first.
1731                                 */
1732                                if (ch == '\n' && msg[i+1] == '\n')
1733                                        in_body = 1;
1734                        }
1735                        else if (non_ascii(ch)) {
1736                                need_8bit_cte = 1;
1737                                break;
1738                        }
1739                }
1740        }
1741
1742        pp_header(pp, encoding, commit, &msg, sb);
1743        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1744                strbuf_addch(sb, '\n');
1745        }
1746
1747        /* Skip excess blank lines at the beginning of body, if any... */
1748        msg = skip_empty_lines(msg);
1749
1750        /* These formats treat the title line specially. */
1751        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1752                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1753
1754        beginning_of_body = sb->len;
1755        if (pp->fmt != CMIT_FMT_ONELINE)
1756                pp_remainder(pp, &msg, sb, indent);
1757        strbuf_rtrim(sb);
1758
1759        /* Make sure there is an EOLN for the non-oneline case */
1760        if (pp->fmt != CMIT_FMT_ONELINE)
1761                strbuf_addch(sb, '\n');
1762
1763        /*
1764         * The caller may append additional body text in e-mail
1765         * format.  Make sure we did not strip the blank line
1766         * between the header and the body.
1767         */
1768        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1769                strbuf_addch(sb, '\n');
1770
1771        logmsg_free(reencoded, commit);
1772}
1773
1774void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1775                    struct strbuf *sb)
1776{
1777        struct pretty_print_context pp = {0};
1778        pp.fmt = fmt;
1779        pretty_print_commit(&pp, commit, sb);
1780}