pretty.con commit pretty: remove intermediate strbufs from pp_user_info() (a0511b3)
   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                      const char *output_encoding)
 589{
 590        static const char *utf8 = "UTF-8";
 591        const char *use_encoding;
 592        char *encoding;
 593        char *msg = commit->buffer;
 594        char *out;
 595
 596        if (!msg) {
 597                enum object_type type;
 598                unsigned long size;
 599
 600                msg = read_sha1_file(commit->object.sha1, &type, &size);
 601                if (!msg)
 602                        die("Cannot read commit object %s",
 603                            sha1_to_hex(commit->object.sha1));
 604                if (type != OBJ_COMMIT)
 605                        die("Expected commit for '%s', got %s",
 606                            sha1_to_hex(commit->object.sha1), typename(type));
 607        }
 608
 609        if (!output_encoding || !*output_encoding)
 610                return msg;
 611        encoding = get_header(commit, msg, "encoding");
 612        use_encoding = encoding ? encoding : utf8;
 613        if (same_encoding(use_encoding, output_encoding)) {
 614                /*
 615                 * No encoding work to be done. If we have no encoding header
 616                 * at all, then there's nothing to do, and we can return the
 617                 * message verbatim (whether newly allocated or not).
 618                 */
 619                if (!encoding)
 620                        return msg;
 621
 622                /*
 623                 * Otherwise, we still want to munge the encoding header in the
 624                 * result, which will be done by modifying the buffer. If we
 625                 * are using a fresh copy, we can reuse it. But if we are using
 626                 * the cached copy from commit->buffer, we need to duplicate it
 627                 * to avoid munging commit->buffer.
 628                 */
 629                out = msg;
 630                if (out == commit->buffer)
 631                        out = xstrdup(out);
 632        }
 633        else {
 634                /*
 635                 * There's actual encoding work to do. Do the reencoding, which
 636                 * still leaves the header to be replaced in the next step. At
 637                 * this point, we are done with msg. If we allocated a fresh
 638                 * copy, we can free it.
 639                 */
 640                out = reencode_string(msg, output_encoding, use_encoding);
 641                if (out && msg != commit->buffer)
 642                        free(msg);
 643        }
 644
 645        /*
 646         * This replacement actually consumes the buffer we hand it, so we do
 647         * not have to worry about freeing the old "out" here.
 648         */
 649        if (out)
 650                out = replace_encoding_header(out, output_encoding);
 651
 652        free(encoding);
 653        /*
 654         * If the re-encoding failed, out might be NULL here; in that
 655         * case we just return the commit message verbatim.
 656         */
 657        return out ? out : msg;
 658}
 659
 660void logmsg_free(char *msg, const struct commit *commit)
 661{
 662        if (msg != commit->buffer)
 663                free(msg);
 664}
 665
 666static int mailmap_name(const char **email, size_t *email_len,
 667                        const char **name, size_t *name_len)
 668{
 669        static struct string_list *mail_map;
 670        if (!mail_map) {
 671                mail_map = xcalloc(1, sizeof(*mail_map));
 672                read_mailmap(mail_map, NULL);
 673        }
 674        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 675}
 676
 677static size_t format_person_part(struct strbuf *sb, char part,
 678                                 const char *msg, int len, enum date_mode dmode)
 679{
 680        /* currently all placeholders have same length */
 681        const int placeholder_len = 2;
 682        struct ident_split s;
 683        const char *name, *mail;
 684        size_t maillen, namelen;
 685
 686        if (split_ident_line(&s, msg, len) < 0)
 687                goto skip;
 688
 689        name = s.name_begin;
 690        namelen = s.name_end - s.name_begin;
 691        mail = s.mail_begin;
 692        maillen = s.mail_end - s.mail_begin;
 693
 694        if (part == 'N' || part == 'E') /* mailmap lookup */
 695                mailmap_name(&mail, &maillen, &name, &namelen);
 696        if (part == 'n' || part == 'N') {       /* name */
 697                strbuf_add(sb, name, namelen);
 698                return placeholder_len;
 699        }
 700        if (part == 'e' || part == 'E') {       /* email */
 701                strbuf_add(sb, mail, maillen);
 702                return placeholder_len;
 703        }
 704
 705        if (!s.date_begin)
 706                goto skip;
 707
 708        if (part == 't') {      /* date, UNIX timestamp */
 709                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 710                return placeholder_len;
 711        }
 712
 713        switch (part) {
 714        case 'd':       /* date */
 715                strbuf_addstr(sb, show_ident_date(&s, dmode));
 716                return placeholder_len;
 717        case 'D':       /* date, RFC2822 style */
 718                strbuf_addstr(sb, show_ident_date(&s, DATE_RFC2822));
 719                return placeholder_len;
 720        case 'r':       /* date, relative */
 721                strbuf_addstr(sb, show_ident_date(&s, DATE_RELATIVE));
 722                return placeholder_len;
 723        case 'i':       /* date, ISO 8601 */
 724                strbuf_addstr(sb, show_ident_date(&s, DATE_ISO8601));
 725                return placeholder_len;
 726        }
 727
 728skip:
 729        /*
 730         * reading from either a bogus commit, or a reflog entry with
 731         * %gn, %ge, etc.; 'sb' cannot be updated, but we still need
 732         * to compute a valid return value.
 733         */
 734        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 735            || part == 'D' || part == 'r' || part == 'i')
 736                return placeholder_len;
 737
 738        return 0; /* unknown placeholder */
 739}
 740
 741struct chunk {
 742        size_t off;
 743        size_t len;
 744};
 745
 746struct format_commit_context {
 747        const struct commit *commit;
 748        const struct pretty_print_context *pretty_ctx;
 749        unsigned commit_header_parsed:1;
 750        unsigned commit_message_parsed:1;
 751        unsigned commit_signature_parsed:1;
 752        struct {
 753                char *gpg_output;
 754                char *gpg_status;
 755                char good_bad;
 756                char *signer;
 757                char *key;
 758        } signature;
 759        char *message;
 760        size_t width, indent1, indent2;
 761
 762        /* These offsets are relative to the start of the commit message. */
 763        struct chunk author;
 764        struct chunk committer;
 765        struct chunk encoding;
 766        size_t message_off;
 767        size_t subject_off;
 768        size_t body_off;
 769
 770        /* The following ones are relative to the result struct strbuf. */
 771        struct chunk abbrev_commit_hash;
 772        struct chunk abbrev_tree_hash;
 773        struct chunk abbrev_parent_hashes;
 774        size_t wrap_start;
 775};
 776
 777static int add_again(struct strbuf *sb, struct chunk *chunk)
 778{
 779        if (chunk->len) {
 780                strbuf_adddup(sb, chunk->off, chunk->len);
 781                return 1;
 782        }
 783
 784        /*
 785         * We haven't seen this chunk before.  Our caller is surely
 786         * going to add it the hard way now.  Remember the most likely
 787         * start of the to-be-added chunk: the current end of the
 788         * struct strbuf.
 789         */
 790        chunk->off = sb->len;
 791        return 0;
 792}
 793
 794static void parse_commit_header(struct format_commit_context *context)
 795{
 796        const char *msg = context->message;
 797        int i;
 798
 799        for (i = 0; msg[i]; i++) {
 800                int eol;
 801                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 802                        ; /* do nothing */
 803
 804                if (i == eol) {
 805                        break;
 806                } else if (!prefixcmp(msg + i, "author ")) {
 807                        context->author.off = i + 7;
 808                        context->author.len = eol - i - 7;
 809                } else if (!prefixcmp(msg + i, "committer ")) {
 810                        context->committer.off = i + 10;
 811                        context->committer.len = eol - i - 10;
 812                } else if (!prefixcmp(msg + i, "encoding ")) {
 813                        context->encoding.off = i + 9;
 814                        context->encoding.len = eol - i - 9;
 815                }
 816                i = eol;
 817        }
 818        context->message_off = i;
 819        context->commit_header_parsed = 1;
 820}
 821
 822static int istitlechar(char c)
 823{
 824        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 825                (c >= '0' && c <= '9') || c == '.' || c == '_';
 826}
 827
 828static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 829{
 830        size_t trimlen;
 831        size_t start_len = sb->len;
 832        int space = 2;
 833
 834        for (; *msg && *msg != '\n'; msg++) {
 835                if (istitlechar(*msg)) {
 836                        if (space == 1)
 837                                strbuf_addch(sb, '-');
 838                        space = 0;
 839                        strbuf_addch(sb, *msg);
 840                        if (*msg == '.')
 841                                while (*(msg+1) == '.')
 842                                        msg++;
 843                } else
 844                        space |= 1;
 845        }
 846
 847        /* trim any trailing '.' or '-' characters */
 848        trimlen = 0;
 849        while (sb->len - trimlen > start_len &&
 850                (sb->buf[sb->len - 1 - trimlen] == '.'
 851                || sb->buf[sb->len - 1 - trimlen] == '-'))
 852                trimlen++;
 853        strbuf_remove(sb, sb->len - trimlen, trimlen);
 854}
 855
 856const char *format_subject(struct strbuf *sb, const char *msg,
 857                           const char *line_separator)
 858{
 859        int first = 1;
 860
 861        for (;;) {
 862                const char *line = msg;
 863                int linelen = get_one_line(line);
 864
 865                msg += linelen;
 866                if (!linelen || is_empty_line(line, &linelen))
 867                        break;
 868
 869                if (!sb)
 870                        continue;
 871                strbuf_grow(sb, linelen + 2);
 872                if (!first)
 873                        strbuf_addstr(sb, line_separator);
 874                strbuf_add(sb, line, linelen);
 875                first = 0;
 876        }
 877        return msg;
 878}
 879
 880static void parse_commit_message(struct format_commit_context *c)
 881{
 882        const char *msg = c->message + c->message_off;
 883        const char *start = c->message;
 884
 885        msg = skip_empty_lines(msg);
 886        c->subject_off = msg - start;
 887
 888        msg = format_subject(NULL, msg, NULL);
 889        msg = skip_empty_lines(msg);
 890        c->body_off = msg - start;
 891
 892        c->commit_message_parsed = 1;
 893}
 894
 895static void format_decoration(struct strbuf *sb, const struct commit *commit)
 896{
 897        struct name_decoration *d;
 898        const char *prefix = " (";
 899
 900        load_ref_decorations(DECORATE_SHORT_REFS);
 901        d = lookup_decoration(&name_decoration, &commit->object);
 902        while (d) {
 903                strbuf_addstr(sb, prefix);
 904                prefix = ", ";
 905                strbuf_addstr(sb, d->name);
 906                d = d->next;
 907        }
 908        if (prefix[0] == ',')
 909                strbuf_addch(sb, ')');
 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 struct {
 942        char result;
 943        const char *check;
 944} signature_check[] = {
 945        { 'G', "\n[GNUPG:] GOODSIG " },
 946        { 'B', "\n[GNUPG:] BADSIG " },
 947};
 948
 949static void parse_signature_lines(struct format_commit_context *ctx)
 950{
 951        const char *buf = ctx->signature.gpg_status;
 952        int i;
 953
 954        for (i = 0; i < ARRAY_SIZE(signature_check); i++) {
 955                const char *found = strstr(buf, signature_check[i].check);
 956                const char *next;
 957                if (!found)
 958                        continue;
 959                ctx->signature.good_bad = signature_check[i].result;
 960                found += strlen(signature_check[i].check);
 961                ctx->signature.key = xmemdupz(found, 16);
 962                found += 17;
 963                next = strchrnul(found, '\n');
 964                ctx->signature.signer = xmemdupz(found, next - found);
 965                break;
 966        }
 967}
 968
 969static void parse_commit_signature(struct format_commit_context *ctx)
 970{
 971        struct strbuf payload = STRBUF_INIT;
 972        struct strbuf signature = STRBUF_INIT;
 973        struct strbuf gpg_output = STRBUF_INIT;
 974        struct strbuf gpg_status = STRBUF_INIT;
 975        int status;
 976
 977        ctx->commit_signature_parsed = 1;
 978
 979        if (parse_signed_commit(ctx->commit->object.sha1,
 980                                &payload, &signature) <= 0)
 981                goto out;
 982        status = verify_signed_buffer(payload.buf, payload.len,
 983                                      signature.buf, signature.len,
 984                                      &gpg_output, &gpg_status);
 985        if (status && !gpg_output.len)
 986                goto out;
 987        ctx->signature.gpg_output = strbuf_detach(&gpg_output, NULL);
 988        ctx->signature.gpg_status = strbuf_detach(&gpg_status, NULL);
 989        parse_signature_lines(ctx);
 990
 991 out:
 992        strbuf_release(&gpg_status);
 993        strbuf_release(&gpg_output);
 994        strbuf_release(&payload);
 995        strbuf_release(&signature);
 996}
 997
 998
 999static int format_reflog_person(struct strbuf *sb,
1000                                char part,
1001                                struct reflog_walk_info *log,
1002                                enum date_mode dmode)
1003{
1004        const char *ident;
1005
1006        if (!log)
1007                return 2;
1008
1009        ident = get_reflog_ident(log);
1010        if (!ident)
1011                return 2;
1012
1013        return format_person_part(sb, part, ident, strlen(ident), dmode);
1014}
1015
1016static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
1017                                void *context)
1018{
1019        struct format_commit_context *c = context;
1020        const struct commit *commit = c->commit;
1021        const char *msg = c->message;
1022        struct commit_list *p;
1023        int h1, h2;
1024
1025        /* these are independent of the commit */
1026        switch (placeholder[0]) {
1027        case 'C':
1028                if (placeholder[1] == '(') {
1029                        const char *begin = placeholder + 2;
1030                        const char *end = strchr(begin, ')');
1031                        char color[COLOR_MAXLEN];
1032
1033                        if (!end)
1034                                return 0;
1035                        if (!prefixcmp(begin, "auto,")) {
1036                                if (!want_color(c->pretty_ctx->color))
1037                                        return end - placeholder + 1;
1038                                begin += 5;
1039                        }
1040                        color_parse_mem(begin,
1041                                        end - begin,
1042                                        "--pretty format", color);
1043                        strbuf_addstr(sb, color);
1044                        return end - placeholder + 1;
1045                }
1046                if (!prefixcmp(placeholder + 1, "red")) {
1047                        strbuf_addstr(sb, GIT_COLOR_RED);
1048                        return 4;
1049                } else if (!prefixcmp(placeholder + 1, "green")) {
1050                        strbuf_addstr(sb, GIT_COLOR_GREEN);
1051                        return 6;
1052                } else if (!prefixcmp(placeholder + 1, "blue")) {
1053                        strbuf_addstr(sb, GIT_COLOR_BLUE);
1054                        return 5;
1055                } else if (!prefixcmp(placeholder + 1, "reset")) {
1056                        strbuf_addstr(sb, GIT_COLOR_RESET);
1057                        return 6;
1058                } else
1059                        return 0;
1060        case 'n':               /* newline */
1061                strbuf_addch(sb, '\n');
1062                return 1;
1063        case 'x':
1064                /* %x00 == NUL, %x0a == LF, etc. */
1065                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1066                    h1 <= 16 &&
1067                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1068                    h2 <= 16) {
1069                        strbuf_addch(sb, (h1<<4)|h2);
1070                        return 3;
1071                } else
1072                        return 0;
1073        case 'w':
1074                if (placeholder[1] == '(') {
1075                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1076                        char *next;
1077                        const char *start = placeholder + 2;
1078                        const char *end = strchr(start, ')');
1079                        if (!end)
1080                                return 0;
1081                        if (end > start) {
1082                                width = strtoul(start, &next, 10);
1083                                if (*next == ',') {
1084                                        indent1 = strtoul(next + 1, &next, 10);
1085                                        if (*next == ',') {
1086                                                indent2 = strtoul(next + 1,
1087                                                                 &next, 10);
1088                                        }
1089                                }
1090                                if (*next != ')')
1091                                        return 0;
1092                        }
1093                        rewrap_message_tail(sb, c, width, indent1, indent2);
1094                        return end - placeholder + 1;
1095                } else
1096                        return 0;
1097        }
1098
1099        /* these depend on the commit */
1100        if (!commit->object.parsed)
1101                parse_object(commit->object.sha1);
1102
1103        switch (placeholder[0]) {
1104        case 'H':               /* commit hash */
1105                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
1106                return 1;
1107        case 'h':               /* abbreviated commit hash */
1108                if (add_again(sb, &c->abbrev_commit_hash))
1109                        return 1;
1110                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
1111                                                     c->pretty_ctx->abbrev));
1112                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1113                return 1;
1114        case 'T':               /* tree hash */
1115                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
1116                return 1;
1117        case 't':               /* abbreviated tree hash */
1118                if (add_again(sb, &c->abbrev_tree_hash))
1119                        return 1;
1120                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
1121                                                     c->pretty_ctx->abbrev));
1122                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1123                return 1;
1124        case 'P':               /* parent hashes */
1125                for (p = commit->parents; p; p = p->next) {
1126                        if (p != commit->parents)
1127                                strbuf_addch(sb, ' ');
1128                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
1129                }
1130                return 1;
1131        case 'p':               /* abbreviated parent hashes */
1132                if (add_again(sb, &c->abbrev_parent_hashes))
1133                        return 1;
1134                for (p = commit->parents; p; p = p->next) {
1135                        if (p != commit->parents)
1136                                strbuf_addch(sb, ' ');
1137                        strbuf_addstr(sb, find_unique_abbrev(
1138                                        p->item->object.sha1,
1139                                        c->pretty_ctx->abbrev));
1140                }
1141                c->abbrev_parent_hashes.len = sb->len -
1142                                              c->abbrev_parent_hashes.off;
1143                return 1;
1144        case 'm':               /* left/right/bottom */
1145                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1146                return 1;
1147        case 'd':
1148                format_decoration(sb, commit);
1149                return 1;
1150        case 'g':               /* reflog info */
1151                switch(placeholder[1]) {
1152                case 'd':       /* reflog selector */
1153                case 'D':
1154                        if (c->pretty_ctx->reflog_info)
1155                                get_reflog_selector(sb,
1156                                                    c->pretty_ctx->reflog_info,
1157                                                    c->pretty_ctx->date_mode,
1158                                                    c->pretty_ctx->date_mode_explicit,
1159                                                    (placeholder[1] == 'd'));
1160                        return 2;
1161                case 's':       /* reflog message */
1162                        if (c->pretty_ctx->reflog_info)
1163                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1164                        return 2;
1165                case 'n':
1166                case 'N':
1167                case 'e':
1168                case 'E':
1169                        return format_reflog_person(sb,
1170                                                    placeholder[1],
1171                                                    c->pretty_ctx->reflog_info,
1172                                                    c->pretty_ctx->date_mode);
1173                }
1174                return 0;       /* unknown %g placeholder */
1175        case 'N':
1176                if (c->pretty_ctx->notes_message) {
1177                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1178                        return 1;
1179                }
1180                return 0;
1181        }
1182
1183        if (placeholder[0] == 'G') {
1184                if (!c->commit_signature_parsed)
1185                        parse_commit_signature(c);
1186                switch (placeholder[1]) {
1187                case 'G':
1188                        if (c->signature.gpg_output)
1189                                strbuf_addstr(sb, c->signature.gpg_output);
1190                        break;
1191                case '?':
1192                        switch (c->signature.good_bad) {
1193                        case 'G':
1194                        case 'B':
1195                                strbuf_addch(sb, c->signature.good_bad);
1196                        }
1197                        break;
1198                case 'S':
1199                        if (c->signature.signer)
1200                                strbuf_addstr(sb, c->signature.signer);
1201                        break;
1202                case 'K':
1203                        if (c->signature.key)
1204                                strbuf_addstr(sb, c->signature.key);
1205                        break;
1206                }
1207                return 2;
1208        }
1209
1210
1211        /* For the rest we have to parse the commit header. */
1212        if (!c->commit_header_parsed)
1213                parse_commit_header(c);
1214
1215        switch (placeholder[0]) {
1216        case 'a':       /* author ... */
1217                return format_person_part(sb, placeholder[1],
1218                                   msg + c->author.off, c->author.len,
1219                                   c->pretty_ctx->date_mode);
1220        case 'c':       /* committer ... */
1221                return format_person_part(sb, placeholder[1],
1222                                   msg + c->committer.off, c->committer.len,
1223                                   c->pretty_ctx->date_mode);
1224        case 'e':       /* encoding */
1225                strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
1226                return 1;
1227        case 'B':       /* raw body */
1228                /* message_off is always left at the initial newline */
1229                strbuf_addstr(sb, msg + c->message_off + 1);
1230                return 1;
1231        }
1232
1233        /* Now we need to parse the commit message. */
1234        if (!c->commit_message_parsed)
1235                parse_commit_message(c);
1236
1237        switch (placeholder[0]) {
1238        case 's':       /* subject */
1239                format_subject(sb, msg + c->subject_off, " ");
1240                return 1;
1241        case 'f':       /* sanitized subject */
1242                format_sanitized_subject(sb, msg + c->subject_off);
1243                return 1;
1244        case 'b':       /* body */
1245                strbuf_addstr(sb, msg + c->body_off);
1246                return 1;
1247        }
1248        return 0;       /* unknown placeholder */
1249}
1250
1251static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
1252                                 void *context)
1253{
1254        int consumed;
1255        size_t orig_len;
1256        enum {
1257                NO_MAGIC,
1258                ADD_LF_BEFORE_NON_EMPTY,
1259                DEL_LF_BEFORE_EMPTY,
1260                ADD_SP_BEFORE_NON_EMPTY
1261        } magic = NO_MAGIC;
1262
1263        switch (placeholder[0]) {
1264        case '-':
1265                magic = DEL_LF_BEFORE_EMPTY;
1266                break;
1267        case '+':
1268                magic = ADD_LF_BEFORE_NON_EMPTY;
1269                break;
1270        case ' ':
1271                magic = ADD_SP_BEFORE_NON_EMPTY;
1272                break;
1273        default:
1274                break;
1275        }
1276        if (magic != NO_MAGIC)
1277                placeholder++;
1278
1279        orig_len = sb->len;
1280        consumed = format_commit_one(sb, placeholder, context);
1281        if (magic == NO_MAGIC)
1282                return consumed;
1283
1284        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1285                while (sb->len && sb->buf[sb->len - 1] == '\n')
1286                        strbuf_setlen(sb, sb->len - 1);
1287        } else if (orig_len != sb->len) {
1288                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1289                        strbuf_insert(sb, orig_len, "\n", 1);
1290                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1291                        strbuf_insert(sb, orig_len, " ", 1);
1292        }
1293        return consumed + 1;
1294}
1295
1296static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1297                                   void *context)
1298{
1299        struct userformat_want *w = context;
1300
1301        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1302                placeholder++;
1303
1304        switch (*placeholder) {
1305        case 'N':
1306                w->notes = 1;
1307                break;
1308        }
1309        return 0;
1310}
1311
1312void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1313{
1314        struct strbuf dummy = STRBUF_INIT;
1315
1316        if (!fmt) {
1317                if (!user_format)
1318                        return;
1319                fmt = user_format;
1320        }
1321        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1322        strbuf_release(&dummy);
1323}
1324
1325void format_commit_message(const struct commit *commit,
1326                           const char *format, struct strbuf *sb,
1327                           const struct pretty_print_context *pretty_ctx)
1328{
1329        struct format_commit_context context;
1330        const char *output_enc = pretty_ctx->output_encoding;
1331
1332        memset(&context, 0, sizeof(context));
1333        context.commit = commit;
1334        context.pretty_ctx = pretty_ctx;
1335        context.wrap_start = sb->len;
1336        context.message = logmsg_reencode(commit, output_enc);
1337
1338        strbuf_expand(sb, format, format_commit_item, &context);
1339        rewrap_message_tail(sb, &context, 0, 0, 0);
1340
1341        logmsg_free(context.message, commit);
1342        free(context.signature.gpg_output);
1343        free(context.signature.signer);
1344}
1345
1346static void pp_header(const struct pretty_print_context *pp,
1347                      const char *encoding,
1348                      const struct commit *commit,
1349                      const char **msg_p,
1350                      struct strbuf *sb)
1351{
1352        int parents_shown = 0;
1353
1354        for (;;) {
1355                const char *line = *msg_p;
1356                int linelen = get_one_line(*msg_p);
1357
1358                if (!linelen)
1359                        return;
1360                *msg_p += linelen;
1361
1362                if (linelen == 1)
1363                        /* End of header */
1364                        return;
1365
1366                if (pp->fmt == CMIT_FMT_RAW) {
1367                        strbuf_add(sb, line, linelen);
1368                        continue;
1369                }
1370
1371                if (!prefixcmp(line, "parent ")) {
1372                        if (linelen != 48)
1373                                die("bad parent line in commit");
1374                        continue;
1375                }
1376
1377                if (!parents_shown) {
1378                        struct commit_list *parent;
1379                        int num;
1380                        for (parent = commit->parents, num = 0;
1381                             parent;
1382                             parent = parent->next, num++)
1383                                ;
1384                        /* with enough slop */
1385                        strbuf_grow(sb, num * 50 + 20);
1386                        add_merge_info(pp, sb, commit);
1387                        parents_shown = 1;
1388                }
1389
1390                /*
1391                 * MEDIUM == DEFAULT shows only author with dates.
1392                 * FULL shows both authors but not dates.
1393                 * FULLER shows both authors and dates.
1394                 */
1395                if (!prefixcmp(line, "author ")) {
1396                        strbuf_grow(sb, linelen + 80);
1397                        pp_user_info(pp, "Author", sb, line + 7, encoding);
1398                }
1399                if (!prefixcmp(line, "committer ") &&
1400                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1401                        strbuf_grow(sb, linelen + 80);
1402                        pp_user_info(pp, "Commit", sb, line + 10, encoding);
1403                }
1404        }
1405}
1406
1407void pp_title_line(const struct pretty_print_context *pp,
1408                   const char **msg_p,
1409                   struct strbuf *sb,
1410                   const char *encoding,
1411                   int need_8bit_cte)
1412{
1413        static const int max_length = 78; /* per rfc2047 */
1414        struct strbuf title;
1415
1416        strbuf_init(&title, 80);
1417        *msg_p = format_subject(&title, *msg_p,
1418                                pp->preserve_subject ? "\n" : " ");
1419
1420        strbuf_grow(sb, title.len + 1024);
1421        if (pp->subject) {
1422                strbuf_addstr(sb, pp->subject);
1423                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1424                        add_rfc2047(sb, title.buf, title.len,
1425                                                encoding, RFC2047_SUBJECT);
1426                else
1427                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1428                                         -last_line_length(sb), 1, max_length);
1429        } else {
1430                strbuf_addbuf(sb, &title);
1431        }
1432        strbuf_addch(sb, '\n');
1433
1434        if (need_8bit_cte > 0) {
1435                const char *header_fmt =
1436                        "MIME-Version: 1.0\n"
1437                        "Content-Type: text/plain; charset=%s\n"
1438                        "Content-Transfer-Encoding: 8bit\n";
1439                strbuf_addf(sb, header_fmt, encoding);
1440        }
1441        if (pp->after_subject) {
1442                strbuf_addstr(sb, pp->after_subject);
1443        }
1444        if (pp->fmt == CMIT_FMT_EMAIL) {
1445                strbuf_addch(sb, '\n');
1446        }
1447        strbuf_release(&title);
1448}
1449
1450void pp_remainder(const struct pretty_print_context *pp,
1451                  const char **msg_p,
1452                  struct strbuf *sb,
1453                  int indent)
1454{
1455        int first = 1;
1456        for (;;) {
1457                const char *line = *msg_p;
1458                int linelen = get_one_line(line);
1459                *msg_p += linelen;
1460
1461                if (!linelen)
1462                        break;
1463
1464                if (is_empty_line(line, &linelen)) {
1465                        if (first)
1466                                continue;
1467                        if (pp->fmt == CMIT_FMT_SHORT)
1468                                break;
1469                }
1470                first = 0;
1471
1472                strbuf_grow(sb, linelen + indent + 20);
1473                if (indent) {
1474                        memset(sb->buf + sb->len, ' ', indent);
1475                        strbuf_setlen(sb, sb->len + indent);
1476                }
1477                strbuf_add(sb, line, linelen);
1478                strbuf_addch(sb, '\n');
1479        }
1480}
1481
1482void pretty_print_commit(const struct pretty_print_context *pp,
1483                         const struct commit *commit,
1484                         struct strbuf *sb)
1485{
1486        unsigned long beginning_of_body;
1487        int indent = 4;
1488        const char *msg;
1489        char *reencoded;
1490        const char *encoding;
1491        int need_8bit_cte = pp->need_8bit_cte;
1492
1493        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1494                format_commit_message(commit, user_format, sb, pp);
1495                return;
1496        }
1497
1498        encoding = get_log_output_encoding();
1499        msg = reencoded = logmsg_reencode(commit, encoding);
1500
1501        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1502                indent = 0;
1503
1504        /*
1505         * We need to check and emit Content-type: to mark it
1506         * as 8-bit if we haven't done so.
1507         */
1508        if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1509                int i, ch, in_body;
1510
1511                for (in_body = i = 0; (ch = msg[i]); i++) {
1512                        if (!in_body) {
1513                                /* author could be non 7-bit ASCII but
1514                                 * the log may be so; skip over the
1515                                 * header part first.
1516                                 */
1517                                if (ch == '\n' && msg[i+1] == '\n')
1518                                        in_body = 1;
1519                        }
1520                        else if (non_ascii(ch)) {
1521                                need_8bit_cte = 1;
1522                                break;
1523                        }
1524                }
1525        }
1526
1527        pp_header(pp, encoding, commit, &msg, sb);
1528        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1529                strbuf_addch(sb, '\n');
1530        }
1531
1532        /* Skip excess blank lines at the beginning of body, if any... */
1533        msg = skip_empty_lines(msg);
1534
1535        /* These formats treat the title line specially. */
1536        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1537                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1538
1539        beginning_of_body = sb->len;
1540        if (pp->fmt != CMIT_FMT_ONELINE)
1541                pp_remainder(pp, &msg, sb, indent);
1542        strbuf_rtrim(sb);
1543
1544        /* Make sure there is an EOLN for the non-oneline case */
1545        if (pp->fmt != CMIT_FMT_ONELINE)
1546                strbuf_addch(sb, '\n');
1547
1548        /*
1549         * The caller may append additional body text in e-mail
1550         * format.  Make sure we did not strip the blank line
1551         * between the header and the body.
1552         */
1553        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1554                strbuf_addch(sb, '\n');
1555
1556        logmsg_free(reencoded, commit);
1557}
1558
1559void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1560                    struct strbuf *sb)
1561{
1562        struct pretty_print_context pp = {0};
1563        pp.fmt = fmt;
1564        pretty_print_commit(&pp, commit, sb);
1565}