pretty.con commit test-sha1: add a binary output mode (54c93cb)
   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 (prefixcmp(var, "pretty."))
  44                return 0;
  45
  46        name = var + strlen("pretty.");
  47        for (i = 0; i < builtin_formats_len; i++) {
  48                if (!strcmp(commit_formats[i].name, name))
  49                        return 0;
  50        }
  51
  52        for (i = builtin_formats_len; i < commit_formats_len; i++) {
  53                if (!strcmp(commit_formats[i].name, name)) {
  54                        commit_format = &commit_formats[i];
  55                        break;
  56                }
  57        }
  58
  59        if (!commit_format) {
  60                ALLOC_GROW(commit_formats, commit_formats_len+1,
  61                           commit_formats_alloc);
  62                commit_format = &commit_formats[commit_formats_len];
  63                memset(commit_format, 0, sizeof(*commit_format));
  64                commit_formats_len++;
  65        }
  66
  67        commit_format->name = xstrdup(name);
  68        commit_format->format = CMIT_FMT_USERFORMAT;
  69        git_config_string(&fmt, var, value);
  70        if (!prefixcmp(fmt, "format:") || !prefixcmp(fmt, "tformat:")) {
  71                commit_format->is_tformat = fmt[0] == 't';
  72                fmt = strchr(fmt, ':') + 1;
  73        } else if (strchr(fmt, '%'))
  74                commit_format->is_tformat = 1;
  75        else
  76                commit_format->is_alias = 1;
  77        commit_format->user_format = fmt;
  78
  79        return 0;
  80}
  81
  82static void setup_commit_formats(void)
  83{
  84        struct cmt_fmt_map builtin_formats[] = {
  85                { "raw",        CMIT_FMT_RAW,           0 },
  86                { "medium",     CMIT_FMT_MEDIUM,        0 },
  87                { "short",      CMIT_FMT_SHORT,         0 },
  88                { "email",      CMIT_FMT_EMAIL,         0 },
  89                { "fuller",     CMIT_FMT_FULLER,        0 },
  90                { "full",       CMIT_FMT_FULL,          0 },
  91                { "oneline",    CMIT_FMT_ONELINE,       1 }
  92        };
  93        commit_formats_len = ARRAY_SIZE(builtin_formats);
  94        builtin_formats_len = commit_formats_len;
  95        ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
  96        memcpy(commit_formats, builtin_formats,
  97               sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  98
  99        git_config(git_pretty_formats_config, NULL);
 100}
 101
 102static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 103                                                        const char *original,
 104                                                        int num_redirections)
 105{
 106        struct cmt_fmt_map *found = NULL;
 107        size_t found_match_len = 0;
 108        int i;
 109
 110        if (num_redirections >= commit_formats_len)
 111                die("invalid --pretty format: "
 112                    "'%s' references an alias which points to itself",
 113                    original);
 114
 115        for (i = 0; i < commit_formats_len; i++) {
 116                size_t match_len;
 117
 118                if (prefixcmp(commit_formats[i].name, sought))
 119                        continue;
 120
 121                match_len = strlen(commit_formats[i].name);
 122                if (found == NULL || found_match_len > match_len) {
 123                        found = &commit_formats[i];
 124                        found_match_len = match_len;
 125                }
 126        }
 127
 128        if (found && found->is_alias) {
 129                found = find_commit_format_recursive(found->user_format,
 130                                                     original,
 131                                                     num_redirections+1);
 132        }
 133
 134        return found;
 135}
 136
 137static struct cmt_fmt_map *find_commit_format(const char *sought)
 138{
 139        if (!commit_formats)
 140                setup_commit_formats();
 141
 142        return find_commit_format_recursive(sought, sought, 0);
 143}
 144
 145void get_commit_format(const char *arg, struct rev_info *rev)
 146{
 147        struct cmt_fmt_map *commit_format;
 148
 149        rev->use_terminator = 0;
 150        if (!arg || !*arg) {
 151                rev->commit_format = CMIT_FMT_DEFAULT;
 152                return;
 153        }
 154        if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
 155                save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 156                return;
 157        }
 158
 159        if (strchr(arg, '%')) {
 160                save_user_format(rev, arg, 1);
 161                return;
 162        }
 163
 164        commit_format = find_commit_format(arg);
 165        if (!commit_format)
 166                die("invalid --pretty format: %s", arg);
 167
 168        rev->commit_format = commit_format->format;
 169        rev->use_terminator = commit_format->is_tformat;
 170        if (commit_format->format == CMIT_FMT_USERFORMAT) {
 171                save_user_format(rev, commit_format->user_format,
 172                                 commit_format->is_tformat);
 173        }
 174}
 175
 176/*
 177 * Generic support for pretty-printing the header
 178 */
 179static int get_one_line(const char *msg)
 180{
 181        int ret = 0;
 182
 183        for (;;) {
 184                char c = *msg++;
 185                if (!c)
 186                        break;
 187                ret++;
 188                if (c == '\n')
 189                        break;
 190        }
 191        return ret;
 192}
 193
 194/* High bit set, or ISO-2022-INT */
 195static int non_ascii(int ch)
 196{
 197        return !isascii(ch) || ch == '\033';
 198}
 199
 200int has_non_ascii(const char *s)
 201{
 202        int ch;
 203        if (!s)
 204                return 0;
 205        while ((ch = *s++) != '\0') {
 206                if (non_ascii(ch))
 207                        return 1;
 208        }
 209        return 0;
 210}
 211
 212static int is_rfc822_special(char ch)
 213{
 214        switch (ch) {
 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        case '\\':
 228                return 1;
 229        default:
 230                return 0;
 231        }
 232}
 233
 234static int needs_rfc822_quoting(const char *s, int len)
 235{
 236        int i;
 237        for (i = 0; i < len; i++)
 238                if (is_rfc822_special(s[i]))
 239                        return 1;
 240        return 0;
 241}
 242
 243static int last_line_length(struct strbuf *sb)
 244{
 245        int i;
 246
 247        /* How many bytes are already used on the last line? */
 248        for (i = sb->len - 1; i >= 0; i--)
 249                if (sb->buf[i] == '\n')
 250                        break;
 251        return sb->len - (i + 1);
 252}
 253
 254static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 255{
 256        int i;
 257
 258        /* just a guess, we may have to also backslash-quote */
 259        strbuf_grow(out, len + 2);
 260
 261        strbuf_addch(out, '"');
 262        for (i = 0; i < len; i++) {
 263                switch (s[i]) {
 264                case '"':
 265                case '\\':
 266                        strbuf_addch(out, '\\');
 267                        /* fall through */
 268                default:
 269                        strbuf_addch(out, s[i]);
 270                }
 271        }
 272        strbuf_addch(out, '"');
 273}
 274
 275enum rfc2047_type {
 276        RFC2047_SUBJECT,
 277        RFC2047_ADDRESS,
 278};
 279
 280static int is_rfc2047_special(char ch, enum rfc2047_type type)
 281{
 282        /*
 283         * rfc2047, section 4.2:
 284         *
 285         *    8-bit values which correspond to printable ASCII characters other
 286         *    than "=", "?", and "_" (underscore), MAY be represented as those
 287         *    characters.  (But see section 5 for restrictions.)  In
 288         *    particular, SPACE and TAB MUST NOT be represented as themselves
 289         *    within encoded words.
 290         */
 291
 292        /*
 293         * rule out non-ASCII characters and non-printable characters (the
 294         * non-ASCII check should be redundant as isprint() is not localized
 295         * and only knows about ASCII, but be defensive about that)
 296         */
 297        if (non_ascii(ch) || !isprint(ch))
 298                return 1;
 299
 300        /*
 301         * rule out special printable characters (' ' should be the only
 302         * whitespace character considered printable, but be defensive and use
 303         * isspace())
 304         */
 305        if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 306                return 1;
 307
 308        /*
 309         * rfc2047, section 5.3:
 310         *
 311         *    As a replacement for a 'word' entity within a 'phrase', for example,
 312         *    one that precedes an address in a From, To, or Cc header.  The ABNF
 313         *    definition for 'phrase' from RFC 822 thus becomes:
 314         *
 315         *    phrase = 1*( encoded-word / word )
 316         *
 317         *    In this case the set of characters that may be used in a "Q"-encoded
 318         *    'encoded-word' is restricted to: <upper and lower case ASCII
 319         *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 320         *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 321         *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 322         *    'special' by 'linear-white-space'.
 323         */
 324
 325        if (type != RFC2047_ADDRESS)
 326                return 0;
 327
 328        /* '=' and '_' are special cases and have been checked above */
 329        return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 330}
 331
 332static int needs_rfc2047_encoding(const char *line, int len,
 333                                  enum rfc2047_type type)
 334{
 335        int i;
 336
 337        for (i = 0; i < len; i++) {
 338                int ch = line[i];
 339                if (non_ascii(ch) || ch == '\n')
 340                        return 1;
 341                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 342                        return 1;
 343        }
 344
 345        return 0;
 346}
 347
 348static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 349                       const char *encoding, enum rfc2047_type type)
 350{
 351        static const int max_encoded_length = 76; /* per rfc2047 */
 352        int i;
 353        int line_len = last_line_length(sb);
 354
 355        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 356        strbuf_addf(sb, "=?%s?q?", encoding);
 357        line_len += strlen(encoding) + 5; /* 5 for =??q? */
 358
 359        while (len) {
 360                /*
 361                 * RFC 2047, section 5 (3):
 362                 *
 363                 * Each 'encoded-word' MUST represent an integral number of
 364                 * characters.  A multi-octet character may not be split across
 365                 * adjacent 'encoded- word's.
 366                 */
 367                const unsigned char *p = (const unsigned char *)line;
 368                int chrlen = mbs_chrlen(&line, &len, encoding);
 369                int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 370
 371                /* "=%02X" * chrlen, or the byte itself */
 372                const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 373                int         encoded_len = is_special ? 3 * chrlen : 1;
 374
 375                /*
 376                 * According to RFC 2047, we could encode the special character
 377                 * ' ' (space) with '_' (underscore) for readability. But many
 378                 * programs do not understand this and just leave the
 379                 * underscore in place. Thus, we do nothing special here, which
 380                 * causes ' ' to be encoded as '=20', avoiding this problem.
 381                 */
 382
 383                if (line_len + encoded_len + 2 > max_encoded_length) {
 384                        /* It won't fit with trailing "?=" --- break the line */
 385                        strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 386                        line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 387                }
 388
 389                for (i = 0; i < chrlen; i++)
 390                        strbuf_addf(sb, encoded_fmt, p[i]);
 391                line_len += encoded_len;
 392        }
 393        strbuf_addstr(sb, "?=");
 394}
 395
 396static const char *show_ident_date(const struct ident_split *ident,
 397                                   enum date_mode mode)
 398{
 399        unsigned long date = 0;
 400        int tz = 0;
 401
 402        if (ident->date_begin && ident->date_end)
 403                date = strtoul(ident->date_begin, NULL, 10);
 404        if (ident->tz_begin && ident->tz_end)
 405                tz = strtol(ident->tz_begin, NULL, 10);
 406        return show_date(date, tz, mode);
 407}
 408
 409void pp_user_info(const struct pretty_print_context *pp,
 410                  const char *what, struct strbuf *sb,
 411                  const char *line, const char *encoding)
 412{
 413        struct ident_split ident;
 414        char *line_end;
 415        const char *mailbuf, *namebuf;
 416        size_t namelen, maillen;
 417        int max_length = 78; /* per rfc2822 */
 418
 419        if (pp->fmt == CMIT_FMT_ONELINE)
 420                return;
 421
 422        line_end = strchrnul(line, '\n');
 423        if (split_ident_line(&ident, line, line_end - line))
 424                return;
 425
 426        mailbuf = ident.mail_begin;
 427        maillen = ident.mail_end - ident.mail_begin;
 428        namebuf = ident.name_begin;
 429        namelen = ident.name_end - ident.name_begin;
 430
 431        if (pp->mailmap)
 432                map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 433
 434        if (pp->fmt == CMIT_FMT_EMAIL) {
 435                strbuf_addstr(sb, "From: ");
 436                if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 437                        add_rfc2047(sb, namebuf, namelen,
 438                                    encoding, RFC2047_ADDRESS);
 439                        max_length = 76; /* per rfc2047 */
 440                } else if (needs_rfc822_quoting(namebuf, namelen)) {
 441                        struct strbuf quoted = STRBUF_INIT;
 442                        add_rfc822_quoted(&quoted, namebuf, namelen);
 443                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 444                                                        -6, 1, max_length);
 445                        strbuf_release(&quoted);
 446                } else {
 447                        strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 448                                                 -6, 1, max_length);
 449                }
 450
 451                if (max_length <
 452                    last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 453                        strbuf_addch(sb, '\n');
 454                strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 455        } else {
 456                strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 457                            (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 458                            (int)namelen, namebuf, (int)maillen, mailbuf);
 459        }
 460
 461        switch (pp->fmt) {
 462        case CMIT_FMT_MEDIUM:
 463                strbuf_addf(sb, "Date:   %s\n",
 464                            show_ident_date(&ident, pp->date_mode));
 465                break;
 466        case CMIT_FMT_EMAIL:
 467                strbuf_addf(sb, "Date: %s\n",
 468                            show_ident_date(&ident, DATE_RFC2822));
 469                break;
 470        case CMIT_FMT_FULLER:
 471                strbuf_addf(sb, "%sDate: %s\n", what,
 472                            show_ident_date(&ident, pp->date_mode));
 473                break;
 474        default:
 475                /* notin' */
 476                break;
 477        }
 478}
 479
 480static int is_empty_line(const char *line, int *len_p)
 481{
 482        int len = *len_p;
 483        while (len && isspace(line[len-1]))
 484                len--;
 485        *len_p = len;
 486        return !len;
 487}
 488
 489static const char *skip_empty_lines(const char *msg)
 490{
 491        for (;;) {
 492                int linelen = get_one_line(msg);
 493                int ll = linelen;
 494                if (!linelen)
 495                        break;
 496                if (!is_empty_line(msg, &ll))
 497                        break;
 498                msg += linelen;
 499        }
 500        return msg;
 501}
 502
 503static void add_merge_info(const struct pretty_print_context *pp,
 504                           struct strbuf *sb, const struct commit *commit)
 505{
 506        struct commit_list *parent = commit->parents;
 507
 508        if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 509            !parent || !parent->next)
 510                return;
 511
 512        strbuf_addstr(sb, "Merge:");
 513
 514        while (parent) {
 515                struct commit *p = parent->item;
 516                const char *hex = NULL;
 517                if (pp->abbrev)
 518                        hex = find_unique_abbrev(p->object.sha1, pp->abbrev);
 519                if (!hex)
 520                        hex = sha1_to_hex(p->object.sha1);
 521                parent = parent->next;
 522
 523                strbuf_addf(sb, " %s", hex);
 524        }
 525        strbuf_addch(sb, '\n');
 526}
 527
 528static char *get_header(const struct commit *commit, const char *msg,
 529                        const char *key)
 530{
 531        int key_len = strlen(key);
 532        const char *line = msg;
 533
 534        while (line) {
 535                const char *eol = strchr(line, '\n'), *next;
 536
 537                if (line == eol)
 538                        return NULL;
 539                if (!eol) {
 540                        warning("malformed commit (header is missing newline): %s",
 541                                sha1_to_hex(commit->object.sha1));
 542                        eol = line + strlen(line);
 543                        next = NULL;
 544                } else
 545                        next = eol + 1;
 546                if (eol - line > key_len &&
 547                    !strncmp(line, key, key_len) &&
 548                    line[key_len] == ' ') {
 549                        return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 550                }
 551                line = next;
 552        }
 553        return NULL;
 554}
 555
 556static char *replace_encoding_header(char *buf, const char *encoding)
 557{
 558        struct strbuf tmp = STRBUF_INIT;
 559        size_t start, len;
 560        char *cp = buf;
 561
 562        /* guess if there is an encoding header before a \n\n */
 563        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 564                cp = strchr(cp, '\n');
 565                if (!cp || *++cp == '\n')
 566                        return buf;
 567        }
 568        start = cp - buf;
 569        cp = strchr(cp, '\n');
 570        if (!cp)
 571                return buf; /* should not happen but be defensive */
 572        len = cp + 1 - (buf + start);
 573
 574        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 575        if (is_encoding_utf8(encoding)) {
 576                /* we have re-coded to UTF-8; drop the header */
 577                strbuf_remove(&tmp, start, len);
 578        } else {
 579                /* just replaces XXXX in 'encoding XXXX\n' */
 580                strbuf_splice(&tmp, start + strlen("encoding "),
 581                                          len - strlen("encoding \n"),
 582                                          encoding, strlen(encoding));
 583        }
 584        return strbuf_detach(&tmp, NULL);
 585}
 586
 587char *logmsg_reencode(const struct commit *commit,
 588                      char **commit_encoding,
 589                      const char *output_encoding)
 590{
 591        static const char *utf8 = "UTF-8";
 592        const char *use_encoding;
 593        char *encoding;
 594        char *msg = commit->buffer;
 595        char *out;
 596
 597        if (!msg) {
 598                enum object_type type;
 599                unsigned long size;
 600
 601                msg = read_sha1_file(commit->object.sha1, &type, &size);
 602                if (!msg)
 603                        die("Cannot read commit object %s",
 604                            sha1_to_hex(commit->object.sha1));
 605                if (type != OBJ_COMMIT)
 606                        die("Expected commit for '%s', got %s",
 607                            sha1_to_hex(commit->object.sha1), typename(type));
 608        }
 609
 610        if (!output_encoding || !*output_encoding) {
 611                if (commit_encoding)
 612                        *commit_encoding =
 613                                get_header(commit, msg, "encoding");
 614                return msg;
 615        }
 616        encoding = get_header(commit, msg, "encoding");
 617        if (commit_encoding)
 618                *commit_encoding = encoding;
 619        use_encoding = encoding ? encoding : utf8;
 620        if (same_encoding(use_encoding, output_encoding)) {
 621                /*
 622                 * No encoding work to be done. If we have no encoding header
 623                 * at all, then there's nothing to do, and we can return the
 624                 * message verbatim (whether newly allocated or not).
 625                 */
 626                if (!encoding)
 627                        return msg;
 628
 629                /*
 630                 * Otherwise, we still want to munge the encoding header in the
 631                 * result, which will be done by modifying the buffer. If we
 632                 * are using a fresh copy, we can reuse it. But if we are using
 633                 * the cached copy from commit->buffer, we need to duplicate it
 634                 * to avoid munging commit->buffer.
 635                 */
 636                out = msg;
 637                if (out == commit->buffer)
 638                        out = xstrdup(out);
 639        }
 640        else {
 641                /*
 642                 * There's actual encoding work to do. Do the reencoding, which
 643                 * still leaves the header to be replaced in the next step. At
 644                 * this point, we are done with msg. If we allocated a fresh
 645                 * copy, we can free it.
 646                 */
 647                out = reencode_string(msg, output_encoding, use_encoding);
 648                if (out && msg != commit->buffer)
 649                        free(msg);
 650        }
 651
 652        /*
 653         * This replacement actually consumes the buffer we hand it, so we do
 654         * not have to worry about freeing the old "out" here.
 655         */
 656        if (out)
 657                out = replace_encoding_header(out, output_encoding);
 658
 659        if (!commit_encoding)
 660                free(encoding);
 661        /*
 662         * If the re-encoding failed, out might be NULL here; in that
 663         * case we just return the commit message verbatim.
 664         */
 665        return out ? out : msg;
 666}
 667
 668void logmsg_free(char *msg, const struct commit *commit)
 669{
 670        if (msg != commit->buffer)
 671                free(msg);
 672}
 673
 674static int mailmap_name(const char **email, size_t *email_len,
 675                        const char **name, size_t *name_len)
 676{
 677        static struct string_list *mail_map;
 678        if (!mail_map) {
 679                mail_map = xcalloc(1, sizeof(*mail_map));
 680                read_mailmap(mail_map, NULL);
 681        }
 682        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 683}
 684
 685static size_t format_person_part(struct strbuf *sb, char part,
 686                                 const char *msg, int len, enum date_mode dmode)
 687{
 688        /* currently all placeholders have same length */
 689        const int placeholder_len = 2;
 690        struct ident_split s;
 691        const char *name, *mail;
 692        size_t maillen, namelen;
 693
 694        if (split_ident_line(&s, msg, len) < 0)
 695                goto skip;
 696
 697        name = s.name_begin;
 698        namelen = s.name_end - s.name_begin;
 699        mail = s.mail_begin;
 700        maillen = s.mail_end - s.mail_begin;
 701
 702        if (part == 'N' || part == 'E') /* mailmap lookup */
 703                mailmap_name(&mail, &maillen, &name, &namelen);
 704        if (part == 'n' || part == 'N') {       /* name */
 705                strbuf_add(sb, name, namelen);
 706                return placeholder_len;
 707        }
 708        if (part == 'e' || part == 'E') {       /* email */
 709                strbuf_add(sb, mail, maillen);
 710                return placeholder_len;
 711        }
 712
 713        if (!s.date_begin)
 714                goto skip;
 715
 716        if (part == 't') {      /* date, UNIX timestamp */
 717                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 718                return placeholder_len;
 719        }
 720
 721        switch (part) {
 722        case 'd':       /* date */
 723                strbuf_addstr(sb, show_ident_date(&s, dmode));
 724                return placeholder_len;
 725        case 'D':       /* date, RFC2822 style */
 726                strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
 727                return placeholder_len;
 728        case 'r':       /* date, relative */
 729                strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
 730                return placeholder_len;
 731        case 'i':       /* date, ISO 8601 */
 732                strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
 733                return placeholder_len;
 734        }
 735
 736skip:
 737        /*
 738         * reading from either a bogus commit, or a reflog entry with
 739         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 740         * to compute a valid return value.
 741         */
 742        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 743            || part == 'D' || part == 'r' || part == 'i')
 744                return placeholder_len;
 745
 746        return 0; /* unknown placeholder */
 747}
 748
 749struct chunk {
 750        size_t off;
 751        size_t len;
 752};
 753
 754enum flush_type {
 755        no_flush,
 756        flush_right,
 757        flush_left,
 758        flush_left_and_steal,
 759        flush_both
 760};
 761
 762enum trunc_type {
 763        trunc_none,
 764        trunc_left,
 765        trunc_middle,
 766        trunc_right
 767};
 768
 769struct format_commit_context {
 770        const struct commit *commit;
 771        const struct pretty_print_context *pretty_ctx;
 772        unsigned commit_header_parsed:1;
 773        unsigned commit_message_parsed:1;
 774        struct signature_check signature_check;
 775        enum flush_type flush_type;
 776        enum trunc_type truncate;
 777        char *message;
 778        char *commit_encoding;
 779        size_t width, indent1, indent2;
 780        int auto_color;
 781        int padding;
 782
 783        /* These offsets are relative to the start of the commit message. */
 784        struct chunk author;
 785        struct chunk committer;
 786        size_t message_off;
 787        size_t subject_off;
 788        size_t body_off;
 789
 790        /* The following ones are relative to the result struct strbuf. */
 791        struct chunk abbrev_commit_hash;
 792        struct chunk abbrev_tree_hash;
 793        struct chunk abbrev_parent_hashes;
 794        size_t wrap_start;
 795};
 796
 797static int add_again(struct strbuf *sb, struct chunk *chunk)
 798{
 799        if (chunk->len) {
 800                strbuf_adddup(sb, chunk->off, chunk->len);
 801                return 1;
 802        }
 803
 804        /*
 805         * We haven't seen this chunk before.  Our caller is surely
 806         * going to add it the hard way now.  Remember the most likely
 807         * start of the to-be-added chunk: the current end of the
 808         * struct strbuf.
 809         */
 810        chunk->off = sb->len;
 811        return 0;
 812}
 813
 814static void parse_commit_header(struct format_commit_context *context)
 815{
 816        const char *msg = context->message;
 817        int i;
 818
 819        for (i = 0; msg[i]; i++) {
 820                int eol;
 821                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 822                        ; /* do nothing */
 823
 824                if (i == eol) {
 825                        break;
 826                } else if (!prefixcmp(msg + i, "author ")) {
 827                        context->author.off = i + 7;
 828                        context->author.len = eol - i - 7;
 829                } else if (!prefixcmp(msg + i, "committer ")) {
 830                        context->committer.off = i + 10;
 831                        context->committer.len = eol - i - 10;
 832                }
 833                i = eol;
 834        }
 835        context->message_off = i;
 836        context->commit_header_parsed = 1;
 837}
 838
 839static int istitlechar(char c)
 840{
 841        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 842                (c >= '0' && c <= '9') || c == '.' || c == '_';
 843}
 844
 845static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 846{
 847        size_t trimlen;
 848        size_t start_len = sb->len;
 849        int space = 2;
 850
 851        for (; *msg && *msg != '\n'; msg++) {
 852                if (istitlechar(*msg)) {
 853                        if (space == 1)
 854                                strbuf_addch(sb, '-');
 855                        space = 0;
 856                        strbuf_addch(sb, *msg);
 857                        if (*msg == '.')
 858                                while (*(msg+1) == '.')
 859                                        msg++;
 860                } else
 861                        space |= 1;
 862        }
 863
 864        /* trim any trailing '.' or '-' characters */
 865        trimlen = 0;
 866        while (sb->len - trimlen > start_len &&
 867                (sb->buf[sb->len - 1 - trimlen] == '.'
 868                || sb->buf[sb->len - 1 - trimlen] == '-'))
 869                trimlen++;
 870        strbuf_remove(sb, sb->len - trimlen, trimlen);
 871}
 872
 873const char *format_subject(struct strbuf *sb, const char *msg,
 874                           const char *line_separator)
 875{
 876        int first = 1;
 877
 878        for (;;) {
 879                const char *line = msg;
 880                int linelen = get_one_line(line);
 881
 882                msg += linelen;
 883                if (!linelen || is_empty_line(line, &linelen))
 884                        break;
 885
 886                if (!sb)
 887                        continue;
 888                strbuf_grow(sb, linelen + 2);
 889                if (!first)
 890                        strbuf_addstr(sb, line_separator);
 891                strbuf_add(sb, line, linelen);
 892                first = 0;
 893        }
 894        return msg;
 895}
 896
 897static void parse_commit_message(struct format_commit_context *c)
 898{
 899        const char *msg = c->message + c->message_off;
 900        const char *start = c->message;
 901
 902        msg = skip_empty_lines(msg);
 903        c->subject_off = msg - start;
 904
 905        msg = format_subject(NULL, msg, NULL);
 906        msg = skip_empty_lines(msg);
 907        c->body_off = msg - start;
 908
 909        c->commit_message_parsed = 1;
 910}
 911
 912static void strbuf_wrap(struct strbuf *sb, size_t pos,
 913                        size_t width, size_t indent1, size_t indent2)
 914{
 915        struct strbuf tmp = STRBUF_INIT;
 916
 917        if (pos)
 918                strbuf_add(&tmp, sb->buf, pos);
 919        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 920                                (int) indent1, (int) indent2, (int) width);
 921        strbuf_swap(&tmp, sb);
 922        strbuf_release(&tmp);
 923}
 924
 925static void rewrap_message_tail(struct strbuf *sb,
 926                                struct format_commit_context *c,
 927                                size_t new_width, size_t new_indent1,
 928                                size_t new_indent2)
 929{
 930        if (c->width == new_width && c->indent1 == new_indent1 &&
 931            c->indent2 == new_indent2)
 932                return;
 933        if (c->wrap_start < sb->len)
 934                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 935        c->wrap_start = sb->len;
 936        c->width = new_width;
 937        c->indent1 = new_indent1;
 938        c->indent2 = new_indent2;
 939}
 940
 941static int format_reflog_person(struct strbuf *sb,
 942                                char part,
 943                                struct reflog_walk_info *log,
 944                                enum date_mode dmode)
 945{
 946        const char *ident;
 947
 948        if (!log)
 949                return 2;
 950
 951        ident = get_reflog_ident(log);
 952        if (!ident)
 953                return 2;
 954
 955        return format_person_part(sb, part, ident, strlen(ident), dmode);
 956}
 957
 958static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 959                          const char *placeholder,
 960                          struct format_commit_context *c)
 961{
 962        if (placeholder[1] == '(') {
 963                const char *begin = placeholder + 2;
 964                const char *end = strchr(begin, ')');
 965                char color[COLOR_MAXLEN];
 966
 967                if (!end)
 968                        return 0;
 969                if (!prefixcmp(begin, "auto,")) {
 970                        if (!want_color(c->pretty_ctx->color))
 971                                return end - placeholder + 1;
 972                        begin += 5;
 973                }
 974                color_parse_mem(begin,
 975                                end - begin,
 976                                "--pretty format", color);
 977                strbuf_addstr(sb, color);
 978                return end - placeholder + 1;
 979        }
 980        if (!prefixcmp(placeholder + 1, "red")) {
 981                strbuf_addstr(sb, GIT_COLOR_RED);
 982                return 4;
 983        } else if (!prefixcmp(placeholder + 1, "green")) {
 984                strbuf_addstr(sb, GIT_COLOR_GREEN);
 985                return 6;
 986        } else if (!prefixcmp(placeholder + 1, "blue")) {
 987                strbuf_addstr(sb, GIT_COLOR_BLUE);
 988                return 5;
 989        } else if (!prefixcmp(placeholder + 1, "reset")) {
 990                strbuf_addstr(sb, GIT_COLOR_RESET);
 991                return 6;
 992        } else
 993                return 0;
 994}
 995
 996static size_t parse_padding_placeholder(struct strbuf *sb,
 997                                        const char *placeholder,
 998                                        struct format_commit_context *c)
 999{
1000        const char *ch = placeholder;
1001        enum flush_type flush_type;
1002        int to_column = 0;
1003
1004        switch (*ch++) {
1005        case '<':
1006                flush_type = flush_right;
1007                break;
1008        case '>':
1009                if (*ch == '<') {
1010                        flush_type = flush_both;
1011                        ch++;
1012                } else if (*ch == '>') {
1013                        flush_type = flush_left_and_steal;
1014                        ch++;
1015                } else
1016                        flush_type = flush_left;
1017                break;
1018        default:
1019                return 0;
1020        }
1021
1022        /* the next value means "wide enough to that column" */
1023        if (*ch == '|') {
1024                to_column = 1;
1025                ch++;
1026        }
1027
1028        if (*ch == '(') {
1029                const char *start = ch + 1;
1030                const char *end = start + strcspn(start, ",)");
1031                char *next;
1032                int width;
1033                if (!end || end == start)
1034                        return 0;
1035                width = strtoul(start, &next, 10);
1036                if (next == start || width == 0)
1037                        return 0;
1038                c->padding = to_column ? -width : width;
1039                c->flush_type = flush_type;
1040
1041                if (*end == ',') {
1042                        start = end + 1;
1043                        end = strchr(start, ')');
1044                        if (!end || end == start)
1045                                return 0;
1046                        if (!prefixcmp(start, "trunc)"))
1047                                c->truncate = trunc_right;
1048                        else if (!prefixcmp(start, "ltrunc)"))
1049                                c->truncate = trunc_left;
1050                        else if (!prefixcmp(start, "mtrunc)"))
1051                                c->truncate = trunc_middle;
1052                        else
1053                                return 0;
1054                } else
1055                        c->truncate = trunc_none;
1056
1057                return end - placeholder + 1;
1058        }
1059        return 0;
1060}
1061
1062static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1063                                const char *placeholder,
1064                                void *context)
1065{
1066        struct format_commit_context *c = context;
1067        const struct commit *commit = c->commit;
1068        const char *msg = c->message;
1069        struct commit_list *p;
1070        int h1, h2;
1071
1072        /* these are independent of the commit */
1073        switch (placeholder[0]) {
1074        case 'C':
1075                if (!prefixcmp(placeholder + 1, "(auto)")) {
1076                        c->auto_color = 1;
1077                        return 7; /* consumed 7 bytes, "C(auto)" */
1078                } else {
1079                        int ret = parse_color(sb, placeholder, c);
1080                        if (ret)
1081                                c->auto_color = 0;
1082                        /*
1083                         * Otherwise, we decided to treat %C<unknown>
1084                         * as a literal string, and the previous
1085                         * %C(auto) is still valid.
1086                         */
1087                        return ret;
1088                }
1089        case 'n':               /* newline */
1090                strbuf_addch(sb, '\n');
1091                return 1;
1092        case 'x':
1093                /* %x00 == NUL, %x0a == LF, etc. */
1094                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1095                    h1 <= 16 &&
1096                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1097                    h2 <= 16) {
1098                        strbuf_addch(sb, (h1<<4)|h2);
1099                        return 3;
1100                } else
1101                        return 0;
1102        case 'w':
1103                if (placeholder[1] == '(') {
1104                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1105                        char *next;
1106                        const char *start = placeholder + 2;
1107                        const char *end = strchr(start, ')');
1108                        if (!end)
1109                                return 0;
1110                        if (end > start) {
1111                                width = strtoul(start, &next, 10);
1112                                if (*next == ',') {
1113                                        indent1 = strtoul(next + 1, &next, 10);
1114                                        if (*next == ',') {
1115                                                indent2 = strtoul(next + 1,
1116                                                                 &next, 10);
1117                                        }
1118                                }
1119                                if (*next != ')')
1120                                        return 0;
1121                        }
1122                        rewrap_message_tail(sb, c, width, indent1, indent2);
1123                        return end - placeholder + 1;
1124                } else
1125                        return 0;
1126
1127        case '<':
1128        case '>':
1129                return parse_padding_placeholder(sb, placeholder, c);
1130        }
1131
1132        /* these depend on the commit */
1133        if (!commit->object.parsed)
1134                parse_object(commit->object.sha1);
1135
1136        switch (placeholder[0]) {
1137        case 'H':               /* commit hash */
1138                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1139                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1140                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1141                return 1;
1142        case 'h':               /* abbreviated commit hash */
1143                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1144                if (add_again(sb, &c->abbrev_commit_hash)) {
1145                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1146                        return 1;
1147                }
1148                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1149                                                     c->pretty_ctx->abbrev));
1150                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1151                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1152                return 1;
1153        case 'T':               /* tree hash */
1154                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1155                return 1;
1156        case 't':               /* abbreviated tree hash */
1157                if (add_again(sb, &c->abbrev_tree_hash))
1158                        return 1;
1159                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1160                                                     c->pretty_ctx->abbrev));
1161                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1162                return 1;
1163        case 'P':               /* parent hashes */
1164                for (p = commit->parents; p; p = p->next) {
1165                        if (p != commit->parents)
1166                                strbuf_addch(sb, ' ');
1167                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1168                }
1169                return 1;
1170        case 'p':               /* abbreviated parent hashes */
1171                if (add_again(sb, &c->abbrev_parent_hashes))
1172                        return 1;
1173                for (p = commit->parents; p; p = p->next) {
1174                        if (p != commit->parents)
1175                                strbuf_addch(sb, ' ');
1176                        strbuf_addstr(sb, find_unique_abbrev(
1177                                        p->item->object.sha1,
1178                                        c->pretty_ctx->abbrev));
1179                }
1180                c->abbrev_parent_hashes.len = sb->len -
1181                                              c->abbrev_parent_hashes.off;
1182                return 1;
1183        case 'm':               /* left/right/bottom */
1184                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1185                return 1;
1186        case 'd':
1187                load_ref_decorations(DECORATE_SHORT_REFS);
1188                format_decorations(sb, commit, c->auto_color);
1189                return 1;
1190        case 'g':               /* reflog info */
1191                switch(placeholder[1]) {
1192                case 'd':       /* reflog selector */
1193                case 'D':
1194                        if (c->pretty_ctx->reflog_info)
1195                                get_reflog_selector(sb,
1196                                                    c->pretty_ctx->reflog_info,
1197                                                    c->pretty_ctx->date_mode,
1198                                                    c->pretty_ctx->date_mode_explicit,
1199                                                    (placeholder[1] == 'd'));
1200                        return 2;
1201                case 's':       /* reflog message */
1202                        if (c->pretty_ctx->reflog_info)
1203                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1204                        return 2;
1205                case 'n':
1206                case 'N':
1207                case 'e':
1208                case 'E':
1209                        return format_reflog_person(sb,
1210                                                    placeholder[1],
1211                                                    c->pretty_ctx->reflog_info,
1212                                                    c->pretty_ctx->date_mode);
1213                }
1214                return 0;       /* unknown %g placeholder */
1215        case 'N':
1216                if (c->pretty_ctx->notes_message) {
1217                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1218                        return 1;
1219                }
1220                return 0;
1221        }
1222
1223        if (placeholder[0] == 'G') {
1224                if (!c->signature_check.result)
1225                        check_commit_signature(c->commit, &(c->signature_check));
1226                switch (placeholder[1]) {
1227                case 'G':
1228                        if (c->signature_check.gpg_output)
1229                                strbuf_addstr(sb, c->signature_check.gpg_output);
1230                        break;
1231                case '?':
1232                        switch (c->signature_check.result) {
1233                        case 'G':
1234                        case 'B':
1235                        case 'U':
1236                        case 'N':
1237                                strbuf_addch(sb, c->signature_check.result);
1238                        }
1239                        break;
1240                case 'S':
1241                        if (c->signature_check.signer)
1242                                strbuf_addstr(sb, c->signature_check.signer);
1243                        break;
1244                case 'K':
1245                        if (c->signature_check.key)
1246                                strbuf_addstr(sb, c->signature_check.key);
1247                        break;
1248                }
1249                return 2;
1250        }
1251
1252
1253        /* For the rest we have to parse the commit header. */
1254        if (!c->commit_header_parsed)
1255                parse_commit_header(c);
1256
1257        switch (placeholder[0]) {
1258        case 'a':       /* author ... */
1259                return format_person_part(sb, placeholder[1],
1260                                   msg + c->author.off, c->author.len,
1261                                   c->pretty_ctx->date_mode);
1262        case 'c':       /* committer ... */
1263                return format_person_part(sb, placeholder[1],
1264                                   msg + c->committer.off, c->committer.len,
1265                                   c->pretty_ctx->date_mode);
1266        case 'e':       /* encoding */
1267                if (c->commit_encoding)
1268                        strbuf_addstr(sb, c->commit_encoding);
1269                return 1;
1270        case 'B':       /* raw body */
1271                /* message_off is always left at the initial newline */
1272                strbuf_addstr(sb, msg + c->message_off + 1);
1273                return 1;
1274        }
1275
1276        /* Now we need to parse the commit message. */
1277        if (!c->commit_message_parsed)
1278                parse_commit_message(c);
1279
1280        switch (placeholder[0]) {
1281        case 's':       /* subject */
1282                format_subject(sb, msg + c->subject_off, " ");
1283                return 1;
1284        case 'f':       /* sanitized subject */
1285                format_sanitized_subject(sb, msg + c->subject_off);
1286                return 1;
1287        case 'b':       /* body */
1288                strbuf_addstr(sb, msg + c->body_off);
1289                return 1;
1290        }
1291        return 0;       /* unknown placeholder */
1292}
1293
1294static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1295                                    const char *placeholder,
1296                                    struct format_commit_context *c)
1297{
1298        struct strbuf local_sb = STRBUF_INIT;
1299        int total_consumed = 0, len, padding = c->padding;
1300        if (padding < 0) {
1301                const char *start = strrchr(sb->buf, '\n');
1302                int occupied;
1303                if (!start)
1304                        start = sb->buf;
1305                occupied = utf8_strnwidth(start, -1, 1);
1306                padding = (-padding) - occupied;
1307        }
1308        while (1) {
1309                int modifier = *placeholder == 'C';
1310                int consumed = format_commit_one(&local_sb, placeholder, c);
1311                total_consumed += consumed;
1312
1313                if (!modifier)
1314                        break;
1315
1316                placeholder += consumed;
1317                if (*placeholder != '%')
1318                        break;
1319                placeholder++;
1320                total_consumed++;
1321        }
1322        len = utf8_strnwidth(local_sb.buf, -1, 1);
1323
1324        if (c->flush_type == flush_left_and_steal) {
1325                const char *ch = sb->buf + sb->len - 1;
1326                while (len > padding && ch > sb->buf) {
1327                        const char *p;
1328                        if (*ch == ' ') {
1329                                ch--;
1330                                padding++;
1331                                continue;
1332                        }
1333                        /* check for trailing ansi sequences */
1334                        if (*ch != 'm')
1335                                break;
1336                        p = ch - 1;
1337                        while (ch - p < 10 && *p != '\033')
1338                                p--;
1339                        if (*p != '\033' ||
1340                            ch + 1 - p != display_mode_esc_sequence_len(p))
1341                                break;
1342                        /*
1343                         * got a good ansi sequence, put it back to
1344                         * local_sb as we're cutting sb
1345                         */
1346                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1347                        ch = p - 1;
1348                }
1349                strbuf_setlen(sb, ch + 1 - sb->buf);
1350                c->flush_type = flush_left;
1351        }
1352
1353        if (len > padding) {
1354                switch (c->truncate) {
1355                case trunc_left:
1356                        strbuf_utf8_replace(&local_sb,
1357                                            0, len - (padding - 2),
1358                                            "..");
1359                        break;
1360                case trunc_middle:
1361                        strbuf_utf8_replace(&local_sb,
1362                                            padding / 2 - 1,
1363                                            len - (padding - 2),
1364                                            "..");
1365                        break;
1366                case trunc_right:
1367                        strbuf_utf8_replace(&local_sb,
1368                                            padding - 2, len - (padding - 2),
1369                                            "..");
1370                        break;
1371                case trunc_none:
1372                        break;
1373                }
1374                strbuf_addstr(sb, local_sb.buf);
1375        } else {
1376                int sb_len = sb->len, offset = 0;
1377                if (c->flush_type == flush_left)
1378                        offset = padding - len;
1379                else if (c->flush_type == flush_both)
1380                        offset = (padding - len) / 2;
1381                /*
1382                 * we calculate padding in columns, now
1383                 * convert it back to chars
1384                 */
1385                padding = padding - len + local_sb.len;
1386                strbuf_grow(sb, padding);
1387                strbuf_setlen(sb, sb_len + padding);
1388                memset(sb->buf + sb_len, ' ', sb->len - sb_len);
1389                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1390                       local_sb.len);
1391        }
1392        strbuf_release(&local_sb);
1393        c->flush_type = no_flush;
1394        return total_consumed;
1395}
1396
1397static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1398                                 const char *placeholder,
1399                                 void *context)
1400{
1401        int consumed;
1402        size_t orig_len;
1403        enum {
1404                NO_MAGIC,
1405                ADD_LF_BEFORE_NON_EMPTY,
1406                DEL_LF_BEFORE_EMPTY,
1407                ADD_SP_BEFORE_NON_EMPTY
1408        } magic = NO_MAGIC;
1409
1410        switch (placeholder[0]) {
1411        case '-':
1412                magic = DEL_LF_BEFORE_EMPTY;
1413                break;
1414        case '+':
1415                magic = ADD_LF_BEFORE_NON_EMPTY;
1416                break;
1417        case ' ':
1418                magic = ADD_SP_BEFORE_NON_EMPTY;
1419                break;
1420        default:
1421                break;
1422        }
1423        if (magic != NO_MAGIC)
1424                placeholder++;
1425
1426        orig_len = sb->len;
1427        if (((struct format_commit_context *)context)->flush_type != no_flush)
1428                consumed = format_and_pad_commit(sb, placeholder, context);
1429        else
1430                consumed = format_commit_one(sb, placeholder, context);
1431        if (magic == NO_MAGIC)
1432                return consumed;
1433
1434        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1435                while (sb->len && sb->buf[sb->len - 1] == '\n')
1436                        strbuf_setlen(sb, sb->len - 1);
1437        } else if (orig_len != sb->len) {
1438                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1439                        strbuf_insert(sb, orig_len, "\n", 1);
1440                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1441                        strbuf_insert(sb, orig_len, " ", 1);
1442        }
1443        return consumed + 1;
1444}
1445
1446static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1447                                   void *context)
1448{
1449        struct userformat_want *w = context;
1450
1451        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1452                placeholder++;
1453
1454        switch (*placeholder) {
1455        case 'N':
1456                w->notes = 1;
1457                break;
1458        }
1459        return 0;
1460}
1461
1462void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1463{
1464        struct strbuf dummy = STRBUF_INIT;
1465
1466        if (!fmt) {
1467                if (!user_format)
1468                        return;
1469                fmt = user_format;
1470        }
1471        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1472        strbuf_release(&dummy);
1473}
1474
1475void format_commit_message(const struct commit *commit,
1476                           const char *format, struct strbuf *sb,
1477                           const struct pretty_print_context *pretty_ctx)
1478{
1479        struct format_commit_context context;
1480        const char *output_enc = pretty_ctx->output_encoding;
1481        const char *utf8 = "UTF-8";
1482
1483        memset(&context, 0, sizeof(context));
1484        context.commit = commit;
1485        context.pretty_ctx = pretty_ctx;
1486        context.wrap_start = sb->len;
1487        context.message = logmsg_reencode(commit,
1488                                          &context.commit_encoding,
1489                                          output_enc);
1490
1491        strbuf_expand(sb, format, format_commit_item, &context);
1492        rewrap_message_tail(sb, &context, 0, 0, 0);
1493
1494        if (output_enc) {
1495                if (same_encoding(utf8, output_enc))
1496                        output_enc = NULL;
1497        } else {
1498                if (context.commit_encoding &&
1499                    !same_encoding(context.commit_encoding, utf8))
1500                        output_enc = context.commit_encoding;
1501        }
1502
1503        if (output_enc) {
1504                int outsz;
1505                char *out = reencode_string_len(sb->buf, sb->len,
1506                                                output_enc, utf8, &outsz);
1507                if (out)
1508                        strbuf_attach(sb, out, outsz, outsz + 1);
1509        }
1510
1511        free(context.commit_encoding);
1512        logmsg_free(context.message, commit);
1513        free(context.signature_check.gpg_output);
1514        free(context.signature_check.signer);
1515}
1516
1517static void pp_header(const struct pretty_print_context *pp,
1518                      const char *encoding,
1519                      const struct commit *commit,
1520                      const char **msg_p,
1521                      struct strbuf *sb)
1522{
1523        int parents_shown = 0;
1524
1525        for (;;) {
1526                const char *line = *msg_p;
1527                int linelen = get_one_line(*msg_p);
1528
1529                if (!linelen)
1530                        return;
1531                *msg_p += linelen;
1532
1533                if (linelen == 1)
1534                        /* End of header */
1535                        return;
1536
1537                if (pp->fmt == CMIT_FMT_RAW) {
1538                        strbuf_add(sb, line, linelen);
1539                        continue;
1540                }
1541
1542                if (!prefixcmp(line, "parent ")) {
1543                        if (linelen != 48)
1544                                die("bad parent line in commit");
1545                        continue;
1546                }
1547
1548                if (!parents_shown) {
1549                        struct commit_list *parent;
1550                        int num;
1551                        for (parent = commit->parents, num = 0;
1552                             parent;
1553                             parent = parent->next, num++)
1554                                ;
1555                        /* with enough slop */
1556                        strbuf_grow(sb, num * 50 + 20);
1557                        add_merge_info(pp, sb, commit);
1558                        parents_shown = 1;
1559                }
1560
1561                /*
1562                 * MEDIUM == DEFAULT shows only author with dates.
1563                 * FULL shows both authors but not dates.
1564                 * FULLER shows both authors and dates.
1565                 */
1566                if (!prefixcmp(line, "author ")) {
1567                        strbuf_grow(sb, linelen + 80);
1568                        pp_user_info(pp, "Author", sb, line + 7, encoding);
1569                }
1570                if (!prefixcmp(line, "committer ") &&
1571                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1572                        strbuf_grow(sb, linelen + 80);
1573                        pp_user_info(pp, "Commit", sb, line + 10, encoding);
1574                }
1575        }
1576}
1577
1578void pp_title_line(const struct pretty_print_context *pp,
1579                   const char **msg_p,
1580                   struct strbuf *sb,
1581                   const char *encoding,
1582                   int need_8bit_cte)
1583{
1584        static const int max_length = 78; /* per rfc2047 */
1585        struct strbuf title;
1586
1587        strbuf_init(&title, 80);
1588        *msg_p = format_subject(&title, *msg_p,
1589                                pp->preserve_subject ? "\n" : " ");
1590
1591        strbuf_grow(sb, title.len + 1024);
1592        if (pp->subject) {
1593                strbuf_addstr(sb, pp->subject);
1594                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1595                        add_rfc2047(sb, title.buf, title.len,
1596                                                encoding, RFC2047_SUBJECT);
1597                else
1598                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1599                                         -last_line_length(sb), 1, max_length);
1600        } else {
1601                strbuf_addbuf(sb, &title);
1602        }
1603        strbuf_addch(sb, '\n');
1604
1605        if (need_8bit_cte > 0) {
1606                const char *header_fmt =
1607                        "MIME-Version: 1.0\n"
1608                        "Content-Type: text/plain; charset=%s\n"
1609                        "Content-Transfer-Encoding: 8bit\n";
1610                strbuf_addf(sb, header_fmt, encoding);
1611        }
1612        if (pp->after_subject) {
1613                strbuf_addstr(sb, pp->after_subject);
1614        }
1615        if (pp->fmt == CMIT_FMT_EMAIL) {
1616                strbuf_addch(sb, '\n');
1617        }
1618        strbuf_release(&title);
1619}
1620
1621void pp_remainder(const struct pretty_print_context *pp,
1622                  const char **msg_p,
1623                  struct strbuf *sb,
1624                  int indent)
1625{
1626        int first = 1;
1627        for (;;) {
1628                const char *line = *msg_p;
1629                int linelen = get_one_line(line);
1630                *msg_p += linelen;
1631
1632                if (!linelen)
1633                        break;
1634
1635                if (is_empty_line(line, &linelen)) {
1636                        if (first)
1637                                continue;
1638                        if (pp->fmt == CMIT_FMT_SHORT)
1639                                break;
1640                }
1641                first = 0;
1642
1643                strbuf_grow(sb, linelen + indent + 20);
1644                if (indent) {
1645                        memset(sb->buf + sb->len, ' ', indent);
1646                        strbuf_setlen(sb, sb->len + indent);
1647                }
1648                strbuf_add(sb, line, linelen);
1649                strbuf_addch(sb, '\n');
1650        }
1651}
1652
1653void pretty_print_commit(const struct pretty_print_context *pp,
1654                         const struct commit *commit,
1655                         struct strbuf *sb)
1656{
1657        unsigned long beginning_of_body;
1658        int indent = 4;
1659        const char *msg;
1660        char *reencoded;
1661        const char *encoding;
1662        int need_8bit_cte = pp->need_8bit_cte;
1663
1664        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1665                format_commit_message(commit, user_format, sb, pp);
1666                return;
1667        }
1668
1669        encoding = get_log_output_encoding();
1670        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1671
1672        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1673                indent = 0;
1674
1675        /*
1676         * We need to check and emit Content-type: to mark it
1677         * as 8-bit if we haven't done so.
1678         */
1679        if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1680                int i, ch, in_body;
1681
1682                for (in_body = i = 0; (ch = msg[i]); i++) {
1683                        if (!in_body) {
1684                                /* author could be non 7-bit ASCII but
1685                                 * the log may be so; skip over the
1686                                 * header part first.
1687                                 */
1688                                if (ch == '\n' && msg[i+1] == '\n')
1689                                        in_body = 1;
1690                        }
1691                        else if (non_ascii(ch)) {
1692                                need_8bit_cte = 1;
1693                                break;
1694                        }
1695                }
1696        }
1697
1698        pp_header(pp, encoding, commit, &msg, sb);
1699        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1700                strbuf_addch(sb, '\n');
1701        }
1702
1703        /* Skip excess blank lines at the beginning of body, if any... */
1704        msg = skip_empty_lines(msg);
1705
1706        /* These formats treat the title line specially. */
1707        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1708                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1709
1710        beginning_of_body = sb->len;
1711        if (pp->fmt != CMIT_FMT_ONELINE)
1712                pp_remainder(pp, &msg, sb, indent);
1713        strbuf_rtrim(sb);
1714
1715        /* Make sure there is an EOLN for the non-oneline case */
1716        if (pp->fmt != CMIT_FMT_ONELINE)
1717                strbuf_addch(sb, '\n');
1718
1719        /*
1720         * The caller may append additional body text in e-mail
1721         * format.  Make sure we did not strip the blank line
1722         * between the header and the body.
1723         */
1724        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1725                strbuf_addch(sb, '\n');
1726
1727        logmsg_free(reencoded, commit);
1728}
1729
1730void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1731                    struct strbuf *sb)
1732{
1733        struct pretty_print_context pp = {0};
1734        pp.fmt = fmt;
1735        pretty_print_commit(&pp, commit, sb);
1736}