pretty.con commit t5537: fix incorrect expectation in test case 10 (3b32a7c)
   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 (prefixcmp(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 (!prefixcmp(fmt, "format:") || !prefixcmp(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 (prefixcmp(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 (!prefixcmp(arg, "format:") || !prefixcmp(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        int tz = 0;
 401
 402        if (ident->date_begin && ident->date_end)
 403                date = strtoul(ident->date_begin, NULL, 10);
 404        if (ident->tz_begin && ident->tz_end)
 405                tz = strtol(ident->tz_begin, NULL, 10);
 406        return show_date(date, tz, mode);
 407}
 408
 409void pp_user_info(struct pretty_print_context *pp,
 410                  const char *what, struct strbuf *sb,
 411                  const char *line, const char *encoding)
 412{
 413        struct ident_split ident;
 414        char *line_end;
 415        const char *mailbuf, *namebuf;
 416        size_t namelen, maillen;
 417        int max_length = 78; /* per rfc2822 */
 418
 419        if (pp->fmt == CMIT_FMT_ONELINE)
 420                return;
 421
 422        line_end = strchrnul(line, '\n');
 423        if (split_ident_line(&ident, line, line_end - line))
 424                return;
 425
 426        mailbuf = ident.mail_begin;
 427        maillen = ident.mail_end - ident.mail_begin;
 428        namebuf = ident.name_begin;
 429        namelen = ident.name_end - ident.name_begin;
 430
 431        if (pp->mailmap)
 432                map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 433
 434        if (pp->fmt == CMIT_FMT_EMAIL) {
 435                if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 436                        struct strbuf buf = STRBUF_INIT;
 437
 438                        strbuf_addstr(&buf, "From: ");
 439                        strbuf_add(&buf, namebuf, namelen);
 440                        strbuf_addstr(&buf, " <");
 441                        strbuf_add(&buf, mailbuf, maillen);
 442                        strbuf_addstr(&buf, ">\n");
 443                        string_list_append(&pp->in_body_headers,
 444                                           strbuf_detach(&buf, NULL));
 445
 446                        mailbuf = pp->from_ident->mail_begin;
 447                        maillen = pp->from_ident->mail_end - mailbuf;
 448                        namebuf = pp->from_ident->name_begin;
 449                        namelen = pp->from_ident->name_end - namebuf;
 450                }
 451
 452                strbuf_addstr(sb, "From: ");
 453                if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 454                        add_rfc2047(sb, namebuf, namelen,
 455                                    encoding, RFC2047_ADDRESS);
 456                        max_length = 76; /* per rfc2047 */
 457                } else if (needs_rfc822_quoting(namebuf, namelen)) {
 458                        struct strbuf quoted = STRBUF_INIT;
 459                        add_rfc822_quoted(&quoted, namebuf, namelen);
 460                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 461                                                        -6, 1, max_length);
 462                        strbuf_release(&quoted);
 463                } else {
 464                        strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 465                                                 -6, 1, max_length);
 466                }
 467
 468                if (max_length <
 469                    last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 470                        strbuf_addch(sb, '\n');
 471                strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 472        } else {
 473                strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 474                            (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 475                            (int)namelen, namebuf, (int)maillen, mailbuf);
 476        }
 477
 478        switch (pp->fmt) {
 479        case CMIT_FMT_MEDIUM:
 480                strbuf_addf(sb, "Date:   %s\n",
 481                            show_ident_date(&ident, pp->date_mode));
 482                break;
 483        case CMIT_FMT_EMAIL:
 484                strbuf_addf(sb, "Date: %s\n",
 485                            show_ident_date(&ident, DATE_RFC2822));
 486                break;
 487        case CMIT_FMT_FULLER:
 488                strbuf_addf(sb, "%sDate: %s\n", what,
 489                            show_ident_date(&ident, pp->date_mode));
 490                break;
 491        default:
 492                /* notin' */
 493                break;
 494        }
 495}
 496
 497static int is_empty_line(const char *line, int *len_p)
 498{
 499        int len = *len_p;
 500        while (len && isspace(line[len - 1]))
 501                len--;
 502        *len_p = len;
 503        return !len;
 504}
 505
 506static const char *skip_empty_lines(const char *msg)
 507{
 508        for (;;) {
 509                int linelen = get_one_line(msg);
 510                int ll = linelen;
 511                if (!linelen)
 512                        break;
 513                if (!is_empty_line(msg, &ll))
 514                        break;
 515                msg += linelen;
 516        }
 517        return msg;
 518}
 519
 520static void add_merge_info(const struct pretty_print_context *pp,
 521                           struct strbuf *sb, const struct commit *commit)
 522{
 523        struct commit_list *parent = commit->parents;
 524
 525        if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 526            !parent || !parent->next)
 527                return;
 528
 529        strbuf_addstr(sb, "Merge:");
 530
 531        while (parent) {
 532                struct commit *p = parent->item;
 533                const char *hex = NULL;
 534                if (pp->abbrev)
 535                        hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
 536                if (!hex)
 537                        hex = sha1_to_hex(p->object.sha1);
 538                parent = parent->next;
 539
 540                strbuf_addf(sb, " %s", hex);
 541        }
 542        strbuf_addch(sb, '\n');
 543}
 544
 545static char *get_header(const struct commit *commit, const char *msg,
 546                        const char *key)
 547{
 548        int key_len = strlen(key);
 549        const char *line = msg;
 550
 551        while (line) {
 552                const char *eol = strchr(line, '\n'), *next;
 553
 554                if (line == eol)
 555                        return NULL;
 556                if (!eol) {
 557                        warning("malformed commit (header is missing newline): %s",
 558                                sha1_to_hex(commit->object.sha1));
 559                        eol = line + strlen(line);
 560                        next = NULL;
 561                } else
 562                        next = eol + 1;
 563                if (eol - line > key_len &&
 564                    !strncmp(line, key, key_len) &&
 565                    line[key_len] == ' ') {
 566                        return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 567                }
 568                line = next;
 569        }
 570        return NULL;
 571}
 572
 573static char *replace_encoding_header(char *buf, const char *encoding)
 574{
 575        struct strbuf tmp = STRBUF_INIT;
 576        size_t start, len;
 577        char *cp = buf;
 578
 579        /* guess if there is an encoding header before a \n\n */
 580        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 581                cp = strchr(cp, '\n');
 582                if (!cp || *++cp == '\n')
 583                        return buf;
 584        }
 585        start = cp - buf;
 586        cp = strchr(cp, '\n');
 587        if (!cp)
 588                return buf; /* should not happen but be defensive */
 589        len = cp + 1 - (buf + start);
 590
 591        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 592        if (is_encoding_utf8(encoding)) {
 593                /* we have re-coded to UTF-8; drop the header */
 594                strbuf_remove(&tmp, start, len);
 595        } else {
 596                /* just replaces XXXX in 'encoding XXXX\n' */
 597                strbuf_splice(&tmp, start + strlen("encoding "),
 598                                          len - strlen("encoding \n"),
 599                                          encoding, strlen(encoding));
 600        }
 601        return strbuf_detach(&tmp, NULL);
 602}
 603
 604char *logmsg_reencode(const struct commit *commit,
 605                      char **commit_encoding,
 606                      const char *output_encoding)
 607{
 608        static const char *utf8 = "UTF-8";
 609        const char *use_encoding;
 610        char *encoding;
 611        char *msg = commit->buffer;
 612        char *out;
 613
 614        if (!msg) {
 615                enum object_type type;
 616                unsigned long size;
 617
 618                msg = read_sha1_file(commit->object.sha1, &type, &size);
 619                if (!msg)
 620                        die("Cannot read commit object %s",
 621                            sha1_to_hex(commit->object.sha1));
 622                if (type != OBJ_COMMIT)
 623                        die("Expected commit for '%s', got %s",
 624                            sha1_to_hex(commit->object.sha1), typename(type));
 625        }
 626
 627        if (!output_encoding || !*output_encoding) {
 628                if (commit_encoding)
 629                        *commit_encoding =
 630                                get_header(commit, msg, "encoding");
 631                return msg;
 632        }
 633        encoding = get_header(commit, msg, "encoding");
 634        if (commit_encoding)
 635                *commit_encoding = encoding;
 636        use_encoding = encoding ? encoding : utf8;
 637        if (same_encoding(use_encoding, output_encoding)) {
 638                /*
 639                 * No encoding work to be done. If we have no encoding header
 640                 * at all, then there's nothing to do, and we can return the
 641                 * message verbatim (whether newly allocated or not).
 642                 */
 643                if (!encoding)
 644                        return msg;
 645
 646                /*
 647                 * Otherwise, we still want to munge the encoding header in the
 648                 * result, which will be done by modifying the buffer. If we
 649                 * are using a fresh copy, we can reuse it. But if we are using
 650                 * the cached copy from commit->buffer, we need to duplicate it
 651                 * to avoid munging commit->buffer.
 652                 */
 653                out = msg;
 654                if (out == commit->buffer)
 655                        out = xstrdup(out);
 656        }
 657        else {
 658                /*
 659                 * There's actual encoding work to do. Do the reencoding, which
 660                 * still leaves the header to be replaced in the next step. At
 661                 * this point, we are done with msg. If we allocated a fresh
 662                 * copy, we can free it.
 663                 */
 664                out = reencode_string(msg, output_encoding, use_encoding);
 665                if (out && msg != commit->buffer)
 666                        free(msg);
 667        }
 668
 669        /*
 670         * This replacement actually consumes the buffer we hand it, so we do
 671         * not have to worry about freeing the old "out" here.
 672         */
 673        if (out)
 674                out = replace_encoding_header(out, output_encoding);
 675
 676        if (!commit_encoding)
 677                free(encoding);
 678        /*
 679         * If the re-encoding failed, out might be NULL here; in that
 680         * case we just return the commit message verbatim.
 681         */
 682        return out ? out : msg;
 683}
 684
 685void logmsg_free(char *msg, const struct commit *commit)
 686{
 687        if (msg != commit->buffer)
 688                free(msg);
 689}
 690
 691static int mailmap_name(const char **email, size_t *email_len,
 692                        const char **name, size_t *name_len)
 693{
 694        static struct string_list *mail_map;
 695        if (!mail_map) {
 696                mail_map = xcalloc(1, sizeof(*mail_map));
 697                read_mailmap(mail_map, NULL);
 698        }
 699        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 700}
 701
 702static size_t format_person_part(struct strbuf *sb, char part,
 703                                 const char *msg, int len, enum date_mode dmode)
 704{
 705        /* currently all placeholders have same length */
 706        const int placeholder_len = 2;
 707        struct ident_split s;
 708        const char *name, *mail;
 709        size_t maillen, namelen;
 710
 711        if (split_ident_line(&s, msg, len) < 0)
 712                goto skip;
 713
 714        name = s.name_begin;
 715        namelen = s.name_end - s.name_begin;
 716        mail = s.mail_begin;
 717        maillen = s.mail_end - s.mail_begin;
 718
 719        if (part == 'N' || part == 'E') /* mailmap lookup */
 720                mailmap_name(&mail, &maillen, &name, &namelen);
 721        if (part == 'n' || part == 'N') {       /* name */
 722                strbuf_add(sb, name, namelen);
 723                return placeholder_len;
 724        }
 725        if (part == 'e' || part == 'E') {       /* email */
 726                strbuf_add(sb, mail, maillen);
 727                return placeholder_len;
 728        }
 729
 730        if (!s.date_begin)
 731                goto skip;
 732
 733        if (part == 't') {      /* date, UNIX timestamp */
 734                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 735                return placeholder_len;
 736        }
 737
 738        switch (part) {
 739        case 'd':       /* date */
 740                strbuf_addstr(sb, show_ident_date(&s, dmode));
 741                return placeholder_len;
 742        case 'D':       /* date, RFC2822 style */
 743                strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
 744                return placeholder_len;
 745        case 'r':       /* date, relative */
 746                strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
 747                return placeholder_len;
 748        case 'i':       /* date, ISO 8601 */
 749                strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
 750                return placeholder_len;
 751        }
 752
 753skip:
 754        /*
 755         * reading from either a bogus commit, or a reflog entry with
 756         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 757         * to compute a valid return value.
 758         */
 759        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 760            || part == 'D' || part == 'r' || part == 'i')
 761                return placeholder_len;
 762
 763        return 0; /* unknown placeholder */
 764}
 765
 766struct chunk {
 767        size_t off;
 768        size_t len;
 769};
 770
 771enum flush_type {
 772        no_flush,
 773        flush_right,
 774        flush_left,
 775        flush_left_and_steal,
 776        flush_both
 777};
 778
 779enum trunc_type {
 780        trunc_none,
 781        trunc_left,
 782        trunc_middle,
 783        trunc_right
 784};
 785
 786struct format_commit_context {
 787        const struct commit *commit;
 788        const struct pretty_print_context *pretty_ctx;
 789        unsigned commit_header_parsed:1;
 790        unsigned commit_message_parsed:1;
 791        struct signature_check signature_check;
 792        enum flush_type flush_type;
 793        enum trunc_type truncate;
 794        char *message;
 795        char *commit_encoding;
 796        size_t width, indent1, indent2;
 797        int auto_color;
 798        int padding;
 799
 800        /* These offsets are relative to the start of the commit message. */
 801        struct chunk author;
 802        struct chunk committer;
 803        size_t message_off;
 804        size_t subject_off;
 805        size_t body_off;
 806
 807        /* The following ones are relative to the result struct strbuf. */
 808        struct chunk abbrev_commit_hash;
 809        struct chunk abbrev_tree_hash;
 810        struct chunk abbrev_parent_hashes;
 811        size_t wrap_start;
 812};
 813
 814static int add_again(struct strbuf *sb, struct chunk *chunk)
 815{
 816        if (chunk->len) {
 817                strbuf_adddup(sb, chunk->off, chunk->len);
 818                return 1;
 819        }
 820
 821        /*
 822         * We haven't seen this chunk before.  Our caller is surely
 823         * going to add it the hard way now.  Remember the most likely
 824         * start of the to-be-added chunk: the current end of the
 825         * struct strbuf.
 826         */
 827        chunk->off = sb->len;
 828        return 0;
 829}
 830
 831static void parse_commit_header(struct format_commit_context *context)
 832{
 833        const char *msg = context->message;
 834        int i;
 835
 836        for (i = 0; msg[i]; i++) {
 837                int eol;
 838                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 839                        ; /* do nothing */
 840
 841                if (i == eol) {
 842                        break;
 843                } else if (!prefixcmp(msg + i, "author ")) {
 844                        context->author.off = i + 7;
 845                        context->author.len = eol - i - 7;
 846                } else if (!prefixcmp(msg + i, "committer ")) {
 847                        context->committer.off = i + 10;
 848                        context->committer.len = eol - i - 10;
 849                }
 850                i = eol;
 851        }
 852        context->message_off = i;
 853        context->commit_header_parsed = 1;
 854}
 855
 856static int istitlechar(char c)
 857{
 858        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 859                (c >= '0' && c <= '9') || c == '.' || c == '_';
 860}
 861
 862static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 863{
 864        size_t trimlen;
 865        size_t start_len = sb->len;
 866        int space = 2;
 867
 868        for (; *msg && *msg != '\n'; msg++) {
 869                if (istitlechar(*msg)) {
 870                        if (space == 1)
 871                                strbuf_addch(sb, '-');
 872                        space = 0;
 873                        strbuf_addch(sb, *msg);
 874                        if (*msg == '.')
 875                                while (*(msg+1) == '.')
 876                                        msg++;
 877                } else
 878                        space |= 1;
 879        }
 880
 881        /* trim any trailing '.' or '-' characters */
 882        trimlen = 0;
 883        while (sb->len - trimlen > start_len &&
 884                (sb->buf[sb->len - 1 - trimlen] == '.'
 885                || sb->buf[sb->len - 1 - trimlen] == '-'))
 886                trimlen++;
 887        strbuf_remove(sb, sb->len - trimlen, trimlen);
 888}
 889
 890const char *format_subject(struct strbuf *sb, const char *msg,
 891                           const char *line_separator)
 892{
 893        int first = 1;
 894
 895        for (;;) {
 896                const char *line = msg;
 897                int linelen = get_one_line(line);
 898
 899                msg += linelen;
 900                if (!linelen || is_empty_line(line, &linelen))
 901                        break;
 902
 903                if (!sb)
 904                        continue;
 905                strbuf_grow(sb, linelen + 2);
 906                if (!first)
 907                        strbuf_addstr(sb, line_separator);
 908                strbuf_add(sb, line, linelen);
 909                first = 0;
 910        }
 911        return msg;
 912}
 913
 914static void parse_commit_message(struct format_commit_context *c)
 915{
 916        const char *msg = c->message + c->message_off;
 917        const char *start = c->message;
 918
 919        msg = skip_empty_lines(msg);
 920        c->subject_off = msg - start;
 921
 922        msg = format_subject(NULL, msg, NULL);
 923        msg = skip_empty_lines(msg);
 924        c->body_off = msg - start;
 925
 926        c->commit_message_parsed = 1;
 927}
 928
 929static void strbuf_wrap(struct strbuf *sb, size_t pos,
 930                        size_t width, size_t indent1, size_t indent2)
 931{
 932        struct strbuf tmp = STRBUF_INIT;
 933
 934        if (pos)
 935                strbuf_add(&tmp, sb->buf, pos);
 936        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 937                                (int) indent1, (int) indent2, (int) width);
 938        strbuf_swap(&tmp, sb);
 939        strbuf_release(&tmp);
 940}
 941
 942static void rewrap_message_tail(struct strbuf *sb,
 943                                struct format_commit_context *c,
 944                                size_t new_width, size_t new_indent1,
 945                                size_t new_indent2)
 946{
 947        if (c->width == new_width && c->indent1 == new_indent1 &&
 948            c->indent2 == new_indent2)
 949                return;
 950        if (c->wrap_start < sb->len)
 951                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 952        c->wrap_start = sb->len;
 953        c->width = new_width;
 954        c->indent1 = new_indent1;
 955        c->indent2 = new_indent2;
 956}
 957
 958static int format_reflog_person(struct strbuf *sb,
 959                                char part,
 960                                struct reflog_walk_info *log,
 961                                enum date_mode dmode)
 962{
 963        const char *ident;
 964
 965        if (!log)
 966                return 2;
 967
 968        ident = get_reflog_ident(log);
 969        if (!ident)
 970                return 2;
 971
 972        return format_person_part(sb, part, ident, strlen(ident), dmode);
 973}
 974
 975static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 976                          const char *placeholder,
 977                          struct format_commit_context *c)
 978{
 979        if (placeholder[1] == '(') {
 980                const char *begin = placeholder + 2;
 981                const char *end = strchr(begin, ')');
 982                char color[COLOR_MAXLEN];
 983
 984                if (!end)
 985                        return 0;
 986                if (!prefixcmp(begin, "auto,")) {
 987                        if (!want_color(c->pretty_ctx->color))
 988                                return end - placeholder + 1;
 989                        begin += 5;
 990                }
 991                color_parse_mem(begin,
 992                                end - begin,
 993                                "--pretty format", color);
 994                strbuf_addstr(sb, color);
 995                return end - placeholder + 1;
 996        }
 997        if (!prefixcmp(placeholder + 1, "red")) {
 998                strbuf_addstr(sb, GIT_COLOR_RED);
 999                return 4;
1000        } else if (!prefixcmp(placeholder + 1, "green")) {
1001                strbuf_addstr(sb, GIT_COLOR_GREEN);
1002                return 6;
1003        } else if (!prefixcmp(placeholder + 1, "blue")) {
1004                strbuf_addstr(sb, GIT_COLOR_BLUE);
1005                return 5;
1006        } else if (!prefixcmp(placeholder + 1, "reset")) {
1007                strbuf_addstr(sb, GIT_COLOR_RESET);
1008                return 6;
1009        } else
1010                return 0;
1011}
1012
1013static size_t parse_padding_placeholder(struct strbuf *sb,
1014                                        const char *placeholder,
1015                                        struct format_commit_context *c)
1016{
1017        const char *ch = placeholder;
1018        enum flush_type flush_type;
1019        int to_column = 0;
1020
1021        switch (*ch++) {
1022        case '<':
1023                flush_type = flush_right;
1024                break;
1025        case '>':
1026                if (*ch == '<') {
1027                        flush_type = flush_both;
1028                        ch++;
1029                } else if (*ch == '>') {
1030                        flush_type = flush_left_and_steal;
1031                        ch++;
1032                } else
1033                        flush_type = flush_left;
1034                break;
1035        default:
1036                return 0;
1037        }
1038
1039        /* the next value means "wide enough to that column" */
1040        if (*ch == '|') {
1041                to_column = 1;
1042                ch++;
1043        }
1044
1045        if (*ch == '(') {
1046                const char *start = ch + 1;
1047                const char *end = start + strcspn(start, ",)");
1048                char *next;
1049                int width;
1050                if (!end || end == start)
1051                        return 0;
1052                width = strtoul(start, &next, 10);
1053                if (next == start || width == 0)
1054                        return 0;
1055                c->padding = to_column ? -width : width;
1056                c->flush_type = flush_type;
1057
1058                if (*end == ',') {
1059                        start = end + 1;
1060                        end = strchr(start, ')');
1061                        if (!end || end == start)
1062                                return 0;
1063                        if (!prefixcmp(start, "trunc)"))
1064                                c->truncate = trunc_right;
1065                        else if (!prefixcmp(start, "ltrunc)"))
1066                                c->truncate = trunc_left;
1067                        else if (!prefixcmp(start, "mtrunc)"))
1068                                c->truncate = trunc_middle;
1069                        else
1070                                return 0;
1071                } else
1072                        c->truncate = trunc_none;
1073
1074                return end - placeholder + 1;
1075        }
1076        return 0;
1077}
1078
1079static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1080                                const char *placeholder,
1081                                void *context)
1082{
1083        struct format_commit_context *c = context;
1084        const struct commit *commit = c->commit;
1085        const char *msg = c->message;
1086        struct commit_list *p;
1087        int h1, h2;
1088
1089        /* these are independent of the commit */
1090        switch (placeholder[0]) {
1091        case 'C':
1092                if (!prefixcmp(placeholder + 1, "(auto)")) {
1093                        c->auto_color = 1;
1094                        return 7; /* consumed 7 bytes, "C(auto)" */
1095                } else {
1096                        int ret = parse_color(sb, placeholder, c);
1097                        if (ret)
1098                                c->auto_color = 0;
1099                        /*
1100                         * Otherwise, we decided to treat %C<unknown>
1101                         * as a literal string, and the previous
1102                         * %C(auto) is still valid.
1103                         */
1104                        return ret;
1105                }
1106        case 'n':               /* newline */
1107                strbuf_addch(sb, '\n');
1108                return 1;
1109        case 'x':
1110                /* %x00 == NUL, %x0a == LF, etc. */
1111                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1112                    h1 <= 16 &&
1113                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1114                    h2 <= 16) {
1115                        strbuf_addch(sb, (h1<<4)|h2);
1116                        return 3;
1117                } else
1118                        return 0;
1119        case 'w':
1120                if (placeholder[1] == '(') {
1121                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1122                        char *next;
1123                        const char *start = placeholder + 2;
1124                        const char *end = strchr(start, ')');
1125                        if (!end)
1126                                return 0;
1127                        if (end > start) {
1128                                width = strtoul(start, &next, 10);
1129                                if (*next == ',') {
1130                                        indent1 = strtoul(next + 1, &next, 10);
1131                                        if (*next == ',') {
1132                                                indent2 = strtoul(next + 1,
1133                                                                 &next, 10);
1134                                        }
1135                                }
1136                                if (*next != ')')
1137                                        return 0;
1138                        }
1139                        rewrap_message_tail(sb, c, width, indent1, indent2);
1140                        return end - placeholder + 1;
1141                } else
1142                        return 0;
1143
1144        case '<':
1145        case '>':
1146                return parse_padding_placeholder(sb, placeholder, c);
1147        }
1148
1149        /* these depend on the commit */
1150        if (!commit->object.parsed)
1151                parse_object(commit->object.sha1);
1152
1153        switch (placeholder[0]) {
1154        case 'H':               /* commit hash */
1155                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1156                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1157                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1158                return 1;
1159        case 'h':               /* abbreviated commit hash */
1160                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1161                if (add_again(sb, &c->abbrev_commit_hash)) {
1162                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1163                        return 1;
1164                }
1165                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1166                                                     c->pretty_ctx->abbrev));
1167                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1168                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1169                return 1;
1170        case 'T':               /* tree hash */
1171                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1172                return 1;
1173        case 't':               /* abbreviated tree hash */
1174                if (add_again(sb, &c->abbrev_tree_hash))
1175                        return 1;
1176                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1177                                                     c->pretty_ctx->abbrev));
1178                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1179                return 1;
1180        case 'P':               /* parent hashes */
1181                for (p = commit->parents; p; p = p->next) {
1182                        if (p != commit->parents)
1183                                strbuf_addch(sb, ' ');
1184                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1185                }
1186                return 1;
1187        case 'p':               /* abbreviated parent hashes */
1188                if (add_again(sb, &c->abbrev_parent_hashes))
1189                        return 1;
1190                for (p = commit->parents; p; p = p->next) {
1191                        if (p != commit->parents)
1192                                strbuf_addch(sb, ' ');
1193                        strbuf_addstr(sb, find_unique_abbrev(
1194                                        p->item->object.sha1,
1195                                        c->pretty_ctx->abbrev));
1196                }
1197                c->abbrev_parent_hashes.len = sb->len -
1198                                              c->abbrev_parent_hashes.off;
1199                return 1;
1200        case 'm':               /* left/right/bottom */
1201                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1202                return 1;
1203        case 'd':
1204                load_ref_decorations(DECORATE_SHORT_REFS);
1205                format_decorations(sb, commit, c->auto_color);
1206                return 1;
1207        case 'g':               /* reflog info */
1208                switch(placeholder[1]) {
1209                case 'd':       /* reflog selector */
1210                case 'D':
1211                        if (c->pretty_ctx->reflog_info)
1212                                get_reflog_selector(sb,
1213                                                    c->pretty_ctx->reflog_info,
1214                                                    c->pretty_ctx->date_mode,
1215                                                    c->pretty_ctx->date_mode_explicit,
1216                                                    (placeholder[1] == 'd'));
1217                        return 2;
1218                case 's':       /* reflog message */
1219                        if (c->pretty_ctx->reflog_info)
1220                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1221                        return 2;
1222                case 'n':
1223                case 'N':
1224                case 'e':
1225                case 'E':
1226                        return format_reflog_person(sb,
1227                                                    placeholder[1],
1228                                                    c->pretty_ctx->reflog_info,
1229                                                    c->pretty_ctx->date_mode);
1230                }
1231                return 0;       /* unknown %g placeholder */
1232        case 'N':
1233                if (c->pretty_ctx->notes_message) {
1234                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1235                        return 1;
1236                }
1237                return 0;
1238        }
1239
1240        if (placeholder[0] == 'G') {
1241                if (!c->signature_check.result)
1242                        check_commit_signature(c->commit, &(c->signature_check));
1243                switch (placeholder[1]) {
1244                case 'G':
1245                        if (c->signature_check.gpg_output)
1246                                strbuf_addstr(sb, c->signature_check.gpg_output);
1247                        break;
1248                case '?':
1249                        switch (c->signature_check.result) {
1250                        case 'G':
1251                        case 'B':
1252                        case 'U':
1253                        case 'N':
1254                                strbuf_addch(sb, c->signature_check.result);
1255                        }
1256                        break;
1257                case 'S':
1258                        if (c->signature_check.signer)
1259                                strbuf_addstr(sb, c->signature_check.signer);
1260                        break;
1261                case 'K':
1262                        if (c->signature_check.key)
1263                                strbuf_addstr(sb, c->signature_check.key);
1264                        break;
1265                }
1266                return 2;
1267        }
1268
1269
1270        /* For the rest we have to parse the commit header. */
1271        if (!c->commit_header_parsed)
1272                parse_commit_header(c);
1273
1274        switch (placeholder[0]) {
1275        case 'a':       /* author ... */
1276                return format_person_part(sb, placeholder[1],
1277                                   msg + c->author.off, c->author.len,
1278                                   c->pretty_ctx->date_mode);
1279        case 'c':       /* committer ... */
1280                return format_person_part(sb, placeholder[1],
1281                                   msg + c->committer.off, c->committer.len,
1282                                   c->pretty_ctx->date_mode);
1283        case 'e':       /* encoding */
1284                if (c->commit_encoding)
1285                        strbuf_addstr(sb, c->commit_encoding);
1286                return 1;
1287        case 'B':       /* raw body */
1288                /* message_off is always left at the initial newline */
1289                strbuf_addstr(sb, msg + c->message_off + 1);
1290                return 1;
1291        }
1292
1293        /* Now we need to parse the commit message. */
1294        if (!c->commit_message_parsed)
1295                parse_commit_message(c);
1296
1297        switch (placeholder[0]) {
1298        case 's':       /* subject */
1299                format_subject(sb, msg + c->subject_off, " ");
1300                return 1;
1301        case 'f':       /* sanitized subject */
1302                format_sanitized_subject(sb, msg + c->subject_off);
1303                return 1;
1304        case 'b':       /* body */
1305                strbuf_addstr(sb, msg + c->body_off);
1306                return 1;
1307        }
1308        return 0;       /* unknown placeholder */
1309}
1310
1311static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1312                                    const char *placeholder,
1313                                    struct format_commit_context *c)
1314{
1315        struct strbuf local_sb = STRBUF_INIT;
1316        int total_consumed = 0, len, padding = c->padding;
1317        if (padding < 0) {
1318                const char *start = strrchr(sb->buf, '\n');
1319                int occupied;
1320                if (!start)
1321                        start = sb->buf;
1322                occupied = utf8_strnwidth(start, -1, 1);
1323                padding = (-padding) - occupied;
1324        }
1325        while (1) {
1326                int modifier = *placeholder == 'C';
1327                int consumed = format_commit_one(&local_sb, placeholder, c);
1328                total_consumed += consumed;
1329
1330                if (!modifier)
1331                        break;
1332
1333                placeholder += consumed;
1334                if (*placeholder != '%')
1335                        break;
1336                placeholder++;
1337                total_consumed++;
1338        }
1339        len = utf8_strnwidth(local_sb.buf, -1, 1);
1340
1341        if (c->flush_type == flush_left_and_steal) {
1342                const char *ch = sb->buf + sb->len - 1;
1343                while (len > padding && ch > sb->buf) {
1344                        const char *p;
1345                        if (*ch == ' ') {
1346                                ch--;
1347                                padding++;
1348                                continue;
1349                        }
1350                        /* check for trailing ansi sequences */
1351                        if (*ch != 'm')
1352                                break;
1353                        p = ch - 1;
1354                        while (ch - p < 10 && *p != '\033')
1355                                p--;
1356                        if (*p != '\033' ||
1357                            ch + 1 - p != display_mode_esc_sequence_len(p))
1358                                break;
1359                        /*
1360                         * got a good ansi sequence, put it back to
1361                         * local_sb as we're cutting sb
1362                         */
1363                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1364                        ch = p - 1;
1365                }
1366                strbuf_setlen(sb, ch + 1 - sb->buf);
1367                c->flush_type = flush_left;
1368        }
1369
1370        if (len > padding) {
1371                switch (c->truncate) {
1372                case trunc_left:
1373                        strbuf_utf8_replace(&local_sb,
1374                                            0, len - (padding - 2),
1375                                            "..");
1376                        break;
1377                case trunc_middle:
1378                        strbuf_utf8_replace(&local_sb,
1379                                            padding / 2 - 1,
1380                                            len - (padding - 2),
1381                                            "..");
1382                        break;
1383                case trunc_right:
1384                        strbuf_utf8_replace(&local_sb,
1385                                            padding - 2, len - (padding - 2),
1386                                            "..");
1387                        break;
1388                case trunc_none:
1389                        break;
1390                }
1391                strbuf_addstr(sb, local_sb.buf);
1392        } else {
1393                int sb_len = sb->len, offset = 0;
1394                if (c->flush_type == flush_left)
1395                        offset = padding - len;
1396                else if (c->flush_type == flush_both)
1397                        offset = (padding - len) / 2;
1398                /*
1399                 * we calculate padding in columns, now
1400                 * convert it back to chars
1401                 */
1402                padding = padding - len + local_sb.len;
1403                strbuf_grow(sb, padding);
1404                strbuf_setlen(sb, sb_len + padding);
1405                memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1406                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1407                       local_sb.len);
1408        }
1409        strbuf_release(&local_sb);
1410        c->flush_type = no_flush;
1411        return total_consumed;
1412}
1413
1414static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1415                                 const char *placeholder,
1416                                 void *context)
1417{
1418        int consumed;
1419        size_t orig_len;
1420        enum {
1421                NO_MAGIC,
1422                ADD_LF_BEFORE_NON_EMPTY,
1423                DEL_LF_BEFORE_EMPTY,
1424                ADD_SP_BEFORE_NON_EMPTY
1425        } magic = NO_MAGIC;
1426
1427        switch (placeholder[0]) {
1428        case '-':
1429                magic = DEL_LF_BEFORE_EMPTY;
1430                break;
1431        case '+':
1432                magic = ADD_LF_BEFORE_NON_EMPTY;
1433                break;
1434        case ' ':
1435                magic = ADD_SP_BEFORE_NON_EMPTY;
1436                break;
1437        default:
1438                break;
1439        }
1440        if (magic != NO_MAGIC)
1441                placeholder++;
1442
1443        orig_len = sb->len;
1444        if (((struct format_commit_context *)context)->flush_type != no_flush)
1445                consumed = format_and_pad_commit(sb, placeholder, context);
1446        else
1447                consumed = format_commit_one(sb, placeholder, context);
1448        if (magic == NO_MAGIC)
1449                return consumed;
1450
1451        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1452                while (sb->len && sb->buf[sb->len - 1] == '\n')
1453                        strbuf_setlen(sb, sb->len - 1);
1454        } else if (orig_len != sb->len) {
1455                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1456                        strbuf_insert(sb, orig_len, "\n", 1);
1457                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1458                        strbuf_insert(sb, orig_len, " ", 1);
1459        }
1460        return consumed + 1;
1461}
1462
1463static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1464                                   void *context)
1465{
1466        struct userformat_want *w = context;
1467
1468        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1469                placeholder++;
1470
1471        switch (*placeholder) {
1472        case 'N':
1473                w->notes = 1;
1474                break;
1475        }
1476        return 0;
1477}
1478
1479void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1480{
1481        struct strbuf dummy = STRBUF_INIT;
1482
1483        if (!fmt) {
1484                if (!user_format)
1485                        return;
1486                fmt = user_format;
1487        }
1488        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1489        strbuf_release(&dummy);
1490}
1491
1492void format_commit_message(const struct commit *commit,
1493                           const char *format, struct strbuf *sb,
1494                           const struct pretty_print_context *pretty_ctx)
1495{
1496        struct format_commit_context context;
1497        const char *output_enc = pretty_ctx->output_encoding;
1498        const char *utf8 = "UTF-8";
1499
1500        memset(&context, 0, sizeof(context));
1501        context.commit = commit;
1502        context.pretty_ctx = pretty_ctx;
1503        context.wrap_start = sb->len;
1504        context.message = logmsg_reencode(commit,
1505                                          &context.commit_encoding,
1506                                          output_enc);
1507
1508        strbuf_expand(sb, format, format_commit_item, &context);
1509        rewrap_message_tail(sb, &context, 0, 0, 0);
1510
1511        if (output_enc) {
1512                if (same_encoding(utf8, output_enc))
1513                        output_enc = NULL;
1514        } else {
1515                if (context.commit_encoding &&
1516                    !same_encoding(context.commit_encoding, utf8))
1517                        output_enc = context.commit_encoding;
1518        }
1519
1520        if (output_enc) {
1521                int outsz;
1522                char *out = reencode_string_len(sb->buf, sb->len,
1523                                                output_enc, utf8, &outsz);
1524                if (out)
1525                        strbuf_attach(sb, out, outsz, outsz + 1);
1526        }
1527
1528        free(context.commit_encoding);
1529        logmsg_free(context.message, commit);
1530        free(context.signature_check.gpg_output);
1531        free(context.signature_check.signer);
1532}
1533
1534static void pp_header(struct pretty_print_context *pp,
1535                      const char *encoding,
1536                      const struct commit *commit,
1537                      const char **msg_p,
1538                      struct strbuf *sb)
1539{
1540        int parents_shown = 0;
1541
1542        for (;;) {
1543                const char *line = *msg_p;
1544                int linelen = get_one_line(*msg_p);
1545
1546                if (!linelen)
1547                        return;
1548                *msg_p += linelen;
1549
1550                if (linelen == 1)
1551                        /* End of header */
1552                        return;
1553
1554                if (pp->fmt == CMIT_FMT_RAW) {
1555                        strbuf_add(sb, line, linelen);
1556                        continue;
1557                }
1558
1559                if (!prefixcmp(line, "parent ")) {
1560                        if (linelen != 48)
1561                                die("bad parent line in commit");
1562                        continue;
1563                }
1564
1565                if (!parents_shown) {
1566                        struct commit_list *parent;
1567                        int num;
1568                        for (parent = commit->parents, num = 0;
1569                             parent;
1570                             parent = parent->next, num++)
1571                                ;
1572                        /* with enough slop */
1573                        strbuf_grow(sb, num * 50 + 20);
1574                        add_merge_info(pp, sb, commit);
1575                        parents_shown = 1;
1576                }
1577
1578                /*
1579                 * MEDIUM == DEFAULT shows only author with dates.
1580                 * FULL shows both authors but not dates.
1581                 * FULLER shows both authors and dates.
1582                 */
1583                if (!prefixcmp(line, "author ")) {
1584                        strbuf_grow(sb, linelen + 80);
1585                        pp_user_info(pp, "Author", sb, line + 7, encoding);
1586                }
1587                if (!prefixcmp(line, "committer ") &&
1588                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1589                        strbuf_grow(sb, linelen + 80);
1590                        pp_user_info(pp, "Commit", sb, line + 10, encoding);
1591                }
1592        }
1593}
1594
1595void pp_title_line(struct pretty_print_context *pp,
1596                   const char **msg_p,
1597                   struct strbuf *sb,
1598                   const char *encoding,
1599                   int need_8bit_cte)
1600{
1601        static const int max_length = 78; /* per rfc2047 */
1602        struct strbuf title;
1603
1604        strbuf_init(&title, 80);
1605        *msg_p = format_subject(&title, *msg_p,
1606                                pp->preserve_subject ? "\n" : " ");
1607
1608        strbuf_grow(sb, title.len + 1024);
1609        if (pp->subject) {
1610                strbuf_addstr(sb, pp->subject);
1611                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1612                        add_rfc2047(sb, title.buf, title.len,
1613                                                encoding, RFC2047_SUBJECT);
1614                else
1615                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1616                                         -last_line_length(sb), 1, max_length);
1617        } else {
1618                strbuf_addbuf(sb, &title);
1619        }
1620        strbuf_addch(sb, '\n');
1621
1622        if (need_8bit_cte == 0) {
1623                int i;
1624                for (i = 0; i < pp->in_body_headers.nr; i++) {
1625                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1626                                need_8bit_cte = 1;
1627                                break;
1628                        }
1629                }
1630        }
1631
1632        if (need_8bit_cte > 0) {
1633                const char *header_fmt =
1634                        "MIME-Version: 1.0\n"
1635                        "Content-Type: text/plain; charset=%s\n"
1636                        "Content-Transfer-Encoding: 8bit\n";
1637                strbuf_addf(sb, header_fmt, encoding);
1638        }
1639        if (pp->after_subject) {
1640                strbuf_addstr(sb, pp->after_subject);
1641        }
1642        if (pp->fmt == CMIT_FMT_EMAIL) {
1643                strbuf_addch(sb, '\n');
1644        }
1645
1646        if (pp->in_body_headers.nr) {
1647                int i;
1648                for (i = 0; i < pp->in_body_headers.nr; i++) {
1649                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1650                        free(pp->in_body_headers.items[i].string);
1651                }
1652                string_list_clear(&pp->in_body_headers, 0);
1653                strbuf_addch(sb, '\n');
1654        }
1655
1656        strbuf_release(&title);
1657}
1658
1659void pp_remainder(struct pretty_print_context *pp,
1660                  const char **msg_p,
1661                  struct strbuf *sb,
1662                  int indent)
1663{
1664        int first = 1;
1665        for (;;) {
1666                const char *line = *msg_p;
1667                int linelen = get_one_line(line);
1668                *msg_p += linelen;
1669
1670                if (!linelen)
1671                        break;
1672
1673                if (is_empty_line(line, &linelen)) {
1674                        if (first)
1675                                continue;
1676                        if (pp->fmt == CMIT_FMT_SHORT)
1677                                break;
1678                }
1679                first = 0;
1680
1681                strbuf_grow(sb, linelen + indent + 20);
1682                if (indent) {
1683                        memset(sb->buf + sb->len, ' ', indent);
1684                        strbuf_setlen(sb, sb->len + indent);
1685                }
1686                strbuf_add(sb, line, linelen);
1687                strbuf_addch(sb, '\n');
1688        }
1689}
1690
1691void pretty_print_commit(struct pretty_print_context *pp,
1692                         const struct commit *commit,
1693                         struct strbuf *sb)
1694{
1695        unsigned long beginning_of_body;
1696        int indent = 4;
1697        const char *msg;
1698        char *reencoded;
1699        const char *encoding;
1700        int need_8bit_cte = pp->need_8bit_cte;
1701
1702        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1703                format_commit_message(commit, user_format, sb, pp);
1704                return;
1705        }
1706
1707        encoding = get_log_output_encoding();
1708        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1709
1710        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1711                indent = 0;
1712
1713        /*
1714         * We need to check and emit Content-type: to mark it
1715         * as 8-bit if we haven't done so.
1716         */
1717        if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1718                int i, ch, in_body;
1719
1720                for (in_body = i = 0; (ch = msg[i]); i++) {
1721                        if (!in_body) {
1722                                /* author could be non 7-bit ASCII but
1723                                 * the log may be so; skip over the
1724                                 * header part first.
1725                                 */
1726                                if (ch == '\n' && msg[i+1] == '\n')
1727                                        in_body = 1;
1728                        }
1729                        else if (non_ascii(ch)) {
1730                                need_8bit_cte = 1;
1731                                break;
1732                        }
1733                }
1734        }
1735
1736        pp_header(pp, encoding, commit, &msg, sb);
1737        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1738                strbuf_addch(sb, '\n');
1739        }
1740
1741        /* Skip excess blank lines at the beginning of body, if any... */
1742        msg = skip_empty_lines(msg);
1743
1744        /* These formats treat the title line specially. */
1745        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1746                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1747
1748        beginning_of_body = sb->len;
1749        if (pp->fmt != CMIT_FMT_ONELINE)
1750                pp_remainder(pp, &msg, sb, indent);
1751        strbuf_rtrim(sb);
1752
1753        /* Make sure there is an EOLN for the non-oneline case */
1754        if (pp->fmt != CMIT_FMT_ONELINE)
1755                strbuf_addch(sb, '\n');
1756
1757        /*
1758         * The caller may append additional body text in e-mail
1759         * format.  Make sure we did not strip the blank line
1760         * between the header and the body.
1761         */
1762        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1763                strbuf_addch(sb, '\n');
1764
1765        logmsg_free(reencoded, commit);
1766}
1767
1768void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1769                    struct strbuf *sb)
1770{
1771        struct pretty_print_context pp = {0};
1772        pp.fmt = fmt;
1773        pretty_print_commit(&pp, commit, sb);
1774}