pretty.con commit diff-lib: allow ita entries treated as "not yet exist in index" (425a28e)
   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 expand_tabs_in_log;
  20        int is_alias;
  21        const char *user_format;
  22} *commit_formats;
  23static size_t builtin_formats_len;
  24static size_t commit_formats_len;
  25static size_t commit_formats_alloc;
  26static struct cmt_fmt_map *find_commit_format(const char *sought);
  27
  28int commit_format_is_empty(enum cmit_fmt fmt)
  29{
  30        return fmt == CMIT_FMT_USERFORMAT && !*user_format;
  31}
  32
  33static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  34{
  35        free(user_format);
  36        user_format = xstrdup(cp);
  37        if (is_tformat)
  38                rev->use_terminator = 1;
  39        rev->commit_format = CMIT_FMT_USERFORMAT;
  40}
  41
  42static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  43{
  44        struct cmt_fmt_map *commit_format = NULL;
  45        const char *name;
  46        const char *fmt;
  47        int i;
  48
  49        if (!skip_prefix(var, "pretty.", &name))
  50                return 0;
  51
  52        for (i = 0; i < builtin_formats_len; i++) {
  53                if (!strcmp(commit_formats[i].name, name))
  54                        return 0;
  55        }
  56
  57        for (i = builtin_formats_len; i < commit_formats_len; i++) {
  58                if (!strcmp(commit_formats[i].name, name)) {
  59                        commit_format = &commit_formats[i];
  60                        break;
  61                }
  62        }
  63
  64        if (!commit_format) {
  65                ALLOC_GROW(commit_formats, commit_formats_len+1,
  66                           commit_formats_alloc);
  67                commit_format = &commit_formats[commit_formats_len];
  68                memset(commit_format, 0, sizeof(*commit_format));
  69                commit_formats_len++;
  70        }
  71
  72        commit_format->name = xstrdup(name);
  73        commit_format->format = CMIT_FMT_USERFORMAT;
  74        if (git_config_string(&fmt, var, value))
  75                return -1;
  76
  77        if (skip_prefix(fmt, "format:", &fmt))
  78                commit_format->is_tformat = 0;
  79        else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
  80                commit_format->is_tformat = 1;
  81        else
  82                commit_format->is_alias = 1;
  83        commit_format->user_format = fmt;
  84
  85        return 0;
  86}
  87
  88static void setup_commit_formats(void)
  89{
  90        struct cmt_fmt_map builtin_formats[] = {
  91                { "raw",        CMIT_FMT_RAW,           0,      0 },
  92                { "medium",     CMIT_FMT_MEDIUM,        0,      8 },
  93                { "short",      CMIT_FMT_SHORT,         0,      0 },
  94                { "email",      CMIT_FMT_EMAIL,         0,      0 },
  95                { "mboxrd",     CMIT_FMT_MBOXRD,        0,      0 },
  96                { "fuller",     CMIT_FMT_FULLER,        0,      8 },
  97                { "full",       CMIT_FMT_FULL,          0,      8 },
  98                { "oneline",    CMIT_FMT_ONELINE,       1,      0 }
  99        };
 100        commit_formats_len = ARRAY_SIZE(builtin_formats);
 101        builtin_formats_len = commit_formats_len;
 102        ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
 103        memcpy(commit_formats, builtin_formats,
 104               sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
 105
 106        git_config(git_pretty_formats_config, NULL);
 107}
 108
 109static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 110                                                        const char *original,
 111                                                        int num_redirections)
 112{
 113        struct cmt_fmt_map *found = NULL;
 114        size_t found_match_len = 0;
 115        int i;
 116
 117        if (num_redirections >= commit_formats_len)
 118                die("invalid --pretty format: "
 119                    "'%s' references an alias which points to itself",
 120                    original);
 121
 122        for (i = 0; i < commit_formats_len; i++) {
 123                size_t match_len;
 124
 125                if (!starts_with(commit_formats[i].name, sought))
 126                        continue;
 127
 128                match_len = strlen(commit_formats[i].name);
 129                if (found == NULL || found_match_len > match_len) {
 130                        found = &commit_formats[i];
 131                        found_match_len = match_len;
 132                }
 133        }
 134
 135        if (found && found->is_alias) {
 136                found = find_commit_format_recursive(found->user_format,
 137                                                     original,
 138                                                     num_redirections+1);
 139        }
 140
 141        return found;
 142}
 143
 144static struct cmt_fmt_map *find_commit_format(const char *sought)
 145{
 146        if (!commit_formats)
 147                setup_commit_formats();
 148
 149        return find_commit_format_recursive(sought, sought, 0);
 150}
 151
 152void get_commit_format(const char *arg, struct rev_info *rev)
 153{
 154        struct cmt_fmt_map *commit_format;
 155
 156        rev->use_terminator = 0;
 157        if (!arg) {
 158                rev->commit_format = CMIT_FMT_DEFAULT;
 159                return;
 160        }
 161        if (skip_prefix(arg, "format:", &arg)) {
 162                save_user_format(rev, arg, 0);
 163                return;
 164        }
 165
 166        if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
 167                save_user_format(rev, arg, 1);
 168                return;
 169        }
 170
 171        commit_format = find_commit_format(arg);
 172        if (!commit_format)
 173                die("invalid --pretty format: %s", arg);
 174
 175        rev->commit_format = commit_format->format;
 176        rev->use_terminator = commit_format->is_tformat;
 177        rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
 178        if (commit_format->format == CMIT_FMT_USERFORMAT) {
 179                save_user_format(rev, commit_format->user_format,
 180                                 commit_format->is_tformat);
 181        }
 182}
 183
 184/*
 185 * Generic support for pretty-printing the header
 186 */
 187static int get_one_line(const char *msg)
 188{
 189        int ret = 0;
 190
 191        for (;;) {
 192                char c = *msg++;
 193                if (!c)
 194                        break;
 195                ret++;
 196                if (c == '\n')
 197                        break;
 198        }
 199        return ret;
 200}
 201
 202/* High bit set, or ISO-2022-INT */
 203static int non_ascii(int ch)
 204{
 205        return !isascii(ch) || ch == '\033';
 206}
 207
 208int has_non_ascii(const char *s)
 209{
 210        int ch;
 211        if (!s)
 212                return 0;
 213        while ((ch = *s++) != '\0') {
 214                if (non_ascii(ch))
 215                        return 1;
 216        }
 217        return 0;
 218}
 219
 220static int is_rfc822_special(char ch)
 221{
 222        switch (ch) {
 223        case '(':
 224        case ')':
 225        case '<':
 226        case '>':
 227        case '[':
 228        case ']':
 229        case ':':
 230        case ';':
 231        case '@':
 232        case ',':
 233        case '.':
 234        case '"':
 235        case '\\':
 236                return 1;
 237        default:
 238                return 0;
 239        }
 240}
 241
 242static int needs_rfc822_quoting(const char *s, int len)
 243{
 244        int i;
 245        for (i = 0; i < len; i++)
 246                if (is_rfc822_special(s[i]))
 247                        return 1;
 248        return 0;
 249}
 250
 251static int last_line_length(struct strbuf *sb)
 252{
 253        int i;
 254
 255        /* How many bytes are already used on the last line? */
 256        for (i = sb->len - 1; i >= 0; i--)
 257                if (sb->buf[i] == '\n')
 258                        break;
 259        return sb->len - (i + 1);
 260}
 261
 262static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 263{
 264        int i;
 265
 266        /* just a guess, we may have to also backslash-quote */
 267        strbuf_grow(out, len + 2);
 268
 269        strbuf_addch(out, '"');
 270        for (i = 0; i < len; i++) {
 271                switch (s[i]) {
 272                case '"':
 273                case '\\':
 274                        strbuf_addch(out, '\\');
 275                        /* fall through */
 276                default:
 277                        strbuf_addch(out, s[i]);
 278                }
 279        }
 280        strbuf_addch(out, '"');
 281}
 282
 283enum rfc2047_type {
 284        RFC2047_SUBJECT,
 285        RFC2047_ADDRESS
 286};
 287
 288static int is_rfc2047_special(char ch, enum rfc2047_type type)
 289{
 290        /*
 291         * rfc2047, section 4.2:
 292         *
 293         *    8-bit values which correspond to printable ASCII characters other
 294         *    than "=", "?", and "_" (underscore), MAY be represented as those
 295         *    characters.  (But see section 5 for restrictions.)  In
 296         *    particular, SPACE and TAB MUST NOT be represented as themselves
 297         *    within encoded words.
 298         */
 299
 300        /*
 301         * rule out non-ASCII characters and non-printable characters (the
 302         * non-ASCII check should be redundant as isprint() is not localized
 303         * and only knows about ASCII, but be defensive about that)
 304         */
 305        if (non_ascii(ch) || !isprint(ch))
 306                return 1;
 307
 308        /*
 309         * rule out special printable characters (' ' should be the only
 310         * whitespace character considered printable, but be defensive and use
 311         * isspace())
 312         */
 313        if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 314                return 1;
 315
 316        /*
 317         * rfc2047, section 5.3:
 318         *
 319         *    As a replacement for a 'word' entity within a 'phrase', for example,
 320         *    one that precedes an address in a From, To, or Cc header.  The ABNF
 321         *    definition for 'phrase' from RFC 822 thus becomes:
 322         *
 323         *    phrase = 1*( encoded-word / word )
 324         *
 325         *    In this case the set of characters that may be used in a "Q"-encoded
 326         *    'encoded-word' is restricted to: <upper and lower case ASCII
 327         *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 328         *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 329         *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 330         *    'special' by 'linear-white-space'.
 331         */
 332
 333        if (type != RFC2047_ADDRESS)
 334                return 0;
 335
 336        /* '=' and '_' are special cases and have been checked above */
 337        return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 338}
 339
 340static int needs_rfc2047_encoding(const char *line, int len,
 341                                  enum rfc2047_type type)
 342{
 343        int i;
 344
 345        for (i = 0; i < len; i++) {
 346                int ch = line[i];
 347                if (non_ascii(ch) || ch == '\n')
 348                        return 1;
 349                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 350                        return 1;
 351        }
 352
 353        return 0;
 354}
 355
 356static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 357                       const char *encoding, enum rfc2047_type type)
 358{
 359        static const int max_encoded_length = 76; /* per rfc2047 */
 360        int i;
 361        int line_len = last_line_length(sb);
 362
 363        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 364        strbuf_addf(sb, "=?%s?q?", encoding);
 365        line_len += strlen(encoding) + 5; /* 5 for =??q? */
 366
 367        while (len) {
 368                /*
 369                 * RFC 2047, section 5 (3):
 370                 *
 371                 * Each 'encoded-word' MUST represent an integral number of
 372                 * characters.  A multi-octet character may not be split across
 373                 * adjacent 'encoded- word's.
 374                 */
 375                const unsigned char *p = (const unsigned char *)line;
 376                int chrlen = mbs_chrlen(&line, &len, encoding);
 377                int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 378
 379                /* "=%02X" * chrlen, or the byte itself */
 380                const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 381                int         encoded_len = is_special ? 3 * chrlen : 1;
 382
 383                /*
 384                 * According to RFC 2047, we could encode the special character
 385                 * ' ' (space) with '_' (underscore) for readability. But many
 386                 * programs do not understand this and just leave the
 387                 * underscore in place. Thus, we do nothing special here, which
 388                 * causes ' ' to be encoded as '=20', avoiding this problem.
 389                 */
 390
 391                if (line_len + encoded_len + 2 > max_encoded_length) {
 392                        /* It won't fit with trailing "?=" --- break the line */
 393                        strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 394                        line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 395                }
 396
 397                for (i = 0; i < chrlen; i++)
 398                        strbuf_addf(sb, encoded_fmt, p[i]);
 399                line_len += encoded_len;
 400        }
 401        strbuf_addstr(sb, "?=");
 402}
 403
 404const char *show_ident_date(const struct ident_split *ident,
 405                            const struct date_mode *mode)
 406{
 407        unsigned long date = 0;
 408        long tz = 0;
 409
 410        if (ident->date_begin && ident->date_end)
 411                date = strtoul(ident->date_begin, NULL, 10);
 412        if (date_overflows(date))
 413                date = 0;
 414        else {
 415                if (ident->tz_begin && ident->tz_end)
 416                        tz = strtol(ident->tz_begin, NULL, 10);
 417                if (tz >= INT_MAX || tz <= INT_MIN)
 418                        tz = 0;
 419        }
 420        return show_date(date, tz, mode);
 421}
 422
 423void pp_user_info(struct pretty_print_context *pp,
 424                  const char *what, struct strbuf *sb,
 425                  const char *line, const char *encoding)
 426{
 427        struct ident_split ident;
 428        char *line_end;
 429        const char *mailbuf, *namebuf;
 430        size_t namelen, maillen;
 431        int max_length = 78; /* per rfc2822 */
 432
 433        if (pp->fmt == CMIT_FMT_ONELINE)
 434                return;
 435
 436        line_end = strchrnul(line, '\n');
 437        if (split_ident_line(&ident, line, line_end - line))
 438                return;
 439
 440        mailbuf = ident.mail_begin;
 441        maillen = ident.mail_end - ident.mail_begin;
 442        namebuf = ident.name_begin;
 443        namelen = ident.name_end - ident.name_begin;
 444
 445        if (pp->mailmap)
 446                map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 447
 448        if (cmit_fmt_is_mail(pp->fmt)) {
 449                if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 450                        struct strbuf buf = STRBUF_INIT;
 451
 452                        strbuf_addstr(&buf, "From: ");
 453                        strbuf_add(&buf, namebuf, namelen);
 454                        strbuf_addstr(&buf, " <");
 455                        strbuf_add(&buf, mailbuf, maillen);
 456                        strbuf_addstr(&buf, ">\n");
 457                        string_list_append(&pp->in_body_headers,
 458                                           strbuf_detach(&buf, NULL));
 459
 460                        mailbuf = pp->from_ident->mail_begin;
 461                        maillen = pp->from_ident->mail_end - mailbuf;
 462                        namebuf = pp->from_ident->name_begin;
 463                        namelen = pp->from_ident->name_end - namebuf;
 464                }
 465
 466                strbuf_addstr(sb, "From: ");
 467                if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 468                        add_rfc2047(sb, namebuf, namelen,
 469                                    encoding, RFC2047_ADDRESS);
 470                        max_length = 76; /* per rfc2047 */
 471                } else if (needs_rfc822_quoting(namebuf, namelen)) {
 472                        struct strbuf quoted = STRBUF_INIT;
 473                        add_rfc822_quoted(&quoted, namebuf, namelen);
 474                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 475                                                        -6, 1, max_length);
 476                        strbuf_release(&quoted);
 477                } else {
 478                        strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 479                                                 -6, 1, max_length);
 480                }
 481
 482                if (max_length <
 483                    last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 484                        strbuf_addch(sb, '\n');
 485                strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 486        } else {
 487                strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 488                            (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 489                            (int)namelen, namebuf, (int)maillen, mailbuf);
 490        }
 491
 492        switch (pp->fmt) {
 493        case CMIT_FMT_MEDIUM:
 494                strbuf_addf(sb, "Date:   %s\n",
 495                            show_ident_date(&ident, &pp->date_mode));
 496                break;
 497        case CMIT_FMT_EMAIL:
 498        case CMIT_FMT_MBOXRD:
 499                strbuf_addf(sb, "Date: %s\n",
 500                            show_ident_date(&ident, DATE_MODE(RFC2822)));
 501                break;
 502        case CMIT_FMT_FULLER:
 503                strbuf_addf(sb, "%sDate: %s\n", what,
 504                            show_ident_date(&ident, &pp->date_mode));
 505                break;
 506        default:
 507                /* notin' */
 508                break;
 509        }
 510}
 511
 512static int is_blank_line(const char *line, int *len_p)
 513{
 514        int len = *len_p;
 515        while (len && isspace(line[len - 1]))
 516                len--;
 517        *len_p = len;
 518        return !len;
 519}
 520
 521const char *skip_blank_lines(const char *msg)
 522{
 523        for (;;) {
 524                int linelen = get_one_line(msg);
 525                int ll = linelen;
 526                if (!linelen)
 527                        break;
 528                if (!is_blank_line(msg, &ll))
 529                        break;
 530                msg += linelen;
 531        }
 532        return msg;
 533}
 534
 535static void add_merge_info(const struct pretty_print_context *pp,
 536                           struct strbuf *sb, const struct commit *commit)
 537{
 538        struct commit_list *parent = commit->parents;
 539
 540        if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
 541            !parent || !parent->next)
 542                return;
 543
 544        strbuf_addstr(sb, "Merge:");
 545
 546        while (parent) {
 547                struct commit *p = parent->item;
 548                const char *hex = NULL;
 549                if (pp->abbrev)
 550                        hex = find_unique_abbrev(p->object.oid.hash, pp->abbrev);
 551                if (!hex)
 552                        hex = oid_to_hex(&p->object.oid);
 553                parent = parent->next;
 554
 555                strbuf_addf(sb, " %s", hex);
 556        }
 557        strbuf_addch(sb, '\n');
 558}
 559
 560static char *get_header(const char *msg, const char *key)
 561{
 562        size_t len;
 563        const char *v = find_commit_header(msg, key, &len);
 564        return v ? xmemdupz(v, len) : NULL;
 565}
 566
 567static char *replace_encoding_header(char *buf, const char *encoding)
 568{
 569        struct strbuf tmp = STRBUF_INIT;
 570        size_t start, len;
 571        char *cp = buf;
 572
 573        /* guess if there is an encoding header before a \n\n */
 574        while (!starts_with(cp, "encoding ")) {
 575                cp = strchr(cp, '\n');
 576                if (!cp || *++cp == '\n')
 577                        return buf;
 578        }
 579        start = cp - buf;
 580        cp = strchr(cp, '\n');
 581        if (!cp)
 582                return buf; /* should not happen but be defensive */
 583        len = cp + 1 - (buf + start);
 584
 585        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 586        if (is_encoding_utf8(encoding)) {
 587                /* we have re-coded to UTF-8; drop the header */
 588                strbuf_remove(&tmp, start, len);
 589        } else {
 590                /* just replaces XXXX in 'encoding XXXX\n' */
 591                strbuf_splice(&tmp, start + strlen("encoding "),
 592                                          len - strlen("encoding \n"),
 593                                          encoding, strlen(encoding));
 594        }
 595        return strbuf_detach(&tmp, NULL);
 596}
 597
 598const char *logmsg_reencode(const struct commit *commit,
 599                            char **commit_encoding,
 600                            const char *output_encoding)
 601{
 602        static const char *utf8 = "UTF-8";
 603        const char *use_encoding;
 604        char *encoding;
 605        const char *msg = get_commit_buffer(commit, NULL);
 606        char *out;
 607
 608        if (!output_encoding || !*output_encoding) {
 609                if (commit_encoding)
 610                        *commit_encoding = get_header(msg, "encoding");
 611                return msg;
 612        }
 613        encoding = get_header(msg, "encoding");
 614        if (commit_encoding)
 615                *commit_encoding = encoding;
 616        use_encoding = encoding ? encoding : utf8;
 617        if (same_encoding(use_encoding, output_encoding)) {
 618                /*
 619                 * No encoding work to be done. If we have no encoding header
 620                 * at all, then there's nothing to do, and we can return the
 621                 * message verbatim (whether newly allocated or not).
 622                 */
 623                if (!encoding)
 624                        return msg;
 625
 626                /*
 627                 * Otherwise, we still want to munge the encoding header in the
 628                 * result, which will be done by modifying the buffer. If we
 629                 * are using a fresh copy, we can reuse it. But if we are using
 630                 * the cached copy from get_commit_buffer, we need to duplicate it
 631                 * to avoid munging the cached copy.
 632                 */
 633                if (msg == get_cached_commit_buffer(commit, NULL))
 634                        out = xstrdup(msg);
 635                else
 636                        out = (char *)msg;
 637        }
 638        else {
 639                /*
 640                 * There's actual encoding work to do. Do the reencoding, which
 641                 * still leaves the header to be replaced in the next step. At
 642                 * this point, we are done with msg. If we allocated a fresh
 643                 * copy, we can free it.
 644                 */
 645                out = reencode_string(msg, output_encoding, use_encoding);
 646                if (out)
 647                        unuse_commit_buffer(commit, msg);
 648        }
 649
 650        /*
 651         * This replacement actually consumes the buffer we hand it, so we do
 652         * not have to worry about freeing the old "out" here.
 653         */
 654        if (out)
 655                out = replace_encoding_header(out, output_encoding);
 656
 657        if (!commit_encoding)
 658                free(encoding);
 659        /*
 660         * If the re-encoding failed, out might be NULL here; in that
 661         * case we just return the commit message verbatim.
 662         */
 663        return out ? out : msg;
 664}
 665
 666static int mailmap_name(const char **email, size_t *email_len,
 667                        const char **name, size_t *name_len)
 668{
 669        static struct string_list *mail_map;
 670        if (!mail_map) {
 671                mail_map = xcalloc(1, sizeof(*mail_map));
 672                read_mailmap(mail_map, NULL);
 673        }
 674        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 675}
 676
 677static size_t format_person_part(struct strbuf *sb, char part,
 678                                 const char *msg, int len,
 679                                 const struct date_mode *dmode)
 680{
 681        /* currently all placeholders have same length */
 682        const int placeholder_len = 2;
 683        struct ident_split s;
 684        const char *name, *mail;
 685        size_t maillen, namelen;
 686
 687        if (split_ident_line(&s, msg, len) < 0)
 688                goto skip;
 689
 690        name = s.name_begin;
 691        namelen = s.name_end - s.name_begin;
 692        mail = s.mail_begin;
 693        maillen = s.mail_end - s.mail_begin;
 694
 695        if (part == 'N' || part == 'E') /* mailmap lookup */
 696                mailmap_name(&mail, &maillen, &name, &namelen);
 697        if (part == 'n' || part == 'N') {       /* name */
 698                strbuf_add(sb, name, namelen);
 699                return placeholder_len;
 700        }
 701        if (part == 'e' || part == 'E') {       /* email */
 702                strbuf_add(sb, mail, maillen);
 703                return placeholder_len;
 704        }
 705
 706        if (!s.date_begin)
 707                goto skip;
 708
 709        if (part == 't') {      /* date, UNIX timestamp */
 710                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 711                return placeholder_len;
 712        }
 713
 714        switch (part) {
 715        case 'd':       /* date */
 716                strbuf_addstr(sb, show_ident_date(&s, dmode));
 717                return placeholder_len;
 718        case 'D':       /* date, RFC2822 style */
 719                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
 720                return placeholder_len;
 721        case 'r':       /* date, relative */
 722                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
 723                return placeholder_len;
 724        case 'i':       /* date, ISO 8601-like */
 725                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
 726                return placeholder_len;
 727        case 'I':       /* date, ISO 8601 strict */
 728                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
 729                return placeholder_len;
 730        }
 731
 732skip:
 733        /*
 734         * reading from either a bogus commit, or a reflog entry with
 735         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 736         * to compute a valid return value.
 737         */
 738        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 739            || part == 'D' || part == 'r' || part == 'i')
 740                return placeholder_len;
 741
 742        return 0; /* unknown placeholder */
 743}
 744
 745struct chunk {
 746        size_t off;
 747        size_t len;
 748};
 749
 750enum flush_type {
 751        no_flush,
 752        flush_right,
 753        flush_left,
 754        flush_left_and_steal,
 755        flush_both
 756};
 757
 758enum trunc_type {
 759        trunc_none,
 760        trunc_left,
 761        trunc_middle,
 762        trunc_right
 763};
 764
 765struct format_commit_context {
 766        const struct commit *commit;
 767        const struct pretty_print_context *pretty_ctx;
 768        unsigned commit_header_parsed:1;
 769        unsigned commit_message_parsed:1;
 770        struct signature_check signature_check;
 771        enum flush_type flush_type;
 772        enum trunc_type truncate;
 773        const char *message;
 774        char *commit_encoding;
 775        size_t width, indent1, indent2;
 776        int auto_color;
 777        int padding;
 778
 779        /* These offsets are relative to the start of the commit message. */
 780        struct chunk author;
 781        struct chunk committer;
 782        size_t message_off;
 783        size_t subject_off;
 784        size_t body_off;
 785
 786        /* The following ones are relative to the result struct strbuf. */
 787        struct chunk abbrev_commit_hash;
 788        struct chunk abbrev_tree_hash;
 789        struct chunk abbrev_parent_hashes;
 790        size_t wrap_start;
 791};
 792
 793static int add_again(struct strbuf *sb, struct chunk *chunk)
 794{
 795        if (chunk->len) {
 796                strbuf_adddup(sb, chunk->off, chunk->len);
 797                return 1;
 798        }
 799
 800        /*
 801         * We haven't seen this chunk before.  Our caller is surely
 802         * going to add it the hard way now.  Remember the most likely
 803         * start of the to-be-added chunk: the current end of the
 804         * struct strbuf.
 805         */
 806        chunk->off = sb->len;
 807        return 0;
 808}
 809
 810static void parse_commit_header(struct format_commit_context *context)
 811{
 812        const char *msg = context->message;
 813        int i;
 814
 815        for (i = 0; msg[i]; i++) {
 816                const char *name;
 817                int eol;
 818                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 819                        ; /* do nothing */
 820
 821                if (i == eol) {
 822                        break;
 823                } else if (skip_prefix(msg + i, "author ", &name)) {
 824                        context->author.off = name - msg;
 825                        context->author.len = msg + eol - name;
 826                } else if (skip_prefix(msg + i, "committer ", &name)) {
 827                        context->committer.off = name - msg;
 828                        context->committer.len = msg + eol - name;
 829                }
 830                i = eol;
 831        }
 832        context->message_off = i;
 833        context->commit_header_parsed = 1;
 834}
 835
 836static int istitlechar(char c)
 837{
 838        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 839                (c >= '0' && c <= '9') || c == '.' || c == '_';
 840}
 841
 842static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 843{
 844        size_t trimlen;
 845        size_t start_len = sb->len;
 846        int space = 2;
 847
 848        for (; *msg && *msg != '\n'; msg++) {
 849                if (istitlechar(*msg)) {
 850                        if (space == 1)
 851                                strbuf_addch(sb, '-');
 852                        space = 0;
 853                        strbuf_addch(sb, *msg);
 854                        if (*msg == '.')
 855                                while (*(msg+1) == '.')
 856                                        msg++;
 857                } else
 858                        space |= 1;
 859        }
 860
 861        /* trim any trailing '.' or '-' characters */
 862        trimlen = 0;
 863        while (sb->len - trimlen > start_len &&
 864                (sb->buf[sb->len - 1 - trimlen] == '.'
 865                || sb->buf[sb->len - 1 - trimlen] == '-'))
 866                trimlen++;
 867        strbuf_remove(sb, sb->len - trimlen, trimlen);
 868}
 869
 870const char *format_subject(struct strbuf *sb, const char *msg,
 871                           const char *line_separator)
 872{
 873        int first = 1;
 874
 875        for (;;) {
 876                const char *line = msg;
 877                int linelen = get_one_line(line);
 878
 879                msg += linelen;
 880                if (!linelen || is_blank_line(line, &linelen))
 881                        break;
 882
 883                if (!sb)
 884                        continue;
 885                strbuf_grow(sb, linelen + 2);
 886                if (!first)
 887                        strbuf_addstr(sb, line_separator);
 888                strbuf_add(sb, line, linelen);
 889                first = 0;
 890        }
 891        return msg;
 892}
 893
 894static void parse_commit_message(struct format_commit_context *c)
 895{
 896        const char *msg = c->message + c->message_off;
 897        const char *start = c->message;
 898
 899        msg = skip_blank_lines(msg);
 900        c->subject_off = msg - start;
 901
 902        msg = format_subject(NULL, msg, NULL);
 903        msg = skip_blank_lines(msg);
 904        c->body_off = msg - start;
 905
 906        c->commit_message_parsed = 1;
 907}
 908
 909static void strbuf_wrap(struct strbuf *sb, size_t pos,
 910                        size_t width, size_t indent1, size_t indent2)
 911{
 912        struct strbuf tmp = STRBUF_INIT;
 913
 914        if (pos)
 915                strbuf_add(&tmp, sb->buf, pos);
 916        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 917                                (int) indent1, (int) indent2, (int) width);
 918        strbuf_swap(&tmp, sb);
 919        strbuf_release(&tmp);
 920}
 921
 922static void rewrap_message_tail(struct strbuf *sb,
 923                                struct format_commit_context *c,
 924                                size_t new_width, size_t new_indent1,
 925                                size_t new_indent2)
 926{
 927        if (c->width == new_width && c->indent1 == new_indent1 &&
 928            c->indent2 == new_indent2)
 929                return;
 930        if (c->wrap_start < sb->len)
 931                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 932        c->wrap_start = sb->len;
 933        c->width = new_width;
 934        c->indent1 = new_indent1;
 935        c->indent2 = new_indent2;
 936}
 937
 938static int format_reflog_person(struct strbuf *sb,
 939                                char part,
 940                                struct reflog_walk_info *log,
 941                                const struct date_mode *dmode)
 942{
 943        const char *ident;
 944
 945        if (!log)
 946                return 2;
 947
 948        ident = get_reflog_ident(log);
 949        if (!ident)
 950                return 2;
 951
 952        return format_person_part(sb, part, ident, strlen(ident), dmode);
 953}
 954
 955static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 956                          const char *placeholder,
 957                          struct format_commit_context *c)
 958{
 959        const char *rest = placeholder;
 960
 961        if (placeholder[1] == '(') {
 962                const char *begin = placeholder + 2;
 963                const char *end = strchr(begin, ')');
 964                char color[COLOR_MAXLEN];
 965
 966                if (!end)
 967                        return 0;
 968                if (skip_prefix(begin, "auto,", &begin)) {
 969                        if (!want_color(c->pretty_ctx->color))
 970                                return end - placeholder + 1;
 971                }
 972                if (color_parse_mem(begin, end - begin, color) < 0)
 973                        die(_("unable to parse --pretty format"));
 974                strbuf_addstr(sb, color);
 975                return end - placeholder + 1;
 976        }
 977        if (skip_prefix(placeholder + 1, "red", &rest))
 978                strbuf_addstr(sb, GIT_COLOR_RED);
 979        else if (skip_prefix(placeholder + 1, "green", &rest))
 980                strbuf_addstr(sb, GIT_COLOR_GREEN);
 981        else if (skip_prefix(placeholder + 1, "blue", &rest))
 982                strbuf_addstr(sb, GIT_COLOR_BLUE);
 983        else if (skip_prefix(placeholder + 1, "reset", &rest))
 984                strbuf_addstr(sb, GIT_COLOR_RESET);
 985        return rest - placeholder;
 986}
 987
 988static size_t parse_padding_placeholder(struct strbuf *sb,
 989                                        const char *placeholder,
 990                                        struct format_commit_context *c)
 991{
 992        const char *ch = placeholder;
 993        enum flush_type flush_type;
 994        int to_column = 0;
 995
 996        switch (*ch++) {
 997        case '<':
 998                flush_type = flush_right;
 999                break;
1000        case '>':
1001                if (*ch == '<') {
1002                        flush_type = flush_both;
1003                        ch++;
1004                } else if (*ch == '>') {
1005                        flush_type = flush_left_and_steal;
1006                        ch++;
1007                } else
1008                        flush_type = flush_left;
1009                break;
1010        default:
1011                return 0;
1012        }
1013
1014        /* the next value means "wide enough to that column" */
1015        if (*ch == '|') {
1016                to_column = 1;
1017                ch++;
1018        }
1019
1020        if (*ch == '(') {
1021                const char *start = ch + 1;
1022                const char *end = start + strcspn(start, ",)");
1023                char *next;
1024                int width;
1025                if (!end || end == start)
1026                        return 0;
1027                width = strtol(start, &next, 10);
1028                if (next == start || width == 0)
1029                        return 0;
1030                if (width < 0) {
1031                        if (to_column)
1032                                width += term_columns();
1033                        if (width < 0)
1034                                return 0;
1035                }
1036                c->padding = to_column ? -width : width;
1037                c->flush_type = flush_type;
1038
1039                if (*end == ',') {
1040                        start = end + 1;
1041                        end = strchr(start, ')');
1042                        if (!end || end == start)
1043                                return 0;
1044                        if (starts_with(start, "trunc)"))
1045                                c->truncate = trunc_right;
1046                        else if (starts_with(start, "ltrunc)"))
1047                                c->truncate = trunc_left;
1048                        else if (starts_with(start, "mtrunc)"))
1049                                c->truncate = trunc_middle;
1050                        else
1051                                return 0;
1052                } else
1053                        c->truncate = trunc_none;
1054
1055                return end - placeholder + 1;
1056        }
1057        return 0;
1058}
1059
1060static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1061                                const char *placeholder,
1062                                void *context)
1063{
1064        struct format_commit_context *c = context;
1065        const struct commit *commit = c->commit;
1066        const char *msg = c->message;
1067        struct commit_list *p;
1068        int ch;
1069
1070        /* these are independent of the commit */
1071        switch (placeholder[0]) {
1072        case 'C':
1073                if (starts_with(placeholder + 1, "(auto)")) {
1074                        c->auto_color = want_color(c->pretty_ctx->color);
1075                        if (c->auto_color)
1076                                strbuf_addstr(sb, GIT_COLOR_RESET);
1077                        return 7; /* consumed 7 bytes, "C(auto)" */
1078                } else {
1079                        int ret = parse_color(sb, placeholder, c);
1080                        if (ret)
1081                                c->auto_color = 0;
1082                        /*
1083                         * Otherwise, we decided to treat %C<unknown>
1084                         * as a literal string, and the previous
1085                         * %C(auto) is still valid.
1086                         */
1087                        return ret;
1088                }
1089        case 'n':               /* newline */
1090                strbuf_addch(sb, '\n');
1091                return 1;
1092        case 'x':
1093                /* %x00 == NUL, %x0a == LF, etc. */
1094                ch = hex2chr(placeholder + 1);
1095                if (ch < 0)
1096                        return 0;
1097                strbuf_addch(sb, ch);
1098                return 3;
1099        case 'w':
1100                if (placeholder[1] == '(') {
1101                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1102                        char *next;
1103                        const char *start = placeholder + 2;
1104                        const char *end = strchr(start, ')');
1105                        if (!end)
1106                                return 0;
1107                        if (end > start) {
1108                                width = strtoul(start, &next, 10);
1109                                if (*next == ',') {
1110                                        indent1 = strtoul(next + 1, &next, 10);
1111                                        if (*next == ',') {
1112                                                indent2 = strtoul(next + 1,
1113                                                                 &next, 10);
1114                                        }
1115                                }
1116                                if (*next != ')')
1117                                        return 0;
1118                        }
1119                        rewrap_message_tail(sb, c, width, indent1, indent2);
1120                        return end - placeholder + 1;
1121                } else
1122                        return 0;
1123
1124        case '<':
1125        case '>':
1126                return parse_padding_placeholder(sb, placeholder, c);
1127        }
1128
1129        /* these depend on the commit */
1130        if (!commit->object.parsed)
1131                parse_object(commit->object.oid.hash);
1132
1133        switch (placeholder[0]) {
1134        case 'H':               /* commit hash */
1135                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1136                strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1137                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1138                return 1;
1139        case 'h':               /* abbreviated commit hash */
1140                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1141                if (add_again(sb, &c->abbrev_commit_hash)) {
1142                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1143                        return 1;
1144                }
1145                strbuf_add_unique_abbrev(sb, commit->object.oid.hash,
1146                                         c->pretty_ctx->abbrev);
1147                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1148                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1149                return 1;
1150        case 'T':               /* tree hash */
1151                strbuf_addstr(sb, oid_to_hex(&commit->tree->object.oid));
1152                return 1;
1153        case 't':               /* abbreviated tree hash */
1154                if (add_again(sb, &c->abbrev_tree_hash))
1155                        return 1;
1156                strbuf_add_unique_abbrev(sb, commit->tree->object.oid.hash,
1157                                         c->pretty_ctx->abbrev);
1158                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1159                return 1;
1160        case 'P':               /* parent hashes */
1161                for (p = commit->parents; p; p = p->next) {
1162                        if (p != commit->parents)
1163                                strbuf_addch(sb, ' ');
1164                        strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1165                }
1166                return 1;
1167        case 'p':               /* abbreviated parent hashes */
1168                if (add_again(sb, &c->abbrev_parent_hashes))
1169                        return 1;
1170                for (p = commit->parents; p; p = p->next) {
1171                        if (p != commit->parents)
1172                                strbuf_addch(sb, ' ');
1173                        strbuf_add_unique_abbrev(sb, p->item->object.oid.hash,
1174                                                 c->pretty_ctx->abbrev);
1175                }
1176                c->abbrev_parent_hashes.len = sb->len -
1177                                              c->abbrev_parent_hashes.off;
1178                return 1;
1179        case 'm':               /* left/right/bottom */
1180                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1181                return 1;
1182        case 'd':
1183                load_ref_decorations(DECORATE_SHORT_REFS);
1184                format_decorations(sb, commit, c->auto_color);
1185                return 1;
1186        case 'D':
1187                load_ref_decorations(DECORATE_SHORT_REFS);
1188                format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1189                return 1;
1190        case 'g':               /* reflog info */
1191                switch(placeholder[1]) {
1192                case 'd':       /* reflog selector */
1193                case 'D':
1194                        if (c->pretty_ctx->reflog_info)
1195                                get_reflog_selector(sb,
1196                                                    c->pretty_ctx->reflog_info,
1197                                                    &c->pretty_ctx->date_mode,
1198                                                    c->pretty_ctx->date_mode_explicit,
1199                                                    (placeholder[1] == 'd'));
1200                        return 2;
1201                case 's':       /* reflog message */
1202                        if (c->pretty_ctx->reflog_info)
1203                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1204                        return 2;
1205                case 'n':
1206                case 'N':
1207                case 'e':
1208                case 'E':
1209                        return format_reflog_person(sb,
1210                                                    placeholder[1],
1211                                                    c->pretty_ctx->reflog_info,
1212                                                    &c->pretty_ctx->date_mode);
1213                }
1214                return 0;       /* unknown %g placeholder */
1215        case 'N':
1216                if (c->pretty_ctx->notes_message) {
1217                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1218                        return 1;
1219                }
1220                return 0;
1221        }
1222
1223        if (placeholder[0] == 'G') {
1224                if (!c->signature_check.result)
1225                        check_commit_signature(c->commit, &(c->signature_check));
1226                switch (placeholder[1]) {
1227                case 'G':
1228                        if (c->signature_check.gpg_output)
1229                                strbuf_addstr(sb, c->signature_check.gpg_output);
1230                        break;
1231                case '?':
1232                        switch (c->signature_check.result) {
1233                        case 'G':
1234                        case 'B':
1235                        case 'U':
1236                        case 'N':
1237                                strbuf_addch(sb, c->signature_check.result);
1238                        }
1239                        break;
1240                case 'S':
1241                        if (c->signature_check.signer)
1242                                strbuf_addstr(sb, c->signature_check.signer);
1243                        break;
1244                case 'K':
1245                        if (c->signature_check.key)
1246                                strbuf_addstr(sb, c->signature_check.key);
1247                        break;
1248                default:
1249                        return 0;
1250                }
1251                return 2;
1252        }
1253
1254
1255        /* For the rest we have to parse the commit header. */
1256        if (!c->commit_header_parsed)
1257                parse_commit_header(c);
1258
1259        switch (placeholder[0]) {
1260        case 'a':       /* author ... */
1261                return format_person_part(sb, placeholder[1],
1262                                   msg + c->author.off, c->author.len,
1263                                   &c->pretty_ctx->date_mode);
1264        case 'c':       /* committer ... */
1265                return format_person_part(sb, placeholder[1],
1266                                   msg + c->committer.off, c->committer.len,
1267                                   &c->pretty_ctx->date_mode);
1268        case 'e':       /* encoding */
1269                if (c->commit_encoding)
1270                        strbuf_addstr(sb, c->commit_encoding);
1271                return 1;
1272        case 'B':       /* raw body */
1273                /* message_off is always left at the initial newline */
1274                strbuf_addstr(sb, msg + c->message_off + 1);
1275                return 1;
1276        }
1277
1278        /* Now we need to parse the commit message. */
1279        if (!c->commit_message_parsed)
1280                parse_commit_message(c);
1281
1282        switch (placeholder[0]) {
1283        case 's':       /* subject */
1284                format_subject(sb, msg + c->subject_off, " ");
1285                return 1;
1286        case 'f':       /* sanitized subject */
1287                format_sanitized_subject(sb, msg + c->subject_off);
1288                return 1;
1289        case 'b':       /* body */
1290                strbuf_addstr(sb, msg + c->body_off);
1291                return 1;
1292        }
1293        return 0;       /* unknown placeholder */
1294}
1295
1296static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1297                                    const char *placeholder,
1298                                    struct format_commit_context *c)
1299{
1300        struct strbuf local_sb = STRBUF_INIT;
1301        int total_consumed = 0, len, padding = c->padding;
1302        if (padding < 0) {
1303                const char *start = strrchr(sb->buf, '\n');
1304                int occupied;
1305                if (!start)
1306                        start = sb->buf;
1307                occupied = utf8_strnwidth(start, -1, 1);
1308                occupied += c->pretty_ctx->graph_width;
1309                padding = (-padding) - occupied;
1310        }
1311        while (1) {
1312                int modifier = *placeholder == 'C';
1313                int consumed = format_commit_one(&local_sb, placeholder, c);
1314                total_consumed += consumed;
1315
1316                if (!modifier)
1317                        break;
1318
1319                placeholder += consumed;
1320                if (*placeholder != '%')
1321                        break;
1322                placeholder++;
1323                total_consumed++;
1324        }
1325        len = utf8_strnwidth(local_sb.buf, -1, 1);
1326
1327        if (c->flush_type == flush_left_and_steal) {
1328                const char *ch = sb->buf + sb->len - 1;
1329                while (len > padding && ch > sb->buf) {
1330                        const char *p;
1331                        if (*ch == ' ') {
1332                                ch--;
1333                                padding++;
1334                                continue;
1335                        }
1336                        /* check for trailing ansi sequences */
1337                        if (*ch != 'm')
1338                                break;
1339                        p = ch - 1;
1340                        while (ch - p < 10 && *p != '\033')
1341                                p--;
1342                        if (*p != '\033' ||
1343                            ch + 1 - p != display_mode_esc_sequence_len(p))
1344                                break;
1345                        /*
1346                         * got a good ansi sequence, put it back to
1347                         * local_sb as we're cutting sb
1348                         */
1349                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1350                        ch = p - 1;
1351                }
1352                strbuf_setlen(sb, ch + 1 - sb->buf);
1353                c->flush_type = flush_left;
1354        }
1355
1356        if (len > padding) {
1357                switch (c->truncate) {
1358                case trunc_left:
1359                        strbuf_utf8_replace(&local_sb,
1360                                            0, len - (padding - 2),
1361                                            "..");
1362                        break;
1363                case trunc_middle:
1364                        strbuf_utf8_replace(&local_sb,
1365                                            padding / 2 - 1,
1366                                            len - (padding - 2),
1367                                            "..");
1368                        break;
1369                case trunc_right:
1370                        strbuf_utf8_replace(&local_sb,
1371                                            padding - 2, len - (padding - 2),
1372                                            "..");
1373                        break;
1374                case trunc_none:
1375                        break;
1376                }
1377                strbuf_addbuf(sb, &local_sb);
1378        } else {
1379                int sb_len = sb->len, offset = 0;
1380                if (c->flush_type == flush_left)
1381                        offset = padding - len;
1382                else if (c->flush_type == flush_both)
1383                        offset = (padding - len) / 2;
1384                /*
1385                 * we calculate padding in columns, now
1386                 * convert it back to chars
1387                 */
1388                padding = padding - len + local_sb.len;
1389                strbuf_addchars(sb, ' ', padding);
1390                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1391                       local_sb.len);
1392        }
1393        strbuf_release(&local_sb);
1394        c->flush_type = no_flush;
1395        return total_consumed;
1396}
1397
1398static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1399                                 const char *placeholder,
1400                                 void *context)
1401{
1402        int consumed;
1403        size_t orig_len;
1404        enum {
1405                NO_MAGIC,
1406                ADD_LF_BEFORE_NON_EMPTY,
1407                DEL_LF_BEFORE_EMPTY,
1408                ADD_SP_BEFORE_NON_EMPTY
1409        } magic = NO_MAGIC;
1410
1411        switch (placeholder[0]) {
1412        case '-':
1413                magic = DEL_LF_BEFORE_EMPTY;
1414                break;
1415        case '+':
1416                magic = ADD_LF_BEFORE_NON_EMPTY;
1417                break;
1418        case ' ':
1419                magic = ADD_SP_BEFORE_NON_EMPTY;
1420                break;
1421        default:
1422                break;
1423        }
1424        if (magic != NO_MAGIC)
1425                placeholder++;
1426
1427        orig_len = sb->len;
1428        if (((struct format_commit_context *)context)->flush_type != no_flush)
1429                consumed = format_and_pad_commit(sb, placeholder, context);
1430        else
1431                consumed = format_commit_one(sb, placeholder, context);
1432        if (magic == NO_MAGIC)
1433                return consumed;
1434
1435        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1436                while (sb->len && sb->buf[sb->len - 1] == '\n')
1437                        strbuf_setlen(sb, sb->len - 1);
1438        } else if (orig_len != sb->len) {
1439                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1440                        strbuf_insert(sb, orig_len, "\n", 1);
1441                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1442                        strbuf_insert(sb, orig_len, " ", 1);
1443        }
1444        return consumed + 1;
1445}
1446
1447static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1448                                   void *context)
1449{
1450        struct userformat_want *w = context;
1451
1452        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1453                placeholder++;
1454
1455        switch (*placeholder) {
1456        case 'N':
1457                w->notes = 1;
1458                break;
1459        }
1460        return 0;
1461}
1462
1463void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1464{
1465        struct strbuf dummy = STRBUF_INIT;
1466
1467        if (!fmt) {
1468                if (!user_format)
1469                        return;
1470                fmt = user_format;
1471        }
1472        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1473        strbuf_release(&dummy);
1474}
1475
1476void format_commit_message(const struct commit *commit,
1477                           const char *format, struct strbuf *sb,
1478                           const struct pretty_print_context *pretty_ctx)
1479{
1480        struct format_commit_context context;
1481        const char *output_enc = pretty_ctx->output_encoding;
1482        const char *utf8 = "UTF-8";
1483
1484        memset(&context, 0, sizeof(context));
1485        context.commit = commit;
1486        context.pretty_ctx = pretty_ctx;
1487        context.wrap_start = sb->len;
1488        /*
1489         * convert a commit message to UTF-8 first
1490         * as far as 'format_commit_item' assumes it in UTF-8
1491         */
1492        context.message = logmsg_reencode(commit,
1493                                          &context.commit_encoding,
1494                                          utf8);
1495
1496        strbuf_expand(sb, format, format_commit_item, &context);
1497        rewrap_message_tail(sb, &context, 0, 0, 0);
1498
1499        /* then convert a commit message to an actual output encoding */
1500        if (output_enc) {
1501                if (same_encoding(utf8, output_enc))
1502                        output_enc = NULL;
1503        } else {
1504                if (context.commit_encoding &&
1505                    !same_encoding(context.commit_encoding, utf8))
1506                        output_enc = context.commit_encoding;
1507        }
1508
1509        if (output_enc) {
1510                int outsz;
1511                char *out = reencode_string_len(sb->buf, sb->len,
1512                                                output_enc, utf8, &outsz);
1513                if (out)
1514                        strbuf_attach(sb, out, outsz, outsz + 1);
1515        }
1516
1517        free(context.commit_encoding);
1518        unuse_commit_buffer(commit, context.message);
1519}
1520
1521static void pp_header(struct pretty_print_context *pp,
1522                      const char *encoding,
1523                      const struct commit *commit,
1524                      const char **msg_p,
1525                      struct strbuf *sb)
1526{
1527        int parents_shown = 0;
1528
1529        for (;;) {
1530                const char *name, *line = *msg_p;
1531                int linelen = get_one_line(*msg_p);
1532
1533                if (!linelen)
1534                        return;
1535                *msg_p += linelen;
1536
1537                if (linelen == 1)
1538                        /* End of header */
1539                        return;
1540
1541                if (pp->fmt == CMIT_FMT_RAW) {
1542                        strbuf_add(sb, line, linelen);
1543                        continue;
1544                }
1545
1546                if (starts_with(line, "parent ")) {
1547                        if (linelen != 48)
1548                                die("bad parent line in commit");
1549                        continue;
1550                }
1551
1552                if (!parents_shown) {
1553                        unsigned num = commit_list_count(commit->parents);
1554                        /* with enough slop */
1555                        strbuf_grow(sb, num * 50 + 20);
1556                        add_merge_info(pp, sb, commit);
1557                        parents_shown = 1;
1558                }
1559
1560                /*
1561                 * MEDIUM == DEFAULT shows only author with dates.
1562                 * FULL shows both authors but not dates.
1563                 * FULLER shows both authors and dates.
1564                 */
1565                if (skip_prefix(line, "author ", &name)) {
1566                        strbuf_grow(sb, linelen + 80);
1567                        pp_user_info(pp, "Author", sb, name, encoding);
1568                }
1569                if (skip_prefix(line, "committer ", &name) &&
1570                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1571                        strbuf_grow(sb, linelen + 80);
1572                        pp_user_info(pp, "Commit", sb, name, encoding);
1573                }
1574        }
1575}
1576
1577void pp_title_line(struct pretty_print_context *pp,
1578                   const char **msg_p,
1579                   struct strbuf *sb,
1580                   const char *encoding,
1581                   int need_8bit_cte)
1582{
1583        static const int max_length = 78; /* per rfc2047 */
1584        struct strbuf title;
1585
1586        strbuf_init(&title, 80);
1587        *msg_p = format_subject(&title, *msg_p,
1588                                pp->preserve_subject ? "\n" : " ");
1589
1590        strbuf_grow(sb, title.len + 1024);
1591        if (pp->subject) {
1592                strbuf_addstr(sb, pp->subject);
1593                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1594                        add_rfc2047(sb, title.buf, title.len,
1595                                                encoding, RFC2047_SUBJECT);
1596                else
1597                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1598                                         -last_line_length(sb), 1, max_length);
1599        } else {
1600                strbuf_addbuf(sb, &title);
1601        }
1602        strbuf_addch(sb, '\n');
1603
1604        if (need_8bit_cte == 0) {
1605                int i;
1606                for (i = 0; i < pp->in_body_headers.nr; i++) {
1607                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1608                                need_8bit_cte = 1;
1609                                break;
1610                        }
1611                }
1612        }
1613
1614        if (need_8bit_cte > 0) {
1615                const char *header_fmt =
1616                        "MIME-Version: 1.0\n"
1617                        "Content-Type: text/plain; charset=%s\n"
1618                        "Content-Transfer-Encoding: 8bit\n";
1619                strbuf_addf(sb, header_fmt, encoding);
1620        }
1621        if (pp->after_subject) {
1622                strbuf_addstr(sb, pp->after_subject);
1623        }
1624        if (cmit_fmt_is_mail(pp->fmt)) {
1625                strbuf_addch(sb, '\n');
1626        }
1627
1628        if (pp->in_body_headers.nr) {
1629                int i;
1630                for (i = 0; i < pp->in_body_headers.nr; i++) {
1631                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1632                        free(pp->in_body_headers.items[i].string);
1633                }
1634                string_list_clear(&pp->in_body_headers, 0);
1635                strbuf_addch(sb, '\n');
1636        }
1637
1638        strbuf_release(&title);
1639}
1640
1641static int pp_utf8_width(const char *start, const char *end)
1642{
1643        int width = 0;
1644        size_t remain = end - start;
1645
1646        while (remain) {
1647                int n = utf8_width(&start, &remain);
1648                if (n < 0 || !start)
1649                        return -1;
1650                width += n;
1651        }
1652        return width;
1653}
1654
1655static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1656                                 const char *line, int linelen)
1657{
1658        const char *tab;
1659
1660        while ((tab = memchr(line, '\t', linelen)) != NULL) {
1661                int width = pp_utf8_width(line, tab);
1662
1663                /*
1664                 * If it wasn't well-formed utf8, or it
1665                 * had characters with badly defined
1666                 * width (control characters etc), just
1667                 * give up on trying to align things.
1668                 */
1669                if (width < 0)
1670                        break;
1671
1672                /* Output the data .. */
1673                strbuf_add(sb, line, tab - line);
1674
1675                /* .. and the de-tabified tab */
1676                strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1677
1678                /* Skip over the printed part .. */
1679                linelen -= tab + 1 - line;
1680                line = tab + 1;
1681        }
1682
1683        /*
1684         * Print out everything after the last tab without
1685         * worrying about width - there's nothing more to
1686         * align.
1687         */
1688        strbuf_add(sb, line, linelen);
1689}
1690
1691/*
1692 * pp_handle_indent() prints out the intendation, and
1693 * the whole line (without the final newline), after
1694 * de-tabifying.
1695 */
1696static void pp_handle_indent(struct pretty_print_context *pp,
1697                             struct strbuf *sb, int indent,
1698                             const char *line, int linelen)
1699{
1700        strbuf_addchars(sb, ' ', indent);
1701        if (pp->expand_tabs_in_log)
1702                strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1703        else
1704                strbuf_add(sb, line, linelen);
1705}
1706
1707static int is_mboxrd_from(const char *line, int len)
1708{
1709        /*
1710         * a line matching /^From $/ here would only have len == 4
1711         * at this point because is_empty_line would've trimmed all
1712         * trailing space
1713         */
1714        return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1715}
1716
1717void pp_remainder(struct pretty_print_context *pp,
1718                  const char **msg_p,
1719                  struct strbuf *sb,
1720                  int indent)
1721{
1722        int first = 1;
1723        for (;;) {
1724                const char *line = *msg_p;
1725                int linelen = get_one_line(line);
1726                *msg_p += linelen;
1727
1728                if (!linelen)
1729                        break;
1730
1731                if (is_blank_line(line, &linelen)) {
1732                        if (first)
1733                                continue;
1734                        if (pp->fmt == CMIT_FMT_SHORT)
1735                                break;
1736                }
1737                first = 0;
1738
1739                strbuf_grow(sb, linelen + indent + 20);
1740                if (indent)
1741                        pp_handle_indent(pp, sb, indent, line, linelen);
1742                else if (pp->expand_tabs_in_log)
1743                        strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1744                                             line, linelen);
1745                else {
1746                        if (pp->fmt == CMIT_FMT_MBOXRD &&
1747                                        is_mboxrd_from(line, linelen))
1748                                strbuf_addch(sb, '>');
1749
1750                        strbuf_add(sb, line, linelen);
1751                }
1752                strbuf_addch(sb, '\n');
1753        }
1754}
1755
1756void pretty_print_commit(struct pretty_print_context *pp,
1757                         const struct commit *commit,
1758                         struct strbuf *sb)
1759{
1760        unsigned long beginning_of_body;
1761        int indent = 4;
1762        const char *msg;
1763        const char *reencoded;
1764        const char *encoding;
1765        int need_8bit_cte = pp->need_8bit_cte;
1766
1767        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1768                format_commit_message(commit, user_format, sb, pp);
1769                return;
1770        }
1771
1772        encoding = get_log_output_encoding();
1773        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1774
1775        if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1776                indent = 0;
1777
1778        /*
1779         * We need to check and emit Content-type: to mark it
1780         * as 8-bit if we haven't done so.
1781         */
1782        if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1783                int i, ch, in_body;
1784
1785                for (in_body = i = 0; (ch = msg[i]); i++) {
1786                        if (!in_body) {
1787                                /* author could be non 7-bit ASCII but
1788                                 * the log may be so; skip over the
1789                                 * header part first.
1790                                 */
1791                                if (ch == '\n' && msg[i+1] == '\n')
1792                                        in_body = 1;
1793                        }
1794                        else if (non_ascii(ch)) {
1795                                need_8bit_cte = 1;
1796                                break;
1797                        }
1798                }
1799        }
1800
1801        pp_header(pp, encoding, commit, &msg, sb);
1802        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1803                strbuf_addch(sb, '\n');
1804        }
1805
1806        /* Skip excess blank lines at the beginning of body, if any... */
1807        msg = skip_blank_lines(msg);
1808
1809        /* These formats treat the title line specially. */
1810        if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1811                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1812
1813        beginning_of_body = sb->len;
1814        if (pp->fmt != CMIT_FMT_ONELINE)
1815                pp_remainder(pp, &msg, sb, indent);
1816        strbuf_rtrim(sb);
1817
1818        /* Make sure there is an EOLN for the non-oneline case */
1819        if (pp->fmt != CMIT_FMT_ONELINE)
1820                strbuf_addch(sb, '\n');
1821
1822        /*
1823         * The caller may append additional body text in e-mail
1824         * format.  Make sure we did not strip the blank line
1825         * between the header and the body.
1826         */
1827        if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1828                strbuf_addch(sb, '\n');
1829
1830        unuse_commit_buffer(commit, reencoded);
1831}
1832
1833void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1834                    struct strbuf *sb)
1835{
1836        struct pretty_print_context pp = {0};
1837        pp.fmt = fmt;
1838        pretty_print_commit(&pp, commit, sb);
1839}