pretty.con commit imap-send: use skip_prefix instead of using magic numbers (59a642f)
   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 struct commit *commit, const char *msg,
 551                        const char *key)
 552{
 553        int key_len = strlen(key);
 554        const char *line = msg;
 555
 556        while (line) {
 557                const char *eol = strchrnul(line, '\n'), *next;
 558
 559                if (line == eol)
 560                        return NULL;
 561                if (!*eol) {
 562                        warning("malformed commit (header is missing newline): %s",
 563                                sha1_to_hex(commit->object.sha1));
 564                        next = NULL;
 565                } else
 566                        next = eol + 1;
 567                if (eol - line > key_len &&
 568                    !strncmp(line, key, key_len) &&
 569                    line[key_len] == ' ') {
 570                        return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 571                }
 572                line = next;
 573        }
 574        return NULL;
 575}
 576
 577static char *replace_encoding_header(char *buf, const char *encoding)
 578{
 579        struct strbuf tmp = STRBUF_INIT;
 580        size_t start, len;
 581        char *cp = buf;
 582
 583        /* guess if there is an encoding header before a \n\n */
 584        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 585                cp = strchr(cp, '\n');
 586                if (!cp || *++cp == '\n')
 587                        return buf;
 588        }
 589        start = cp - buf;
 590        cp = strchr(cp, '\n');
 591        if (!cp)
 592                return buf; /* should not happen but be defensive */
 593        len = cp + 1 - (buf + start);
 594
 595        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 596        if (is_encoding_utf8(encoding)) {
 597                /* we have re-coded to UTF-8; drop the header */
 598                strbuf_remove(&tmp, start, len);
 599        } else {
 600                /* just replaces XXXX in 'encoding XXXX\n' */
 601                strbuf_splice(&tmp, start + strlen("encoding "),
 602                                          len - strlen("encoding \n"),
 603                                          encoding, strlen(encoding));
 604        }
 605        return strbuf_detach(&tmp, NULL);
 606}
 607
 608char *logmsg_reencode(const struct commit *commit,
 609                      char **commit_encoding,
 610                      const char *output_encoding)
 611{
 612        static const char *utf8 = "UTF-8";
 613        const char *use_encoding;
 614        char *encoding;
 615        char *msg = commit->buffer;
 616        char *out;
 617
 618        if (!msg) {
 619                enum object_type type;
 620                unsigned long size;
 621
 622                msg = read_sha1_file(commit->object.sha1, &type, &size);
 623                if (!msg)
 624                        die("Cannot read commit object %s",
 625                            sha1_to_hex(commit->object.sha1));
 626                if (type != OBJ_COMMIT)
 627                        die("Expected commit for '%s', got %s",
 628                            sha1_to_hex(commit->object.sha1), typename(type));
 629        }
 630
 631        if (!output_encoding || !*output_encoding) {
 632                if (commit_encoding)
 633                        *commit_encoding =
 634                                get_header(commit, msg, "encoding");
 635                return msg;
 636        }
 637        encoding = get_header(commit, msg, "encoding");
 638        if (commit_encoding)
 639                *commit_encoding = encoding;
 640        use_encoding = encoding ? encoding : utf8;
 641        if (same_encoding(use_encoding, output_encoding)) {
 642                /*
 643                 * No encoding work to be done. If we have no encoding header
 644                 * at all, then there's nothing to do, and we can return the
 645                 * message verbatim (whether newly allocated or not).
 646                 */
 647                if (!encoding)
 648                        return msg;
 649
 650                /*
 651                 * Otherwise, we still want to munge the encoding header in the
 652                 * result, which will be done by modifying the buffer. If we
 653                 * are using a fresh copy, we can reuse it. But if we are using
 654                 * the cached copy from commit->buffer, we need to duplicate it
 655                 * to avoid munging commit->buffer.
 656                 */
 657                out = msg;
 658                if (out == commit->buffer)
 659                        out = xstrdup(out);
 660        }
 661        else {
 662                /*
 663                 * There's actual encoding work to do. Do the reencoding, which
 664                 * still leaves the header to be replaced in the next step. At
 665                 * this point, we are done with msg. If we allocated a fresh
 666                 * copy, we can free it.
 667                 */
 668                out = reencode_string(msg, output_encoding, use_encoding);
 669                if (out && msg != commit->buffer)
 670                        free(msg);
 671        }
 672
 673        /*
 674         * This replacement actually consumes the buffer we hand it, so we do
 675         * not have to worry about freeing the old "out" here.
 676         */
 677        if (out)
 678                out = replace_encoding_header(out, output_encoding);
 679
 680        if (!commit_encoding)
 681                free(encoding);
 682        /*
 683         * If the re-encoding failed, out might be NULL here; in that
 684         * case we just return the commit message verbatim.
 685         */
 686        return out ? out : msg;
 687}
 688
 689void logmsg_free(char *msg, const struct commit *commit)
 690{
 691        if (msg != commit->buffer)
 692                free(msg);
 693}
 694
 695static int mailmap_name(const char **email, size_t *email_len,
 696                        const char **name, size_t *name_len)
 697{
 698        static struct string_list *mail_map;
 699        if (!mail_map) {
 700                mail_map = xcalloc(1, sizeof(*mail_map));
 701                read_mailmap(mail_map, NULL);
 702        }
 703        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 704}
 705
 706static size_t format_person_part(struct strbuf *sb, char part,
 707                                 const char *msg, int len, enum date_mode dmode)
 708{
 709        /* currently all placeholders have same length */
 710        const int placeholder_len = 2;
 711        struct ident_split s;
 712        const char *name, *mail;
 713        size_t maillen, namelen;
 714
 715        if (split_ident_line(&s, msg, len) < 0)
 716                goto skip;
 717
 718        name = s.name_begin;
 719        namelen = s.name_end - s.name_begin;
 720        mail = s.mail_begin;
 721        maillen = s.mail_end - s.mail_begin;
 722
 723        if (part == 'N' || part == 'E') /* mailmap lookup */
 724                mailmap_name(&mail, &maillen, &name, &namelen);
 725        if (part == 'n' || part == 'N') {       /* name */
 726                strbuf_add(sb, name, namelen);
 727                return placeholder_len;
 728        }
 729        if (part == 'e' || part == 'E') {       /* email */
 730                strbuf_add(sb, mail, maillen);
 731                return placeholder_len;
 732        }
 733
 734        if (!s.date_begin)
 735                goto skip;
 736
 737        if (part == 't') {      /* date, UNIX timestamp */
 738                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 739                return placeholder_len;
 740        }
 741
 742        switch (part) {
 743        case 'd':       /* date */
 744                strbuf_addstr(sb, show_ident_date(&s, dmode));
 745                return placeholder_len;
 746        case 'D':       /* date, RFC2822 style */
 747                strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
 748                return placeholder_len;
 749        case 'r':       /* date, relative */
 750                strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
 751                return placeholder_len;
 752        case 'i':       /* date, ISO 8601 */
 753                strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
 754                return placeholder_len;
 755        }
 756
 757skip:
 758        /*
 759         * reading from either a bogus commit, or a reflog entry with
 760         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 761         * to compute a valid return value.
 762         */
 763        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 764            || part == 'D' || part == 'r' || part == 'i')
 765                return placeholder_len;
 766
 767        return 0; /* unknown placeholder */
 768}
 769
 770struct chunk {
 771        size_t off;
 772        size_t len;
 773};
 774
 775enum flush_type {
 776        no_flush,
 777        flush_right,
 778        flush_left,
 779        flush_left_and_steal,
 780        flush_both
 781};
 782
 783enum trunc_type {
 784        trunc_none,
 785        trunc_left,
 786        trunc_middle,
 787        trunc_right
 788};
 789
 790struct format_commit_context {
 791        const struct commit *commit;
 792        const struct pretty_print_context *pretty_ctx;
 793        unsigned commit_header_parsed:1;
 794        unsigned commit_message_parsed:1;
 795        struct signature_check signature_check;
 796        enum flush_type flush_type;
 797        enum trunc_type truncate;
 798        char *message;
 799        char *commit_encoding;
 800        size_t width, indent1, indent2;
 801        int auto_color;
 802        int padding;
 803
 804        /* These offsets are relative to the start of the commit message. */
 805        struct chunk author;
 806        struct chunk committer;
 807        size_t message_off;
 808        size_t subject_off;
 809        size_t body_off;
 810
 811        /* The following ones are relative to the result struct strbuf. */
 812        struct chunk abbrev_commit_hash;
 813        struct chunk abbrev_tree_hash;
 814        struct chunk abbrev_parent_hashes;
 815        size_t wrap_start;
 816};
 817
 818static int add_again(struct strbuf *sb, struct chunk *chunk)
 819{
 820        if (chunk->len) {
 821                strbuf_adddup(sb, chunk->off, chunk->len);
 822                return 1;
 823        }
 824
 825        /*
 826         * We haven't seen this chunk before.  Our caller is surely
 827         * going to add it the hard way now.  Remember the most likely
 828         * start of the to-be-added chunk: the current end of the
 829         * struct strbuf.
 830         */
 831        chunk->off = sb->len;
 832        return 0;
 833}
 834
 835static void parse_commit_header(struct format_commit_context *context)
 836{
 837        const char *msg = context->message;
 838        int i;
 839
 840        for (i = 0; msg[i]; i++) {
 841                int eol;
 842                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 843                        ; /* do nothing */
 844
 845                if (i == eol) {
 846                        break;
 847                } else if (starts_with(msg + i, "author ")) {
 848                        context->author.off = i + 7;
 849                        context->author.len = eol - i - 7;
 850                } else if (starts_with(msg + i, "committer ")) {
 851                        context->committer.off = i + 10;
 852                        context->committer.len = eol - i - 10;
 853                }
 854                i = eol;
 855        }
 856        context->message_off = i;
 857        context->commit_header_parsed = 1;
 858}
 859
 860static int istitlechar(char c)
 861{
 862        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 863                (c >= '0' && c <= '9') || c == '.' || c == '_';
 864}
 865
 866static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 867{
 868        size_t trimlen;
 869        size_t start_len = sb->len;
 870        int space = 2;
 871
 872        for (; *msg && *msg != '\n'; msg++) {
 873                if (istitlechar(*msg)) {
 874                        if (space == 1)
 875                                strbuf_addch(sb, '-');
 876                        space = 0;
 877                        strbuf_addch(sb, *msg);
 878                        if (*msg == '.')
 879                                while (*(msg+1) == '.')
 880                                        msg++;
 881                } else
 882                        space |= 1;
 883        }
 884
 885        /* trim any trailing '.' or '-' characters */
 886        trimlen = 0;
 887        while (sb->len - trimlen > start_len &&
 888                (sb->buf[sb->len - 1 - trimlen] == '.'
 889                || sb->buf[sb->len - 1 - trimlen] == '-'))
 890                trimlen++;
 891        strbuf_remove(sb, sb->len - trimlen, trimlen);
 892}
 893
 894const char *format_subject(struct strbuf *sb, const char *msg,
 895                           const char *line_separator)
 896{
 897        int first = 1;
 898
 899        for (;;) {
 900                const char *line = msg;
 901                int linelen = get_one_line(line);
 902
 903                msg += linelen;
 904                if (!linelen || is_empty_line(line, &linelen))
 905                        break;
 906
 907                if (!sb)
 908                        continue;
 909                strbuf_grow(sb, linelen + 2);
 910                if (!first)
 911                        strbuf_addstr(sb, line_separator);
 912                strbuf_add(sb, line, linelen);
 913                first = 0;
 914        }
 915        return msg;
 916}
 917
 918static void parse_commit_message(struct format_commit_context *c)
 919{
 920        const char *msg = c->message + c->message_off;
 921        const char *start = c->message;
 922
 923        msg = skip_empty_lines(msg);
 924        c->subject_off = msg - start;
 925
 926        msg = format_subject(NULL, msg, NULL);
 927        msg = skip_empty_lines(msg);
 928        c->body_off = msg - start;
 929
 930        c->commit_message_parsed = 1;
 931}
 932
 933static void strbuf_wrap(struct strbuf *sb, size_t pos,
 934                        size_t width, size_t indent1, size_t indent2)
 935{
 936        struct strbuf tmp = STRBUF_INIT;
 937
 938        if (pos)
 939                strbuf_add(&tmp, sb->buf, pos);
 940        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 941                                (int) indent1, (int) indent2, (int) width);
 942        strbuf_swap(&tmp, sb);
 943        strbuf_release(&tmp);
 944}
 945
 946static void rewrap_message_tail(struct strbuf *sb,
 947                                struct format_commit_context *c,
 948                                size_t new_width, size_t new_indent1,
 949                                size_t new_indent2)
 950{
 951        if (c->width == new_width && c->indent1 == new_indent1 &&
 952            c->indent2 == new_indent2)
 953                return;
 954        if (c->wrap_start < sb->len)
 955                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 956        c->wrap_start = sb->len;
 957        c->width = new_width;
 958        c->indent1 = new_indent1;
 959        c->indent2 = new_indent2;
 960}
 961
 962static int format_reflog_person(struct strbuf *sb,
 963                                char part,
 964                                struct reflog_walk_info *log,
 965                                enum date_mode dmode)
 966{
 967        const char *ident;
 968
 969        if (!log)
 970                return 2;
 971
 972        ident = get_reflog_ident(log);
 973        if (!ident)
 974                return 2;
 975
 976        return format_person_part(sb, part, ident, strlen(ident), dmode);
 977}
 978
 979static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 980                          const char *placeholder,
 981                          struct format_commit_context *c)
 982{
 983        if (placeholder[1] == '(') {
 984                const char *begin = placeholder + 2;
 985                const char *end = strchr(begin, ')');
 986                char color[COLOR_MAXLEN];
 987
 988                if (!end)
 989                        return 0;
 990                if (starts_with(begin, "auto,")) {
 991                        if (!want_color(c->pretty_ctx->color))
 992                                return end - placeholder + 1;
 993                        begin += 5;
 994                }
 995                color_parse_mem(begin,
 996                                end - begin,
 997                                "--pretty format", color);
 998                strbuf_addstr(sb, color);
 999                return end - placeholder + 1;
1000        }
1001        if (starts_with(placeholder + 1, "red")) {
1002                strbuf_addstr(sb, GIT_COLOR_RED);
1003                return 4;
1004        } else if (starts_with(placeholder + 1, "green")) {
1005                strbuf_addstr(sb, GIT_COLOR_GREEN);
1006                return 6;
1007        } else if (starts_with(placeholder + 1, "blue")) {
1008                strbuf_addstr(sb, GIT_COLOR_BLUE);
1009                return 5;
1010        } else if (starts_with(placeholder + 1, "reset")) {
1011                strbuf_addstr(sb, GIT_COLOR_RESET);
1012                return 6;
1013        } else
1014                return 0;
1015}
1016
1017static size_t parse_padding_placeholder(struct strbuf *sb,
1018                                        const char *placeholder,
1019                                        struct format_commit_context *c)
1020{
1021        const char *ch = placeholder;
1022        enum flush_type flush_type;
1023        int to_column = 0;
1024
1025        switch (*ch++) {
1026        case '<':
1027                flush_type = flush_right;
1028                break;
1029        case '>':
1030                if (*ch == '<') {
1031                        flush_type = flush_both;
1032                        ch++;
1033                } else if (*ch == '>') {
1034                        flush_type = flush_left_and_steal;
1035                        ch++;
1036                } else
1037                        flush_type = flush_left;
1038                break;
1039        default:
1040                return 0;
1041        }
1042
1043        /* the next value means "wide enough to that column" */
1044        if (*ch == '|') {
1045                to_column = 1;
1046                ch++;
1047        }
1048
1049        if (*ch == '(') {
1050                const char *start = ch + 1;
1051                const char *end = start + strcspn(start, ",)");
1052                char *next;
1053                int width;
1054                if (!end || end == start)
1055                        return 0;
1056                width = strtoul(start, &next, 10);
1057                if (next == start || width == 0)
1058                        return 0;
1059                c->padding = to_column ? -width : width;
1060                c->flush_type = flush_type;
1061
1062                if (*end == ',') {
1063                        start = end + 1;
1064                        end = strchr(start, ')');
1065                        if (!end || end == start)
1066                                return 0;
1067                        if (starts_with(start, "trunc)"))
1068                                c->truncate = trunc_right;
1069                        else if (starts_with(start, "ltrunc)"))
1070                                c->truncate = trunc_left;
1071                        else if (starts_with(start, "mtrunc)"))
1072                                c->truncate = trunc_middle;
1073                        else
1074                                return 0;
1075                } else
1076                        c->truncate = trunc_none;
1077
1078                return end - placeholder + 1;
1079        }
1080        return 0;
1081}
1082
1083static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1084                                const char *placeholder,
1085                                void *context)
1086{
1087        struct format_commit_context *c = context;
1088        const struct commit *commit = c->commit;
1089        const char *msg = c->message;
1090        struct commit_list *p;
1091        int h1, h2;
1092
1093        /* these are independent of the commit */
1094        switch (placeholder[0]) {
1095        case 'C':
1096                if (starts_with(placeholder + 1, "(auto)")) {
1097                        c->auto_color = 1;
1098                        return 7; /* consumed 7 bytes, "C(auto)" */
1099                } else {
1100                        int ret = parse_color(sb, placeholder, c);
1101                        if (ret)
1102                                c->auto_color = 0;
1103                        /*
1104                         * Otherwise, we decided to treat %C<unknown>
1105                         * as a literal string, and the previous
1106                         * %C(auto) is still valid.
1107                         */
1108                        return ret;
1109                }
1110        case 'n':               /* newline */
1111                strbuf_addch(sb, '\n');
1112                return 1;
1113        case 'x':
1114                /* %x00 == NUL, %x0a == LF, etc. */
1115                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1116                    h1 <= 16 &&
1117                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1118                    h2 <= 16) {
1119                        strbuf_addch(sb, (h1<<4)|h2);
1120                        return 3;
1121                } else
1122                        return 0;
1123        case 'w':
1124                if (placeholder[1] == '(') {
1125                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1126                        char *next;
1127                        const char *start = placeholder + 2;
1128                        const char *end = strchr(start, ')');
1129                        if (!end)
1130                                return 0;
1131                        if (end > start) {
1132                                width = strtoul(start, &next, 10);
1133                                if (*next == ',') {
1134                                        indent1 = strtoul(next + 1, &next, 10);
1135                                        if (*next == ',') {
1136                                                indent2 = strtoul(next + 1,
1137                                                                 &next, 10);
1138                                        }
1139                                }
1140                                if (*next != ')')
1141                                        return 0;
1142                        }
1143                        rewrap_message_tail(sb, c, width, indent1, indent2);
1144                        return end - placeholder + 1;
1145                } else
1146                        return 0;
1147
1148        case '<':
1149        case '>':
1150                return parse_padding_placeholder(sb, placeholder, c);
1151        }
1152
1153        /* these depend on the commit */
1154        if (!commit->object.parsed)
1155                parse_object(commit->object.sha1);
1156
1157        switch (placeholder[0]) {
1158        case 'H':               /* commit hash */
1159                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1160                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1161                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1162                return 1;
1163        case 'h':               /* abbreviated commit hash */
1164                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1165                if (add_again(sb, &c->abbrev_commit_hash)) {
1166                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1167                        return 1;
1168                }
1169                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1170                                                     c->pretty_ctx->abbrev));
1171                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1172                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1173                return 1;
1174        case 'T':               /* tree hash */
1175                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1176                return 1;
1177        case 't':               /* abbreviated tree hash */
1178                if (add_again(sb, &c->abbrev_tree_hash))
1179                        return 1;
1180                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1181                                                     c->pretty_ctx->abbrev));
1182                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1183                return 1;
1184        case 'P':               /* parent hashes */
1185                for (p = commit->parents; p; p = p->next) {
1186                        if (p != commit->parents)
1187                                strbuf_addch(sb, ' ');
1188                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1189                }
1190                return 1;
1191        case 'p':               /* abbreviated parent hashes */
1192                if (add_again(sb, &c->abbrev_parent_hashes))
1193                        return 1;
1194                for (p = commit->parents; p; p = p->next) {
1195                        if (p != commit->parents)
1196                                strbuf_addch(sb, ' ');
1197                        strbuf_addstr(sb, find_unique_abbrev(
1198                                        p->item->object.sha1,
1199                                        c->pretty_ctx->abbrev));
1200                }
1201                c->abbrev_parent_hashes.len = sb->len -
1202                                              c->abbrev_parent_hashes.off;
1203                return 1;
1204        case 'm':               /* left/right/bottom */
1205                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1206                return 1;
1207        case 'd':
1208                load_ref_decorations(DECORATE_SHORT_REFS);
1209                format_decorations(sb, commit, c->auto_color);
1210                return 1;
1211        case 'g':               /* reflog info */
1212                switch(placeholder[1]) {
1213                case 'd':       /* reflog selector */
1214                case 'D':
1215                        if (c->pretty_ctx->reflog_info)
1216                                get_reflog_selector(sb,
1217                                                    c->pretty_ctx->reflog_info,
1218                                                    c->pretty_ctx->date_mode,
1219                                                    c->pretty_ctx->date_mode_explicit,
1220                                                    (placeholder[1] == 'd'));
1221                        return 2;
1222                case 's':       /* reflog message */
1223                        if (c->pretty_ctx->reflog_info)
1224                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1225                        return 2;
1226                case 'n':
1227                case 'N':
1228                case 'e':
1229                case 'E':
1230                        return format_reflog_person(sb,
1231                                                    placeholder[1],
1232                                                    c->pretty_ctx->reflog_info,
1233                                                    c->pretty_ctx->date_mode);
1234                }
1235                return 0;       /* unknown %g placeholder */
1236        case 'N':
1237                if (c->pretty_ctx->notes_message) {
1238                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1239                        return 1;
1240                }
1241                return 0;
1242        }
1243
1244        if (placeholder[0] == 'G') {
1245                if (!c->signature_check.result)
1246                        check_commit_signature(c->commit, &(c->signature_check));
1247                switch (placeholder[1]) {
1248                case 'G':
1249                        if (c->signature_check.gpg_output)
1250                                strbuf_addstr(sb, c->signature_check.gpg_output);
1251                        break;
1252                case '?':
1253                        switch (c->signature_check.result) {
1254                        case 'G':
1255                        case 'B':
1256                        case 'U':
1257                        case 'N':
1258                                strbuf_addch(sb, c->signature_check.result);
1259                        }
1260                        break;
1261                case 'S':
1262                        if (c->signature_check.signer)
1263                                strbuf_addstr(sb, c->signature_check.signer);
1264                        break;
1265                case 'K':
1266                        if (c->signature_check.key)
1267                                strbuf_addstr(sb, c->signature_check.key);
1268                        break;
1269                }
1270                return 2;
1271        }
1272
1273
1274        /* For the rest we have to parse the commit header. */
1275        if (!c->commit_header_parsed)
1276                parse_commit_header(c);
1277
1278        switch (placeholder[0]) {
1279        case 'a':       /* author ... */
1280                return format_person_part(sb, placeholder[1],
1281                                   msg + c->author.off, c->author.len,
1282                                   c->pretty_ctx->date_mode);
1283        case 'c':       /* committer ... */
1284                return format_person_part(sb, placeholder[1],
1285                                   msg + c->committer.off, c->committer.len,
1286                                   c->pretty_ctx->date_mode);
1287        case 'e':       /* encoding */
1288                if (c->commit_encoding)
1289                        strbuf_addstr(sb, c->commit_encoding);
1290                return 1;
1291        case 'B':       /* raw body */
1292                /* message_off is always left at the initial newline */
1293                strbuf_addstr(sb, msg + c->message_off + 1);
1294                return 1;
1295        }
1296
1297        /* Now we need to parse the commit message. */
1298        if (!c->commit_message_parsed)
1299                parse_commit_message(c);
1300
1301        switch (placeholder[0]) {
1302        case 's':       /* subject */
1303                format_subject(sb, msg + c->subject_off, " ");
1304                return 1;
1305        case 'f':       /* sanitized subject */
1306                format_sanitized_subject(sb, msg + c->subject_off);
1307                return 1;
1308        case 'b':       /* body */
1309                strbuf_addstr(sb, msg + c->body_off);
1310                return 1;
1311        }
1312        return 0;       /* unknown placeholder */
1313}
1314
1315static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1316                                    const char *placeholder,
1317                                    struct format_commit_context *c)
1318{
1319        struct strbuf local_sb = STRBUF_INIT;
1320        int total_consumed = 0, len, padding = c->padding;
1321        if (padding < 0) {
1322                const char *start = strrchr(sb->buf, '\n');
1323                int occupied;
1324                if (!start)
1325                        start = sb->buf;
1326                occupied = utf8_strnwidth(start, -1, 1);
1327                padding = (-padding) - occupied;
1328        }
1329        while (1) {
1330                int modifier = *placeholder == 'C';
1331                int consumed = format_commit_one(&local_sb, placeholder, c);
1332                total_consumed += consumed;
1333
1334                if (!modifier)
1335                        break;
1336
1337                placeholder += consumed;
1338                if (*placeholder != '%')
1339                        break;
1340                placeholder++;
1341                total_consumed++;
1342        }
1343        len = utf8_strnwidth(local_sb.buf, -1, 1);
1344
1345        if (c->flush_type == flush_left_and_steal) {
1346                const char *ch = sb->buf + sb->len - 1;
1347                while (len > padding && ch > sb->buf) {
1348                        const char *p;
1349                        if (*ch == ' ') {
1350                                ch--;
1351                                padding++;
1352                                continue;
1353                        }
1354                        /* check for trailing ansi sequences */
1355                        if (*ch != 'm')
1356                                break;
1357                        p = ch - 1;
1358                        while (ch - p < 10 && *p != '\033')
1359                                p--;
1360                        if (*p != '\033' ||
1361                            ch + 1 - p != display_mode_esc_sequence_len(p))
1362                                break;
1363                        /*
1364                         * got a good ansi sequence, put it back to
1365                         * local_sb as we're cutting sb
1366                         */
1367                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1368                        ch = p - 1;
1369                }
1370                strbuf_setlen(sb, ch + 1 - sb->buf);
1371                c->flush_type = flush_left;
1372        }
1373
1374        if (len > padding) {
1375                switch (c->truncate) {
1376                case trunc_left:
1377                        strbuf_utf8_replace(&local_sb,
1378                                            0, len - (padding - 2),
1379                                            "..");
1380                        break;
1381                case trunc_middle:
1382                        strbuf_utf8_replace(&local_sb,
1383                                            padding / 2 - 1,
1384                                            len - (padding - 2),
1385                                            "..");
1386                        break;
1387                case trunc_right:
1388                        strbuf_utf8_replace(&local_sb,
1389                                            padding - 2, len - (padding - 2),
1390                                            "..");
1391                        break;
1392                case trunc_none:
1393                        break;
1394                }
1395                strbuf_addstr(sb, local_sb.buf);
1396        } else {
1397                int sb_len = sb->len, offset = 0;
1398                if (c->flush_type == flush_left)
1399                        offset = padding - len;
1400                else if (c->flush_type == flush_both)
1401                        offset = (padding - len) / 2;
1402                /*
1403                 * we calculate padding in columns, now
1404                 * convert it back to chars
1405                 */
1406                padding = padding - len + local_sb.len;
1407                strbuf_grow(sb, padding);
1408                strbuf_setlen(sb, sb_len + padding);
1409                memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1410                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1411                       local_sb.len);
1412        }
1413        strbuf_release(&local_sb);
1414        c->flush_type = no_flush;
1415        return total_consumed;
1416}
1417
1418static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1419                                 const char *placeholder,
1420                                 void *context)
1421{
1422        int consumed;
1423        size_t orig_len;
1424        enum {
1425                NO_MAGIC,
1426                ADD_LF_BEFORE_NON_EMPTY,
1427                DEL_LF_BEFORE_EMPTY,
1428                ADD_SP_BEFORE_NON_EMPTY
1429        } magic = NO_MAGIC;
1430
1431        switch (placeholder[0]) {
1432        case '-':
1433                magic = DEL_LF_BEFORE_EMPTY;
1434                break;
1435        case '+':
1436                magic = ADD_LF_BEFORE_NON_EMPTY;
1437                break;
1438        case ' ':
1439                magic = ADD_SP_BEFORE_NON_EMPTY;
1440                break;
1441        default:
1442                break;
1443        }
1444        if (magic != NO_MAGIC)
1445                placeholder++;
1446
1447        orig_len = sb->len;
1448        if (((struct format_commit_context *)context)->flush_type != no_flush)
1449                consumed = format_and_pad_commit(sb, placeholder, context);
1450        else
1451                consumed = format_commit_one(sb, placeholder, context);
1452        if (magic == NO_MAGIC)
1453                return consumed;
1454
1455        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1456                while (sb->len && sb->buf[sb->len - 1] == '\n')
1457                        strbuf_setlen(sb, sb->len - 1);
1458        } else if (orig_len != sb->len) {
1459                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1460                        strbuf_insert(sb, orig_len, "\n", 1);
1461                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1462                        strbuf_insert(sb, orig_len, " ", 1);
1463        }
1464        return consumed + 1;
1465}
1466
1467static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1468                                   void *context)
1469{
1470        struct userformat_want *w = context;
1471
1472        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1473                placeholder++;
1474
1475        switch (*placeholder) {
1476        case 'N':
1477                w->notes = 1;
1478                break;
1479        }
1480        return 0;
1481}
1482
1483void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1484{
1485        struct strbuf dummy = STRBUF_INIT;
1486
1487        if (!fmt) {
1488                if (!user_format)
1489                        return;
1490                fmt = user_format;
1491        }
1492        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1493        strbuf_release(&dummy);
1494}
1495
1496void format_commit_message(const struct commit *commit,
1497                           const char *format, struct strbuf *sb,
1498                           const struct pretty_print_context *pretty_ctx)
1499{
1500        struct format_commit_context context;
1501        const char *output_enc = pretty_ctx->output_encoding;
1502        const char *utf8 = "UTF-8";
1503
1504        memset(&context, 0, sizeof(context));
1505        context.commit = commit;
1506        context.pretty_ctx = pretty_ctx;
1507        context.wrap_start = sb->len;
1508        /*
1509         * convert a commit message to UTF-8 first
1510         * as far as 'format_commit_item' assumes it in UTF-8
1511         */
1512        context.message = logmsg_reencode(commit,
1513                                          &context.commit_encoding,
1514                                          utf8);
1515
1516        strbuf_expand(sb, format, format_commit_item, &context);
1517        rewrap_message_tail(sb, &context, 0, 0, 0);
1518
1519        /* then convert a commit message to an actual output encoding */
1520        if (output_enc) {
1521                if (same_encoding(utf8, output_enc))
1522                        output_enc = NULL;
1523        } else {
1524                if (context.commit_encoding &&
1525                    !same_encoding(context.commit_encoding, utf8))
1526                        output_enc = context.commit_encoding;
1527        }
1528
1529        if (output_enc) {
1530                int outsz;
1531                char *out = reencode_string_len(sb->buf, sb->len,
1532                                                output_enc, utf8, &outsz);
1533                if (out)
1534                        strbuf_attach(sb, out, outsz, outsz + 1);
1535        }
1536
1537        free(context.commit_encoding);
1538        logmsg_free(context.message, commit);
1539        free(context.signature_check.gpg_output);
1540        free(context.signature_check.signer);
1541}
1542
1543static void pp_header(struct pretty_print_context *pp,
1544                      const char *encoding,
1545                      const struct commit *commit,
1546                      const char **msg_p,
1547                      struct strbuf *sb)
1548{
1549        int parents_shown = 0;
1550
1551        for (;;) {
1552                const char *line = *msg_p;
1553                int linelen = get_one_line(*msg_p);
1554
1555                if (!linelen)
1556                        return;
1557                *msg_p += linelen;
1558
1559                if (linelen == 1)
1560                        /* End of header */
1561                        return;
1562
1563                if (pp->fmt == CMIT_FMT_RAW) {
1564                        strbuf_add(sb, line, linelen);
1565                        continue;
1566                }
1567
1568                if (starts_with(line, "parent ")) {
1569                        if (linelen != 48)
1570                                die("bad parent line in commit");
1571                        continue;
1572                }
1573
1574                if (!parents_shown) {
1575                        struct commit_list *parent;
1576                        int num;
1577                        for (parent = commit->parents, num = 0;
1578                             parent;
1579                             parent = parent->next, num++)
1580                                ;
1581                        /* with enough slop */
1582                        strbuf_grow(sb, num * 50 + 20);
1583                        add_merge_info(pp, sb, commit);
1584                        parents_shown = 1;
1585                }
1586
1587                /*
1588                 * MEDIUM == DEFAULT shows only author with dates.
1589                 * FULL shows both authors but not dates.
1590                 * FULLER shows both authors and dates.
1591                 */
1592                if (starts_with(line, "author ")) {
1593                        strbuf_grow(sb, linelen + 80);
1594                        pp_user_info(pp, "Author", sb, line + 7, encoding);
1595                }
1596                if (starts_with(line, "committer ") &&
1597                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1598                        strbuf_grow(sb, linelen + 80);
1599                        pp_user_info(pp, "Commit", sb, line + 10, encoding);
1600                }
1601        }
1602}
1603
1604void pp_title_line(struct pretty_print_context *pp,
1605                   const char **msg_p,
1606                   struct strbuf *sb,
1607                   const char *encoding,
1608                   int need_8bit_cte)
1609{
1610        static const int max_length = 78; /* per rfc2047 */
1611        struct strbuf title;
1612
1613        strbuf_init(&title, 80);
1614        *msg_p = format_subject(&title, *msg_p,
1615                                pp->preserve_subject ? "\n" : " ");
1616
1617        strbuf_grow(sb, title.len + 1024);
1618        if (pp->subject) {
1619                strbuf_addstr(sb, pp->subject);
1620                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1621                        add_rfc2047(sb, title.buf, title.len,
1622                                                encoding, RFC2047_SUBJECT);
1623                else
1624                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1625                                         -last_line_length(sb), 1, max_length);
1626        } else {
1627                strbuf_addbuf(sb, &title);
1628        }
1629        strbuf_addch(sb, '\n');
1630
1631        if (need_8bit_cte == 0) {
1632                int i;
1633                for (i = 0; i < pp->in_body_headers.nr; i++) {
1634                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1635                                need_8bit_cte = 1;
1636                                break;
1637                        }
1638                }
1639        }
1640
1641        if (need_8bit_cte > 0) {
1642                const char *header_fmt =
1643                        "MIME-Version: 1.0\n"
1644                        "Content-Type: text/plain; charset=%s\n"
1645                        "Content-Transfer-Encoding: 8bit\n";
1646                strbuf_addf(sb, header_fmt, encoding);
1647        }
1648        if (pp->after_subject) {
1649                strbuf_addstr(sb, pp->after_subject);
1650        }
1651        if (pp->fmt == CMIT_FMT_EMAIL) {
1652                strbuf_addch(sb, '\n');
1653        }
1654
1655        if (pp->in_body_headers.nr) {
1656                int i;
1657                for (i = 0; i < pp->in_body_headers.nr; i++) {
1658                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1659                        free(pp->in_body_headers.items[i].string);
1660                }
1661                string_list_clear(&pp->in_body_headers, 0);
1662                strbuf_addch(sb, '\n');
1663        }
1664
1665        strbuf_release(&title);
1666}
1667
1668void pp_remainder(struct pretty_print_context *pp,
1669                  const char **msg_p,
1670                  struct strbuf *sb,
1671                  int indent)
1672{
1673        int first = 1;
1674        for (;;) {
1675                const char *line = *msg_p;
1676                int linelen = get_one_line(line);
1677                *msg_p += linelen;
1678
1679                if (!linelen)
1680                        break;
1681
1682                if (is_empty_line(line, &linelen)) {
1683                        if (first)
1684                                continue;
1685                        if (pp->fmt == CMIT_FMT_SHORT)
1686                                break;
1687                }
1688                first = 0;
1689
1690                strbuf_grow(sb, linelen + indent + 20);
1691                if (indent) {
1692                        memset(sb->buf + sb->len, ' ', indent);
1693                        strbuf_setlen(sb, sb->len + indent);
1694                }
1695                strbuf_add(sb, line, linelen);
1696                strbuf_addch(sb, '\n');
1697        }
1698}
1699
1700void pretty_print_commit(struct pretty_print_context *pp,
1701                         const struct commit *commit,
1702                         struct strbuf *sb)
1703{
1704        unsigned long beginning_of_body;
1705        int indent = 4;
1706        const char *msg;
1707        char *reencoded;
1708        const char *encoding;
1709        int need_8bit_cte = pp->need_8bit_cte;
1710
1711        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1712                format_commit_message(commit, user_format, sb, pp);
1713                return;
1714        }
1715
1716        encoding = get_log_output_encoding();
1717        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1718
1719        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1720                indent = 0;
1721
1722        /*
1723         * We need to check and emit Content-type: to mark it
1724         * as 8-bit if we haven't done so.
1725         */
1726        if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1727                int i, ch, in_body;
1728
1729                for (in_body = i = 0; (ch = msg[i]); i++) {
1730                        if (!in_body) {
1731                                /* author could be non 7-bit ASCII but
1732                                 * the log may be so; skip over the
1733                                 * header part first.
1734                                 */
1735                                if (ch == '\n' && msg[i+1] == '\n')
1736                                        in_body = 1;
1737                        }
1738                        else if (non_ascii(ch)) {
1739                                need_8bit_cte = 1;
1740                                break;
1741                        }
1742                }
1743        }
1744
1745        pp_header(pp, encoding, commit, &msg, sb);
1746        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1747                strbuf_addch(sb, '\n');
1748        }
1749
1750        /* Skip excess blank lines at the beginning of body, if any... */
1751        msg = skip_empty_lines(msg);
1752
1753        /* These formats treat the title line specially. */
1754        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1755                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1756
1757        beginning_of_body = sb->len;
1758        if (pp->fmt != CMIT_FMT_ONELINE)
1759                pp_remainder(pp, &msg, sb, indent);
1760        strbuf_rtrim(sb);
1761
1762        /* Make sure there is an EOLN for the non-oneline case */
1763        if (pp->fmt != CMIT_FMT_ONELINE)
1764                strbuf_addch(sb, '\n');
1765
1766        /*
1767         * The caller may append additional body text in e-mail
1768         * format.  Make sure we did not strip the blank line
1769         * between the header and the body.
1770         */
1771        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1772                strbuf_addch(sb, '\n');
1773
1774        logmsg_free(reencoded, commit);
1775}
1776
1777void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1778                    struct strbuf *sb)
1779{
1780        struct pretty_print_context pp = {0};
1781        pp.fmt = fmt;
1782        pretty_print_commit(&pp, commit, sb);
1783}