pretty.con commit t4205: refactor %(trailers) tests (cc1735c)
   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#include "trailer.h"
  14
  15static char *user_format;
  16static struct cmt_fmt_map {
  17        const char *name;
  18        enum cmit_fmt format;
  19        int is_tformat;
  20        int expand_tabs_in_log;
  21        int is_alias;
  22        const char *user_format;
  23} *commit_formats;
  24static size_t builtin_formats_len;
  25static size_t commit_formats_len;
  26static size_t commit_formats_alloc;
  27static struct cmt_fmt_map *find_commit_format(const char *sought);
  28
  29int commit_format_is_empty(enum cmit_fmt fmt)
  30{
  31        return fmt == CMIT_FMT_USERFORMAT && !*user_format;
  32}
  33
  34static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  35{
  36        free(user_format);
  37        user_format = xstrdup(cp);
  38        if (is_tformat)
  39                rev->use_terminator = 1;
  40        rev->commit_format = CMIT_FMT_USERFORMAT;
  41}
  42
  43static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  44{
  45        struct cmt_fmt_map *commit_format = NULL;
  46        const char *name;
  47        const char *fmt;
  48        int i;
  49
  50        if (!skip_prefix(var, "pretty.", &name))
  51                return 0;
  52
  53        for (i = 0; i < builtin_formats_len; i++) {
  54                if (!strcmp(commit_formats[i].name, name))
  55                        return 0;
  56        }
  57
  58        for (i = builtin_formats_len; i < commit_formats_len; i++) {
  59                if (!strcmp(commit_formats[i].name, name)) {
  60                        commit_format = &commit_formats[i];
  61                        break;
  62                }
  63        }
  64
  65        if (!commit_format) {
  66                ALLOC_GROW(commit_formats, commit_formats_len+1,
  67                           commit_formats_alloc);
  68                commit_format = &commit_formats[commit_formats_len];
  69                memset(commit_format, 0, sizeof(*commit_format));
  70                commit_formats_len++;
  71        }
  72
  73        commit_format->name = xstrdup(name);
  74        commit_format->format = CMIT_FMT_USERFORMAT;
  75        if (git_config_string(&fmt, var, value))
  76                return -1;
  77
  78        if (skip_prefix(fmt, "format:", &fmt))
  79                commit_format->is_tformat = 0;
  80        else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
  81                commit_format->is_tformat = 1;
  82        else
  83                commit_format->is_alias = 1;
  84        commit_format->user_format = fmt;
  85
  86        return 0;
  87}
  88
  89static void setup_commit_formats(void)
  90{
  91        struct cmt_fmt_map builtin_formats[] = {
  92                { "raw",        CMIT_FMT_RAW,           0,      0 },
  93                { "medium",     CMIT_FMT_MEDIUM,        0,      8 },
  94                { "short",      CMIT_FMT_SHORT,         0,      0 },
  95                { "email",      CMIT_FMT_EMAIL,         0,      0 },
  96                { "mboxrd",     CMIT_FMT_MBOXRD,        0,      0 },
  97                { "fuller",     CMIT_FMT_FULLER,        0,      8 },
  98                { "full",       CMIT_FMT_FULL,          0,      8 },
  99                { "oneline",    CMIT_FMT_ONELINE,       1,      0 }
 100        };
 101        commit_formats_len = ARRAY_SIZE(builtin_formats);
 102        builtin_formats_len = commit_formats_len;
 103        ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
 104        memcpy(commit_formats, builtin_formats,
 105               sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
 106
 107        git_config(git_pretty_formats_config, NULL);
 108}
 109
 110static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 111                                                        const char *original,
 112                                                        int num_redirections)
 113{
 114        struct cmt_fmt_map *found = NULL;
 115        size_t found_match_len = 0;
 116        int i;
 117
 118        if (num_redirections >= commit_formats_len)
 119                die("invalid --pretty format: "
 120                    "'%s' references an alias which points to itself",
 121                    original);
 122
 123        for (i = 0; i < commit_formats_len; i++) {
 124                size_t match_len;
 125
 126                if (!starts_with(commit_formats[i].name, sought))
 127                        continue;
 128
 129                match_len = strlen(commit_formats[i].name);
 130                if (found == NULL || found_match_len > match_len) {
 131                        found = &commit_formats[i];
 132                        found_match_len = match_len;
 133                }
 134        }
 135
 136        if (found && found->is_alias) {
 137                found = find_commit_format_recursive(found->user_format,
 138                                                     original,
 139                                                     num_redirections+1);
 140        }
 141
 142        return found;
 143}
 144
 145static struct cmt_fmt_map *find_commit_format(const char *sought)
 146{
 147        if (!commit_formats)
 148                setup_commit_formats();
 149
 150        return find_commit_format_recursive(sought, sought, 0);
 151}
 152
 153void get_commit_format(const char *arg, struct rev_info *rev)
 154{
 155        struct cmt_fmt_map *commit_format;
 156
 157        rev->use_terminator = 0;
 158        if (!arg) {
 159                rev->commit_format = CMIT_FMT_DEFAULT;
 160                return;
 161        }
 162        if (skip_prefix(arg, "format:", &arg)) {
 163                save_user_format(rev, arg, 0);
 164                return;
 165        }
 166
 167        if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
 168                save_user_format(rev, arg, 1);
 169                return;
 170        }
 171
 172        commit_format = find_commit_format(arg);
 173        if (!commit_format)
 174                die("invalid --pretty format: %s", arg);
 175
 176        rev->commit_format = commit_format->format;
 177        rev->use_terminator = commit_format->is_tformat;
 178        rev->expand_tabs_in_log_default = commit_format->expand_tabs_in_log;
 179        if (commit_format->format == CMIT_FMT_USERFORMAT) {
 180                save_user_format(rev, commit_format->user_format,
 181                                 commit_format->is_tformat);
 182        }
 183}
 184
 185/*
 186 * Generic support for pretty-printing the header
 187 */
 188static int get_one_line(const char *msg)
 189{
 190        int ret = 0;
 191
 192        for (;;) {
 193                char c = *msg++;
 194                if (!c)
 195                        break;
 196                ret++;
 197                if (c == '\n')
 198                        break;
 199        }
 200        return ret;
 201}
 202
 203/* High bit set, or ISO-2022-INT */
 204static int non_ascii(int ch)
 205{
 206        return !isascii(ch) || ch == '\033';
 207}
 208
 209int has_non_ascii(const char *s)
 210{
 211        int ch;
 212        if (!s)
 213                return 0;
 214        while ((ch = *s++) != '\0') {
 215                if (non_ascii(ch))
 216                        return 1;
 217        }
 218        return 0;
 219}
 220
 221static int is_rfc822_special(char ch)
 222{
 223        switch (ch) {
 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        case '\\':
 237                return 1;
 238        default:
 239                return 0;
 240        }
 241}
 242
 243static int needs_rfc822_quoting(const char *s, int len)
 244{
 245        int i;
 246        for (i = 0; i < len; i++)
 247                if (is_rfc822_special(s[i]))
 248                        return 1;
 249        return 0;
 250}
 251
 252static int last_line_length(struct strbuf *sb)
 253{
 254        int i;
 255
 256        /* How many bytes are already used on the last line? */
 257        for (i = sb->len - 1; i >= 0; i--)
 258                if (sb->buf[i] == '\n')
 259                        break;
 260        return sb->len - (i + 1);
 261}
 262
 263static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 264{
 265        int i;
 266
 267        /* just a guess, we may have to also backslash-quote */
 268        strbuf_grow(out, len + 2);
 269
 270        strbuf_addch(out, '"');
 271        for (i = 0; i < len; i++) {
 272                switch (s[i]) {
 273                case '"':
 274                case '\\':
 275                        strbuf_addch(out, '\\');
 276                        /* fall through */
 277                default:
 278                        strbuf_addch(out, s[i]);
 279                }
 280        }
 281        strbuf_addch(out, '"');
 282}
 283
 284enum rfc2047_type {
 285        RFC2047_SUBJECT,
 286        RFC2047_ADDRESS
 287};
 288
 289static int is_rfc2047_special(char ch, enum rfc2047_type type)
 290{
 291        /*
 292         * rfc2047, section 4.2:
 293         *
 294         *    8-bit values which correspond to printable ASCII characters other
 295         *    than "=", "?", and "_" (underscore), MAY be represented as those
 296         *    characters.  (But see section 5 for restrictions.)  In
 297         *    particular, SPACE and TAB MUST NOT be represented as themselves
 298         *    within encoded words.
 299         */
 300
 301        /*
 302         * rule out non-ASCII characters and non-printable characters (the
 303         * non-ASCII check should be redundant as isprint() is not localized
 304         * and only knows about ASCII, but be defensive about that)
 305         */
 306        if (non_ascii(ch) || !isprint(ch))
 307                return 1;
 308
 309        /*
 310         * rule out special printable characters (' ' should be the only
 311         * whitespace character considered printable, but be defensive and use
 312         * isspace())
 313         */
 314        if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 315                return 1;
 316
 317        /*
 318         * rfc2047, section 5.3:
 319         *
 320         *    As a replacement for a 'word' entity within a 'phrase', for example,
 321         *    one that precedes an address in a From, To, or Cc header.  The ABNF
 322         *    definition for 'phrase' from RFC 822 thus becomes:
 323         *
 324         *    phrase = 1*( encoded-word / word )
 325         *
 326         *    In this case the set of characters that may be used in a "Q"-encoded
 327         *    'encoded-word' is restricted to: <upper and lower case ASCII
 328         *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 329         *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 330         *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 331         *    'special' by 'linear-white-space'.
 332         */
 333
 334        if (type != RFC2047_ADDRESS)
 335                return 0;
 336
 337        /* '=' and '_' are special cases and have been checked above */
 338        return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 339}
 340
 341static int needs_rfc2047_encoding(const char *line, int len,
 342                                  enum rfc2047_type type)
 343{
 344        int i;
 345
 346        for (i = 0; i < len; i++) {
 347                int ch = line[i];
 348                if (non_ascii(ch) || ch == '\n')
 349                        return 1;
 350                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 351                        return 1;
 352        }
 353
 354        return 0;
 355}
 356
 357static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 358                       const char *encoding, enum rfc2047_type type)
 359{
 360        static const int max_encoded_length = 76; /* per rfc2047 */
 361        int i;
 362        int line_len = last_line_length(sb);
 363
 364        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 365        strbuf_addf(sb, "=?%s?q?", encoding);
 366        line_len += strlen(encoding) + 5; /* 5 for =??q? */
 367
 368        while (len) {
 369                /*
 370                 * RFC 2047, section 5 (3):
 371                 *
 372                 * Each 'encoded-word' MUST represent an integral number of
 373                 * characters.  A multi-octet character may not be split across
 374                 * adjacent 'encoded- word's.
 375                 */
 376                const unsigned char *p = (const unsigned char *)line;
 377                int chrlen = mbs_chrlen(&line, &len, encoding);
 378                int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 379
 380                /* "=%02X" * chrlen, or the byte itself */
 381                const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 382                int         encoded_len = is_special ? 3 * chrlen : 1;
 383
 384                /*
 385                 * According to RFC 2047, we could encode the special character
 386                 * ' ' (space) with '_' (underscore) for readability. But many
 387                 * programs do not understand this and just leave the
 388                 * underscore in place. Thus, we do nothing special here, which
 389                 * causes ' ' to be encoded as '=20', avoiding this problem.
 390                 */
 391
 392                if (line_len + encoded_len + 2 > max_encoded_length) {
 393                        /* It won't fit with trailing "?=" --- break the line */
 394                        strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 395                        line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 396                }
 397
 398                for (i = 0; i < chrlen; i++)
 399                        strbuf_addf(sb, encoded_fmt, p[i]);
 400                line_len += encoded_len;
 401        }
 402        strbuf_addstr(sb, "?=");
 403}
 404
 405const char *show_ident_date(const struct ident_split *ident,
 406                            const struct date_mode *mode)
 407{
 408        unsigned long date = 0;
 409        long tz = 0;
 410
 411        if (ident->date_begin && ident->date_end)
 412                date = strtoul(ident->date_begin, NULL, 10);
 413        if (date_overflows(date))
 414                date = 0;
 415        else {
 416                if (ident->tz_begin && ident->tz_end)
 417                        tz = strtol(ident->tz_begin, NULL, 10);
 418                if (tz >= INT_MAX || tz <= INT_MIN)
 419                        tz = 0;
 420        }
 421        return show_date(date, tz, mode);
 422}
 423
 424void pp_user_info(struct pretty_print_context *pp,
 425                  const char *what, struct strbuf *sb,
 426                  const char *line, const char *encoding)
 427{
 428        struct ident_split ident;
 429        char *line_end;
 430        const char *mailbuf, *namebuf;
 431        size_t namelen, maillen;
 432        int max_length = 78; /* per rfc2822 */
 433
 434        if (pp->fmt == CMIT_FMT_ONELINE)
 435                return;
 436
 437        line_end = strchrnul(line, '\n');
 438        if (split_ident_line(&ident, line, line_end - line))
 439                return;
 440
 441        mailbuf = ident.mail_begin;
 442        maillen = ident.mail_end - ident.mail_begin;
 443        namebuf = ident.name_begin;
 444        namelen = ident.name_end - ident.name_begin;
 445
 446        if (pp->mailmap)
 447                map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 448
 449        if (cmit_fmt_is_mail(pp->fmt)) {
 450                if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 451                        struct strbuf buf = STRBUF_INIT;
 452
 453                        strbuf_addstr(&buf, "From: ");
 454                        strbuf_add(&buf, namebuf, namelen);
 455                        strbuf_addstr(&buf, " <");
 456                        strbuf_add(&buf, mailbuf, maillen);
 457                        strbuf_addstr(&buf, ">\n");
 458                        string_list_append(&pp->in_body_headers,
 459                                           strbuf_detach(&buf, NULL));
 460
 461                        mailbuf = pp->from_ident->mail_begin;
 462                        maillen = pp->from_ident->mail_end - mailbuf;
 463                        namebuf = pp->from_ident->name_begin;
 464                        namelen = pp->from_ident->name_end - namebuf;
 465                }
 466
 467                strbuf_addstr(sb, "From: ");
 468                if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 469                        add_rfc2047(sb, namebuf, namelen,
 470                                    encoding, RFC2047_ADDRESS);
 471                        max_length = 76; /* per rfc2047 */
 472                } else if (needs_rfc822_quoting(namebuf, namelen)) {
 473                        struct strbuf quoted = STRBUF_INIT;
 474                        add_rfc822_quoted(&quoted, namebuf, namelen);
 475                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 476                                                        -6, 1, max_length);
 477                        strbuf_release(&quoted);
 478                } else {
 479                        strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 480                                                 -6, 1, max_length);
 481                }
 482
 483                if (max_length <
 484                    last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 485                        strbuf_addch(sb, '\n');
 486                strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 487        } else {
 488                strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 489                            (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 490                            (int)namelen, namebuf, (int)maillen, mailbuf);
 491        }
 492
 493        switch (pp->fmt) {
 494        case CMIT_FMT_MEDIUM:
 495                strbuf_addf(sb, "Date:   %s\n",
 496                            show_ident_date(&ident, &pp->date_mode));
 497                break;
 498        case CMIT_FMT_EMAIL:
 499        case CMIT_FMT_MBOXRD:
 500                strbuf_addf(sb, "Date: %s\n",
 501                            show_ident_date(&ident, DATE_MODE(RFC2822)));
 502                break;
 503        case CMIT_FMT_FULLER:
 504                strbuf_addf(sb, "%sDate: %s\n", what,
 505                            show_ident_date(&ident, &pp->date_mode));
 506                break;
 507        default:
 508                /* notin' */
 509                break;
 510        }
 511}
 512
 513static int is_blank_line(const char *line, int *len_p)
 514{
 515        int len = *len_p;
 516        while (len && isspace(line[len - 1]))
 517                len--;
 518        *len_p = len;
 519        return !len;
 520}
 521
 522const char *skip_blank_lines(const char *msg)
 523{
 524        for (;;) {
 525                int linelen = get_one_line(msg);
 526                int ll = linelen;
 527                if (!linelen)
 528                        break;
 529                if (!is_blank_line(msg, &ll))
 530                        break;
 531                msg += linelen;
 532        }
 533        return msg;
 534}
 535
 536static void add_merge_info(const struct pretty_print_context *pp,
 537                           struct strbuf *sb, const struct commit *commit)
 538{
 539        struct commit_list *parent = commit->parents;
 540
 541        if ((pp->fmt == CMIT_FMT_ONELINE) || (cmit_fmt_is_mail(pp->fmt)) ||
 542            !parent || !parent->next)
 543                return;
 544
 545        strbuf_addstr(sb, "Merge:");
 546
 547        while (parent) {
 548                struct object_id *oidp = &parent->item->object.oid;
 549                strbuf_addch(sb, ' ');
 550                if (pp->abbrev)
 551                        strbuf_add_unique_abbrev(sb, oidp->hash, pp->abbrev);
 552                else
 553                        strbuf_addstr(sb, oid_to_hex(oidp));
 554                parent = parent->next;
 555        }
 556        strbuf_addch(sb, '\n');
 557}
 558
 559static char *get_header(const char *msg, const char *key)
 560{
 561        size_t len;
 562        const char *v = find_commit_header(msg, key, &len);
 563        return v ? xmemdupz(v, len) : NULL;
 564}
 565
 566static char *replace_encoding_header(char *buf, const char *encoding)
 567{
 568        struct strbuf tmp = STRBUF_INIT;
 569        size_t start, len;
 570        char *cp = buf;
 571
 572        /* guess if there is an encoding header before a \n\n */
 573        while (!starts_with(cp, "encoding ")) {
 574                cp = strchr(cp, '\n');
 575                if (!cp || *++cp == '\n')
 576                        return buf;
 577        }
 578        start = cp - buf;
 579        cp = strchr(cp, '\n');
 580        if (!cp)
 581                return buf; /* should not happen but be defensive */
 582        len = cp + 1 - (buf + start);
 583
 584        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 585        if (is_encoding_utf8(encoding)) {
 586                /* we have re-coded to UTF-8; drop the header */
 587                strbuf_remove(&tmp, start, len);
 588        } else {
 589                /* just replaces XXXX in 'encoding XXXX\n' */
 590                strbuf_splice(&tmp, start + strlen("encoding "),
 591                                          len - strlen("encoding \n"),
 592                                          encoding, strlen(encoding));
 593        }
 594        return strbuf_detach(&tmp, NULL);
 595}
 596
 597const char *logmsg_reencode(const struct commit *commit,
 598                            char **commit_encoding,
 599                            const char *output_encoding)
 600{
 601        static const char *utf8 = "UTF-8";
 602        const char *use_encoding;
 603        char *encoding;
 604        const char *msg = get_commit_buffer(commit, NULL);
 605        char *out;
 606
 607        if (!output_encoding || !*output_encoding) {
 608                if (commit_encoding)
 609                        *commit_encoding = get_header(msg, "encoding");
 610                return msg;
 611        }
 612        encoding = get_header(msg, "encoding");
 613        if (commit_encoding)
 614                *commit_encoding = encoding;
 615        use_encoding = encoding ? encoding : utf8;
 616        if (same_encoding(use_encoding, output_encoding)) {
 617                /*
 618                 * No encoding work to be done. If we have no encoding header
 619                 * at all, then there's nothing to do, and we can return the
 620                 * message verbatim (whether newly allocated or not).
 621                 */
 622                if (!encoding)
 623                        return msg;
 624
 625                /*
 626                 * Otherwise, we still want to munge the encoding header in the
 627                 * result, which will be done by modifying the buffer. If we
 628                 * are using a fresh copy, we can reuse it. But if we are using
 629                 * the cached copy from get_commit_buffer, we need to duplicate it
 630                 * to avoid munging the cached copy.
 631                 */
 632                if (msg == get_cached_commit_buffer(commit, NULL))
 633                        out = xstrdup(msg);
 634                else
 635                        out = (char *)msg;
 636        }
 637        else {
 638                /*
 639                 * There's actual encoding work to do. Do the reencoding, which
 640                 * still leaves the header to be replaced in the next step. At
 641                 * this point, we are done with msg. If we allocated a fresh
 642                 * copy, we can free it.
 643                 */
 644                out = reencode_string(msg, output_encoding, use_encoding);
 645                if (out)
 646                        unuse_commit_buffer(commit, msg);
 647        }
 648
 649        /*
 650         * This replacement actually consumes the buffer we hand it, so we do
 651         * not have to worry about freeing the old "out" here.
 652         */
 653        if (out)
 654                out = replace_encoding_header(out, output_encoding);
 655
 656        if (!commit_encoding)
 657                free(encoding);
 658        /*
 659         * If the re-encoding failed, out might be NULL here; in that
 660         * case we just return the commit message verbatim.
 661         */
 662        return out ? out : msg;
 663}
 664
 665static int mailmap_name(const char **email, size_t *email_len,
 666                        const char **name, size_t *name_len)
 667{
 668        static struct string_list *mail_map;
 669        if (!mail_map) {
 670                mail_map = xcalloc(1, sizeof(*mail_map));
 671                read_mailmap(mail_map, NULL);
 672        }
 673        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 674}
 675
 676static size_t format_person_part(struct strbuf *sb, char part,
 677                                 const char *msg, int len,
 678                                 const struct date_mode *dmode)
 679{
 680        /* currently all placeholders have same length */
 681        const int placeholder_len = 2;
 682        struct ident_split s;
 683        const char *name, *mail;
 684        size_t maillen, namelen;
 685
 686        if (split_ident_line(&s, msg, len) < 0)
 687                goto skip;
 688
 689        name = s.name_begin;
 690        namelen = s.name_end - s.name_begin;
 691        mail = s.mail_begin;
 692        maillen = s.mail_end - s.mail_begin;
 693
 694        if (part == 'N' || part == 'E') /* mailmap lookup */
 695                mailmap_name(&mail, &maillen, &name, &namelen);
 696        if (part == 'n' || part == 'N') {       /* name */
 697                strbuf_add(sb, name, namelen);
 698                return placeholder_len;
 699        }
 700        if (part == 'e' || part == 'E') {       /* email */
 701                strbuf_add(sb, mail, maillen);
 702                return placeholder_len;
 703        }
 704
 705        if (!s.date_begin)
 706                goto skip;
 707
 708        if (part == 't') {      /* date, UNIX timestamp */
 709                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 710                return placeholder_len;
 711        }
 712
 713        switch (part) {
 714        case 'd':       /* date */
 715                strbuf_addstr(sb, show_ident_date(&s, dmode));
 716                return placeholder_len;
 717        case 'D':       /* date, RFC2822 style */
 718                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
 719                return placeholder_len;
 720        case 'r':       /* date, relative */
 721                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
 722                return placeholder_len;
 723        case 'i':       /* date, ISO 8601-like */
 724                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
 725                return placeholder_len;
 726        case 'I':       /* date, ISO 8601 strict */
 727                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
 728                return placeholder_len;
 729        }
 730
 731skip:
 732        /*
 733         * reading from either a bogus commit, or a reflog entry with
 734         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 735         * to compute a valid return value.
 736         */
 737        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 738            || part == 'D' || part == 'r' || part == 'i')
 739                return placeholder_len;
 740
 741        return 0; /* unknown placeholder */
 742}
 743
 744struct chunk {
 745        size_t off;
 746        size_t len;
 747};
 748
 749enum flush_type {
 750        no_flush,
 751        flush_right,
 752        flush_left,
 753        flush_left_and_steal,
 754        flush_both
 755};
 756
 757enum trunc_type {
 758        trunc_none,
 759        trunc_left,
 760        trunc_middle,
 761        trunc_right
 762};
 763
 764struct format_commit_context {
 765        const struct commit *commit;
 766        const struct pretty_print_context *pretty_ctx;
 767        unsigned commit_header_parsed:1;
 768        unsigned commit_message_parsed:1;
 769        struct signature_check signature_check;
 770        enum flush_type flush_type;
 771        enum trunc_type truncate;
 772        const char *message;
 773        char *commit_encoding;
 774        size_t width, indent1, indent2;
 775        int auto_color;
 776        int padding;
 777
 778        /* These offsets are relative to the start of the commit message. */
 779        struct chunk author;
 780        struct chunk committer;
 781        size_t message_off;
 782        size_t subject_off;
 783        size_t body_off;
 784
 785        /* The following ones are relative to the result struct strbuf. */
 786        size_t wrap_start;
 787};
 788
 789static void parse_commit_header(struct format_commit_context *context)
 790{
 791        const char *msg = context->message;
 792        int i;
 793
 794        for (i = 0; msg[i]; i++) {
 795                const char *name;
 796                int eol;
 797                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 798                        ; /* do nothing */
 799
 800                if (i == eol) {
 801                        break;
 802                } else if (skip_prefix(msg + i, "author ", &name)) {
 803                        context->author.off = name - msg;
 804                        context->author.len = msg + eol - name;
 805                } else if (skip_prefix(msg + i, "committer ", &name)) {
 806                        context->committer.off = name - msg;
 807                        context->committer.len = msg + eol - name;
 808                }
 809                i = eol;
 810        }
 811        context->message_off = i;
 812        context->commit_header_parsed = 1;
 813}
 814
 815static int istitlechar(char c)
 816{
 817        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 818                (c >= '0' && c <= '9') || c == '.' || c == '_';
 819}
 820
 821static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 822{
 823        size_t trimlen;
 824        size_t start_len = sb->len;
 825        int space = 2;
 826
 827        for (; *msg && *msg != '\n'; msg++) {
 828                if (istitlechar(*msg)) {
 829                        if (space == 1)
 830                                strbuf_addch(sb, '-');
 831                        space = 0;
 832                        strbuf_addch(sb, *msg);
 833                        if (*msg == '.')
 834                                while (*(msg+1) == '.')
 835                                        msg++;
 836                } else
 837                        space |= 1;
 838        }
 839
 840        /* trim any trailing '.' or '-' characters */
 841        trimlen = 0;
 842        while (sb->len - trimlen > start_len &&
 843                (sb->buf[sb->len - 1 - trimlen] == '.'
 844                || sb->buf[sb->len - 1 - trimlen] == '-'))
 845                trimlen++;
 846        strbuf_remove(sb, sb->len - trimlen, trimlen);
 847}
 848
 849const char *format_subject(struct strbuf *sb, const char *msg,
 850                           const char *line_separator)
 851{
 852        int first = 1;
 853
 854        for (;;) {
 855                const char *line = msg;
 856                int linelen = get_one_line(line);
 857
 858                msg += linelen;
 859                if (!linelen || is_blank_line(line, &linelen))
 860                        break;
 861
 862                if (!sb)
 863                        continue;
 864                strbuf_grow(sb, linelen + 2);
 865                if (!first)
 866                        strbuf_addstr(sb, line_separator);
 867                strbuf_add(sb, line, linelen);
 868                first = 0;
 869        }
 870        return msg;
 871}
 872
 873static void parse_commit_message(struct format_commit_context *c)
 874{
 875        const char *msg = c->message + c->message_off;
 876        const char *start = c->message;
 877
 878        msg = skip_blank_lines(msg);
 879        c->subject_off = msg - start;
 880
 881        msg = format_subject(NULL, msg, NULL);
 882        msg = skip_blank_lines(msg);
 883        c->body_off = msg - start;
 884
 885        c->commit_message_parsed = 1;
 886}
 887
 888static void strbuf_wrap(struct strbuf *sb, size_t pos,
 889                        size_t width, size_t indent1, size_t indent2)
 890{
 891        struct strbuf tmp = STRBUF_INIT;
 892
 893        if (pos)
 894                strbuf_add(&tmp, sb->buf, pos);
 895        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 896                                (int) indent1, (int) indent2, (int) width);
 897        strbuf_swap(&tmp, sb);
 898        strbuf_release(&tmp);
 899}
 900
 901static void rewrap_message_tail(struct strbuf *sb,
 902                                struct format_commit_context *c,
 903                                size_t new_width, size_t new_indent1,
 904                                size_t new_indent2)
 905{
 906        if (c->width == new_width && c->indent1 == new_indent1 &&
 907            c->indent2 == new_indent2)
 908                return;
 909        if (c->wrap_start < sb->len)
 910                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 911        c->wrap_start = sb->len;
 912        c->width = new_width;
 913        c->indent1 = new_indent1;
 914        c->indent2 = new_indent2;
 915}
 916
 917static int format_reflog_person(struct strbuf *sb,
 918                                char part,
 919                                struct reflog_walk_info *log,
 920                                const struct date_mode *dmode)
 921{
 922        const char *ident;
 923
 924        if (!log)
 925                return 2;
 926
 927        ident = get_reflog_ident(log);
 928        if (!ident)
 929                return 2;
 930
 931        return format_person_part(sb, part, ident, strlen(ident), dmode);
 932}
 933
 934static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 935                          const char *placeholder,
 936                          struct format_commit_context *c)
 937{
 938        const char *rest = placeholder;
 939
 940        if (placeholder[1] == '(') {
 941                const char *begin = placeholder + 2;
 942                const char *end = strchr(begin, ')');
 943                char color[COLOR_MAXLEN];
 944
 945                if (!end)
 946                        return 0;
 947                if (skip_prefix(begin, "auto,", &begin)) {
 948                        if (!want_color(c->pretty_ctx->color))
 949                                return end - placeholder + 1;
 950                }
 951                if (color_parse_mem(begin, end - begin, color) < 0)
 952                        die(_("unable to parse --pretty format"));
 953                strbuf_addstr(sb, color);
 954                return end - placeholder + 1;
 955        }
 956        if (skip_prefix(placeholder + 1, "red", &rest))
 957                strbuf_addstr(sb, GIT_COLOR_RED);
 958        else if (skip_prefix(placeholder + 1, "green", &rest))
 959                strbuf_addstr(sb, GIT_COLOR_GREEN);
 960        else if (skip_prefix(placeholder + 1, "blue", &rest))
 961                strbuf_addstr(sb, GIT_COLOR_BLUE);
 962        else if (skip_prefix(placeholder + 1, "reset", &rest))
 963                strbuf_addstr(sb, GIT_COLOR_RESET);
 964        return rest - placeholder;
 965}
 966
 967static size_t parse_padding_placeholder(struct strbuf *sb,
 968                                        const char *placeholder,
 969                                        struct format_commit_context *c)
 970{
 971        const char *ch = placeholder;
 972        enum flush_type flush_type;
 973        int to_column = 0;
 974
 975        switch (*ch++) {
 976        case '<':
 977                flush_type = flush_right;
 978                break;
 979        case '>':
 980                if (*ch == '<') {
 981                        flush_type = flush_both;
 982                        ch++;
 983                } else if (*ch == '>') {
 984                        flush_type = flush_left_and_steal;
 985                        ch++;
 986                } else
 987                        flush_type = flush_left;
 988                break;
 989        default:
 990                return 0;
 991        }
 992
 993        /* the next value means "wide enough to that column" */
 994        if (*ch == '|') {
 995                to_column = 1;
 996                ch++;
 997        }
 998
 999        if (*ch == '(') {
1000                const char *start = ch + 1;
1001                const char *end = start + strcspn(start, ",)");
1002                char *next;
1003                int width;
1004                if (!end || end == start)
1005                        return 0;
1006                width = strtol(start, &next, 10);
1007                if (next == start || width == 0)
1008                        return 0;
1009                if (width < 0) {
1010                        if (to_column)
1011                                width += term_columns();
1012                        if (width < 0)
1013                                return 0;
1014                }
1015                c->padding = to_column ? -width : width;
1016                c->flush_type = flush_type;
1017
1018                if (*end == ',') {
1019                        start = end + 1;
1020                        end = strchr(start, ')');
1021                        if (!end || end == start)
1022                                return 0;
1023                        if (starts_with(start, "trunc)"))
1024                                c->truncate = trunc_right;
1025                        else if (starts_with(start, "ltrunc)"))
1026                                c->truncate = trunc_left;
1027                        else if (starts_with(start, "mtrunc)"))
1028                                c->truncate = trunc_middle;
1029                        else
1030                                return 0;
1031                } else
1032                        c->truncate = trunc_none;
1033
1034                return end - placeholder + 1;
1035        }
1036        return 0;
1037}
1038
1039static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1040                                const char *placeholder,
1041                                void *context)
1042{
1043        struct format_commit_context *c = context;
1044        const struct commit *commit = c->commit;
1045        const char *msg = c->message;
1046        struct commit_list *p;
1047        int ch;
1048
1049        /* these are independent of the commit */
1050        switch (placeholder[0]) {
1051        case 'C':
1052                if (starts_with(placeholder + 1, "(auto)")) {
1053                        c->auto_color = want_color(c->pretty_ctx->color);
1054                        if (c->auto_color && sb->len)
1055                                strbuf_addstr(sb, GIT_COLOR_RESET);
1056                        return 7; /* consumed 7 bytes, "C(auto)" */
1057                } else {
1058                        int ret = parse_color(sb, placeholder, c);
1059                        if (ret)
1060                                c->auto_color = 0;
1061                        /*
1062                         * Otherwise, we decided to treat %C<unknown>
1063                         * as a literal string, and the previous
1064                         * %C(auto) is still valid.
1065                         */
1066                        return ret;
1067                }
1068        case 'n':               /* newline */
1069                strbuf_addch(sb, '\n');
1070                return 1;
1071        case 'x':
1072                /* %x00 == NUL, %x0a == LF, etc. */
1073                ch = hex2chr(placeholder + 1);
1074                if (ch < 0)
1075                        return 0;
1076                strbuf_addch(sb, ch);
1077                return 3;
1078        case 'w':
1079                if (placeholder[1] == '(') {
1080                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1081                        char *next;
1082                        const char *start = placeholder + 2;
1083                        const char *end = strchr(start, ')');
1084                        if (!end)
1085                                return 0;
1086                        if (end > start) {
1087                                width = strtoul(start, &next, 10);
1088                                if (*next == ',') {
1089                                        indent1 = strtoul(next + 1, &next, 10);
1090                                        if (*next == ',') {
1091                                                indent2 = strtoul(next + 1,
1092                                                                 &next, 10);
1093                                        }
1094                                }
1095                                if (*next != ')')
1096                                        return 0;
1097                        }
1098                        rewrap_message_tail(sb, c, width, indent1, indent2);
1099                        return end - placeholder + 1;
1100                } else
1101                        return 0;
1102
1103        case '<':
1104        case '>':
1105                return parse_padding_placeholder(sb, placeholder, c);
1106        }
1107
1108        /* these depend on the commit */
1109        if (!commit->object.parsed)
1110                parse_object(commit->object.oid.hash);
1111
1112        switch (placeholder[0]) {
1113        case 'H':               /* commit hash */
1114                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1115                strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1116                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1117                return 1;
1118        case 'h':               /* abbreviated commit hash */
1119                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1120                strbuf_add_unique_abbrev(sb, commit->object.oid.hash,
1121                                         c->pretty_ctx->abbrev);
1122                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1123                return 1;
1124        case 'T':               /* tree hash */
1125                strbuf_addstr(sb, oid_to_hex(&commit->tree->object.oid));
1126                return 1;
1127        case 't':               /* abbreviated tree hash */
1128                strbuf_add_unique_abbrev(sb, commit->tree->object.oid.hash,
1129                                         c->pretty_ctx->abbrev);
1130                return 1;
1131        case 'P':               /* parent hashes */
1132                for (p = commit->parents; p; p = p->next) {
1133                        if (p != commit->parents)
1134                                strbuf_addch(sb, ' ');
1135                        strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1136                }
1137                return 1;
1138        case 'p':               /* abbreviated parent hashes */
1139                for (p = commit->parents; p; p = p->next) {
1140                        if (p != commit->parents)
1141                                strbuf_addch(sb, ' ');
1142                        strbuf_add_unique_abbrev(sb, p->item->object.oid.hash,
1143                                                 c->pretty_ctx->abbrev);
1144                }
1145                return 1;
1146        case 'm':               /* left/right/bottom */
1147                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1148                return 1;
1149        case 'd':
1150                load_ref_decorations(DECORATE_SHORT_REFS);
1151                format_decorations(sb, commit, c->auto_color);
1152                return 1;
1153        case 'D':
1154                load_ref_decorations(DECORATE_SHORT_REFS);
1155                format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1156                return 1;
1157        case 'g':               /* reflog info */
1158                switch(placeholder[1]) {
1159                case 'd':       /* reflog selector */
1160                case 'D':
1161                        if (c->pretty_ctx->reflog_info)
1162                                get_reflog_selector(sb,
1163                                                    c->pretty_ctx->reflog_info,
1164                                                    &c->pretty_ctx->date_mode,
1165                                                    c->pretty_ctx->date_mode_explicit,
1166                                                    (placeholder[1] == 'd'));
1167                        return 2;
1168                case 's':       /* reflog message */
1169                        if (c->pretty_ctx->reflog_info)
1170                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1171                        return 2;
1172                case 'n':
1173                case 'N':
1174                case 'e':
1175                case 'E':
1176                        return format_reflog_person(sb,
1177                                                    placeholder[1],
1178                                                    c->pretty_ctx->reflog_info,
1179                                                    &c->pretty_ctx->date_mode);
1180                }
1181                return 0;       /* unknown %g placeholder */
1182        case 'N':
1183                if (c->pretty_ctx->notes_message) {
1184                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1185                        return 1;
1186                }
1187                return 0;
1188        }
1189
1190        if (placeholder[0] == 'G') {
1191                if (!c->signature_check.result)
1192                        check_commit_signature(c->commit, &(c->signature_check));
1193                switch (placeholder[1]) {
1194                case 'G':
1195                        if (c->signature_check.gpg_output)
1196                                strbuf_addstr(sb, c->signature_check.gpg_output);
1197                        break;
1198                case '?':
1199                        switch (c->signature_check.result) {
1200                        case 'G':
1201                        case 'B':
1202                        case 'E':
1203                        case 'U':
1204                        case 'N':
1205                        case 'X':
1206                        case 'Y':
1207                        case 'R':
1208                                strbuf_addch(sb, c->signature_check.result);
1209                        }
1210                        break;
1211                case 'S':
1212                        if (c->signature_check.signer)
1213                                strbuf_addstr(sb, c->signature_check.signer);
1214                        break;
1215                case 'K':
1216                        if (c->signature_check.key)
1217                                strbuf_addstr(sb, c->signature_check.key);
1218                        break;
1219                default:
1220                        return 0;
1221                }
1222                return 2;
1223        }
1224
1225
1226        /* For the rest we have to parse the commit header. */
1227        if (!c->commit_header_parsed)
1228                parse_commit_header(c);
1229
1230        switch (placeholder[0]) {
1231        case 'a':       /* author ... */
1232                return format_person_part(sb, placeholder[1],
1233                                   msg + c->author.off, c->author.len,
1234                                   &c->pretty_ctx->date_mode);
1235        case 'c':       /* committer ... */
1236                return format_person_part(sb, placeholder[1],
1237                                   msg + c->committer.off, c->committer.len,
1238                                   &c->pretty_ctx->date_mode);
1239        case 'e':       /* encoding */
1240                if (c->commit_encoding)
1241                        strbuf_addstr(sb, c->commit_encoding);
1242                return 1;
1243        case 'B':       /* raw body */
1244                /* message_off is always left at the initial newline */
1245                strbuf_addstr(sb, msg + c->message_off + 1);
1246                return 1;
1247        }
1248
1249        /* Now we need to parse the commit message. */
1250        if (!c->commit_message_parsed)
1251                parse_commit_message(c);
1252
1253        switch (placeholder[0]) {
1254        case 's':       /* subject */
1255                format_subject(sb, msg + c->subject_off, " ");
1256                return 1;
1257        case 'f':       /* sanitized subject */
1258                format_sanitized_subject(sb, msg + c->subject_off);
1259                return 1;
1260        case 'b':       /* body */
1261                strbuf_addstr(sb, msg + c->body_off);
1262                return 1;
1263        }
1264
1265        if (starts_with(placeholder, "(trailers)")) {
1266                struct process_trailer_options opts = PROCESS_TRAILER_OPTIONS_INIT;
1267                format_trailers_from_commit(sb, msg + c->subject_off, &opts);
1268                return strlen("(trailers)");
1269        }
1270
1271        return 0;       /* unknown placeholder */
1272}
1273
1274static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1275                                    const char *placeholder,
1276                                    struct format_commit_context *c)
1277{
1278        struct strbuf local_sb = STRBUF_INIT;
1279        int total_consumed = 0, len, padding = c->padding;
1280        if (padding < 0) {
1281                const char *start = strrchr(sb->buf, '\n');
1282                int occupied;
1283                if (!start)
1284                        start = sb->buf;
1285                occupied = utf8_strnwidth(start, -1, 1);
1286                occupied += c->pretty_ctx->graph_width;
1287                padding = (-padding) - occupied;
1288        }
1289        while (1) {
1290                int modifier = *placeholder == 'C';
1291                int consumed = format_commit_one(&local_sb, placeholder, c);
1292                total_consumed += consumed;
1293
1294                if (!modifier)
1295                        break;
1296
1297                placeholder += consumed;
1298                if (*placeholder != '%')
1299                        break;
1300                placeholder++;
1301                total_consumed++;
1302        }
1303        len = utf8_strnwidth(local_sb.buf, -1, 1);
1304
1305        if (c->flush_type == flush_left_and_steal) {
1306                const char *ch = sb->buf + sb->len - 1;
1307                while (len > padding && ch > sb->buf) {
1308                        const char *p;
1309                        if (*ch == ' ') {
1310                                ch--;
1311                                padding++;
1312                                continue;
1313                        }
1314                        /* check for trailing ansi sequences */
1315                        if (*ch != 'm')
1316                                break;
1317                        p = ch - 1;
1318                        while (ch - p < 10 && *p != '\033')
1319                                p--;
1320                        if (*p != '\033' ||
1321                            ch + 1 - p != display_mode_esc_sequence_len(p))
1322                                break;
1323                        /*
1324                         * got a good ansi sequence, put it back to
1325                         * local_sb as we're cutting sb
1326                         */
1327                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1328                        ch = p - 1;
1329                }
1330                strbuf_setlen(sb, ch + 1 - sb->buf);
1331                c->flush_type = flush_left;
1332        }
1333
1334        if (len > padding) {
1335                switch (c->truncate) {
1336                case trunc_left:
1337                        strbuf_utf8_replace(&local_sb,
1338                                            0, len - (padding - 2),
1339                                            "..");
1340                        break;
1341                case trunc_middle:
1342                        strbuf_utf8_replace(&local_sb,
1343                                            padding / 2 - 1,
1344                                            len - (padding - 2),
1345                                            "..");
1346                        break;
1347                case trunc_right:
1348                        strbuf_utf8_replace(&local_sb,
1349                                            padding - 2, len - (padding - 2),
1350                                            "..");
1351                        break;
1352                case trunc_none:
1353                        break;
1354                }
1355                strbuf_addbuf(sb, &local_sb);
1356        } else {
1357                int sb_len = sb->len, offset = 0;
1358                if (c->flush_type == flush_left)
1359                        offset = padding - len;
1360                else if (c->flush_type == flush_both)
1361                        offset = (padding - len) / 2;
1362                /*
1363                 * we calculate padding in columns, now
1364                 * convert it back to chars
1365                 */
1366                padding = padding - len + local_sb.len;
1367                strbuf_addchars(sb, ' ', padding);
1368                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1369                       local_sb.len);
1370        }
1371        strbuf_release(&local_sb);
1372        c->flush_type = no_flush;
1373        return total_consumed;
1374}
1375
1376static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1377                                 const char *placeholder,
1378                                 void *context)
1379{
1380        int consumed;
1381        size_t orig_len;
1382        enum {
1383                NO_MAGIC,
1384                ADD_LF_BEFORE_NON_EMPTY,
1385                DEL_LF_BEFORE_EMPTY,
1386                ADD_SP_BEFORE_NON_EMPTY
1387        } magic = NO_MAGIC;
1388
1389        switch (placeholder[0]) {
1390        case '-':
1391                magic = DEL_LF_BEFORE_EMPTY;
1392                break;
1393        case '+':
1394                magic = ADD_LF_BEFORE_NON_EMPTY;
1395                break;
1396        case ' ':
1397                magic = ADD_SP_BEFORE_NON_EMPTY;
1398                break;
1399        default:
1400                break;
1401        }
1402        if (magic != NO_MAGIC)
1403                placeholder++;
1404
1405        orig_len = sb->len;
1406        if (((struct format_commit_context *)context)->flush_type != no_flush)
1407                consumed = format_and_pad_commit(sb, placeholder, context);
1408        else
1409                consumed = format_commit_one(sb, placeholder, context);
1410        if (magic == NO_MAGIC)
1411                return consumed;
1412
1413        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1414                while (sb->len && sb->buf[sb->len - 1] == '\n')
1415                        strbuf_setlen(sb, sb->len - 1);
1416        } else if (orig_len != sb->len) {
1417                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1418                        strbuf_insert(sb, orig_len, "\n", 1);
1419                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1420                        strbuf_insert(sb, orig_len, " ", 1);
1421        }
1422        return consumed + 1;
1423}
1424
1425static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1426                                   void *context)
1427{
1428        struct userformat_want *w = context;
1429
1430        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1431                placeholder++;
1432
1433        switch (*placeholder) {
1434        case 'N':
1435                w->notes = 1;
1436                break;
1437        }
1438        return 0;
1439}
1440
1441void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1442{
1443        struct strbuf dummy = STRBUF_INIT;
1444
1445        if (!fmt) {
1446                if (!user_format)
1447                        return;
1448                fmt = user_format;
1449        }
1450        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1451        strbuf_release(&dummy);
1452}
1453
1454void format_commit_message(const struct commit *commit,
1455                           const char *format, struct strbuf *sb,
1456                           const struct pretty_print_context *pretty_ctx)
1457{
1458        struct format_commit_context context;
1459        const char *output_enc = pretty_ctx->output_encoding;
1460        const char *utf8 = "UTF-8";
1461
1462        memset(&context, 0, sizeof(context));
1463        context.commit = commit;
1464        context.pretty_ctx = pretty_ctx;
1465        context.wrap_start = sb->len;
1466        /*
1467         * convert a commit message to UTF-8 first
1468         * as far as 'format_commit_item' assumes it in UTF-8
1469         */
1470        context.message = logmsg_reencode(commit,
1471                                          &context.commit_encoding,
1472                                          utf8);
1473
1474        strbuf_expand(sb, format, format_commit_item, &context);
1475        rewrap_message_tail(sb, &context, 0, 0, 0);
1476
1477        /* then convert a commit message to an actual output encoding */
1478        if (output_enc) {
1479                if (same_encoding(utf8, output_enc))
1480                        output_enc = NULL;
1481        } else {
1482                if (context.commit_encoding &&
1483                    !same_encoding(context.commit_encoding, utf8))
1484                        output_enc = context.commit_encoding;
1485        }
1486
1487        if (output_enc) {
1488                int outsz;
1489                char *out = reencode_string_len(sb->buf, sb->len,
1490                                                output_enc, utf8, &outsz);
1491                if (out)
1492                        strbuf_attach(sb, out, outsz, outsz + 1);
1493        }
1494
1495        free(context.commit_encoding);
1496        unuse_commit_buffer(commit, context.message);
1497}
1498
1499static void pp_header(struct pretty_print_context *pp,
1500                      const char *encoding,
1501                      const struct commit *commit,
1502                      const char **msg_p,
1503                      struct strbuf *sb)
1504{
1505        int parents_shown = 0;
1506
1507        for (;;) {
1508                const char *name, *line = *msg_p;
1509                int linelen = get_one_line(*msg_p);
1510
1511                if (!linelen)
1512                        return;
1513                *msg_p += linelen;
1514
1515                if (linelen == 1)
1516                        /* End of header */
1517                        return;
1518
1519                if (pp->fmt == CMIT_FMT_RAW) {
1520                        strbuf_add(sb, line, linelen);
1521                        continue;
1522                }
1523
1524                if (starts_with(line, "parent ")) {
1525                        if (linelen != 48)
1526                                die("bad parent line in commit");
1527                        continue;
1528                }
1529
1530                if (!parents_shown) {
1531                        unsigned num = commit_list_count(commit->parents);
1532                        /* with enough slop */
1533                        strbuf_grow(sb, num * 50 + 20);
1534                        add_merge_info(pp, sb, commit);
1535                        parents_shown = 1;
1536                }
1537
1538                /*
1539                 * MEDIUM == DEFAULT shows only author with dates.
1540                 * FULL shows both authors but not dates.
1541                 * FULLER shows both authors and dates.
1542                 */
1543                if (skip_prefix(line, "author ", &name)) {
1544                        strbuf_grow(sb, linelen + 80);
1545                        pp_user_info(pp, "Author", sb, name, encoding);
1546                }
1547                if (skip_prefix(line, "committer ", &name) &&
1548                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1549                        strbuf_grow(sb, linelen + 80);
1550                        pp_user_info(pp, "Commit", sb, name, encoding);
1551                }
1552        }
1553}
1554
1555void pp_title_line(struct pretty_print_context *pp,
1556                   const char **msg_p,
1557                   struct strbuf *sb,
1558                   const char *encoding,
1559                   int need_8bit_cte)
1560{
1561        static const int max_length = 78; /* per rfc2047 */
1562        struct strbuf title;
1563
1564        strbuf_init(&title, 80);
1565        *msg_p = format_subject(&title, *msg_p,
1566                                pp->preserve_subject ? "\n" : " ");
1567
1568        strbuf_grow(sb, title.len + 1024);
1569        if (pp->print_email_subject) {
1570                if (pp->rev)
1571                        fmt_output_email_subject(sb, pp->rev);
1572                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1573                        add_rfc2047(sb, title.buf, title.len,
1574                                                encoding, RFC2047_SUBJECT);
1575                else
1576                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1577                                         -last_line_length(sb), 1, max_length);
1578        } else {
1579                strbuf_addbuf(sb, &title);
1580        }
1581        strbuf_addch(sb, '\n');
1582
1583        if (need_8bit_cte == 0) {
1584                int i;
1585                for (i = 0; i < pp->in_body_headers.nr; i++) {
1586                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1587                                need_8bit_cte = 1;
1588                                break;
1589                        }
1590                }
1591        }
1592
1593        if (need_8bit_cte > 0) {
1594                const char *header_fmt =
1595                        "MIME-Version: 1.0\n"
1596                        "Content-Type: text/plain; charset=%s\n"
1597                        "Content-Transfer-Encoding: 8bit\n";
1598                strbuf_addf(sb, header_fmt, encoding);
1599        }
1600        if (pp->after_subject) {
1601                strbuf_addstr(sb, pp->after_subject);
1602        }
1603        if (cmit_fmt_is_mail(pp->fmt)) {
1604                strbuf_addch(sb, '\n');
1605        }
1606
1607        if (pp->in_body_headers.nr) {
1608                int i;
1609                for (i = 0; i < pp->in_body_headers.nr; i++) {
1610                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1611                        free(pp->in_body_headers.items[i].string);
1612                }
1613                string_list_clear(&pp->in_body_headers, 0);
1614                strbuf_addch(sb, '\n');
1615        }
1616
1617        strbuf_release(&title);
1618}
1619
1620static int pp_utf8_width(const char *start, const char *end)
1621{
1622        int width = 0;
1623        size_t remain = end - start;
1624
1625        while (remain) {
1626                int n = utf8_width(&start, &remain);
1627                if (n < 0 || !start)
1628                        return -1;
1629                width += n;
1630        }
1631        return width;
1632}
1633
1634static void strbuf_add_tabexpand(struct strbuf *sb, int tabwidth,
1635                                 const char *line, int linelen)
1636{
1637        const char *tab;
1638
1639        while ((tab = memchr(line, '\t', linelen)) != NULL) {
1640                int width = pp_utf8_width(line, tab);
1641
1642                /*
1643                 * If it wasn't well-formed utf8, or it
1644                 * had characters with badly defined
1645                 * width (control characters etc), just
1646                 * give up on trying to align things.
1647                 */
1648                if (width < 0)
1649                        break;
1650
1651                /* Output the data .. */
1652                strbuf_add(sb, line, tab - line);
1653
1654                /* .. and the de-tabified tab */
1655                strbuf_addchars(sb, ' ', tabwidth - (width % tabwidth));
1656
1657                /* Skip over the printed part .. */
1658                linelen -= tab + 1 - line;
1659                line = tab + 1;
1660        }
1661
1662        /*
1663         * Print out everything after the last tab without
1664         * worrying about width - there's nothing more to
1665         * align.
1666         */
1667        strbuf_add(sb, line, linelen);
1668}
1669
1670/*
1671 * pp_handle_indent() prints out the intendation, and
1672 * the whole line (without the final newline), after
1673 * de-tabifying.
1674 */
1675static void pp_handle_indent(struct pretty_print_context *pp,
1676                             struct strbuf *sb, int indent,
1677                             const char *line, int linelen)
1678{
1679        strbuf_addchars(sb, ' ', indent);
1680        if (pp->expand_tabs_in_log)
1681                strbuf_add_tabexpand(sb, pp->expand_tabs_in_log, line, linelen);
1682        else
1683                strbuf_add(sb, line, linelen);
1684}
1685
1686static int is_mboxrd_from(const char *line, int len)
1687{
1688        /*
1689         * a line matching /^From $/ here would only have len == 4
1690         * at this point because is_empty_line would've trimmed all
1691         * trailing space
1692         */
1693        return len > 4 && starts_with(line + strspn(line, ">"), "From ");
1694}
1695
1696void pp_remainder(struct pretty_print_context *pp,
1697                  const char **msg_p,
1698                  struct strbuf *sb,
1699                  int indent)
1700{
1701        int first = 1;
1702        for (;;) {
1703                const char *line = *msg_p;
1704                int linelen = get_one_line(line);
1705                *msg_p += linelen;
1706
1707                if (!linelen)
1708                        break;
1709
1710                if (is_blank_line(line, &linelen)) {
1711                        if (first)
1712                                continue;
1713                        if (pp->fmt == CMIT_FMT_SHORT)
1714                                break;
1715                }
1716                first = 0;
1717
1718                strbuf_grow(sb, linelen + indent + 20);
1719                if (indent)
1720                        pp_handle_indent(pp, sb, indent, line, linelen);
1721                else if (pp->expand_tabs_in_log)
1722                        strbuf_add_tabexpand(sb, pp->expand_tabs_in_log,
1723                                             line, linelen);
1724                else {
1725                        if (pp->fmt == CMIT_FMT_MBOXRD &&
1726                                        is_mboxrd_from(line, linelen))
1727                                strbuf_addch(sb, '>');
1728
1729                        strbuf_add(sb, line, linelen);
1730                }
1731                strbuf_addch(sb, '\n');
1732        }
1733}
1734
1735void pretty_print_commit(struct pretty_print_context *pp,
1736                         const struct commit *commit,
1737                         struct strbuf *sb)
1738{
1739        unsigned long beginning_of_body;
1740        int indent = 4;
1741        const char *msg;
1742        const char *reencoded;
1743        const char *encoding;
1744        int need_8bit_cte = pp->need_8bit_cte;
1745
1746        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1747                format_commit_message(commit, user_format, sb, pp);
1748                return;
1749        }
1750
1751        encoding = get_log_output_encoding();
1752        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1753
1754        if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1755                indent = 0;
1756
1757        /*
1758         * We need to check and emit Content-type: to mark it
1759         * as 8-bit if we haven't done so.
1760         */
1761        if (cmit_fmt_is_mail(pp->fmt) && need_8bit_cte == 0) {
1762                int i, ch, in_body;
1763
1764                for (in_body = i = 0; (ch = msg[i]); i++) {
1765                        if (!in_body) {
1766                                /* author could be non 7-bit ASCII but
1767                                 * the log may be so; skip over the
1768                                 * header part first.
1769                                 */
1770                                if (ch == '\n' && msg[i+1] == '\n')
1771                                        in_body = 1;
1772                        }
1773                        else if (non_ascii(ch)) {
1774                                need_8bit_cte = 1;
1775                                break;
1776                        }
1777                }
1778        }
1779
1780        pp_header(pp, encoding, commit, &msg, sb);
1781        if (pp->fmt != CMIT_FMT_ONELINE && !pp->print_email_subject) {
1782                strbuf_addch(sb, '\n');
1783        }
1784
1785        /* Skip excess blank lines at the beginning of body, if any... */
1786        msg = skip_blank_lines(msg);
1787
1788        /* These formats treat the title line specially. */
1789        if (pp->fmt == CMIT_FMT_ONELINE || cmit_fmt_is_mail(pp->fmt))
1790                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1791
1792        beginning_of_body = sb->len;
1793        if (pp->fmt != CMIT_FMT_ONELINE)
1794                pp_remainder(pp, &msg, sb, indent);
1795        strbuf_rtrim(sb);
1796
1797        /* Make sure there is an EOLN for the non-oneline case */
1798        if (pp->fmt != CMIT_FMT_ONELINE)
1799                strbuf_addch(sb, '\n');
1800
1801        /*
1802         * The caller may append additional body text in e-mail
1803         * format.  Make sure we did not strip the blank line
1804         * between the header and the body.
1805         */
1806        if (cmit_fmt_is_mail(pp->fmt) && sb->len <= beginning_of_body)
1807                strbuf_addch(sb, '\n');
1808
1809        unuse_commit_buffer(commit, reencoded);
1810}
1811
1812void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1813                    struct strbuf *sb)
1814{
1815        struct pretty_print_context pp = {0};
1816        pp.fmt = fmt;
1817        pretty_print_commit(&pp, commit, sb);
1818}