pretty.con commit determine_author_info(): reuse parsing functions (f0f9662)
   1#include "cache.h"
   2#include "commit.h"
   3#include "utf8.h"
   4#include "diff.h"
   5#include "revision.h"
   6#include "string-list.h"
   7#include "mailmap.h"
   8#include "log-tree.h"
   9#include "notes.h"
  10#include "color.h"
  11#include "reflog-walk.h"
  12#include "gpg-interface.h"
  13
  14static char *user_format;
  15static struct cmt_fmt_map {
  16        const char *name;
  17        enum cmit_fmt format;
  18        int is_tformat;
  19        int is_alias;
  20        const char *user_format;
  21} *commit_formats;
  22static size_t builtin_formats_len;
  23static size_t commit_formats_len;
  24static size_t commit_formats_alloc;
  25static struct cmt_fmt_map *find_commit_format(const char *sought);
  26
  27static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  28{
  29        free(user_format);
  30        user_format = xstrdup(cp);
  31        if (is_tformat)
  32                rev->use_terminator = 1;
  33        rev->commit_format = CMIT_FMT_USERFORMAT;
  34}
  35
  36static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  37{
  38        struct cmt_fmt_map *commit_format = NULL;
  39        const char *name;
  40        const char *fmt;
  41        int i;
  42
  43        if (!skip_prefix(var, "pretty.", &name))
  44                return 0;
  45
  46        for (i = 0; i < builtin_formats_len; i++) {
  47                if (!strcmp(commit_formats[i].name, name))
  48                        return 0;
  49        }
  50
  51        for (i = builtin_formats_len; i < commit_formats_len; i++) {
  52                if (!strcmp(commit_formats[i].name, name)) {
  53                        commit_format = &commit_formats[i];
  54                        break;
  55                }
  56        }
  57
  58        if (!commit_format) {
  59                ALLOC_GROW(commit_formats, commit_formats_len+1,
  60                           commit_formats_alloc);
  61                commit_format = &commit_formats[commit_formats_len];
  62                memset(commit_format, 0, sizeof(*commit_format));
  63                commit_formats_len++;
  64        }
  65
  66        commit_format->name = xstrdup(name);
  67        commit_format->format = CMIT_FMT_USERFORMAT;
  68        git_config_string(&fmt, var, value);
  69        if (starts_with(fmt, "format:") || starts_with(fmt, "tformat:")) {
  70                commit_format->is_tformat = fmt[0] == 't';
  71                fmt = strchr(fmt, ':') + 1;
  72        } else if (strchr(fmt, '%'))
  73                commit_format->is_tformat = 1;
  74        else
  75                commit_format->is_alias = 1;
  76        commit_format->user_format = fmt;
  77
  78        return 0;
  79}
  80
  81static void setup_commit_formats(void)
  82{
  83        struct cmt_fmt_map builtin_formats[] = {
  84                { "raw",        CMIT_FMT_RAW,           0 },
  85                { "medium",     CMIT_FMT_MEDIUM,        0 },
  86                { "short",      CMIT_FMT_SHORT,         0 },
  87                { "email",      CMIT_FMT_EMAIL,         0 },
  88                { "fuller",     CMIT_FMT_FULLER,        0 },
  89                { "full",       CMIT_FMT_FULL,          0 },
  90                { "oneline",    CMIT_FMT_ONELINE,       1 }
  91        };
  92        commit_formats_len = ARRAY_SIZE(builtin_formats);
  93        builtin_formats_len = commit_formats_len;
  94        ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
  95        memcpy(commit_formats, builtin_formats,
  96               sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  97
  98        git_config(git_pretty_formats_config, NULL);
  99}
 100
 101static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 102                                                        const char *original,
 103                                                        int num_redirections)
 104{
 105        struct cmt_fmt_map *found = NULL;
 106        size_t found_match_len = 0;
 107        int i;
 108
 109        if (num_redirections >= commit_formats_len)
 110                die("invalid --pretty format: "
 111                    "'%s' references an alias which points to itself",
 112                    original);
 113
 114        for (i = 0; i < commit_formats_len; i++) {
 115                size_t match_len;
 116
 117                if (!starts_with(commit_formats[i].name, sought))
 118                        continue;
 119
 120                match_len = strlen(commit_formats[i].name);
 121                if (found == NULL || found_match_len > match_len) {
 122                        found = &commit_formats[i];
 123                        found_match_len = match_len;
 124                }
 125        }
 126
 127        if (found && found->is_alias) {
 128                found = find_commit_format_recursive(found->user_format,
 129                                                     original,
 130                                                     num_redirections+1);
 131        }
 132
 133        return found;
 134}
 135
 136static struct cmt_fmt_map *find_commit_format(const char *sought)
 137{
 138        if (!commit_formats)
 139                setup_commit_formats();
 140
 141        return find_commit_format_recursive(sought, sought, 0);
 142}
 143
 144void get_commit_format(const char *arg, struct rev_info *rev)
 145{
 146        struct cmt_fmt_map *commit_format;
 147
 148        rev->use_terminator = 0;
 149        if (!arg || !*arg) {
 150                rev->commit_format = CMIT_FMT_DEFAULT;
 151                return;
 152        }
 153        if (starts_with(arg, "format:") || starts_with(arg, "tformat:")) {
 154                save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 155                return;
 156        }
 157
 158        if (strchr(arg, '%')) {
 159                save_user_format(rev, arg, 1);
 160                return;
 161        }
 162
 163        commit_format = find_commit_format(arg);
 164        if (!commit_format)
 165                die("invalid --pretty format: %s", arg);
 166
 167        rev->commit_format = commit_format->format;
 168        rev->use_terminator = commit_format->is_tformat;
 169        if (commit_format->format == CMIT_FMT_USERFORMAT) {
 170                save_user_format(rev, commit_format->user_format,
 171                                 commit_format->is_tformat);
 172        }
 173}
 174
 175/*
 176 * Generic support for pretty-printing the header
 177 */
 178static int get_one_line(const char *msg)
 179{
 180        int ret = 0;
 181
 182        for (;;) {
 183                char c = *msg++;
 184                if (!c)
 185                        break;
 186                ret++;
 187                if (c == '\n')
 188                        break;
 189        }
 190        return ret;
 191}
 192
 193/* High bit set, or ISO-2022-INT */
 194static int non_ascii(int ch)
 195{
 196        return !isascii(ch) || ch == '\033';
 197}
 198
 199int has_non_ascii(const char *s)
 200{
 201        int ch;
 202        if (!s)
 203                return 0;
 204        while ((ch = *s++) != '\0') {
 205                if (non_ascii(ch))
 206                        return 1;
 207        }
 208        return 0;
 209}
 210
 211static int is_rfc822_special(char ch)
 212{
 213        switch (ch) {
 214        case '(':
 215        case ')':
 216        case '<':
 217        case '>':
 218        case '[':
 219        case ']':
 220        case ':':
 221        case ';':
 222        case '@':
 223        case ',':
 224        case '.':
 225        case '"':
 226        case '\\':
 227                return 1;
 228        default:
 229                return 0;
 230        }
 231}
 232
 233static int needs_rfc822_quoting(const char *s, int len)
 234{
 235        int i;
 236        for (i = 0; i < len; i++)
 237                if (is_rfc822_special(s[i]))
 238                        return 1;
 239        return 0;
 240}
 241
 242static int last_line_length(struct strbuf *sb)
 243{
 244        int i;
 245
 246        /* How many bytes are already used on the last line? */
 247        for (i = sb->len - 1; i >= 0; i--)
 248                if (sb->buf[i] == '\n')
 249                        break;
 250        return sb->len - (i + 1);
 251}
 252
 253static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 254{
 255        int i;
 256
 257        /* just a guess, we may have to also backslash-quote */
 258        strbuf_grow(out, len + 2);
 259
 260        strbuf_addch(out, '"');
 261        for (i = 0; i < len; i++) {
 262                switch (s[i]) {
 263                case '"':
 264                case '\\':
 265                        strbuf_addch(out, '\\');
 266                        /* fall through */
 267                default:
 268                        strbuf_addch(out, s[i]);
 269                }
 270        }
 271        strbuf_addch(out, '"');
 272}
 273
 274enum rfc2047_type {
 275        RFC2047_SUBJECT,
 276        RFC2047_ADDRESS
 277};
 278
 279static int is_rfc2047_special(char ch, enum rfc2047_type type)
 280{
 281        /*
 282         * rfc2047, section 4.2:
 283         *
 284         *    8-bit values which correspond to printable ASCII characters other
 285         *    than "=", "?", and "_" (underscore), MAY be represented as those
 286         *    characters.  (But see section 5 for restrictions.)  In
 287         *    particular, SPACE and TAB MUST NOT be represented as themselves
 288         *    within encoded words.
 289         */
 290
 291        /*
 292         * rule out non-ASCII characters and non-printable characters (the
 293         * non-ASCII check should be redundant as isprint() is not localized
 294         * and only knows about ASCII, but be defensive about that)
 295         */
 296        if (non_ascii(ch) || !isprint(ch))
 297                return 1;
 298
 299        /*
 300         * rule out special printable characters (' ' should be the only
 301         * whitespace character considered printable, but be defensive and use
 302         * isspace())
 303         */
 304        if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 305                return 1;
 306
 307        /*
 308         * rfc2047, section 5.3:
 309         *
 310         *    As a replacement for a 'word' entity within a 'phrase', for example,
 311         *    one that precedes an address in a From, To, or Cc header.  The ABNF
 312         *    definition for 'phrase' from RFC 822 thus becomes:
 313         *
 314         *    phrase = 1*( encoded-word / word )
 315         *
 316         *    In this case the set of characters that may be used in a "Q"-encoded
 317         *    'encoded-word' is restricted to: <upper and lower case ASCII
 318         *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 319         *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 320         *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 321         *    'special' by 'linear-white-space'.
 322         */
 323
 324        if (type != RFC2047_ADDRESS)
 325                return 0;
 326
 327        /* '=' and '_' are special cases and have been checked above */
 328        return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 329}
 330
 331static int needs_rfc2047_encoding(const char *line, int len,
 332                                  enum rfc2047_type type)
 333{
 334        int i;
 335
 336        for (i = 0; i < len; i++) {
 337                int ch = line[i];
 338                if (non_ascii(ch) || ch == '\n')
 339                        return 1;
 340                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 341                        return 1;
 342        }
 343
 344        return 0;
 345}
 346
 347static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 348                       const char *encoding, enum rfc2047_type type)
 349{
 350        static const int max_encoded_length = 76; /* per rfc2047 */
 351        int i;
 352        int line_len = last_line_length(sb);
 353
 354        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 355        strbuf_addf(sb, "=?%s?q?", encoding);
 356        line_len += strlen(encoding) + 5; /* 5 for =??q? */
 357
 358        while (len) {
 359                /*
 360                 * RFC 2047, section 5 (3):
 361                 *
 362                 * Each 'encoded-word' MUST represent an integral number of
 363                 * characters.  A multi-octet character may not be split across
 364                 * adjacent 'encoded- word's.
 365                 */
 366                const unsigned char *p = (const unsigned char *)line;
 367                int chrlen = mbs_chrlen(&line, &len, encoding);
 368                int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 369
 370                /* "=%02X" * chrlen, or the byte itself */
 371                const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 372                int         encoded_len = is_special ? 3 * chrlen : 1;
 373
 374                /*
 375                 * According to RFC 2047, we could encode the special character
 376                 * ' ' (space) with '_' (underscore) for readability. But many
 377                 * programs do not understand this and just leave the
 378                 * underscore in place. Thus, we do nothing special here, which
 379                 * causes ' ' to be encoded as '=20', avoiding this problem.
 380                 */
 381
 382                if (line_len + encoded_len + 2 > max_encoded_length) {
 383                        /* It won't fit with trailing "?=" --- break the line */
 384                        strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 385                        line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 386                }
 387
 388                for (i = 0; i < chrlen; i++)
 389                        strbuf_addf(sb, encoded_fmt, p[i]);
 390                line_len += encoded_len;
 391        }
 392        strbuf_addstr(sb, "?=");
 393}
 394
 395const char *show_ident_date(const struct ident_split *ident,
 396                            enum date_mode mode)
 397{
 398        unsigned long date = 0;
 399        long tz = 0;
 400
 401        if (ident->date_begin && ident->date_end)
 402                date = strtoul(ident->date_begin, NULL, 10);
 403        if (date_overflows(date))
 404                date = 0;
 405        else {
 406                if (ident->tz_begin && ident->tz_end)
 407                        tz = strtol(ident->tz_begin, NULL, 10);
 408                if (tz >= INT_MAX || tz <= INT_MIN)
 409                        tz = 0;
 410        }
 411        return show_date(date, tz, mode);
 412}
 413
 414void pp_user_info(struct pretty_print_context *pp,
 415                  const char *what, struct strbuf *sb,
 416                  const char *line, const char *encoding)
 417{
 418        struct ident_split ident;
 419        char *line_end;
 420        const char *mailbuf, *namebuf;
 421        size_t namelen, maillen;
 422        int max_length = 78; /* per rfc2822 */
 423
 424        if (pp->fmt == CMIT_FMT_ONELINE)
 425                return;
 426
 427        line_end = strchrnul(line, '\n');
 428        if (split_ident_line(&ident, line, line_end - line))
 429                return;
 430
 431        mailbuf = ident.mail_begin;
 432        maillen = ident.mail_end - ident.mail_begin;
 433        namebuf = ident.name_begin;
 434        namelen = ident.name_end - ident.name_begin;
 435
 436        if (pp->mailmap)
 437                map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 438
 439        if (pp->fmt == CMIT_FMT_EMAIL) {
 440                if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 441                        struct strbuf buf = STRBUF_INIT;
 442
 443                        strbuf_addstr(&buf, "From: ");
 444                        strbuf_add(&buf, namebuf, namelen);
 445                        strbuf_addstr(&buf, " <");
 446                        strbuf_add(&buf, mailbuf, maillen);
 447                        strbuf_addstr(&buf, ">\n");
 448                        string_list_append(&pp->in_body_headers,
 449                                           strbuf_detach(&buf, NULL));
 450
 451                        mailbuf = pp->from_ident->mail_begin;
 452                        maillen = pp->from_ident->mail_end - mailbuf;
 453                        namebuf = pp->from_ident->name_begin;
 454                        namelen = pp->from_ident->name_end - namebuf;
 455                }
 456
 457                strbuf_addstr(sb, "From: ");
 458                if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 459                        add_rfc2047(sb, namebuf, namelen,
 460                                    encoding, RFC2047_ADDRESS);
 461                        max_length = 76; /* per rfc2047 */
 462                } else if (needs_rfc822_quoting(namebuf, namelen)) {
 463                        struct strbuf quoted = STRBUF_INIT;
 464                        add_rfc822_quoted(&quoted, namebuf, namelen);
 465                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 466                                                        -6, 1, max_length);
 467                        strbuf_release(&quoted);
 468                } else {
 469                        strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 470                                                 -6, 1, max_length);
 471                }
 472
 473                if (max_length <
 474                    last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 475                        strbuf_addch(sb, '\n');
 476                strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 477        } else {
 478                strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 479                            (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 480                            (int)namelen, namebuf, (int)maillen, mailbuf);
 481        }
 482
 483        switch (pp->fmt) {
 484        case CMIT_FMT_MEDIUM:
 485                strbuf_addf(sb, "Date:   %s\n",
 486                            show_ident_date(&ident, pp->date_mode));
 487                break;
 488        case CMIT_FMT_EMAIL:
 489                strbuf_addf(sb, "Date: %s\n",
 490                            show_ident_date(&ident, DATE_RFC2822));
 491                break;
 492        case CMIT_FMT_FULLER:
 493                strbuf_addf(sb, "%sDate: %s\n", what,
 494                            show_ident_date(&ident, pp->date_mode));
 495                break;
 496        default:
 497                /* notin' */
 498                break;
 499        }
 500}
 501
 502static int is_empty_line(const char *line, int *len_p)
 503{
 504        int len = *len_p;
 505        while (len && isspace(line[len - 1]))
 506                len--;
 507        *len_p = len;
 508        return !len;
 509}
 510
 511static const char *skip_empty_lines(const char *msg)
 512{
 513        for (;;) {
 514                int linelen = get_one_line(msg);
 515                int ll = linelen;
 516                if (!linelen)
 517                        break;
 518                if (!is_empty_line(msg, &ll))
 519                        break;
 520                msg += linelen;
 521        }
 522        return msg;
 523}
 524
 525static void add_merge_info(const struct pretty_print_context *pp,
 526                           struct strbuf *sb, const struct commit *commit)
 527{
 528        struct commit_list *parent = commit->parents;
 529
 530        if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 531            !parent || !parent->next)
 532                return;
 533
 534        strbuf_addstr(sb, "Merge:");
 535
 536        while (parent) {
 537                struct commit *p = parent->item;
 538                const char *hex = NULL;
 539                if (pp->abbrev)
 540                        hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
 541                if (!hex)
 542                        hex = sha1_to_hex(p->object.sha1);
 543                parent = parent->next;
 544
 545                strbuf_addf(sb, " %s", hex);
 546        }
 547        strbuf_addch(sb, '\n');
 548}
 549
 550static char *get_header(const char *msg, const char *key)
 551{
 552        size_t len;
 553        const char *v = find_commit_header(msg, key, &len);
 554        return v ? xmemdupz(v, len) : NULL;
 555}
 556
 557static char *replace_encoding_header(char *buf, const char *encoding)
 558{
 559        struct strbuf tmp = STRBUF_INIT;
 560        size_t start, len;
 561        char *cp = buf;
 562
 563        /* guess if there is an encoding header before a \n\n */
 564        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 565                cp = strchr(cp, '\n');
 566                if (!cp || *++cp == '\n')
 567                        return buf;
 568        }
 569        start = cp - buf;
 570        cp = strchr(cp, '\n');
 571        if (!cp)
 572                return buf; /* should not happen but be defensive */
 573        len = cp + 1 - (buf + start);
 574
 575        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 576        if (is_encoding_utf8(encoding)) {
 577                /* we have re-coded to UTF-8; drop the header */
 578                strbuf_remove(&tmp, start, len);
 579        } else {
 580                /* just replaces XXXX in 'encoding XXXX\n' */
 581                strbuf_splice(&tmp, start + strlen("encoding "),
 582                                          len - strlen("encoding \n"),
 583                                          encoding, strlen(encoding));
 584        }
 585        return strbuf_detach(&tmp, NULL);
 586}
 587
 588const char *logmsg_reencode(const struct commit *commit,
 589                            char **commit_encoding,
 590                            const char *output_encoding)
 591{
 592        static const char *utf8 = "UTF-8";
 593        const char *use_encoding;
 594        char *encoding;
 595        const char *msg = get_commit_buffer(commit, NULL);
 596        char *out;
 597
 598        if (!output_encoding || !*output_encoding) {
 599                if (commit_encoding)
 600                        *commit_encoding = get_header(msg, "encoding");
 601                return msg;
 602        }
 603        encoding = get_header(msg, "encoding");
 604        if (commit_encoding)
 605                *commit_encoding = encoding;
 606        use_encoding = encoding ? encoding : utf8;
 607        if (same_encoding(use_encoding, output_encoding)) {
 608                /*
 609                 * No encoding work to be done. If we have no encoding header
 610                 * at all, then there's nothing to do, and we can return the
 611                 * message verbatim (whether newly allocated or not).
 612                 */
 613                if (!encoding)
 614                        return msg;
 615
 616                /*
 617                 * Otherwise, we still want to munge the encoding header in the
 618                 * result, which will be done by modifying the buffer. If we
 619                 * are using a fresh copy, we can reuse it. But if we are using
 620                 * the cached copy from get_commit_buffer, we need to duplicate it
 621                 * to avoid munging the cached copy.
 622                 */
 623                if (msg == get_cached_commit_buffer(commit, NULL))
 624                        out = xstrdup(msg);
 625                else
 626                        out = (char *)msg;
 627        }
 628        else {
 629                /*
 630                 * There's actual encoding work to do. Do the reencoding, which
 631                 * still leaves the header to be replaced in the next step. At
 632                 * this point, we are done with msg. If we allocated a fresh
 633                 * copy, we can free it.
 634                 */
 635                out = reencode_string(msg, output_encoding, use_encoding);
 636                if (out)
 637                        unuse_commit_buffer(commit, msg);
 638        }
 639
 640        /*
 641         * This replacement actually consumes the buffer we hand it, so we do
 642         * not have to worry about freeing the old "out" here.
 643         */
 644        if (out)
 645                out = replace_encoding_header(out, output_encoding);
 646
 647        if (!commit_encoding)
 648                free(encoding);
 649        /*
 650         * If the re-encoding failed, out might be NULL here; in that
 651         * case we just return the commit message verbatim.
 652         */
 653        return out ? out : msg;
 654}
 655
 656static int mailmap_name(const char **email, size_t *email_len,
 657                        const char **name, size_t *name_len)
 658{
 659        static struct string_list *mail_map;
 660        if (!mail_map) {
 661                mail_map = xcalloc(1, sizeof(*mail_map));
 662                read_mailmap(mail_map, NULL);
 663        }
 664        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 665}
 666
 667static size_t format_person_part(struct strbuf *sb, char part,
 668                                 const char *msg, int len, enum date_mode dmode)
 669{
 670        /* currently all placeholders have same length */
 671        const int placeholder_len = 2;
 672        struct ident_split s;
 673        const char *name, *mail;
 674        size_t maillen, namelen;
 675
 676        if (split_ident_line(&s, msg, len) < 0)
 677                goto skip;
 678
 679        name = s.name_begin;
 680        namelen = s.name_end - s.name_begin;
 681        mail = s.mail_begin;
 682        maillen = s.mail_end - s.mail_begin;
 683
 684        if (part == 'N' || part == 'E') /* mailmap lookup */
 685                mailmap_name(&mail, &maillen, &name, &namelen);
 686        if (part == 'n' || part == 'N') {       /* name */
 687                strbuf_add(sb, name, namelen);
 688                return placeholder_len;
 689        }
 690        if (part == 'e' || part == 'E') {       /* email */
 691                strbuf_add(sb, mail, maillen);
 692                return placeholder_len;
 693        }
 694
 695        if (!s.date_begin)
 696                goto skip;
 697
 698        if (part == 't') {      /* date, UNIX timestamp */
 699                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 700                return placeholder_len;
 701        }
 702
 703        switch (part) {
 704        case 'd':       /* date */
 705                strbuf_addstr(sb, show_ident_date(&s, dmode));
 706                return placeholder_len;
 707        case 'D':       /* date, RFC2822 style */
 708                strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
 709                return placeholder_len;
 710        case 'r':       /* date, relative */
 711                strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
 712                return placeholder_len;
 713        case 'i':       /* date, ISO 8601 */
 714                strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
 715                return placeholder_len;
 716        }
 717
 718skip:
 719        /*
 720         * reading from either a bogus commit, or a reflog entry with
 721         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 722         * to compute a valid return value.
 723         */
 724        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 725            || part == 'D' || part == 'r' || part == 'i')
 726                return placeholder_len;
 727
 728        return 0; /* unknown placeholder */
 729}
 730
 731struct chunk {
 732        size_t off;
 733        size_t len;
 734};
 735
 736enum flush_type {
 737        no_flush,
 738        flush_right,
 739        flush_left,
 740        flush_left_and_steal,
 741        flush_both
 742};
 743
 744enum trunc_type {
 745        trunc_none,
 746        trunc_left,
 747        trunc_middle,
 748        trunc_right
 749};
 750
 751struct format_commit_context {
 752        const struct commit *commit;
 753        const struct pretty_print_context *pretty_ctx;
 754        unsigned commit_header_parsed:1;
 755        unsigned commit_message_parsed:1;
 756        struct signature_check signature_check;
 757        enum flush_type flush_type;
 758        enum trunc_type truncate;
 759        const char *message;
 760        char *commit_encoding;
 761        size_t width, indent1, indent2;
 762        int auto_color;
 763        int padding;
 764
 765        /* These offsets are relative to the start of the commit message. */
 766        struct chunk author;
 767        struct chunk committer;
 768        size_t message_off;
 769        size_t subject_off;
 770        size_t body_off;
 771
 772        /* The following ones are relative to the result struct strbuf. */
 773        struct chunk abbrev_commit_hash;
 774        struct chunk abbrev_tree_hash;
 775        struct chunk abbrev_parent_hashes;
 776        size_t wrap_start;
 777};
 778
 779static int add_again(struct strbuf *sb, struct chunk *chunk)
 780{
 781        if (chunk->len) {
 782                strbuf_adddup(sb, chunk->off, chunk->len);
 783                return 1;
 784        }
 785
 786        /*
 787         * We haven't seen this chunk before.  Our caller is surely
 788         * going to add it the hard way now.  Remember the most likely
 789         * start of the to-be-added chunk: the current end of the
 790         * struct strbuf.
 791         */
 792        chunk->off = sb->len;
 793        return 0;
 794}
 795
 796static void parse_commit_header(struct format_commit_context *context)
 797{
 798        const char *msg = context->message;
 799        int i;
 800
 801        for (i = 0; msg[i]; i++) {
 802                int eol;
 803                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 804                        ; /* do nothing */
 805
 806                if (i == eol) {
 807                        break;
 808                } else if (starts_with(msg + i, "author ")) {
 809                        context->author.off = i + 7;
 810                        context->author.len = eol - i - 7;
 811                } else if (starts_with(msg + i, "committer ")) {
 812                        context->committer.off = i + 10;
 813                        context->committer.len = eol - i - 10;
 814                }
 815                i = eol;
 816        }
 817        context->message_off = i;
 818        context->commit_header_parsed = 1;
 819}
 820
 821static int istitlechar(char c)
 822{
 823        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 824                (c >= '0' && c <= '9') || c == '.' || c == '_';
 825}
 826
 827static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 828{
 829        size_t trimlen;
 830        size_t start_len = sb->len;
 831        int space = 2;
 832
 833        for (; *msg && *msg != '\n'; msg++) {
 834                if (istitlechar(*msg)) {
 835                        if (space == 1)
 836                                strbuf_addch(sb, '-');
 837                        space = 0;
 838                        strbuf_addch(sb, *msg);
 839                        if (*msg == '.')
 840                                while (*(msg+1) == '.')
 841                                        msg++;
 842                } else
 843                        space |= 1;
 844        }
 845
 846        /* trim any trailing '.' or '-' characters */
 847        trimlen = 0;
 848        while (sb->len - trimlen > start_len &&
 849                (sb->buf[sb->len - 1 - trimlen] == '.'
 850                || sb->buf[sb->len - 1 - trimlen] == '-'))
 851                trimlen++;
 852        strbuf_remove(sb, sb->len - trimlen, trimlen);
 853}
 854
 855const char *format_subject(struct strbuf *sb, const char *msg,
 856                           const char *line_separator)
 857{
 858        int first = 1;
 859
 860        for (;;) {
 861                const char *line = msg;
 862                int linelen = get_one_line(line);
 863
 864                msg += linelen;
 865                if (!linelen || is_empty_line(line, &linelen))
 866                        break;
 867
 868                if (!sb)
 869                        continue;
 870                strbuf_grow(sb, linelen + 2);
 871                if (!first)
 872                        strbuf_addstr(sb, line_separator);
 873                strbuf_add(sb, line, linelen);
 874                first = 0;
 875        }
 876        return msg;
 877}
 878
 879static void parse_commit_message(struct format_commit_context *c)
 880{
 881        const char *msg = c->message + c->message_off;
 882        const char *start = c->message;
 883
 884        msg = skip_empty_lines(msg);
 885        c->subject_off = msg - start;
 886
 887        msg = format_subject(NULL, msg, NULL);
 888        msg = skip_empty_lines(msg);
 889        c->body_off = msg - start;
 890
 891        c->commit_message_parsed = 1;
 892}
 893
 894static void strbuf_wrap(struct strbuf *sb, size_t pos,
 895                        size_t width, size_t indent1, size_t indent2)
 896{
 897        struct strbuf tmp = STRBUF_INIT;
 898
 899        if (pos)
 900                strbuf_add(&tmp, sb->buf, pos);
 901        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 902                                (int) indent1, (int) indent2, (int) width);
 903        strbuf_swap(&tmp, sb);
 904        strbuf_release(&tmp);
 905}
 906
 907static void rewrap_message_tail(struct strbuf *sb,
 908                                struct format_commit_context *c,
 909                                size_t new_width, size_t new_indent1,
 910                                size_t new_indent2)
 911{
 912        if (c->width == new_width && c->indent1 == new_indent1 &&
 913            c->indent2 == new_indent2)
 914                return;
 915        if (c->wrap_start < sb->len)
 916                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 917        c->wrap_start = sb->len;
 918        c->width = new_width;
 919        c->indent1 = new_indent1;
 920        c->indent2 = new_indent2;
 921}
 922
 923static int format_reflog_person(struct strbuf *sb,
 924                                char part,
 925                                struct reflog_walk_info *log,
 926                                enum date_mode dmode)
 927{
 928        const char *ident;
 929
 930        if (!log)
 931                return 2;
 932
 933        ident = get_reflog_ident(log);
 934        if (!ident)
 935                return 2;
 936
 937        return format_person_part(sb, part, ident, strlen(ident), dmode);
 938}
 939
 940static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 941                          const char *placeholder,
 942                          struct format_commit_context *c)
 943{
 944        if (placeholder[1] == '(') {
 945                const char *begin = placeholder + 2;
 946                const char *end = strchr(begin, ')');
 947                char color[COLOR_MAXLEN];
 948
 949                if (!end)
 950                        return 0;
 951                if (starts_with(begin, "auto,")) {
 952                        if (!want_color(c->pretty_ctx->color))
 953                                return end - placeholder + 1;
 954                        begin += 5;
 955                }
 956                color_parse_mem(begin,
 957                                end - begin,
 958                                "--pretty format", color);
 959                strbuf_addstr(sb, color);
 960                return end - placeholder + 1;
 961        }
 962        if (starts_with(placeholder + 1, "red")) {
 963                strbuf_addstr(sb, GIT_COLOR_RED);
 964                return 4;
 965        } else if (starts_with(placeholder + 1, "green")) {
 966                strbuf_addstr(sb, GIT_COLOR_GREEN);
 967                return 6;
 968        } else if (starts_with(placeholder + 1, "blue")) {
 969                strbuf_addstr(sb, GIT_COLOR_BLUE);
 970                return 5;
 971        } else if (starts_with(placeholder + 1, "reset")) {
 972                strbuf_addstr(sb, GIT_COLOR_RESET);
 973                return 6;
 974        } else
 975                return 0;
 976}
 977
 978static size_t parse_padding_placeholder(struct strbuf *sb,
 979                                        const char *placeholder,
 980                                        struct format_commit_context *c)
 981{
 982        const char *ch = placeholder;
 983        enum flush_type flush_type;
 984        int to_column = 0;
 985
 986        switch (*ch++) {
 987        case '<':
 988                flush_type = flush_right;
 989                break;
 990        case '>':
 991                if (*ch == '<') {
 992                        flush_type = flush_both;
 993                        ch++;
 994                } else if (*ch == '>') {
 995                        flush_type = flush_left_and_steal;
 996                        ch++;
 997                } else
 998                        flush_type = flush_left;
 999                break;
1000        default:
1001                return 0;
1002        }
1003
1004        /* the next value means "wide enough to that column" */
1005        if (*ch == '|') {
1006                to_column = 1;
1007                ch++;
1008        }
1009
1010        if (*ch == '(') {
1011                const char *start = ch + 1;
1012                const char *end = start + strcspn(start, ",)");
1013                char *next;
1014                int width;
1015                if (!end || end == start)
1016                        return 0;
1017                width = strtoul(start, &next, 10);
1018                if (next == start || width == 0)
1019                        return 0;
1020                c->padding = to_column ? -width : width;
1021                c->flush_type = flush_type;
1022
1023                if (*end == ',') {
1024                        start = end + 1;
1025                        end = strchr(start, ')');
1026                        if (!end || end == start)
1027                                return 0;
1028                        if (starts_with(start, "trunc)"))
1029                                c->truncate = trunc_right;
1030                        else if (starts_with(start, "ltrunc)"))
1031                                c->truncate = trunc_left;
1032                        else if (starts_with(start, "mtrunc)"))
1033                                c->truncate = trunc_middle;
1034                        else
1035                                return 0;
1036                } else
1037                        c->truncate = trunc_none;
1038
1039                return end - placeholder + 1;
1040        }
1041        return 0;
1042}
1043
1044static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1045                                const char *placeholder,
1046                                void *context)
1047{
1048        struct format_commit_context *c = context;
1049        const struct commit *commit = c->commit;
1050        const char *msg = c->message;
1051        struct commit_list *p;
1052        int h1, h2;
1053
1054        /* these are independent of the commit */
1055        switch (placeholder[0]) {
1056        case 'C':
1057                if (starts_with(placeholder + 1, "(auto)")) {
1058                        c->auto_color = 1;
1059                        return 7; /* consumed 7 bytes, "C(auto)" */
1060                } else {
1061                        int ret = parse_color(sb, placeholder, c);
1062                        if (ret)
1063                                c->auto_color = 0;
1064                        /*
1065                         * Otherwise, we decided to treat %C<unknown>
1066                         * as a literal string, and the previous
1067                         * %C(auto) is still valid.
1068                         */
1069                        return ret;
1070                }
1071        case 'n':               /* newline */
1072                strbuf_addch(sb, '\n');
1073                return 1;
1074        case 'x':
1075                /* %x00 == NUL, %x0a == LF, etc. */
1076                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1077                    h1 <= 16 &&
1078                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1079                    h2 <= 16) {
1080                        strbuf_addch(sb, (h1<<4)|h2);
1081                        return 3;
1082                } else
1083                        return 0;
1084        case 'w':
1085                if (placeholder[1] == '(') {
1086                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1087                        char *next;
1088                        const char *start = placeholder + 2;
1089                        const char *end = strchr(start, ')');
1090                        if (!end)
1091                                return 0;
1092                        if (end > start) {
1093                                width = strtoul(start, &next, 10);
1094                                if (*next == ',') {
1095                                        indent1 = strtoul(next + 1, &next, 10);
1096                                        if (*next == ',') {
1097                                                indent2 = strtoul(next + 1,
1098                                                                 &next, 10);
1099                                        }
1100                                }
1101                                if (*next != ')')
1102                                        return 0;
1103                        }
1104                        rewrap_message_tail(sb, c, width, indent1, indent2);
1105                        return end - placeholder + 1;
1106                } else
1107                        return 0;
1108
1109        case '<':
1110        case '>':
1111                return parse_padding_placeholder(sb, placeholder, c);
1112        }
1113
1114        /* these depend on the commit */
1115        if (!commit->object.parsed)
1116                parse_object(commit->object.sha1);
1117
1118        switch (placeholder[0]) {
1119        case 'H':               /* commit hash */
1120                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1121                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1122                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1123                return 1;
1124        case 'h':               /* abbreviated commit hash */
1125                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1126                if (add_again(sb, &c->abbrev_commit_hash)) {
1127                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1128                        return 1;
1129                }
1130                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1131                                                     c->pretty_ctx->abbrev));
1132                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1133                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1134                return 1;
1135        case 'T':               /* tree hash */
1136                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1137                return 1;
1138        case 't':               /* abbreviated tree hash */
1139                if (add_again(sb, &c->abbrev_tree_hash))
1140                        return 1;
1141                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1142                                                     c->pretty_ctx->abbrev));
1143                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1144                return 1;
1145        case 'P':               /* parent hashes */
1146                for (p = commit->parents; p; p = p->next) {
1147                        if (p != commit->parents)
1148                                strbuf_addch(sb, ' ');
1149                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1150                }
1151                return 1;
1152        case 'p':               /* abbreviated parent hashes */
1153                if (add_again(sb, &c->abbrev_parent_hashes))
1154                        return 1;
1155                for (p = commit->parents; p; p = p->next) {
1156                        if (p != commit->parents)
1157                                strbuf_addch(sb, ' ');
1158                        strbuf_addstr(sb, find_unique_abbrev(
1159                                        p->item->object.sha1,
1160                                        c->pretty_ctx->abbrev));
1161                }
1162                c->abbrev_parent_hashes.len = sb->len -
1163                                              c->abbrev_parent_hashes.off;
1164                return 1;
1165        case 'm':               /* left/right/bottom */
1166                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1167                return 1;
1168        case 'd':
1169                load_ref_decorations(DECORATE_SHORT_REFS);
1170                format_decorations(sb, commit, c->auto_color);
1171                return 1;
1172        case 'g':               /* reflog info */
1173                switch(placeholder[1]) {
1174                case 'd':       /* reflog selector */
1175                case 'D':
1176                        if (c->pretty_ctx->reflog_info)
1177                                get_reflog_selector(sb,
1178                                                    c->pretty_ctx->reflog_info,
1179                                                    c->pretty_ctx->date_mode,
1180                                                    c->pretty_ctx->date_mode_explicit,
1181                                                    (placeholder[1] == 'd'));
1182                        return 2;
1183                case 's':       /* reflog message */
1184                        if (c->pretty_ctx->reflog_info)
1185                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1186                        return 2;
1187                case 'n':
1188                case 'N':
1189                case 'e':
1190                case 'E':
1191                        return format_reflog_person(sb,
1192                                                    placeholder[1],
1193                                                    c->pretty_ctx->reflog_info,
1194                                                    c->pretty_ctx->date_mode);
1195                }
1196                return 0;       /* unknown %g placeholder */
1197        case 'N':
1198                if (c->pretty_ctx->notes_message) {
1199                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1200                        return 1;
1201                }
1202                return 0;
1203        }
1204
1205        if (placeholder[0] == 'G') {
1206                if (!c->signature_check.result)
1207                        check_commit_signature(c->commit, &(c->signature_check));
1208                switch (placeholder[1]) {
1209                case 'G':
1210                        if (c->signature_check.gpg_output)
1211                                strbuf_addstr(sb, c->signature_check.gpg_output);
1212                        break;
1213                case '?':
1214                        switch (c->signature_check.result) {
1215                        case 'G':
1216                        case 'B':
1217                        case 'U':
1218                        case 'N':
1219                                strbuf_addch(sb, c->signature_check.result);
1220                        }
1221                        break;
1222                case 'S':
1223                        if (c->signature_check.signer)
1224                                strbuf_addstr(sb, c->signature_check.signer);
1225                        break;
1226                case 'K':
1227                        if (c->signature_check.key)
1228                                strbuf_addstr(sb, c->signature_check.key);
1229                        break;
1230                default:
1231                        return 0;
1232                }
1233                return 2;
1234        }
1235
1236
1237        /* For the rest we have to parse the commit header. */
1238        if (!c->commit_header_parsed)
1239                parse_commit_header(c);
1240
1241        switch (placeholder[0]) {
1242        case 'a':       /* author ... */
1243                return format_person_part(sb, placeholder[1],
1244                                   msg + c->author.off, c->author.len,
1245                                   c->pretty_ctx->date_mode);
1246        case 'c':       /* committer ... */
1247                return format_person_part(sb, placeholder[1],
1248                                   msg + c->committer.off, c->committer.len,
1249                                   c->pretty_ctx->date_mode);
1250        case 'e':       /* encoding */
1251                if (c->commit_encoding)
1252                        strbuf_addstr(sb, c->commit_encoding);
1253                return 1;
1254        case 'B':       /* raw body */
1255                /* message_off is always left at the initial newline */
1256                strbuf_addstr(sb, msg + c->message_off + 1);
1257                return 1;
1258        }
1259
1260        /* Now we need to parse the commit message. */
1261        if (!c->commit_message_parsed)
1262                parse_commit_message(c);
1263
1264        switch (placeholder[0]) {
1265        case 's':       /* subject */
1266                format_subject(sb, msg + c->subject_off, " ");
1267                return 1;
1268        case 'f':       /* sanitized subject */
1269                format_sanitized_subject(sb, msg + c->subject_off);
1270                return 1;
1271        case 'b':       /* body */
1272                strbuf_addstr(sb, msg + c->body_off);
1273                return 1;
1274        }
1275        return 0;       /* unknown placeholder */
1276}
1277
1278static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1279                                    const char *placeholder,
1280                                    struct format_commit_context *c)
1281{
1282        struct strbuf local_sb = STRBUF_INIT;
1283        int total_consumed = 0, len, padding = c->padding;
1284        if (padding < 0) {
1285                const char *start = strrchr(sb->buf, '\n');
1286                int occupied;
1287                if (!start)
1288                        start = sb->buf;
1289                occupied = utf8_strnwidth(start, -1, 1);
1290                padding = (-padding) - occupied;
1291        }
1292        while (1) {
1293                int modifier = *placeholder == 'C';
1294                int consumed = format_commit_one(&local_sb, placeholder, c);
1295                total_consumed += consumed;
1296
1297                if (!modifier)
1298                        break;
1299
1300                placeholder += consumed;
1301                if (*placeholder != '%')
1302                        break;
1303                placeholder++;
1304                total_consumed++;
1305        }
1306        len = utf8_strnwidth(local_sb.buf, -1, 1);
1307
1308        if (c->flush_type == flush_left_and_steal) {
1309                const char *ch = sb->buf + sb->len - 1;
1310                while (len > padding && ch > sb->buf) {
1311                        const char *p;
1312                        if (*ch == ' ') {
1313                                ch--;
1314                                padding++;
1315                                continue;
1316                        }
1317                        /* check for trailing ansi sequences */
1318                        if (*ch != 'm')
1319                                break;
1320                        p = ch - 1;
1321                        while (ch - p < 10 && *p != '\033')
1322                                p--;
1323                        if (*p != '\033' ||
1324                            ch + 1 - p != display_mode_esc_sequence_len(p))
1325                                break;
1326                        /*
1327                         * got a good ansi sequence, put it back to
1328                         * local_sb as we're cutting sb
1329                         */
1330                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1331                        ch = p - 1;
1332                }
1333                strbuf_setlen(sb, ch + 1 - sb->buf);
1334                c->flush_type = flush_left;
1335        }
1336
1337        if (len > padding) {
1338                switch (c->truncate) {
1339                case trunc_left:
1340                        strbuf_utf8_replace(&local_sb,
1341                                            0, len - (padding - 2),
1342                                            "..");
1343                        break;
1344                case trunc_middle:
1345                        strbuf_utf8_replace(&local_sb,
1346                                            padding / 2 - 1,
1347                                            len - (padding - 2),
1348                                            "..");
1349                        break;
1350                case trunc_right:
1351                        strbuf_utf8_replace(&local_sb,
1352                                            padding - 2, len - (padding - 2),
1353                                            "..");
1354                        break;
1355                case trunc_none:
1356                        break;
1357                }
1358                strbuf_addbuf(sb, &local_sb);
1359        } else {
1360                int sb_len = sb->len, offset = 0;
1361                if (c->flush_type == flush_left)
1362                        offset = padding - len;
1363                else if (c->flush_type == flush_both)
1364                        offset = (padding - len) / 2;
1365                /*
1366                 * we calculate padding in columns, now
1367                 * convert it back to chars
1368                 */
1369                padding = padding - len + local_sb.len;
1370                strbuf_grow(sb, padding);
1371                strbuf_setlen(sb, sb_len + padding);
1372                memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1373                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1374                       local_sb.len);
1375        }
1376        strbuf_release(&local_sb);
1377        c->flush_type = no_flush;
1378        return total_consumed;
1379}
1380
1381static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1382                                 const char *placeholder,
1383                                 void *context)
1384{
1385        int consumed;
1386        size_t orig_len;
1387        enum {
1388                NO_MAGIC,
1389                ADD_LF_BEFORE_NON_EMPTY,
1390                DEL_LF_BEFORE_EMPTY,
1391                ADD_SP_BEFORE_NON_EMPTY
1392        } magic = NO_MAGIC;
1393
1394        switch (placeholder[0]) {
1395        case '-':
1396                magic = DEL_LF_BEFORE_EMPTY;
1397                break;
1398        case '+':
1399                magic = ADD_LF_BEFORE_NON_EMPTY;
1400                break;
1401        case ' ':
1402                magic = ADD_SP_BEFORE_NON_EMPTY;
1403                break;
1404        default:
1405                break;
1406        }
1407        if (magic != NO_MAGIC)
1408                placeholder++;
1409
1410        orig_len = sb->len;
1411        if (((struct format_commit_context *)context)->flush_type != no_flush)
1412                consumed = format_and_pad_commit(sb, placeholder, context);
1413        else
1414                consumed = format_commit_one(sb, placeholder, context);
1415        if (magic == NO_MAGIC)
1416                return consumed;
1417
1418        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1419                while (sb->len && sb->buf[sb->len - 1] == '\n')
1420                        strbuf_setlen(sb, sb->len - 1);
1421        } else if (orig_len != sb->len) {
1422                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1423                        strbuf_insert(sb, orig_len, "\n", 1);
1424                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1425                        strbuf_insert(sb, orig_len, " ", 1);
1426        }
1427        return consumed + 1;
1428}
1429
1430static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1431                                   void *context)
1432{
1433        struct userformat_want *w = context;
1434
1435        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1436                placeholder++;
1437
1438        switch (*placeholder) {
1439        case 'N':
1440                w->notes = 1;
1441                break;
1442        }
1443        return 0;
1444}
1445
1446void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1447{
1448        struct strbuf dummy = STRBUF_INIT;
1449
1450        if (!fmt) {
1451                if (!user_format)
1452                        return;
1453                fmt = user_format;
1454        }
1455        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1456        strbuf_release(&dummy);
1457}
1458
1459void format_commit_message(const struct commit *commit,
1460                           const char *format, struct strbuf *sb,
1461                           const struct pretty_print_context *pretty_ctx)
1462{
1463        struct format_commit_context context;
1464        const char *output_enc = pretty_ctx->output_encoding;
1465        const char *utf8 = "UTF-8";
1466
1467        memset(&context, 0, sizeof(context));
1468        context.commit = commit;
1469        context.pretty_ctx = pretty_ctx;
1470        context.wrap_start = sb->len;
1471        /*
1472         * convert a commit message to UTF-8 first
1473         * as far as 'format_commit_item' assumes it in UTF-8
1474         */
1475        context.message = logmsg_reencode(commit,
1476                                          &context.commit_encoding,
1477                                          utf8);
1478
1479        strbuf_expand(sb, format, format_commit_item, &context);
1480        rewrap_message_tail(sb, &context, 0, 0, 0);
1481
1482        /* then convert a commit message to an actual output encoding */
1483        if (output_enc) {
1484                if (same_encoding(utf8, output_enc))
1485                        output_enc = NULL;
1486        } else {
1487                if (context.commit_encoding &&
1488                    !same_encoding(context.commit_encoding, utf8))
1489                        output_enc = context.commit_encoding;
1490        }
1491
1492        if (output_enc) {
1493                int outsz;
1494                char *out = reencode_string_len(sb->buf, sb->len,
1495                                                output_enc, utf8, &outsz);
1496                if (out)
1497                        strbuf_attach(sb, out, outsz, outsz + 1);
1498        }
1499
1500        free(context.commit_encoding);
1501        unuse_commit_buffer(commit, context.message);
1502}
1503
1504static void pp_header(struct pretty_print_context *pp,
1505                      const char *encoding,
1506                      const struct commit *commit,
1507                      const char **msg_p,
1508                      struct strbuf *sb)
1509{
1510        int parents_shown = 0;
1511
1512        for (;;) {
1513                const char *line = *msg_p;
1514                int linelen = get_one_line(*msg_p);
1515
1516                if (!linelen)
1517                        return;
1518                *msg_p += linelen;
1519
1520                if (linelen == 1)
1521                        /* End of header */
1522                        return;
1523
1524                if (pp->fmt == CMIT_FMT_RAW) {
1525                        strbuf_add(sb, line, linelen);
1526                        continue;
1527                }
1528
1529                if (starts_with(line, "parent ")) {
1530                        if (linelen != 48)
1531                                die("bad parent line in commit");
1532                        continue;
1533                }
1534
1535                if (!parents_shown) {
1536                        unsigned num = commit_list_count(commit->parents);
1537                        /* with enough slop */
1538                        strbuf_grow(sb, num * 50 + 20);
1539                        add_merge_info(pp, sb, commit);
1540                        parents_shown = 1;
1541                }
1542
1543                /*
1544                 * MEDIUM == DEFAULT shows only author with dates.
1545                 * FULL shows both authors but not dates.
1546                 * FULLER shows both authors and dates.
1547                 */
1548                if (starts_with(line, "author ")) {
1549                        strbuf_grow(sb, linelen + 80);
1550                        pp_user_info(pp, "Author", sb, line + 7, encoding);
1551                }
1552                if (starts_with(line, "committer ") &&
1553                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1554                        strbuf_grow(sb, linelen + 80);
1555                        pp_user_info(pp, "Commit", sb, line + 10, encoding);
1556                }
1557        }
1558}
1559
1560void pp_title_line(struct pretty_print_context *pp,
1561                   const char **msg_p,
1562                   struct strbuf *sb,
1563                   const char *encoding,
1564                   int need_8bit_cte)
1565{
1566        static const int max_length = 78; /* per rfc2047 */
1567        struct strbuf title;
1568
1569        strbuf_init(&title, 80);
1570        *msg_p = format_subject(&title, *msg_p,
1571                                pp->preserve_subject ? "\n" : " ");
1572
1573        strbuf_grow(sb, title.len + 1024);
1574        if (pp->subject) {
1575                strbuf_addstr(sb, pp->subject);
1576                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1577                        add_rfc2047(sb, title.buf, title.len,
1578                                                encoding, RFC2047_SUBJECT);
1579                else
1580                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1581                                         -last_line_length(sb), 1, max_length);
1582        } else {
1583                strbuf_addbuf(sb, &title);
1584        }
1585        strbuf_addch(sb, '\n');
1586
1587        if (need_8bit_cte == 0) {
1588                int i;
1589                for (i = 0; i < pp->in_body_headers.nr; i++) {
1590                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1591                                need_8bit_cte = 1;
1592                                break;
1593                        }
1594                }
1595        }
1596
1597        if (need_8bit_cte > 0) {
1598                const char *header_fmt =
1599                        "MIME-Version: 1.0\n"
1600                        "Content-Type: text/plain; charset=%s\n"
1601                        "Content-Transfer-Encoding: 8bit\n";
1602                strbuf_addf(sb, header_fmt, encoding);
1603        }
1604        if (pp->after_subject) {
1605                strbuf_addstr(sb, pp->after_subject);
1606        }
1607        if (pp->fmt == CMIT_FMT_EMAIL) {
1608                strbuf_addch(sb, '\n');
1609        }
1610
1611        if (pp->in_body_headers.nr) {
1612                int i;
1613                for (i = 0; i < pp->in_body_headers.nr; i++) {
1614                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1615                        free(pp->in_body_headers.items[i].string);
1616                }
1617                string_list_clear(&pp->in_body_headers, 0);
1618                strbuf_addch(sb, '\n');
1619        }
1620
1621        strbuf_release(&title);
1622}
1623
1624void pp_remainder(struct pretty_print_context *pp,
1625                  const char **msg_p,
1626                  struct strbuf *sb,
1627                  int indent)
1628{
1629        int first = 1;
1630        for (;;) {
1631                const char *line = *msg_p;
1632                int linelen = get_one_line(line);
1633                *msg_p += linelen;
1634
1635                if (!linelen)
1636                        break;
1637
1638                if (is_empty_line(line, &linelen)) {
1639                        if (first)
1640                                continue;
1641                        if (pp->fmt == CMIT_FMT_SHORT)
1642                                break;
1643                }
1644                first = 0;
1645
1646                strbuf_grow(sb, linelen + indent + 20);
1647                if (indent) {
1648                        memset(sb->buf + sb->len, ' ', indent);
1649                        strbuf_setlen(sb, sb->len + indent);
1650                }
1651                strbuf_add(sb, line, linelen);
1652                strbuf_addch(sb, '\n');
1653        }
1654}
1655
1656void pretty_print_commit(struct pretty_print_context *pp,
1657                         const struct commit *commit,
1658                         struct strbuf *sb)
1659{
1660        unsigned long beginning_of_body;
1661        int indent = 4;
1662        const char *msg;
1663        const char *reencoded;
1664        const char *encoding;
1665        int need_8bit_cte = pp->need_8bit_cte;
1666
1667        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1668                format_commit_message(commit, user_format, sb, pp);
1669                return;
1670        }
1671
1672        encoding = get_log_output_encoding();
1673        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1674
1675        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1676                indent = 0;
1677
1678        /*
1679         * We need to check and emit Content-type: to mark it
1680         * as 8-bit if we haven't done so.
1681         */
1682        if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1683                int i, ch, in_body;
1684
1685                for (in_body = i = 0; (ch = msg[i]); i++) {
1686                        if (!in_body) {
1687                                /* author could be non 7-bit ASCII but
1688                                 * the log may be so; skip over the
1689                                 * header part first.
1690                                 */
1691                                if (ch == '\n' && msg[i+1] == '\n')
1692                                        in_body = 1;
1693                        }
1694                        else if (non_ascii(ch)) {
1695                                need_8bit_cte = 1;
1696                                break;
1697                        }
1698                }
1699        }
1700
1701        pp_header(pp, encoding, commit, &msg, sb);
1702        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1703                strbuf_addch(sb, '\n');
1704        }
1705
1706        /* Skip excess blank lines at the beginning of body, if any... */
1707        msg = skip_empty_lines(msg);
1708
1709        /* These formats treat the title line specially. */
1710        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1711                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1712
1713        beginning_of_body = sb->len;
1714        if (pp->fmt != CMIT_FMT_ONELINE)
1715                pp_remainder(pp, &msg, sb, indent);
1716        strbuf_rtrim(sb);
1717
1718        /* Make sure there is an EOLN for the non-oneline case */
1719        if (pp->fmt != CMIT_FMT_ONELINE)
1720                strbuf_addch(sb, '\n');
1721
1722        /*
1723         * The caller may append additional body text in e-mail
1724         * format.  Make sure we did not strip the blank line
1725         * between the header and the body.
1726         */
1727        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1728                strbuf_addch(sb, '\n');
1729
1730        unuse_commit_buffer(commit, reencoded);
1731}
1732
1733void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1734                    struct strbuf *sb)
1735{
1736        struct pretty_print_context pp = {0};
1737        pp.fmt = fmt;
1738        pretty_print_commit(&pp, commit, sb);
1739}