pretty.con commit blame: improve diagnosis for "--reverse NEW" (d993ce1)
   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
  27int commit_format_is_empty(enum cmit_fmt fmt)
  28{
  29        return fmt == CMIT_FMT_USERFORMAT && !*user_format;
  30}
  31
  32static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  33{
  34        free(user_format);
  35        user_format = xstrdup(cp);
  36        if (is_tformat)
  37                rev->use_terminator = 1;
  38        rev->commit_format = CMIT_FMT_USERFORMAT;
  39}
  40
  41static int git_pretty_formats_config(const char *var, const char *value, void *cb)
  42{
  43        struct cmt_fmt_map *commit_format = NULL;
  44        const char *name;
  45        const char *fmt;
  46        int i;
  47
  48        if (!skip_prefix(var, "pretty.", &name))
  49                return 0;
  50
  51        for (i = 0; i < builtin_formats_len; i++) {
  52                if (!strcmp(commit_formats[i].name, name))
  53                        return 0;
  54        }
  55
  56        for (i = builtin_formats_len; i < commit_formats_len; i++) {
  57                if (!strcmp(commit_formats[i].name, name)) {
  58                        commit_format = &commit_formats[i];
  59                        break;
  60                }
  61        }
  62
  63        if (!commit_format) {
  64                ALLOC_GROW(commit_formats, commit_formats_len+1,
  65                           commit_formats_alloc);
  66                commit_format = &commit_formats[commit_formats_len];
  67                memset(commit_format, 0, sizeof(*commit_format));
  68                commit_formats_len++;
  69        }
  70
  71        commit_format->name = xstrdup(name);
  72        commit_format->format = CMIT_FMT_USERFORMAT;
  73        if (git_config_string(&fmt, var, value))
  74                return -1;
  75
  76        if (skip_prefix(fmt, "format:", &fmt))
  77                commit_format->is_tformat = 0;
  78        else if (skip_prefix(fmt, "tformat:", &fmt) || strchr(fmt, '%'))
  79                commit_format->is_tformat = 1;
  80        else
  81                commit_format->is_alias = 1;
  82        commit_format->user_format = fmt;
  83
  84        return 0;
  85}
  86
  87static void setup_commit_formats(void)
  88{
  89        struct cmt_fmt_map builtin_formats[] = {
  90                { "raw",        CMIT_FMT_RAW,           0 },
  91                { "medium",     CMIT_FMT_MEDIUM,        0 },
  92                { "short",      CMIT_FMT_SHORT,         0 },
  93                { "email",      CMIT_FMT_EMAIL,         0 },
  94                { "fuller",     CMIT_FMT_FULLER,        0 },
  95                { "full",       CMIT_FMT_FULL,          0 },
  96                { "oneline",    CMIT_FMT_ONELINE,       1 }
  97        };
  98        commit_formats_len = ARRAY_SIZE(builtin_formats);
  99        builtin_formats_len = commit_formats_len;
 100        ALLOC_GROW(commit_formats, commit_formats_len, commit_formats_alloc);
 101        memcpy(commit_formats, builtin_formats,
 102               sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
 103
 104        git_config(git_pretty_formats_config, NULL);
 105}
 106
 107static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
 108                                                        const char *original,
 109                                                        int num_redirections)
 110{
 111        struct cmt_fmt_map *found = NULL;
 112        size_t found_match_len = 0;
 113        int i;
 114
 115        if (num_redirections >= commit_formats_len)
 116                die("invalid --pretty format: "
 117                    "'%s' references an alias which points to itself",
 118                    original);
 119
 120        for (i = 0; i < commit_formats_len; i++) {
 121                size_t match_len;
 122
 123                if (!starts_with(commit_formats[i].name, sought))
 124                        continue;
 125
 126                match_len = strlen(commit_formats[i].name);
 127                if (found == NULL || found_match_len > match_len) {
 128                        found = &commit_formats[i];
 129                        found_match_len = match_len;
 130                }
 131        }
 132
 133        if (found && found->is_alias) {
 134                found = find_commit_format_recursive(found->user_format,
 135                                                     original,
 136                                                     num_redirections+1);
 137        }
 138
 139        return found;
 140}
 141
 142static struct cmt_fmt_map *find_commit_format(const char *sought)
 143{
 144        if (!commit_formats)
 145                setup_commit_formats();
 146
 147        return find_commit_format_recursive(sought, sought, 0);
 148}
 149
 150void get_commit_format(const char *arg, struct rev_info *rev)
 151{
 152        struct cmt_fmt_map *commit_format;
 153
 154        rev->use_terminator = 0;
 155        if (!arg) {
 156                rev->commit_format = CMIT_FMT_DEFAULT;
 157                return;
 158        }
 159        if (skip_prefix(arg, "format:", &arg)) {
 160                save_user_format(rev, arg, 0);
 161                return;
 162        }
 163
 164        if (!*arg || skip_prefix(arg, "tformat:", &arg) || strchr(arg, '%')) {
 165                save_user_format(rev, arg, 1);
 166                return;
 167        }
 168
 169        commit_format = find_commit_format(arg);
 170        if (!commit_format)
 171                die("invalid --pretty format: %s", arg);
 172
 173        rev->commit_format = commit_format->format;
 174        rev->use_terminator = commit_format->is_tformat;
 175        if (commit_format->format == CMIT_FMT_USERFORMAT) {
 176                save_user_format(rev, commit_format->user_format,
 177                                 commit_format->is_tformat);
 178        }
 179}
 180
 181/*
 182 * Generic support for pretty-printing the header
 183 */
 184static int get_one_line(const char *msg)
 185{
 186        int ret = 0;
 187
 188        for (;;) {
 189                char c = *msg++;
 190                if (!c)
 191                        break;
 192                ret++;
 193                if (c == '\n')
 194                        break;
 195        }
 196        return ret;
 197}
 198
 199/* High bit set, or ISO-2022-INT */
 200static int non_ascii(int ch)
 201{
 202        return !isascii(ch) || ch == '\033';
 203}
 204
 205int has_non_ascii(const char *s)
 206{
 207        int ch;
 208        if (!s)
 209                return 0;
 210        while ((ch = *s++) != '\0') {
 211                if (non_ascii(ch))
 212                        return 1;
 213        }
 214        return 0;
 215}
 216
 217static int is_rfc822_special(char ch)
 218{
 219        switch (ch) {
 220        case '(':
 221        case ')':
 222        case '<':
 223        case '>':
 224        case '[':
 225        case ']':
 226        case ':':
 227        case ';':
 228        case '@':
 229        case ',':
 230        case '.':
 231        case '"':
 232        case '\\':
 233                return 1;
 234        default:
 235                return 0;
 236        }
 237}
 238
 239static int needs_rfc822_quoting(const char *s, int len)
 240{
 241        int i;
 242        for (i = 0; i < len; i++)
 243                if (is_rfc822_special(s[i]))
 244                        return 1;
 245        return 0;
 246}
 247
 248static int last_line_length(struct strbuf *sb)
 249{
 250        int i;
 251
 252        /* How many bytes are already used on the last line? */
 253        for (i = sb->len - 1; i >= 0; i--)
 254                if (sb->buf[i] == '\n')
 255                        break;
 256        return sb->len - (i + 1);
 257}
 258
 259static void add_rfc822_quoted(struct strbuf *out, const char *s, int len)
 260{
 261        int i;
 262
 263        /* just a guess, we may have to also backslash-quote */
 264        strbuf_grow(out, len + 2);
 265
 266        strbuf_addch(out, '"');
 267        for (i = 0; i < len; i++) {
 268                switch (s[i]) {
 269                case '"':
 270                case '\\':
 271                        strbuf_addch(out, '\\');
 272                        /* fall through */
 273                default:
 274                        strbuf_addch(out, s[i]);
 275                }
 276        }
 277        strbuf_addch(out, '"');
 278}
 279
 280enum rfc2047_type {
 281        RFC2047_SUBJECT,
 282        RFC2047_ADDRESS
 283};
 284
 285static int is_rfc2047_special(char ch, enum rfc2047_type type)
 286{
 287        /*
 288         * rfc2047, section 4.2:
 289         *
 290         *    8-bit values which correspond to printable ASCII characters other
 291         *    than "=", "?", and "_" (underscore), MAY be represented as those
 292         *    characters.  (But see section 5 for restrictions.)  In
 293         *    particular, SPACE and TAB MUST NOT be represented as themselves
 294         *    within encoded words.
 295         */
 296
 297        /*
 298         * rule out non-ASCII characters and non-printable characters (the
 299         * non-ASCII check should be redundant as isprint() is not localized
 300         * and only knows about ASCII, but be defensive about that)
 301         */
 302        if (non_ascii(ch) || !isprint(ch))
 303                return 1;
 304
 305        /*
 306         * rule out special printable characters (' ' should be the only
 307         * whitespace character considered printable, but be defensive and use
 308         * isspace())
 309         */
 310        if (isspace(ch) || ch == '=' || ch == '?' || ch == '_')
 311                return 1;
 312
 313        /*
 314         * rfc2047, section 5.3:
 315         *
 316         *    As a replacement for a 'word' entity within a 'phrase', for example,
 317         *    one that precedes an address in a From, To, or Cc header.  The ABNF
 318         *    definition for 'phrase' from RFC 822 thus becomes:
 319         *
 320         *    phrase = 1*( encoded-word / word )
 321         *
 322         *    In this case the set of characters that may be used in a "Q"-encoded
 323         *    'encoded-word' is restricted to: <upper and lower case ASCII
 324         *    letters, decimal digits, "!", "*", "+", "-", "/", "=", and "_"
 325         *    (underscore, ASCII 95.)>.  An 'encoded-word' that appears within a
 326         *    'phrase' MUST be separated from any adjacent 'word', 'text' or
 327         *    'special' by 'linear-white-space'.
 328         */
 329
 330        if (type != RFC2047_ADDRESS)
 331                return 0;
 332
 333        /* '=' and '_' are special cases and have been checked above */
 334        return !(isalnum(ch) || ch == '!' || ch == '*' || ch == '+' || ch == '-' || ch == '/');
 335}
 336
 337static int needs_rfc2047_encoding(const char *line, int len,
 338                                  enum rfc2047_type type)
 339{
 340        int i;
 341
 342        for (i = 0; i < len; i++) {
 343                int ch = line[i];
 344                if (non_ascii(ch) || ch == '\n')
 345                        return 1;
 346                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 347                        return 1;
 348        }
 349
 350        return 0;
 351}
 352
 353static void add_rfc2047(struct strbuf *sb, const char *line, size_t len,
 354                       const char *encoding, enum rfc2047_type type)
 355{
 356        static const int max_encoded_length = 76; /* per rfc2047 */
 357        int i;
 358        int line_len = last_line_length(sb);
 359
 360        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 361        strbuf_addf(sb, "=?%s?q?", encoding);
 362        line_len += strlen(encoding) + 5; /* 5 for =??q? */
 363
 364        while (len) {
 365                /*
 366                 * RFC 2047, section 5 (3):
 367                 *
 368                 * Each 'encoded-word' MUST represent an integral number of
 369                 * characters.  A multi-octet character may not be split across
 370                 * adjacent 'encoded- word's.
 371                 */
 372                const unsigned char *p = (const unsigned char *)line;
 373                int chrlen = mbs_chrlen(&line, &len, encoding);
 374                int is_special = (chrlen > 1) || is_rfc2047_special(*p, type);
 375
 376                /* "=%02X" * chrlen, or the byte itself */
 377                const char *encoded_fmt = is_special ? "=%02X"    : "%c";
 378                int         encoded_len = is_special ? 3 * chrlen : 1;
 379
 380                /*
 381                 * According to RFC 2047, we could encode the special character
 382                 * ' ' (space) with '_' (underscore) for readability. But many
 383                 * programs do not understand this and just leave the
 384                 * underscore in place. Thus, we do nothing special here, which
 385                 * causes ' ' to be encoded as '=20', avoiding this problem.
 386                 */
 387
 388                if (line_len + encoded_len + 2 > max_encoded_length) {
 389                        /* It won't fit with trailing "?=" --- break the line */
 390                        strbuf_addf(sb, "?=\n =?%s?q?", encoding);
 391                        line_len = strlen(encoding) + 5 + 1; /* =??q? plus SP */
 392                }
 393
 394                for (i = 0; i < chrlen; i++)
 395                        strbuf_addf(sb, encoded_fmt, p[i]);
 396                line_len += encoded_len;
 397        }
 398        strbuf_addstr(sb, "?=");
 399}
 400
 401const char *show_ident_date(const struct ident_split *ident,
 402                            const struct date_mode *mode)
 403{
 404        unsigned long date = 0;
 405        long tz = 0;
 406
 407        if (ident->date_begin && ident->date_end)
 408                date = strtoul(ident->date_begin, NULL, 10);
 409        if (date_overflows(date))
 410                date = 0;
 411        else {
 412                if (ident->tz_begin && ident->tz_end)
 413                        tz = strtol(ident->tz_begin, NULL, 10);
 414                if (tz >= INT_MAX || tz <= INT_MIN)
 415                        tz = 0;
 416        }
 417        return show_date(date, tz, mode);
 418}
 419
 420void pp_user_info(struct pretty_print_context *pp,
 421                  const char *what, struct strbuf *sb,
 422                  const char *line, const char *encoding)
 423{
 424        struct ident_split ident;
 425        char *line_end;
 426        const char *mailbuf, *namebuf;
 427        size_t namelen, maillen;
 428        int max_length = 78; /* per rfc2822 */
 429
 430        if (pp->fmt == CMIT_FMT_ONELINE)
 431                return;
 432
 433        line_end = strchrnul(line, '\n');
 434        if (split_ident_line(&ident, line, line_end - line))
 435                return;
 436
 437        mailbuf = ident.mail_begin;
 438        maillen = ident.mail_end - ident.mail_begin;
 439        namebuf = ident.name_begin;
 440        namelen = ident.name_end - ident.name_begin;
 441
 442        if (pp->mailmap)
 443                map_user(pp->mailmap, &mailbuf, &maillen, &namebuf, &namelen);
 444
 445        if (pp->fmt == CMIT_FMT_EMAIL) {
 446                if (pp->from_ident && ident_cmp(pp->from_ident, &ident)) {
 447                        struct strbuf buf = STRBUF_INIT;
 448
 449                        strbuf_addstr(&buf, "From: ");
 450                        strbuf_add(&buf, namebuf, namelen);
 451                        strbuf_addstr(&buf, " <");
 452                        strbuf_add(&buf, mailbuf, maillen);
 453                        strbuf_addstr(&buf, ">\n");
 454                        string_list_append(&pp->in_body_headers,
 455                                           strbuf_detach(&buf, NULL));
 456
 457                        mailbuf = pp->from_ident->mail_begin;
 458                        maillen = pp->from_ident->mail_end - mailbuf;
 459                        namebuf = pp->from_ident->name_begin;
 460                        namelen = pp->from_ident->name_end - namebuf;
 461                }
 462
 463                strbuf_addstr(sb, "From: ");
 464                if (needs_rfc2047_encoding(namebuf, namelen, RFC2047_ADDRESS)) {
 465                        add_rfc2047(sb, namebuf, namelen,
 466                                    encoding, RFC2047_ADDRESS);
 467                        max_length = 76; /* per rfc2047 */
 468                } else if (needs_rfc822_quoting(namebuf, namelen)) {
 469                        struct strbuf quoted = STRBUF_INIT;
 470                        add_rfc822_quoted(&quoted, namebuf, namelen);
 471                        strbuf_add_wrapped_bytes(sb, quoted.buf, quoted.len,
 472                                                        -6, 1, max_length);
 473                        strbuf_release(&quoted);
 474                } else {
 475                        strbuf_add_wrapped_bytes(sb, namebuf, namelen,
 476                                                 -6, 1, max_length);
 477                }
 478
 479                if (max_length <
 480                    last_line_length(sb) + strlen(" <") + maillen + strlen(">"))
 481                        strbuf_addch(sb, '\n');
 482                strbuf_addf(sb, " <%.*s>\n", (int)maillen, mailbuf);
 483        } else {
 484                strbuf_addf(sb, "%s: %.*s%.*s <%.*s>\n", what,
 485                            (pp->fmt == CMIT_FMT_FULLER) ? 4 : 0, "    ",
 486                            (int)namelen, namebuf, (int)maillen, mailbuf);
 487        }
 488
 489        switch (pp->fmt) {
 490        case CMIT_FMT_MEDIUM:
 491                strbuf_addf(sb, "Date:   %s\n",
 492                            show_ident_date(&ident, &pp->date_mode));
 493                break;
 494        case CMIT_FMT_EMAIL:
 495                strbuf_addf(sb, "Date: %s\n",
 496                            show_ident_date(&ident, DATE_MODE(RFC2822)));
 497                break;
 498        case CMIT_FMT_FULLER:
 499                strbuf_addf(sb, "%sDate: %s\n", what,
 500                            show_ident_date(&ident, &pp->date_mode));
 501                break;
 502        default:
 503                /* notin' */
 504                break;
 505        }
 506}
 507
 508static int is_empty_line(const char *line, int *len_p)
 509{
 510        int len = *len_p;
 511        while (len && isspace(line[len - 1]))
 512                len--;
 513        *len_p = len;
 514        return !len;
 515}
 516
 517static const char *skip_empty_lines(const char *msg)
 518{
 519        for (;;) {
 520                int linelen = get_one_line(msg);
 521                int ll = linelen;
 522                if (!linelen)
 523                        break;
 524                if (!is_empty_line(msg, &ll))
 525                        break;
 526                msg += linelen;
 527        }
 528        return msg;
 529}
 530
 531static void add_merge_info(const struct pretty_print_context *pp,
 532                           struct strbuf *sb, const struct commit *commit)
 533{
 534        struct commit_list *parent = commit->parents;
 535
 536        if ((pp->fmt == CMIT_FMT_ONELINE) || (pp->fmt == CMIT_FMT_EMAIL) ||
 537            !parent || !parent->next)
 538                return;
 539
 540        strbuf_addstr(sb, "Merge:");
 541
 542        while (parent) {
 543                struct commit *p = parent->item;
 544                const char *hex = NULL;
 545                if (pp->abbrev)
 546                        hex = find_unique_abbrev(p->object.oid.hash, pp->abbrev);
 547                if (!hex)
 548                        hex = oid_to_hex(&p->object.oid);
 549                parent = parent->next;
 550
 551                strbuf_addf(sb, " %s", hex);
 552        }
 553        strbuf_addch(sb, '\n');
 554}
 555
 556static char *get_header(const char *msg, const char *key)
 557{
 558        size_t len;
 559        const char *v = find_commit_header(msg, key, &len);
 560        return v ? xmemdupz(v, len) : NULL;
 561}
 562
 563static char *replace_encoding_header(char *buf, const char *encoding)
 564{
 565        struct strbuf tmp = STRBUF_INIT;
 566        size_t start, len;
 567        char *cp = buf;
 568
 569        /* guess if there is an encoding header before a \n\n */
 570        while (!starts_with(cp, "encoding ")) {
 571                cp = strchr(cp, '\n');
 572                if (!cp || *++cp == '\n')
 573                        return buf;
 574        }
 575        start = cp - buf;
 576        cp = strchr(cp, '\n');
 577        if (!cp)
 578                return buf; /* should not happen but be defensive */
 579        len = cp + 1 - (buf + start);
 580
 581        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 582        if (is_encoding_utf8(encoding)) {
 583                /* we have re-coded to UTF-8; drop the header */
 584                strbuf_remove(&tmp, start, len);
 585        } else {
 586                /* just replaces XXXX in 'encoding XXXX\n' */
 587                strbuf_splice(&tmp, start + strlen("encoding "),
 588                                          len - strlen("encoding \n"),
 589                                          encoding, strlen(encoding));
 590        }
 591        return strbuf_detach(&tmp, NULL);
 592}
 593
 594const char *logmsg_reencode(const struct commit *commit,
 595                            char **commit_encoding,
 596                            const char *output_encoding)
 597{
 598        static const char *utf8 = "UTF-8";
 599        const char *use_encoding;
 600        char *encoding;
 601        const char *msg = get_commit_buffer(commit, NULL);
 602        char *out;
 603
 604        if (!output_encoding || !*output_encoding) {
 605                if (commit_encoding)
 606                        *commit_encoding = get_header(msg, "encoding");
 607                return msg;
 608        }
 609        encoding = get_header(msg, "encoding");
 610        if (commit_encoding)
 611                *commit_encoding = 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 get_commit_buffer, we need to duplicate it
 627                 * to avoid munging the cached copy.
 628                 */
 629                if (msg == get_cached_commit_buffer(commit, NULL))
 630                        out = xstrdup(msg);
 631                else
 632                        out = (char *)msg;
 633        }
 634        else {
 635                /*
 636                 * There's actual encoding work to do. Do the reencoding, which
 637                 * still leaves the header to be replaced in the next step. At
 638                 * this point, we are done with msg. If we allocated a fresh
 639                 * copy, we can free it.
 640                 */
 641                out = reencode_string(msg, output_encoding, use_encoding);
 642                if (out)
 643                        unuse_commit_buffer(commit, msg);
 644        }
 645
 646        /*
 647         * This replacement actually consumes the buffer we hand it, so we do
 648         * not have to worry about freeing the old "out" here.
 649         */
 650        if (out)
 651                out = replace_encoding_header(out, output_encoding);
 652
 653        if (!commit_encoding)
 654                free(encoding);
 655        /*
 656         * If the re-encoding failed, out might be NULL here; in that
 657         * case we just return the commit message verbatim.
 658         */
 659        return out ? out : msg;
 660}
 661
 662static int mailmap_name(const char **email, size_t *email_len,
 663                        const char **name, size_t *name_len)
 664{
 665        static struct string_list *mail_map;
 666        if (!mail_map) {
 667                mail_map = xcalloc(1, sizeof(*mail_map));
 668                read_mailmap(mail_map, NULL);
 669        }
 670        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 671}
 672
 673static size_t format_person_part(struct strbuf *sb, char part,
 674                                 const char *msg, int len,
 675                                 const struct date_mode *dmode)
 676{
 677        /* currently all placeholders have same length */
 678        const int placeholder_len = 2;
 679        struct ident_split s;
 680        const char *name, *mail;
 681        size_t maillen, namelen;
 682
 683        if (split_ident_line(&s, msg, len) < 0)
 684                goto skip;
 685
 686        name = s.name_begin;
 687        namelen = s.name_end - s.name_begin;
 688        mail = s.mail_begin;
 689        maillen = s.mail_end - s.mail_begin;
 690
 691        if (part == 'N' || part == 'E') /* mailmap lookup */
 692                mailmap_name(&mail, &maillen, &name, &namelen);
 693        if (part == 'n' || part == 'N') {       /* name */
 694                strbuf_add(sb, name, namelen);
 695                return placeholder_len;
 696        }
 697        if (part == 'e' || part == 'E') {       /* email */
 698                strbuf_add(sb, mail, maillen);
 699                return placeholder_len;
 700        }
 701
 702        if (!s.date_begin)
 703                goto skip;
 704
 705        if (part == 't') {      /* date, UNIX timestamp */
 706                strbuf_add(sb, s.date_begin, s.date_end - s.date_begin);
 707                return placeholder_len;
 708        }
 709
 710        switch (part) {
 711        case 'd':       /* date */
 712                strbuf_addstr(sb, show_ident_date(&s, dmode));
 713                return placeholder_len;
 714        case 'D':       /* date, RFC2822 style */
 715                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RFC2822)));
 716                return placeholder_len;
 717        case 'r':       /* date, relative */
 718                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(RELATIVE)));
 719                return placeholder_len;
 720        case 'i':       /* date, ISO 8601-like */
 721                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601)));
 722                return placeholder_len;
 723        case 'I':       /* date, ISO 8601 strict */
 724                strbuf_addstr(sb, show_ident_date(&s, DATE_MODE(ISO8601_STRICT)));
 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
 746enum flush_type {
 747        no_flush,
 748        flush_right,
 749        flush_left,
 750        flush_left_and_steal,
 751        flush_both
 752};
 753
 754enum trunc_type {
 755        trunc_none,
 756        trunc_left,
 757        trunc_middle,
 758        trunc_right
 759};
 760
 761struct format_commit_context {
 762        const struct commit *commit;
 763        const struct pretty_print_context *pretty_ctx;
 764        unsigned commit_header_parsed:1;
 765        unsigned commit_message_parsed:1;
 766        struct signature_check signature_check;
 767        enum flush_type flush_type;
 768        enum trunc_type truncate;
 769        const char *message;
 770        char *commit_encoding;
 771        size_t width, indent1, indent2;
 772        int auto_color;
 773        int padding;
 774
 775        /* These offsets are relative to the start of the commit message. */
 776        struct chunk author;
 777        struct chunk committer;
 778        size_t message_off;
 779        size_t subject_off;
 780        size_t body_off;
 781
 782        /* The following ones are relative to the result struct strbuf. */
 783        struct chunk abbrev_commit_hash;
 784        struct chunk abbrev_tree_hash;
 785        struct chunk abbrev_parent_hashes;
 786        size_t wrap_start;
 787};
 788
 789static int add_again(struct strbuf *sb, struct chunk *chunk)
 790{
 791        if (chunk->len) {
 792                strbuf_adddup(sb, chunk->off, chunk->len);
 793                return 1;
 794        }
 795
 796        /*
 797         * We haven't seen this chunk before.  Our caller is surely
 798         * going to add it the hard way now.  Remember the most likely
 799         * start of the to-be-added chunk: the current end of the
 800         * struct strbuf.
 801         */
 802        chunk->off = sb->len;
 803        return 0;
 804}
 805
 806static void parse_commit_header(struct format_commit_context *context)
 807{
 808        const char *msg = context->message;
 809        int i;
 810
 811        for (i = 0; msg[i]; i++) {
 812                const char *name;
 813                int eol;
 814                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 815                        ; /* do nothing */
 816
 817                if (i == eol) {
 818                        break;
 819                } else if (skip_prefix(msg + i, "author ", &name)) {
 820                        context->author.off = name - msg;
 821                        context->author.len = msg + eol - name;
 822                } else if (skip_prefix(msg + i, "committer ", &name)) {
 823                        context->committer.off = name - msg;
 824                        context->committer.len = msg + eol - name;
 825                }
 826                i = eol;
 827        }
 828        context->message_off = i;
 829        context->commit_header_parsed = 1;
 830}
 831
 832static int istitlechar(char c)
 833{
 834        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 835                (c >= '0' && c <= '9') || c == '.' || c == '_';
 836}
 837
 838static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 839{
 840        size_t trimlen;
 841        size_t start_len = sb->len;
 842        int space = 2;
 843
 844        for (; *msg && *msg != '\n'; msg++) {
 845                if (istitlechar(*msg)) {
 846                        if (space == 1)
 847                                strbuf_addch(sb, '-');
 848                        space = 0;
 849                        strbuf_addch(sb, *msg);
 850                        if (*msg == '.')
 851                                while (*(msg+1) == '.')
 852                                        msg++;
 853                } else
 854                        space |= 1;
 855        }
 856
 857        /* trim any trailing '.' or '-' characters */
 858        trimlen = 0;
 859        while (sb->len - trimlen > start_len &&
 860                (sb->buf[sb->len - 1 - trimlen] == '.'
 861                || sb->buf[sb->len - 1 - trimlen] == '-'))
 862                trimlen++;
 863        strbuf_remove(sb, sb->len - trimlen, trimlen);
 864}
 865
 866const char *format_subject(struct strbuf *sb, const char *msg,
 867                           const char *line_separator)
 868{
 869        int first = 1;
 870
 871        for (;;) {
 872                const char *line = msg;
 873                int linelen = get_one_line(line);
 874
 875                msg += linelen;
 876                if (!linelen || is_empty_line(line, &linelen))
 877                        break;
 878
 879                if (!sb)
 880                        continue;
 881                strbuf_grow(sb, linelen + 2);
 882                if (!first)
 883                        strbuf_addstr(sb, line_separator);
 884                strbuf_add(sb, line, linelen);
 885                first = 0;
 886        }
 887        return msg;
 888}
 889
 890static void parse_commit_message(struct format_commit_context *c)
 891{
 892        const char *msg = c->message + c->message_off;
 893        const char *start = c->message;
 894
 895        msg = skip_empty_lines(msg);
 896        c->subject_off = msg - start;
 897
 898        msg = format_subject(NULL, msg, NULL);
 899        msg = skip_empty_lines(msg);
 900        c->body_off = msg - start;
 901
 902        c->commit_message_parsed = 1;
 903}
 904
 905static void strbuf_wrap(struct strbuf *sb, size_t pos,
 906                        size_t width, size_t indent1, size_t indent2)
 907{
 908        struct strbuf tmp = STRBUF_INIT;
 909
 910        if (pos)
 911                strbuf_add(&tmp, sb->buf, pos);
 912        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 913                                (int) indent1, (int) indent2, (int) width);
 914        strbuf_swap(&tmp, sb);
 915        strbuf_release(&tmp);
 916}
 917
 918static void rewrap_message_tail(struct strbuf *sb,
 919                                struct format_commit_context *c,
 920                                size_t new_width, size_t new_indent1,
 921                                size_t new_indent2)
 922{
 923        if (c->width == new_width && c->indent1 == new_indent1 &&
 924            c->indent2 == new_indent2)
 925                return;
 926        if (c->wrap_start < sb->len)
 927                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 928        c->wrap_start = sb->len;
 929        c->width = new_width;
 930        c->indent1 = new_indent1;
 931        c->indent2 = new_indent2;
 932}
 933
 934static int format_reflog_person(struct strbuf *sb,
 935                                char part,
 936                                struct reflog_walk_info *log,
 937                                const struct date_mode *dmode)
 938{
 939        const char *ident;
 940
 941        if (!log)
 942                return 2;
 943
 944        ident = get_reflog_ident(log);
 945        if (!ident)
 946                return 2;
 947
 948        return format_person_part(sb, part, ident, strlen(ident), dmode);
 949}
 950
 951static size_t parse_color(struct strbuf *sb, /* in UTF-8 */
 952                          const char *placeholder,
 953                          struct format_commit_context *c)
 954{
 955        const char *rest = placeholder;
 956
 957        if (placeholder[1] == '(') {
 958                const char *begin = placeholder + 2;
 959                const char *end = strchr(begin, ')');
 960                char color[COLOR_MAXLEN];
 961
 962                if (!end)
 963                        return 0;
 964                if (skip_prefix(begin, "auto,", &begin)) {
 965                        if (!want_color(c->pretty_ctx->color))
 966                                return end - placeholder + 1;
 967                }
 968                if (color_parse_mem(begin, end - begin, color) < 0)
 969                        die(_("unable to parse --pretty format"));
 970                strbuf_addstr(sb, color);
 971                return end - placeholder + 1;
 972        }
 973        if (skip_prefix(placeholder + 1, "red", &rest))
 974                strbuf_addstr(sb, GIT_COLOR_RED);
 975        else if (skip_prefix(placeholder + 1, "green", &rest))
 976                strbuf_addstr(sb, GIT_COLOR_GREEN);
 977        else if (skip_prefix(placeholder + 1, "blue", &rest))
 978                strbuf_addstr(sb, GIT_COLOR_BLUE);
 979        else if (skip_prefix(placeholder + 1, "reset", &rest))
 980                strbuf_addstr(sb, GIT_COLOR_RESET);
 981        return rest - placeholder;
 982}
 983
 984static size_t parse_padding_placeholder(struct strbuf *sb,
 985                                        const char *placeholder,
 986                                        struct format_commit_context *c)
 987{
 988        const char *ch = placeholder;
 989        enum flush_type flush_type;
 990        int to_column = 0;
 991
 992        switch (*ch++) {
 993        case '<':
 994                flush_type = flush_right;
 995                break;
 996        case '>':
 997                if (*ch == '<') {
 998                        flush_type = flush_both;
 999                        ch++;
1000                } else if (*ch == '>') {
1001                        flush_type = flush_left_and_steal;
1002                        ch++;
1003                } else
1004                        flush_type = flush_left;
1005                break;
1006        default:
1007                return 0;
1008        }
1009
1010        /* the next value means "wide enough to that column" */
1011        if (*ch == '|') {
1012                to_column = 1;
1013                ch++;
1014        }
1015
1016        if (*ch == '(') {
1017                const char *start = ch + 1;
1018                const char *end = start + strcspn(start, ",)");
1019                char *next;
1020                int width;
1021                if (!end || end == start)
1022                        return 0;
1023                width = strtoul(start, &next, 10);
1024                if (next == start || width == 0)
1025                        return 0;
1026                c->padding = to_column ? -width : width;
1027                c->flush_type = flush_type;
1028
1029                if (*end == ',') {
1030                        start = end + 1;
1031                        end = strchr(start, ')');
1032                        if (!end || end == start)
1033                                return 0;
1034                        if (starts_with(start, "trunc)"))
1035                                c->truncate = trunc_right;
1036                        else if (starts_with(start, "ltrunc)"))
1037                                c->truncate = trunc_left;
1038                        else if (starts_with(start, "mtrunc)"))
1039                                c->truncate = trunc_middle;
1040                        else
1041                                return 0;
1042                } else
1043                        c->truncate = trunc_none;
1044
1045                return end - placeholder + 1;
1046        }
1047        return 0;
1048}
1049
1050static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
1051                                const char *placeholder,
1052                                void *context)
1053{
1054        struct format_commit_context *c = context;
1055        const struct commit *commit = c->commit;
1056        const char *msg = c->message;
1057        struct commit_list *p;
1058        int h1, h2;
1059
1060        /* these are independent of the commit */
1061        switch (placeholder[0]) {
1062        case 'C':
1063                if (starts_with(placeholder + 1, "(auto)")) {
1064                        c->auto_color = 1;
1065                        return 7; /* consumed 7 bytes, "C(auto)" */
1066                } else {
1067                        int ret = parse_color(sb, placeholder, c);
1068                        if (ret)
1069                                c->auto_color = 0;
1070                        /*
1071                         * Otherwise, we decided to treat %C<unknown>
1072                         * as a literal string, and the previous
1073                         * %C(auto) is still valid.
1074                         */
1075                        return ret;
1076                }
1077        case 'n':               /* newline */
1078                strbuf_addch(sb, '\n');
1079                return 1;
1080        case 'x':
1081                /* %x00 == NUL, %x0a == LF, etc. */
1082                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
1083                    h1 <= 16 &&
1084                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
1085                    h2 <= 16) {
1086                        strbuf_addch(sb, (h1<<4)|h2);
1087                        return 3;
1088                } else
1089                        return 0;
1090        case 'w':
1091                if (placeholder[1] == '(') {
1092                        unsigned long width = 0, indent1 = 0, indent2 = 0;
1093                        char *next;
1094                        const char *start = placeholder + 2;
1095                        const char *end = strchr(start, ')');
1096                        if (!end)
1097                                return 0;
1098                        if (end > start) {
1099                                width = strtoul(start, &next, 10);
1100                                if (*next == ',') {
1101                                        indent1 = strtoul(next + 1, &next, 10);
1102                                        if (*next == ',') {
1103                                                indent2 = strtoul(next + 1,
1104                                                                 &next, 10);
1105                                        }
1106                                }
1107                                if (*next != ')')
1108                                        return 0;
1109                        }
1110                        rewrap_message_tail(sb, c, width, indent1, indent2);
1111                        return end - placeholder + 1;
1112                } else
1113                        return 0;
1114
1115        case '<':
1116        case '>':
1117                return parse_padding_placeholder(sb, placeholder, c);
1118        }
1119
1120        /* these depend on the commit */
1121        if (!commit->object.parsed)
1122                parse_object(commit->object.oid.hash);
1123
1124        switch (placeholder[0]) {
1125        case 'H':               /* commit hash */
1126                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1127                strbuf_addstr(sb, oid_to_hex(&commit->object.oid));
1128                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1129                return 1;
1130        case 'h':               /* abbreviated commit hash */
1131                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
1132                if (add_again(sb, &c->abbrev_commit_hash)) {
1133                        strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1134                        return 1;
1135                }
1136                strbuf_addstr(sb, find_unique_abbrev(commit->object.oid.hash,
1137                                                     c->pretty_ctx->abbrev));
1138                strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_RESET));
1139                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
1140                return 1;
1141        case 'T':               /* tree hash */
1142                strbuf_addstr(sb, oid_to_hex(&commit->tree->object.oid));
1143                return 1;
1144        case 't':               /* abbreviated tree hash */
1145                if (add_again(sb, &c->abbrev_tree_hash))
1146                        return 1;
1147                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.oid.hash,
1148                                                     c->pretty_ctx->abbrev));
1149                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
1150                return 1;
1151        case 'P':               /* parent hashes */
1152                for (p = commit->parents; p; p = p->next) {
1153                        if (p != commit->parents)
1154                                strbuf_addch(sb, ' ');
1155                        strbuf_addstr(sb, oid_to_hex(&p->item->object.oid));
1156                }
1157                return 1;
1158        case 'p':               /* abbreviated parent hashes */
1159                if (add_again(sb, &c->abbrev_parent_hashes))
1160                        return 1;
1161                for (p = commit->parents; p; p = p->next) {
1162                        if (p != commit->parents)
1163                                strbuf_addch(sb, ' ');
1164                        strbuf_addstr(sb, find_unique_abbrev(
1165                                        p->item->object.oid.hash,
1166                                        c->pretty_ctx->abbrev));
1167                }
1168                c->abbrev_parent_hashes.len = sb->len -
1169                                              c->abbrev_parent_hashes.off;
1170                return 1;
1171        case 'm':               /* left/right/bottom */
1172                strbuf_addstr(sb, get_revision_mark(NULL, commit));
1173                return 1;
1174        case 'd':
1175                load_ref_decorations(DECORATE_SHORT_REFS);
1176                format_decorations(sb, commit, c->auto_color);
1177                return 1;
1178        case 'D':
1179                load_ref_decorations(DECORATE_SHORT_REFS);
1180                format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
1181                return 1;
1182        case 'g':               /* reflog info */
1183                switch(placeholder[1]) {
1184                case 'd':       /* reflog selector */
1185                case 'D':
1186                        if (c->pretty_ctx->reflog_info)
1187                                get_reflog_selector(sb,
1188                                                    c->pretty_ctx->reflog_info,
1189                                                    &c->pretty_ctx->date_mode,
1190                                                    c->pretty_ctx->date_mode_explicit,
1191                                                    (placeholder[1] == 'd'));
1192                        return 2;
1193                case 's':       /* reflog message */
1194                        if (c->pretty_ctx->reflog_info)
1195                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
1196                        return 2;
1197                case 'n':
1198                case 'N':
1199                case 'e':
1200                case 'E':
1201                        return format_reflog_person(sb,
1202                                                    placeholder[1],
1203                                                    c->pretty_ctx->reflog_info,
1204                                                    &c->pretty_ctx->date_mode);
1205                }
1206                return 0;       /* unknown %g placeholder */
1207        case 'N':
1208                if (c->pretty_ctx->notes_message) {
1209                        strbuf_addstr(sb, c->pretty_ctx->notes_message);
1210                        return 1;
1211                }
1212                return 0;
1213        }
1214
1215        if (placeholder[0] == 'G') {
1216                if (!c->signature_check.result)
1217                        check_commit_signature(c->commit, &(c->signature_check));
1218                switch (placeholder[1]) {
1219                case 'G':
1220                        if (c->signature_check.gpg_output)
1221                                strbuf_addstr(sb, c->signature_check.gpg_output);
1222                        break;
1223                case '?':
1224                        switch (c->signature_check.result) {
1225                        case 'G':
1226                        case 'B':
1227                        case 'U':
1228                        case 'N':
1229                                strbuf_addch(sb, c->signature_check.result);
1230                        }
1231                        break;
1232                case 'S':
1233                        if (c->signature_check.signer)
1234                                strbuf_addstr(sb, c->signature_check.signer);
1235                        break;
1236                case 'K':
1237                        if (c->signature_check.key)
1238                                strbuf_addstr(sb, c->signature_check.key);
1239                        break;
1240                default:
1241                        return 0;
1242                }
1243                return 2;
1244        }
1245
1246
1247        /* For the rest we have to parse the commit header. */
1248        if (!c->commit_header_parsed)
1249                parse_commit_header(c);
1250
1251        switch (placeholder[0]) {
1252        case 'a':       /* author ... */
1253                return format_person_part(sb, placeholder[1],
1254                                   msg + c->author.off, c->author.len,
1255                                   &c->pretty_ctx->date_mode);
1256        case 'c':       /* committer ... */
1257                return format_person_part(sb, placeholder[1],
1258                                   msg + c->committer.off, c->committer.len,
1259                                   &c->pretty_ctx->date_mode);
1260        case 'e':       /* encoding */
1261                if (c->commit_encoding)
1262                        strbuf_addstr(sb, c->commit_encoding);
1263                return 1;
1264        case 'B':       /* raw body */
1265                /* message_off is always left at the initial newline */
1266                strbuf_addstr(sb, msg + c->message_off + 1);
1267                return 1;
1268        }
1269
1270        /* Now we need to parse the commit message. */
1271        if (!c->commit_message_parsed)
1272                parse_commit_message(c);
1273
1274        switch (placeholder[0]) {
1275        case 's':       /* subject */
1276                format_subject(sb, msg + c->subject_off, " ");
1277                return 1;
1278        case 'f':       /* sanitized subject */
1279                format_sanitized_subject(sb, msg + c->subject_off);
1280                return 1;
1281        case 'b':       /* body */
1282                strbuf_addstr(sb, msg + c->body_off);
1283                return 1;
1284        }
1285        return 0;       /* unknown placeholder */
1286}
1287
1288static size_t format_and_pad_commit(struct strbuf *sb, /* in UTF-8 */
1289                                    const char *placeholder,
1290                                    struct format_commit_context *c)
1291{
1292        struct strbuf local_sb = STRBUF_INIT;
1293        int total_consumed = 0, len, padding = c->padding;
1294        if (padding < 0) {
1295                const char *start = strrchr(sb->buf, '\n');
1296                int occupied;
1297                if (!start)
1298                        start = sb->buf;
1299                occupied = utf8_strnwidth(start, -1, 1);
1300                padding = (-padding) - occupied;
1301        }
1302        while (1) {
1303                int modifier = *placeholder == 'C';
1304                int consumed = format_commit_one(&local_sb, placeholder, c);
1305                total_consumed += consumed;
1306
1307                if (!modifier)
1308                        break;
1309
1310                placeholder += consumed;
1311                if (*placeholder != '%')
1312                        break;
1313                placeholder++;
1314                total_consumed++;
1315        }
1316        len = utf8_strnwidth(local_sb.buf, -1, 1);
1317
1318        if (c->flush_type == flush_left_and_steal) {
1319                const char *ch = sb->buf + sb->len - 1;
1320                while (len > padding && ch > sb->buf) {
1321                        const char *p;
1322                        if (*ch == ' ') {
1323                                ch--;
1324                                padding++;
1325                                continue;
1326                        }
1327                        /* check for trailing ansi sequences */
1328                        if (*ch != 'm')
1329                                break;
1330                        p = ch - 1;
1331                        while (ch - p < 10 && *p != '\033')
1332                                p--;
1333                        if (*p != '\033' ||
1334                            ch + 1 - p != display_mode_esc_sequence_len(p))
1335                                break;
1336                        /*
1337                         * got a good ansi sequence, put it back to
1338                         * local_sb as we're cutting sb
1339                         */
1340                        strbuf_insert(&local_sb, 0, p, ch + 1 - p);
1341                        ch = p - 1;
1342                }
1343                strbuf_setlen(sb, ch + 1 - sb->buf);
1344                c->flush_type = flush_left;
1345        }
1346
1347        if (len > padding) {
1348                switch (c->truncate) {
1349                case trunc_left:
1350                        strbuf_utf8_replace(&local_sb,
1351                                            0, len - (padding - 2),
1352                                            "..");
1353                        break;
1354                case trunc_middle:
1355                        strbuf_utf8_replace(&local_sb,
1356                                            padding / 2 - 1,
1357                                            len - (padding - 2),
1358                                            "..");
1359                        break;
1360                case trunc_right:
1361                        strbuf_utf8_replace(&local_sb,
1362                                            padding - 2, len - (padding - 2),
1363                                            "..");
1364                        break;
1365                case trunc_none:
1366                        break;
1367                }
1368                strbuf_addbuf(sb, &local_sb);
1369        } else {
1370                int sb_len = sb->len, offset = 0;
1371                if (c->flush_type == flush_left)
1372                        offset = padding - len;
1373                else if (c->flush_type == flush_both)
1374                        offset = (padding - len) / 2;
1375                /*
1376                 * we calculate padding in columns, now
1377                 * convert it back to chars
1378                 */
1379                padding = padding - len + local_sb.len;
1380                strbuf_addchars(sb, ' ', padding);
1381                memcpy(sb->buf + sb_len + offset, local_sb.buf,
1382                       local_sb.len);
1383        }
1384        strbuf_release(&local_sb);
1385        c->flush_type = no_flush;
1386        return total_consumed;
1387}
1388
1389static size_t format_commit_item(struct strbuf *sb, /* in UTF-8 */
1390                                 const char *placeholder,
1391                                 void *context)
1392{
1393        int consumed;
1394        size_t orig_len;
1395        enum {
1396                NO_MAGIC,
1397                ADD_LF_BEFORE_NON_EMPTY,
1398                DEL_LF_BEFORE_EMPTY,
1399                ADD_SP_BEFORE_NON_EMPTY
1400        } magic = NO_MAGIC;
1401
1402        switch (placeholder[0]) {
1403        case '-':
1404                magic = DEL_LF_BEFORE_EMPTY;
1405                break;
1406        case '+':
1407                magic = ADD_LF_BEFORE_NON_EMPTY;
1408                break;
1409        case ' ':
1410                magic = ADD_SP_BEFORE_NON_EMPTY;
1411                break;
1412        default:
1413                break;
1414        }
1415        if (magic != NO_MAGIC)
1416                placeholder++;
1417
1418        orig_len = sb->len;
1419        if (((struct format_commit_context *)context)->flush_type != no_flush)
1420                consumed = format_and_pad_commit(sb, placeholder, context);
1421        else
1422                consumed = format_commit_one(sb, placeholder, context);
1423        if (magic == NO_MAGIC)
1424                return consumed;
1425
1426        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
1427                while (sb->len && sb->buf[sb->len - 1] == '\n')
1428                        strbuf_setlen(sb, sb->len - 1);
1429        } else if (orig_len != sb->len) {
1430                if (magic == ADD_LF_BEFORE_NON_EMPTY)
1431                        strbuf_insert(sb, orig_len, "\n", 1);
1432                else if (magic == ADD_SP_BEFORE_NON_EMPTY)
1433                        strbuf_insert(sb, orig_len, " ", 1);
1434        }
1435        return consumed + 1;
1436}
1437
1438static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
1439                                   void *context)
1440{
1441        struct userformat_want *w = context;
1442
1443        if (*placeholder == '+' || *placeholder == '-' || *placeholder == ' ')
1444                placeholder++;
1445
1446        switch (*placeholder) {
1447        case 'N':
1448                w->notes = 1;
1449                break;
1450        }
1451        return 0;
1452}
1453
1454void userformat_find_requirements(const char *fmt, struct userformat_want *w)
1455{
1456        struct strbuf dummy = STRBUF_INIT;
1457
1458        if (!fmt) {
1459                if (!user_format)
1460                        return;
1461                fmt = user_format;
1462        }
1463        strbuf_expand(&dummy, fmt, userformat_want_item, w);
1464        strbuf_release(&dummy);
1465}
1466
1467void format_commit_message(const struct commit *commit,
1468                           const char *format, struct strbuf *sb,
1469                           const struct pretty_print_context *pretty_ctx)
1470{
1471        struct format_commit_context context;
1472        const char *output_enc = pretty_ctx->output_encoding;
1473        const char *utf8 = "UTF-8";
1474
1475        memset(&context, 0, sizeof(context));
1476        context.commit = commit;
1477        context.pretty_ctx = pretty_ctx;
1478        context.wrap_start = sb->len;
1479        /*
1480         * convert a commit message to UTF-8 first
1481         * as far as 'format_commit_item' assumes it in UTF-8
1482         */
1483        context.message = logmsg_reencode(commit,
1484                                          &context.commit_encoding,
1485                                          utf8);
1486
1487        strbuf_expand(sb, format, format_commit_item, &context);
1488        rewrap_message_tail(sb, &context, 0, 0, 0);
1489
1490        /* then convert a commit message to an actual output encoding */
1491        if (output_enc) {
1492                if (same_encoding(utf8, output_enc))
1493                        output_enc = NULL;
1494        } else {
1495                if (context.commit_encoding &&
1496                    !same_encoding(context.commit_encoding, utf8))
1497                        output_enc = context.commit_encoding;
1498        }
1499
1500        if (output_enc) {
1501                int outsz;
1502                char *out = reencode_string_len(sb->buf, sb->len,
1503                                                output_enc, utf8, &outsz);
1504                if (out)
1505                        strbuf_attach(sb, out, outsz, outsz + 1);
1506        }
1507
1508        free(context.commit_encoding);
1509        unuse_commit_buffer(commit, context.message);
1510}
1511
1512static void pp_header(struct pretty_print_context *pp,
1513                      const char *encoding,
1514                      const struct commit *commit,
1515                      const char **msg_p,
1516                      struct strbuf *sb)
1517{
1518        int parents_shown = 0;
1519
1520        for (;;) {
1521                const char *name, *line = *msg_p;
1522                int linelen = get_one_line(*msg_p);
1523
1524                if (!linelen)
1525                        return;
1526                *msg_p += linelen;
1527
1528                if (linelen == 1)
1529                        /* End of header */
1530                        return;
1531
1532                if (pp->fmt == CMIT_FMT_RAW) {
1533                        strbuf_add(sb, line, linelen);
1534                        continue;
1535                }
1536
1537                if (starts_with(line, "parent ")) {
1538                        if (linelen != 48)
1539                                die("bad parent line in commit");
1540                        continue;
1541                }
1542
1543                if (!parents_shown) {
1544                        unsigned num = commit_list_count(commit->parents);
1545                        /* with enough slop */
1546                        strbuf_grow(sb, num * 50 + 20);
1547                        add_merge_info(pp, sb, commit);
1548                        parents_shown = 1;
1549                }
1550
1551                /*
1552                 * MEDIUM == DEFAULT shows only author with dates.
1553                 * FULL shows both authors but not dates.
1554                 * FULLER shows both authors and dates.
1555                 */
1556                if (skip_prefix(line, "author ", &name)) {
1557                        strbuf_grow(sb, linelen + 80);
1558                        pp_user_info(pp, "Author", sb, name, encoding);
1559                }
1560                if (skip_prefix(line, "committer ", &name) &&
1561                    (pp->fmt == CMIT_FMT_FULL || pp->fmt == CMIT_FMT_FULLER)) {
1562                        strbuf_grow(sb, linelen + 80);
1563                        pp_user_info(pp, "Commit", sb, name, encoding);
1564                }
1565        }
1566}
1567
1568void pp_title_line(struct pretty_print_context *pp,
1569                   const char **msg_p,
1570                   struct strbuf *sb,
1571                   const char *encoding,
1572                   int need_8bit_cte)
1573{
1574        static const int max_length = 78; /* per rfc2047 */
1575        struct strbuf title;
1576
1577        strbuf_init(&title, 80);
1578        *msg_p = format_subject(&title, *msg_p,
1579                                pp->preserve_subject ? "\n" : " ");
1580
1581        strbuf_grow(sb, title.len + 1024);
1582        if (pp->subject) {
1583                strbuf_addstr(sb, pp->subject);
1584                if (needs_rfc2047_encoding(title.buf, title.len, RFC2047_SUBJECT))
1585                        add_rfc2047(sb, title.buf, title.len,
1586                                                encoding, RFC2047_SUBJECT);
1587                else
1588                        strbuf_add_wrapped_bytes(sb, title.buf, title.len,
1589                                         -last_line_length(sb), 1, max_length);
1590        } else {
1591                strbuf_addbuf(sb, &title);
1592        }
1593        strbuf_addch(sb, '\n');
1594
1595        if (need_8bit_cte == 0) {
1596                int i;
1597                for (i = 0; i < pp->in_body_headers.nr; i++) {
1598                        if (has_non_ascii(pp->in_body_headers.items[i].string)) {
1599                                need_8bit_cte = 1;
1600                                break;
1601                        }
1602                }
1603        }
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
1619        if (pp->in_body_headers.nr) {
1620                int i;
1621                for (i = 0; i < pp->in_body_headers.nr; i++) {
1622                        strbuf_addstr(sb, pp->in_body_headers.items[i].string);
1623                        free(pp->in_body_headers.items[i].string);
1624                }
1625                string_list_clear(&pp->in_body_headers, 0);
1626                strbuf_addch(sb, '\n');
1627        }
1628
1629        strbuf_release(&title);
1630}
1631
1632void pp_remainder(struct pretty_print_context *pp,
1633                  const char **msg_p,
1634                  struct strbuf *sb,
1635                  int indent)
1636{
1637        int first = 1;
1638        for (;;) {
1639                const char *line = *msg_p;
1640                int linelen = get_one_line(line);
1641                *msg_p += linelen;
1642
1643                if (!linelen)
1644                        break;
1645
1646                if (is_empty_line(line, &linelen)) {
1647                        if (first)
1648                                continue;
1649                        if (pp->fmt == CMIT_FMT_SHORT)
1650                                break;
1651                }
1652                first = 0;
1653
1654                strbuf_grow(sb, linelen + indent + 20);
1655                if (indent)
1656                        strbuf_addchars(sb, ' ', indent);
1657                strbuf_add(sb, line, linelen);
1658                strbuf_addch(sb, '\n');
1659        }
1660}
1661
1662void pretty_print_commit(struct pretty_print_context *pp,
1663                         const struct commit *commit,
1664                         struct strbuf *sb)
1665{
1666        unsigned long beginning_of_body;
1667        int indent = 4;
1668        const char *msg;
1669        const char *reencoded;
1670        const char *encoding;
1671        int need_8bit_cte = pp->need_8bit_cte;
1672
1673        if (pp->fmt == CMIT_FMT_USERFORMAT) {
1674                format_commit_message(commit, user_format, sb, pp);
1675                return;
1676        }
1677
1678        encoding = get_log_output_encoding();
1679        msg = reencoded = logmsg_reencode(commit, NULL, encoding);
1680
1681        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1682                indent = 0;
1683
1684        /*
1685         * We need to check and emit Content-type: to mark it
1686         * as 8-bit if we haven't done so.
1687         */
1688        if (pp->fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1689                int i, ch, in_body;
1690
1691                for (in_body = i = 0; (ch = msg[i]); i++) {
1692                        if (!in_body) {
1693                                /* author could be non 7-bit ASCII but
1694                                 * the log may be so; skip over the
1695                                 * header part first.
1696                                 */
1697                                if (ch == '\n' && msg[i+1] == '\n')
1698                                        in_body = 1;
1699                        }
1700                        else if (non_ascii(ch)) {
1701                                need_8bit_cte = 1;
1702                                break;
1703                        }
1704                }
1705        }
1706
1707        pp_header(pp, encoding, commit, &msg, sb);
1708        if (pp->fmt != CMIT_FMT_ONELINE && !pp->subject) {
1709                strbuf_addch(sb, '\n');
1710        }
1711
1712        /* Skip excess blank lines at the beginning of body, if any... */
1713        msg = skip_empty_lines(msg);
1714
1715        /* These formats treat the title line specially. */
1716        if (pp->fmt == CMIT_FMT_ONELINE || pp->fmt == CMIT_FMT_EMAIL)
1717                pp_title_line(pp, &msg, sb, encoding, need_8bit_cte);
1718
1719        beginning_of_body = sb->len;
1720        if (pp->fmt != CMIT_FMT_ONELINE)
1721                pp_remainder(pp, &msg, sb, indent);
1722        strbuf_rtrim(sb);
1723
1724        /* Make sure there is an EOLN for the non-oneline case */
1725        if (pp->fmt != CMIT_FMT_ONELINE)
1726                strbuf_addch(sb, '\n');
1727
1728        /*
1729         * The caller may append additional body text in e-mail
1730         * format.  Make sure we did not strip the blank line
1731         * between the header and the body.
1732         */
1733        if (pp->fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1734                strbuf_addch(sb, '\n');
1735
1736        unuse_commit_buffer(commit, reencoded);
1737}
1738
1739void pp_commit_easy(enum cmit_fmt fmt, const struct commit *commit,
1740                    struct strbuf *sb)
1741{
1742        struct pretty_print_context pp = {0};
1743        pp.fmt = fmt;
1744        pretty_print_commit(&pp, commit, sb);
1745}