pretty.con commit Sync with maint (6598894)
   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 object_id *oidp = &parent->item->object.oid;
 548                strbuf_addch(sb, ' ');
 549                if (pp->abbrev)
 550                        strbuf_add_unique_abbrev(sb, oidp->hash, pp->abbrev);
 551                else
 552                        strbuf_addstr(sb, oid_to_hex(oidp));
 553                parent = parent->next;
 554        }
 555        strbuf_addch(sb, '\n');
 556}
 557
 558static char *get_header(const char *msg, const char *key)
 559{
 560        size_t len;
 561        const char *v = find_commit_header(msg, key, &len);
 562        return v ? xmemdupz(v, len) : NULL;
 563}
 564
 565static char *replace_encoding_header(char *buf, const char *encoding)
 566{
 567        struct strbuf tmp = STRBUF_INIT;
 568        size_t start, len;
 569        char *cp = buf;
 570
 571        /* guess if there is an encoding header before a \n\n */
 572        while (!starts_with(cp, "encoding ")) {
 573                cp = strchr(cp, '\n');
 574                if (!cp || *++cp == '\n')
 575                        return buf;
 576        }
 577        start = cp - buf;
 578        cp = strchr(cp, '\n');
 579        if (!cp)
 580                return buf; /* should not happen but be defensive */
 581        len = cp + 1 - (buf + start);
 582
 583        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 584        if (is_encoding_utf8(encoding)) {
 585                /* we have re-coded to UTF-8; drop the header */
 586                strbuf_remove(&tmp, start, len);
 587        } else {
 588                /* just replaces XXXX in 'encoding XXXX\n' */
 589                strbuf_splice(&tmp, start + strlen("encoding "),
 590                                          len - strlen("encoding \n"),
 591                                          encoding, strlen(encoding));
 592        }
 593        return strbuf_detach(&tmp, NULL);
 594}
 595
 596const char *logmsg_reencode(const struct commit *commit,
 597                            char **commit_encoding,
 598                            const char *output_encoding)
 599{
 600        static const char *utf8 = "UTF-8";
 601        const char *use_encoding;
 602        char *encoding;
 603        const char *msg = get_commit_buffer(commit, NULL);
 604        char *out;
 605
 606        if (!output_encoding || !*output_encoding) {
 607                if (commit_encoding)
 608                        *commit_encoding = get_header(msg, "encoding");
 609                return msg;
 610        }
 611        encoding = get_header(msg, "encoding");
 612        if (commit_encoding)
 613                *commit_encoding = encoding;
 614        use_encoding = encoding ? encoding : utf8;
 615        if (same_encoding(use_encoding, output_encoding)) {
 616                /*
 617                 * No encoding work to be done. If we have no encoding header
 618                 * at all, then there's nothing to do, and we can return the
 619                 * message verbatim (whether newly allocated or not).
 620                 */
 621                if (!encoding)
 622                        return msg;
 623
 624                /*
 625                 * Otherwise, we still want to munge the encoding header in the
 626                 * result, which will be done by modifying the buffer. If we
 627                 * are using a fresh copy, we can reuse it. But if we are using
 628                 * the cached copy from get_commit_buffer, we need to duplicate it
 629                 * to avoid munging the cached copy.
 630                 */
 631                if (msg == get_cached_commit_buffer(commit, NULL))
 632                        out = xstrdup(msg);
 633                else
 634                        out = (char *)msg;
 635        }
 636        else {
 637                /*
 638                 * There's actual encoding work to do. Do the reencoding, which
 639                 * still leaves the header to be replaced in the next step. At
 640                 * this point, we are done with msg. If we allocated a fresh
 641                 * copy, we can free it.
 642                 */
 643                out = reencode_string(msg, output_encoding, use_encoding);
 644                if (out)
 645                        unuse_commit_buffer(commit, msg);
 646        }
 647
 648        /*
 649         * This replacement actually consumes the buffer we hand it, so we do
 650         * not have to worry about freeing the old "out" here.
 651         */
 652        if (out)
 653                out = replace_encoding_header(out, output_encoding);
 654
 655        if (!commit_encoding)
 656                free(encoding);
 657        /*
 658         * If the re-encoding failed, out might be NULL here; in that
 659         * case we just return the commit message verbatim.
 660         */
 661        return out ? out : msg;
 662}
 663
 664static int mailmap_name(const char **email, size_t *email_len,
 665                        const char **name, size_t *name_len)
 666{
 667        static struct string_list *mail_map;
 668        if (!mail_map) {
 669                mail_map = xcalloc(1, sizeof(*mail_map));
 670                read_mailmap(mail_map, NULL);
 671        }
 672        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 673}
 674
 675static size_t format_person_part(struct strbuf *sb, char part,
 676                                 const char *msg, int len,
 677                                 const struct date_mode *dmode)
 678{
 679        /* currently all placeholders have same length */
 680        const int placeholder_len = 2;
 681        struct ident_split s;
 682        const char *name, *mail;
 683        size_t maillen, namelen;
 684
 685        if (split_ident_line(&s, msg, len) < 0)
 686                goto skip;
 687
 688        name = s.name_begin;
 689        namelen = s.name_end - s.name_begin;
 690        mail = s.mail_begin;
 691        maillen = s.mail_end - s.mail_begin;
 692
 693        if (part == 'N' || part == 'E') /* mailmap lookup */
 694                mailmap_name(&mail, &maillen, &name, &namelen);
 695        if (part == 'n' || part == 'N') {       /* name */
 696                strbuf_add(sb, name, namelen);
 697                return placeholder_len;
 698        }
 699        if (part == 'e' || part == 'E') {       /* email */
 700                strbuf_add(sb, mail, maillen);
 701                return placeholder_len;
 702        }
 703
 704        if (!s.date_begin)
 705                goto skip;
 706
 707        if (part == 't') {      /* date, UNIX timestamp */
 708                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 709                return placeholder_len;
 710        }
 711
 712        switch (part) {
 713        case 'd':       /* date */
 714                strbuf_addstr(sb, show_ident_date(&s, dmode));
 715                return placeholder_len;
 716        case 'D':       /* date, RFC2822 style */
 717                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
 718                return placeholder_len;
 719        case 'r':       /* date, relative */
 720                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
 721                return placeholder_len;
 722        case 'i':       /* date, ISO 8601-like */
 723                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
 724                return placeholder_len;
 725        case 'I':       /* date, ISO 8601 strict */
 726                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
 727                return placeholder_len;
 728        }
 729
 730skip:
 731        /*
 732         * reading from either a bogus commit, or a reflog entry with
 733         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 734         * to compute a valid return value.
 735         */
 736        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 737            || part == 'D' || part == 'r' || part == 'i')
 738                return placeholder_len;
 739
 740        return 0; /* unknown placeholder */
 741}
 742
 743struct chunk {
 744        size_t off;
 745        size_t len;
 746};
 747
 748enum flush_type {
 749        no_flush,
 750        flush_right,
 751        flush_left,
 752        flush_left_and_steal,
 753        flush_both
 754};
 755
 756enum trunc_type {
 757        trunc_none,
 758        trunc_left,
 759        trunc_middle,
 760        trunc_right
 761};
 762
 763struct format_commit_context {
 764        const struct commit *commit;
 765        const struct pretty_print_context *pretty_ctx;
 766        unsigned commit_header_parsed:1;
 767        unsigned commit_message_parsed:1;
 768        struct signature_check signature_check;
 769        enum flush_type flush_type;
 770        enum trunc_type truncate;
 771        const char *message;
 772        char *commit_encoding;
 773        size_t width, indent1, indent2;
 774        int auto_color;
 775        int padding;
 776
 777        /* These offsets are relative to the start of the commit message. */
 778        struct chunk author;
 779        struct chunk committer;
 780        size_t message_off;
 781        size_t subject_off;
 782        size_t body_off;
 783
 784        /* The following ones are relative to the result struct strbuf. */
 785        struct chunk abbrev_commit_hash;
 786        struct chunk abbrev_tree_hash;
 787        struct chunk abbrev_parent_hashes;
 788        size_t wrap_start;
 789};
 790
 791static int add_again(struct strbuf *sb, struct chunk *chunk)
 792{
 793        if (chunk->len) {
 794                strbuf_adddup(sb, chunk->off, chunk->len);
 795                return 1;
 796        }
 797
 798        /*
 799         * We haven't seen this chunk before.  Our caller is surely
 800         * going to add it the hard way now.  Remember the most likely
 801         * start of the to-be-added chunk: the current end of the
 802         * struct strbuf.
 803         */
 804        chunk->off = sb->len;
 805        return 0;
 806}
 807
 808static void parse_commit_header(struct format_commit_context *context)
 809{
 810        const char *msg = context->message;
 811        int i;
 812
 813        for (i = 0; msg[i]; i++) {
 814                const char *name;
 815                int eol;
 816                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 817                        ; /* do nothing */
 818
 819                if (i == eol) {
 820                        break;
 821                } else if (skip_prefix(msg + i, "author ", &name)) {
 822                        context->author.off = name - msg;
 823                        context->author.len = msg + eol - name;
 824                } else if (skip_prefix(msg + i, "committer ", &name)) {
 825                        context->committer.off = name - msg;
 826                        context->committer.len = msg + eol - name;
 827                }
 828                i = eol;
 829        }
 830        context->message_off = i;
 831        context->commit_header_parsed = 1;
 832}
 833
 834static int istitlechar(char c)
 835{
 836        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 837                (c >= '0' && c <= '9') || c == '.' || c == '_';
 838}
 839
 840static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 841{
 842        size_t trimlen;
 843        size_t start_len = sb->len;
 844        int space = 2;
 845
 846        for (; *msg && *msg != '\n'; msg++) {
 847                if (istitlechar(*msg)) {
 848                        if (space == 1)
 849                                strbuf_addch(sb, '-');
 850                        space = 0;
 851                        strbuf_addch(sb, *msg);
 852                        if (*msg == '.')
 853                                while (*(msg+1) == '.')
 854                                        msg++;
 855                } else
 856                        space |= 1;
 857        }
 858
 859        /* trim any trailing '.' or '-' characters */
 860        trimlen = 0;
 861        while (sb->len - trimlen > start_len &&
 862                (sb->buf[sb->len - 1 - trimlen] == '.'
 863                || sb->buf[sb->len - 1 - trimlen] == '-'))
 864                trimlen++;
 865        strbuf_remove(sb, sb->len - trimlen, trimlen);
 866}
 867
 868const char *format_subject(struct strbuf *sb, const char *msg,
 869                           const char *line_separator)
 870{
 871        int first = 1;
 872
 873        for (;;) {
 874                const char *line = msg;
 875                int linelen = get_one_line(line);
 876
 877                msg += linelen;
 878                if (!linelen || is_blank_line(line, &linelen))
 879                        break;
 880
 881                if (!sb)
 882                        continue;
 883                strbuf_grow(sb, linelen + 2);
 884                if (!first)
 885                        strbuf_addstr(sb, line_separator);
 886                strbuf_add(sb, line, linelen);
 887                first = 0;
 888        }
 889        return msg;
 890}
 891
 892static void parse_commit_message(struct format_commit_context *c)
 893{
 894        const char *msg = c->message + c->message_off;
 895        const char *start = c->message;
 896
 897        msg = skip_blank_lines(msg);
 898        c->subject_off = msg - start;
 899
 900        msg = format_subject(NULL, msg, NULL);
 901        msg = skip_blank_lines(msg);
 902        c->body_off = msg - start;
 903
 904        c->commit_message_parsed = 1;
 905}
 906
 907static void strbuf_wrap(struct strbuf *sb, size_t pos,
 908                        size_t width, size_t indent1, size_t indent2)
 909{
 910        struct strbuf tmp = STRBUF_INIT;
 911
 912        if (pos)
 913                strbuf_add(&tmp, sb->buf, pos);
 914        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 915                                (int) indent1, (int) indent2, (int) width);
 916        strbuf_swap(&tmp, sb);
 917        strbuf_release(&tmp);
 918}
 919
 920static void rewrap_message_tail(struct strbuf *sb,
 921                                struct format_commit_context *c,
 922                                size_t new_width, size_t new_indent1,
 923                                size_t new_indent2)
 924{
 925        if (c->width == new_width && c->indent1 == new_indent1 &&
 926            c->indent2 == new_indent2)
 927                return;
 928        if (c->wrap_start < sb->len)
 929                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 930        c->wrap_start = sb->len;
 931        c->width = new_width;
 932        c->indent1 = new_indent1;
 933        c->indent2 = new_indent2;
 934}
 935
 936static int format_reflog_person(struct strbuf *sb,
 937                                char part,
 938                                struct reflog_walk_info *log,
 939                                const struct date_mode *dmode)
 940{
 941        const char *ident;
 942
 943        if (!log)
 944                return 2;
 945
 946        ident = get_reflog_ident(log);
 947        if (!ident)
 948                return 2;
 949
 950        return format_person_part(sb, part, ident, strlen(ident), dmode);
 951}
 952
 953static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 954                          const char *placeholder,
 955                          struct format_commit_context *c)
 956{
 957        const char *rest = placeholder;
 958
 959        if (placeholder[1] == '(') {
 960                const char *begin = placeholder + 2;
 961                const char *end = strchr(begin, ')');
 962                char color[COLOR_MAXLEN];
 963
 964                if (!end)
 965                        return 0;
 966                if (skip_prefix(begin, "auto,", &begin)) {
 967                        if (!want_color(c->pretty_ctx->color))
 968                                return end - placeholder + 1;
 969                }
 970                if (color_parse_mem(begin, end - begin, color) < 0)
 971                        die(_("unable to parse --pretty format"));
 972                strbuf_addstr(sb, color);
 973                return end - placeholder + 1;
 974        }
 975        if (skip_prefix(placeholder + 1, "red", &rest))
 976                strbuf_addstr(sb, GIT_COLOR_RED);
 977        else if (skip_prefix(placeholder + 1, "green", &rest))
 978                strbuf_addstr(sb, GIT_COLOR_GREEN);
 979        else if (skip_prefix(placeholder + 1, "blue", &rest))
 980                strbuf_addstr(sb, GIT_COLOR_BLUE);
 981        else if (skip_prefix(placeholder + 1, "reset", &rest))
 982                strbuf_addstr(sb, GIT_COLOR_RESET);
 983        return rest - placeholder;
 984}
 985
 986static size_t parse_padding_placeholder(struct strbuf *sb,
 987                                        const char *placeholder,
 988                                        struct format_commit_context *c)
 989{
 990        const char *ch = placeholder;
 991        enum flush_type flush_type;
 992        int to_column = 0;
 993
 994        switch (*ch++) {
 995        case '<':
 996                flush_type = flush_right;
 997                break;
 998        case '>':
 999                if (*ch == '<') {
1000                        flush_type = flush_both;
1001                        ch++;
1002                } else if (*ch == '>') {
1003                        flush_type = flush_left_and_steal;
1004                        ch++;
1005                } else
1006                        flush_type = flush_left;
1007                break;
1008        default:
1009                return 0;
1010        }
1011
1012        /* the next value means "wide enough to that column" */
1013        if (*ch == '|') {
1014                to_column = 1;
1015                ch++;
1016        }
1017
1018        if (*ch == '(') {
1019                const char *start = ch + 1;
1020                const char *end = start + strcspn(start, ",)");
1021                char *next;
1022                int width;
1023                if (!end || end == start)
1024                        return 0;
1025                width = strtol(start, &next, 10);
1026                if (next == start || width == 0)
1027                        return 0;
1028                if (width < 0) {
1029                        if (to_column)
1030                                width += term_columns();
1031                        if (width < 0)
1032                                return 0;
1033                }
1034                c->padding = to_column ? -width : width;
1035                c->flush_type = flush_type;
1036
1037                if (*end == ',') {
1038                        start = end + 1;
1039                        end = strchr(start, ')');
1040                        if (!end || end == start)
1041                                return 0;
1042                        if (starts_with(start, "trunc)"))
1043                                c->truncate = trunc_right;
1044                        else if (starts_with(start, "ltrunc)"))
1045                                c->truncate = trunc_left;
1046                        else if (starts_with(start, "mtrunc)"))
1047                                c->truncate = trunc_middle;
1048                        else
1049                                return 0;
1050                } else
1051                        c->truncate = trunc_none;
1052
1053                return end - placeholder + 1;
1054        }
1055        return 0;
1056}
1057
1058static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1059                                const char *placeholder,
1060                                void *context)
1061{
1062        struct format_commit_context *c = context;
1063        const struct commit *commit = c->commit;
1064        const char *msg = c->message;
1065        struct commit_list *p;
1066        int ch;
1067
1068        /* these are independent of the commit */
1069        switch (placeholder[0]) {
1070        case 'C':
1071                if (starts_with(placeholder + 1, "(auto)")) {
1072                        c->auto_color = want_color(c->pretty_ctx->color);
1073                        if (c->auto_color && sb->len)
1074                                strbuf_addstr(sb, GIT_COLOR_RESET);
1075                        return 7; /* consumed 7 bytes, "C(auto)" */
1076                } else {
1077                        int ret = parse_color(sb, placeholder, c);
1078                        if (ret)
1079                                c->auto_color = 0;
1080                        /*
1081                         * Otherwise, we decided to treat %C<unknown>
1082                         * as a literal string, and the previous
1083                         * %C(auto) is still valid.
1084                         */
1085                        return ret;
1086                }
1087        case 'n':               /* newline */
1088                strbuf_addch(sb, '\n');
1089                return 1;
1090        case 'x':
1091                /* %x00 == NUL, %x0a == LF, etc. */
1092                ch = hex2chr(placeholder + 1);
1093                if (ch < 0)
1094                        return 0;
1095                strbuf_addch(sb, ch);
1096                return 3;
1097        case 'w':
1098                if (placeholder[1] == '(') {
1099                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1100                        char *next;
1101                        const char *start = placeholder + 2;
1102                        const char *end = strchr(start, ')');
1103                        if (!end)
1104                                return 0;
1105                        if (end > start) {
1106                                width = strtoul(start, &next, 10);
1107                                if (*next == ',') {
1108                                        indent1 = strtoul(next + 1, &next, 10);
1109                                        if (*next == ',') {
1110                                                indent2 = strtoul(next + 1,
1111                                                                 &next, 10);
1112                                        }
1113                                }
1114                                if (*next != ')')
1115                                        return 0;
1116                        }
1117                        rewrap_message_tail(sb, c, width, indent1, indent2);
1118                        return end - placeholder + 1;
1119                } else
1120                        return 0;
1121
1122        case '<':
1123        case '>':
1124                return parse_padding_placeholder(sb, placeholder, c);
1125        }
1126
1127        /* these depend on the commit */
1128        if (!commit->object.parsed)
1129                parse_object(commit->object.oid.hash);
1130
1131        switch (placeholder[0]) {
1132        case 'H':               /* commit hash */
1133                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1134                strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1135                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1136                return 1;
1137        case 'h':               /* abbreviated commit hash */
1138                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1139                if (add_again(sb, &c->abbrev_commit_hash)) {
1140                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1141                        return 1;
1142                }
1143                strbuf_add_unique_abbrev(sb, commit->object.oid.hash,
1144                                         c->pretty_ctx->abbrev);
1145                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1146                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1147                return 1;
1148        case 'T':               /* tree hash */
1149                strbuf_addstr(sb, oid_to_hex(&commit->tree->object.oid));
1150                return 1;
1151        case 't':               /* abbreviated tree hash */
1152                if (add_again(sb, &c->abbrev_tree_hash))
1153                        return 1;
1154                strbuf_add_unique_abbrev(sb, commit->tree->object.oid.hash,
1155                                         c->pretty_ctx->abbrev);
1156                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1157                return 1;
1158        case 'P':               /* parent hashes */
1159                for (p = commit->parents; p; p = p->next) {
1160                        if (p != commit->parents)
1161                                strbuf_addch(sb, ' ');
1162                        strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1163                }
1164                return 1;
1165        case 'p':               /* abbreviated parent hashes */
1166                if (add_again(sb, &c->abbrev_parent_hashes))
1167                        return 1;
1168                for (p = commit->parents; p; p = p->next) {
1169                        if (p != commit->parents)
1170                                strbuf_addch(sb, ' ');
1171                        strbuf_add_unique_abbrev(sb, p->item->object.oid.hash,
1172                                                 c->pretty_ctx->abbrev);
1173                }
1174                c->abbrev_parent_hashes.len = sb->len -
1175                                              c->abbrev_parent_hashes.off;
1176                return 1;
1177        case 'm':               /* left/right/bottom */
1178                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1179                return 1;
1180        case 'd':
1181                load_ref_decorations(DECORATE_SHORT_REFS);
1182                format_decorations(sb, commit, c->auto_color);
1183                return 1;
1184        case 'D':
1185                load_ref_decorations(DECORATE_SHORT_REFS);
1186                format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1187                return 1;
1188        case 'g':               /* reflog info */
1189                switch(placeholder[1]) {
1190                case 'd':       /* reflog selector */
1191                case 'D':
1192                        if (c->pretty_ctx->reflog_info)
1193                                get_reflog_selector(sb,
1194                                                    c->pretty_ctx->reflog_info,
1195                                                    &c->pretty_ctx->date_mode,
1196                                                    c->pretty_ctx->date_mode_explicit,
1197                                                    (placeholder[1] == 'd'));
1198                        return 2;
1199                case 's':       /* reflog message */
1200                        if (c->pretty_ctx->reflog_info)
1201                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1202                        return 2;
1203                case 'n':
1204                case 'N':
1205                case 'e':
1206                case 'E':
1207                        return format_reflog_person(sb,
1208                                                    placeholder[1],
1209                                                    c->pretty_ctx->reflog_info,
1210                                                    &c->pretty_ctx->date_mode);
1211                }
1212                return 0;       /* unknown %g placeholder */
1213        case 'N':
1214                if (c->pretty_ctx->notes_message) {
1215                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1216                        return 1;
1217                }
1218                return 0;
1219        }
1220
1221        if (placeholder[0] == 'G') {
1222                if (!c->signature_check.result)
1223                        check_commit_signature(c->commit, &(c->signature_check));
1224                switch (placeholder[1]) {
1225                case 'G':
1226                        if (c->signature_check.gpg_output)
1227                                strbuf_addstr(sb, c->signature_check.gpg_output);
1228                        break;
1229                case '?':
1230                        switch (c->signature_check.result) {
1231                        case 'G':
1232                        case 'B':
1233                        case 'U':
1234                        case 'N':
1235                                strbuf_addch(sb, c->signature_check.result);
1236                        }
1237                        break;
1238                case 'S':
1239                        if (c->signature_check.signer)
1240                                strbuf_addstr(sb, c->signature_check.signer);
1241                        break;
1242                case 'K':
1243                        if (c->signature_check.key)
1244                                strbuf_addstr(sb, c->signature_check.key);
1245                        break;
1246                default:
1247                        return 0;
1248                }
1249                return 2;
1250        }
1251
1252
1253        /* For the rest we have to parse the commit header. */
1254        if (!c->commit_header_parsed)
1255                parse_commit_header(c);
1256
1257        switch (placeholder[0]) {
1258        case 'a':       /* author ... */
1259                return format_person_part(sb, placeholder[1],
1260                                   msg + c->author.off, c->author.len,
1261                                   &c->pretty_ctx->date_mode);
1262        case 'c':       /* committer ... */
1263                return format_person_part(sb, placeholder[1],
1264                                   msg + c->committer.off, c->committer.len,
1265                                   &c->pretty_ctx->date_mode);
1266        case 'e':       /* encoding */
1267                if (c->commit_encoding)
1268                        strbuf_addstr(sb, c->commit_encoding);
1269                return 1;
1270        case 'B':       /* raw body */
1271                /* message_off is always left at the initial newline */
1272                strbuf_addstr(sb, msg + c->message_off + 1);
1273                return 1;
1274        }
1275
1276        /* Now we need to parse the commit message. */
1277        if (!c->commit_message_parsed)
1278                parse_commit_message(c);
1279
1280        switch (placeholder[0]) {
1281        case 's':       /* subject */
1282                format_subject(sb, msg + c->subject_off, " ");
1283                return 1;
1284        case 'f':       /* sanitized subject */
1285                format_sanitized_subject(sb, msg + c->subject_off);
1286                return 1;
1287        case 'b':       /* body */
1288                strbuf_addstr(sb, msg + c->body_off);
1289                return 1;
1290        }
1291        return 0;       /* unknown placeholder */
1292}
1293
1294static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1295                                    const char *placeholder,
1296                                    struct format_commit_context *c)
1297{
1298        struct strbuf local_sb = STRBUF_INIT;
1299        int total_consumed = 0, len, padding = c->padding;
1300        if (padding < 0) {
1301                const char *start = strrchr(sb->buf, '\n');
1302                int occupied;
1303                if (!start)
1304                        start = sb->buf;
1305                occupied = utf8_strnwidth(start, -1, 1);
1306                occupied += c->pretty_ctx->graph_width;
1307                padding = (-padding) - occupied;
1308        }
1309        while (1) {
1310                int modifier = *placeholder == 'C';
1311                int consumed = format_commit_one(&local_sb, placeholder, c);
1312                total_consumed += consumed;
1313
1314                if (!modifier)
1315                        break;
1316
1317                placeholder += consumed;
1318                if (*placeholder != '%')
1319                        break;
1320                placeholder++;
1321                total_consumed++;
1322        }
1323        len = utf8_strnwidth(local_sb.buf, -1, 1);
1324
1325        if (c->flush_type == flush_left_and_steal) {
1326                const char *ch = sb->buf + sb->len - 1;
1327                while (len > padding && ch > sb->buf) {
1328                        const char *p;
1329                        if (*ch == ' ') {
1330                                ch--;
1331                                padding++;
1332                                continue;
1333                        }
1334                        /* check for trailing ansi sequences */
1335                        if (*ch != 'm')
1336                                break;
1337                        p = ch - 1;
1338                        while (ch - p < 10 && *p != '\033')
1339                                p--;
1340                        if (*p != '\033' ||
1341                            ch + 1 - p != display_mode_esc_sequence_len(p))
1342                                break;
1343                        /*
1344                         * got a good ansi sequence, put it back to
1345                         * local_sb as we're cutting sb
1346                         */
1347                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1348                        ch = p - 1;
1349                }
1350                strbuf_setlen(sb, ch + 1 - sb->buf);
1351                c->flush_type = flush_left;
1352        }
1353
1354        if (len > padding) {
1355                switch (c->truncate) {
1356                case trunc_left:
1357                        strbuf_utf8_replace(&local_sb,
1358                                            0, len - (padding - 2),
1359                                            "..");
1360                        break;
1361                case trunc_middle:
1362                        strbuf_utf8_replace(&local_sb,
1363                                            padding / 2 - 1,
1364                                            len - (padding - 2),
1365                                            "..");
1366                        break;
1367                case trunc_right:
1368                        strbuf_utf8_replace(&local_sb,
1369                                            padding - 2, len - (padding - 2),
1370                                            "..");
1371                        break;
1372                case trunc_none:
1373                        break;
1374                }
1375                strbuf_addbuf(sb, &local_sb);
1376        } else {
1377                int sb_len = sb->len, offset = 0;
1378                if (c->flush_type == flush_left)
1379                        offset = padding - len;
1380                else if (c->flush_type == flush_both)
1381                        offset = (padding - len) / 2;
1382                /*
1383                 * we calculate padding in columns, now
1384                 * convert it back to chars
1385                 */
1386                padding = padding - len + local_sb.len;
1387                strbuf_addchars(sb, ' ', padding);
1388                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1389                       local_sb.len);
1390        }
1391        strbuf_release(&local_sb);
1392        c->flush_type = no_flush;
1393        return total_consumed;
1394}
1395
1396static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1397                                 const char *placeholder,
1398                                 void *context)
1399{
1400        int consumed;
1401        size_t orig_len;
1402        enum {
1403                NO_MAGIC,
1404                ADD_LF_BEFORE_NON_EMPTY,
1405                DEL_LF_BEFORE_EMPTY,
1406                ADD_SP_BEFORE_NON_EMPTY
1407        } magic = NO_MAGIC;
1408
1409        switch (placeholder[0]) {
1410        case '-':
1411                magic = DEL_LF_BEFORE_EMPTY;
1412                break;
1413        case '+':
1414                magic = ADD_LF_BEFORE_NON_EMPTY;
1415                break;
1416        case ' ':
1417                magic = ADD_SP_BEFORE_NON_EMPTY;
1418                break;
1419        default:
1420                break;
1421        }
1422        if (magic != NO_MAGIC)
1423                placeholder++;
1424
1425        orig_len = sb->len;
1426        if (((struct format_commit_context *)context)->flush_type != no_flush)
1427                consumed = format_and_pad_commit(sb, placeholder, context);
1428        else
1429                consumed = format_commit_one(sb, placeholder, context);
1430        if (magic == NO_MAGIC)
1431                return consumed;
1432
1433        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1434                while (sb->len && sb->buf[sb->len - 1] == '\n')
1435                        strbuf_setlen(sb, sb->len - 1);
1436        } else if (orig_len != sb->len) {
1437                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1438                        strbuf_insert(sb, orig_len, "\n", 1);
1439                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1440                        strbuf_insert(sb, orig_len, " ", 1);
1441        }
1442        return consumed + 1;
1443}
1444
1445static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1446                                   void *context)
1447{
1448        struct userformat_want *w = context;
1449
1450        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1451                placeholder++;
1452
1453        switch (*placeholder) {
1454        case 'N':
1455                w->notes = 1;
1456                break;
1457        }
1458        return 0;
1459}
1460
1461void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1462{
1463        struct strbuf dummy = STRBUF_INIT;
1464
1465        if (!fmt) {
1466                if (!user_format)
1467                        return;
1468                fmt = user_format;
1469        }
1470        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1471        strbuf_release(&dummy);
1472}
1473
1474void format_commit_message(const struct commit *commit,
1475                           const char *format, struct strbuf *sb,
1476                           const struct pretty_print_context *pretty_ctx)
1477{
1478        struct format_commit_context context;
1479        const char *output_enc = pretty_ctx->output_encoding;
1480        const char *utf8 = "UTF-8";
1481
1482        memset(&context, 0, sizeof(context));
1483        context.commit = commit;
1484        context.pretty_ctx = pretty_ctx;
1485        context.wrap_start = sb->len;
1486        /*
1487         * convert a commit message to UTF-8 first
1488         * as far as 'format_commit_item' assumes it in UTF-8
1489         */
1490        context.message = logmsg_reencode(commit,
1491                                          &context.commit_encoding,
1492                                          utf8);
1493
1494        strbuf_expand(sb, format, format_commit_item, &context);
1495        rewrap_message_tail(sb, &context, 0, 0, 0);
1496
1497        /* then convert a commit message to an actual output encoding */
1498        if (output_enc) {
1499                if (same_encoding(utf8, output_enc))
1500                        output_enc = NULL;
1501        } else {
1502                if (context.commit_encoding &&
1503                    !same_encoding(context.commit_encoding, utf8))
1504                        output_enc = context.commit_encoding;
1505        }
1506
1507        if (output_enc) {
1508                int outsz;
1509                char *out = reencode_string_len(sb->buf, sb->len,
1510                                                output_enc, utf8, &outsz);
1511                if (out)
1512                        strbuf_attach(sb, out, outsz, outsz + 1);
1513        }
1514
1515        free(context.commit_encoding);
1516        unuse_commit_buffer(commit, context.message);
1517}
1518
1519static void pp_header(struct pretty_print_context *pp,
1520                      const char *encoding,
1521                      const struct commit *commit,
1522                      const char **msg_p,
1523                      struct strbuf *sb)
1524{
1525        int parents_shown = 0;
1526
1527        for (;;) {
1528                const char *name, *line = *msg_p;
1529                int linelen = get_one_line(*msg_p);
1530
1531                if (!linelen)
1532                        return;
1533                *msg_p += linelen;
1534
1535                if (linelen == 1)
1536                        /* End of header */
1537                        return;
1538
1539                if (pp->fmt == CMIT_FMT_RAW) {
1540                        strbuf_add(sb, line, linelen);
1541                        continue;
1542                }
1543
1544                if (starts_with(line, "parent ")) {
1545                        if (linelen != 48)
1546                                die("bad parent line in commit");
1547                        continue;
1548                }
1549
1550                if (!parents_shown) {
1551                        unsigned num = commit_list_count(commit->parents);
1552                        /* with enough slop */
1553                        strbuf_grow(sb, num * 50 + 20);
1554                        add_merge_info(pp, sb, commit);
1555                        parents_shown = 1;
1556                }
1557
1558                /*
1559                 * MEDIUM == DEFAULT shows only author with dates.
1560                 * FULL shows both authors but not dates.
1561                 * FULLER shows both authors and dates.
1562                 */
1563                if (skip_prefix(line, "author ", &name)) {
1564                        strbuf_grow(sb, linelen + 80);
1565                        pp_user_info(pp, "Author", sb, name, encoding);
1566                }
1567                if (skip_prefix(line, "committer ", &name) &&
1568                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1569                        strbuf_grow(sb, linelen + 80);
1570                        pp_user_info(pp, "Commit", sb, name, encoding);
1571                }
1572        }
1573}
1574
1575void pp_title_line(struct pretty_print_context *pp,
1576                   const char **msg_p,
1577                   struct strbuf *sb,
1578                   const char *encoding,
1579                   int need_8bit_cte)
1580{
1581        static const int max_length = 78; /* per rfc2047 */
1582        struct strbuf title;
1583
1584        strbuf_init(&title, 80);
1585        *msg_p = format_subject(&title, *msg_p,
1586                                pp->preserve_subject ? "\n" : " ");
1587
1588        strbuf_grow(sb, title.len + 1024);
1589        if (pp->subject) {
1590                strbuf_addstr(sb, pp->subject);
1591                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1592                        add_rfc2047(sb, title.buf, title.len,
1593                                                encoding, RFC2047_SUBJECT);
1594                else
1595                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1596                                         -last_line_length(sb), 1, max_length);
1597        } else {
1598                strbuf_addbuf(sb, &title);
1599        }
1600        strbuf_addch(sb, '\n');
1601
1602        if (need_8bit_cte == 0) {
1603                int i;
1604                for (i = 0; i < pp->in_body_headers.nr; i++) {
1605                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1606                                need_8bit_cte = 1;
1607                                break;
1608                        }
1609                }
1610        }
1611
1612        if (need_8bit_cte > 0) {
1613                const char *header_fmt =
1614                        "MIME-Version: 1.0\n"
1615                        "Content-Type: text/plain; charset=%s\n"
1616                        "Content-Transfer-Encoding: 8bit\n";
1617                strbuf_addf(sb, header_fmt, encoding);
1618        }
1619        if (pp->after_subject) {
1620                strbuf_addstr(sb, pp->after_subject);
1621        }
1622        if (cmit_fmt_is_mail(pp->fmt)) {
1623                strbuf_addch(sb, '\n');
1624        }
1625
1626        if (pp->in_body_headers.nr) {
1627                int i;
1628                for (i = 0; i < pp->in_body_headers.nr; i++) {
1629                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1630                        free(pp->in_body_headers.items[i].string);
1631                }
1632                string_list_clear(&pp->in_body_headers, 0);
1633                strbuf_addch(sb, '\n');
1634        }
1635
1636        strbuf_release(&title);
1637}
1638
1639static int pp_utf8_width(const char *start, const char *end)
1640{
1641        int width = 0;
1642        size_t remain = end - start;
1643
1644        while (remain) {
1645                int n = utf8_width(&start, &remain);
1646                if (n < 0 || !start)
1647                        return -1;
1648                width += n;
1649        }
1650        return width;
1651}
1652
1653static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1654                                 const char *line, int linelen)
1655{
1656        const char *tab;
1657
1658        while ((tab = memchr(line, '\t', linelen)) != NULL) {
1659                int width = pp_utf8_width(line, tab);
1660
1661                /*
1662                 * If it wasn't well-formed utf8, or it
1663                 * had characters with badly defined
1664                 * width (control characters etc), just
1665                 * give up on trying to align things.
1666                 */
1667                if (width < 0)
1668                        break;
1669
1670                /* Output the data .. */
1671                strbuf_add(sb, line, tab - line);
1672
1673                /* .. and the de-tabified tab */
1674                strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1675
1676                /* Skip over the printed part .. */
1677                linelen -= tab + 1 - line;
1678                line = tab + 1;
1679        }
1680
1681        /*
1682         * Print out everything after the last tab without
1683         * worrying about width - there's nothing more to
1684         * align.
1685         */
1686        strbuf_add(sb, line, linelen);
1687}
1688
1689/*
1690 * pp_handle_indent() prints out the intendation, and
1691 * the whole line (without the final newline), after
1692 * de-tabifying.
1693 */
1694static void pp_handle_indent(struct pretty_print_context *pp,
1695                             struct strbuf *sb, int indent,
1696                             const char *line, int linelen)
1697{
1698        strbuf_addchars(sb, ' ', indent);
1699        if (pp->expand_tabs_in_log)
1700                strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1701        else
1702                strbuf_add(sb, line, linelen);
1703}
1704
1705static int is_mboxrd_from(const char *line, int len)
1706{
1707        /*
1708         * a line matching /^From $/ here would only have len == 4
1709         * at this point because is_empty_line would've trimmed all
1710         * trailing space
1711         */
1712        return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1713}
1714
1715void pp_remainder(struct pretty_print_context *pp,
1716                  const char **msg_p,
1717                  struct strbuf *sb,
1718                  int indent)
1719{
1720        int first = 1;
1721        for (;;) {
1722                const char *line = *msg_p;
1723                int linelen = get_one_line(line);
1724                *msg_p += linelen;
1725
1726                if (!linelen)
1727                        break;
1728
1729                if (is_blank_line(line, &linelen)) {
1730                        if (first)
1731                                continue;
1732                        if (pp->fmt == CMIT_FMT_SHORT)
1733                                break;
1734                }
1735                first = 0;
1736
1737                strbuf_grow(sb, linelen + indent + 20);
1738                if (indent)
1739                        pp_handle_indent(pp, sb, indent, line, linelen);
1740                else if (pp->expand_tabs_in_log)
1741                        strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1742                                             line, linelen);
1743                else {
1744                        if (pp->fmt == CMIT_FMT_MBOXRD &&
1745                                        is_mboxrd_from(line, linelen))
1746                                strbuf_addch(sb, '>');
1747
1748                        strbuf_add(sb, line, linelen);
1749                }
1750                strbuf_addch(sb, '\n');
1751        }
1752}
1753
1754void pretty_print_commit(struct pretty_print_context *pp,
1755                         const struct commit *commit,
1756                         struct strbuf *sb)
1757{
1758        unsigned long beginning_of_body;
1759        int indent = 4;
1760        const char *msg;
1761        const char *reencoded;
1762        const char *encoding;
1763        int need_8bit_cte = pp->need_8bit_cte;
1764
1765        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1766                format_commit_message(commit, user_format, sb, pp);
1767                return;
1768        }
1769
1770        encoding = get_log_output_encoding();
1771        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1772
1773        if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1774                indent = 0;
1775
1776        /*
1777         * We need to check and emit Content-type: to mark it
1778         * as 8-bit if we haven't done so.
1779         */
1780        if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1781                int i, ch, in_body;
1782
1783                for (in_body = i = 0; (ch = msg[i]); i++) {
1784                        if (!in_body) {
1785                                /* author could be non 7-bit ASCII but
1786                                 * the log may be so; skip over the
1787                                 * header part first.
1788                                 */
1789                                if (ch == '\n' && msg[i+1] == '\n')
1790                                        in_body = 1;
1791                        }
1792                        else if (non_ascii(ch)) {
1793                                need_8bit_cte = 1;
1794                                break;
1795                        }
1796                }
1797        }
1798
1799        pp_header(pp, encoding, commit, &msg, sb);
1800        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1801                strbuf_addch(sb, '\n');
1802        }
1803
1804        /* Skip excess blank lines at the beginning of body, if any... */
1805        msg = skip_blank_lines(msg);
1806
1807        /* These formats treat the title line specially. */
1808        if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1809                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1810
1811        beginning_of_body = sb->len;
1812        if (pp->fmt != CMIT_FMT_ONELINE)
1813                pp_remainder(pp, &msg, sb, indent);
1814        strbuf_rtrim(sb);
1815
1816        /* Make sure there is an EOLN for the non-oneline case */
1817        if (pp->fmt != CMIT_FMT_ONELINE)
1818                strbuf_addch(sb, '\n');
1819
1820        /*
1821         * The caller may append additional body text in e-mail
1822         * format.  Make sure we did not strip the blank line
1823         * between the header and the body.
1824         */
1825        if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1826                strbuf_addch(sb, '\n');
1827
1828        unuse_commit_buffer(commit, reencoded);
1829}
1830
1831void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1832                    struct strbuf *sb)
1833{
1834        struct pretty_print_context pp = {0};
1835        pp.fmt = fmt;
1836        pretty_print_commit(&pp, commit, sb);
1837}