Andrew's git - gitweb.git/blob

c2c8901f8f826ea2d81e7c0185abd81964288112
   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
  13static char *user_format;
  14static struct cmt_fmt_map {
  15        const char *name;
  16        enum cmit_fmt format;
  17        int is_tformat;
  18        int is_alias;
  19        const char *user_format;
  20} *commit_formats;
  21static size_t commit_formats_len;
  22static struct cmt_fmt_map *find_commit_format(const char *sought);
  23
  24static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  25{
  26        free(user_format);
  27        user_format = xstrdup(cp);
  28        if (is_tformat)
  29                rev->use_terminator = 1;
  30        rev->commit_format = CMIT_FMT_USERFORMAT;
  31}
  32
  33static void setup_commit_formats(void)
  34{
  35        struct cmt_fmt_map builtin_formats[] = {
  36                { "raw",        CMIT_FMT_RAW,           0 },
  37                { "medium",     CMIT_FMT_MEDIUM,        0 },
  38                { "short",      CMIT_FMT_SHORT,         0 },
  39                { "email",      CMIT_FMT_EMAIL,         0 },
  40                { "fuller",     CMIT_FMT_FULLER,        0 },
  41                { "full",       CMIT_FMT_FULL,          0 },
  42                { "oneline",    CMIT_FMT_ONELINE,       1 }
  43        };
  44        commit_formats_len = ARRAY_SIZE(builtin_formats);
  45        commit_formats = xmalloc(commit_formats_len *
  46                                 sizeof(*builtin_formats));
  47        memcpy(commit_formats, builtin_formats,
  48               sizeof(*builtin_formats)*ARRAY_SIZE(builtin_formats));
  49}
  50
  51static struct cmt_fmt_map *find_commit_format_recursive(const char *sought,
  52                                                        const char *original,
  53                                                        int num_redirections)
  54{
  55        struct cmt_fmt_map *found = NULL;
  56        size_t found_match_len = 0;
  57        int i;
  58
  59        if (num_redirections >= commit_formats_len)
  60                die("invalid --pretty format: "
  61                    "'%s' references an alias which points to itself",
  62                    original);
  63
  64        for (i = 0; i < commit_formats_len; i++) {
  65                size_t match_len;
  66
  67                if (prefixcmp(commit_formats[i].name, sought))
  68                        continue;
  69
  70                match_len = strlen(commit_formats[i].name);
  71                if (found == NULL || found_match_len > match_len) {
  72                        found = &commit_formats[i];
  73                        found_match_len = match_len;
  74                }
  75        }
  76
  77        if (found && found->is_alias) {
  78                found = find_commit_format_recursive(found->user_format,
  79                                                     original,
  80                                                     num_redirections+1);
  81        }
  82
  83        return found;
  84}
  85
  86static struct cmt_fmt_map *find_commit_format(const char *sought)
  87{
  88        if (!commit_formats)
  89                setup_commit_formats();
  90
  91        return find_commit_format_recursive(sought, sought, 0);
  92}
  93
  94void get_commit_format(const char *arg, struct rev_info *rev)
  95{
  96        struct cmt_fmt_map *commit_format;
  97
  98        rev->use_terminator = 0;
  99        if (!arg || !*arg) {
 100                rev->commit_format = CMIT_FMT_DEFAULT;
 101                return;
 102        }
 103        if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
 104                save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
 105                return;
 106        }
 107
 108        if (strchr(arg, '%')) {
 109                save_user_format(rev, arg, 1);
 110                return;
 111        }
 112
 113        commit_format = find_commit_format(arg);
 114        if (!commit_format)
 115                die("invalid --pretty format: %s", arg);
 116
 117        rev->commit_format = commit_format->format;
 118        rev->use_terminator = commit_format->is_tformat;
 119}
 120
 121/*
 122 * Generic support for pretty-printing the header
 123 */
 124static int get_one_line(const char *msg)
 125{
 126        int ret = 0;
 127
 128        for (;;) {
 129                char c = *msg++;
 130                if (!c)
 131                        break;
 132                ret++;
 133                if (c == '\n')
 134                        break;
 135        }
 136        return ret;
 137}
 138
 139/* High bit set, or ISO-2022-INT */
 140static int non_ascii(int ch)
 141{
 142        return !isascii(ch) || ch == '\033';
 143}
 144
 145int has_non_ascii(const char *s)
 146{
 147        int ch;
 148        if (!s)
 149                return 0;
 150        while ((ch = *s++) != '\0') {
 151                if (non_ascii(ch))
 152                        return 1;
 153        }
 154        return 0;
 155}
 156
 157static int is_rfc2047_special(char ch)
 158{
 159        return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 160}
 161
 162static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 163                       const char *encoding)
 164{
 165        int i, last;
 166
 167        for (i = 0; i < len; i++) {
 168                int ch = line[i];
 169                if (non_ascii(ch))
 170                        goto needquote;
 171                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 172                        goto needquote;
 173        }
 174        strbuf_add(sb, line, len);
 175        return;
 176
 177needquote:
 178        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 179        strbuf_addf(sb, "=?%s?q?", encoding);
 180        for (i = last = 0; i < len; i++) {
 181                unsigned ch = line[i] & 0xFF;
 182                /*
 183                 * We encode ' ' using '=20' even though rfc2047
 184                 * allows using '_' for readability.  Unfortunately,
 185                 * many programs do not understand this and just
 186                 * leave the underscore in place.
 187                 */
 188                if (is_rfc2047_special(ch) || ch == ' ') {
 189                        strbuf_add(sb, line + last, i - last);
 190                        strbuf_addf(sb, "=%02X", ch);
 191                        last = i + 1;
 192                }
 193        }
 194        strbuf_add(sb, line + last, len - last);
 195        strbuf_addstr(sb, "?=");
 196}
 197
 198void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 199                  const char *line, enum date_mode dmode,
 200                  const char *encoding)
 201{
 202        char *date;
 203        int namelen;
 204        unsigned long time;
 205        int tz;
 206
 207        if (fmt == CMIT_FMT_ONELINE)
 208                return;
 209        date = strchr(line, '>');
 210        if (!date)
 211                return;
 212        namelen = ++date - line;
 213        time = strtoul(date, &date, 10);
 214        tz = strtol(date, NULL, 10);
 215
 216        if (fmt == CMIT_FMT_EMAIL) {
 217                char *name_tail = strchr(line, '<');
 218                int display_name_length;
 219                if (!name_tail)
 220                        return;
 221                while (line < name_tail && isspace(name_tail[-1]))
 222                        name_tail--;
 223                display_name_length = name_tail - line;
 224                strbuf_addstr(sb, "From: ");
 225                add_rfc2047(sb, line, display_name_length, encoding);
 226                strbuf_add(sb, name_tail, namelen - display_name_length);
 227                strbuf_addch(sb, '\n');
 228        } else {
 229                strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 230                              (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 231                              "    ", namelen, line);
 232        }
 233        switch (fmt) {
 234        case CMIT_FMT_MEDIUM:
 235                strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 236                break;
 237        case CMIT_FMT_EMAIL:
 238                strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 239                break;
 240        case CMIT_FMT_FULLER:
 241                strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 242                break;
 243        default:
 244                /* notin' */
 245                break;
 246        }
 247}
 248
 249static int is_empty_line(const char *line, int *len_p)
 250{
 251        int len = *len_p;
 252        while (len && isspace(line[len-1]))
 253                len--;
 254        *len_p = len;
 255        return !len;
 256}
 257
 258static const char *skip_empty_lines(const char *msg)
 259{
 260        for (;;) {
 261                int linelen = get_one_line(msg);
 262                int ll = linelen;
 263                if (!linelen)
 264                        break;
 265                if (!is_empty_line(msg, &ll))
 266                        break;
 267                msg += linelen;
 268        }
 269        return msg;
 270}
 271
 272static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 273                        const struct commit *commit, int abbrev)
 274{
 275        struct commit_list *parent = commit->parents;
 276
 277        if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 278            !parent || !parent->next)
 279                return;
 280
 281        strbuf_addstr(sb, "Merge:");
 282
 283        while (parent) {
 284                struct commit *p = parent->item;
 285                const char *hex = NULL;
 286                if (abbrev)
 287                        hex = find_unique_abbrev(p->object.sha1, abbrev);
 288                if (!hex)
 289                        hex = sha1_to_hex(p->object.sha1);
 290                parent = parent->next;
 291
 292                strbuf_addf(sb, " %s", hex);
 293        }
 294        strbuf_addch(sb, '\n');
 295}
 296
 297static char *get_header(const struct commit *commit, const char *key)
 298{
 299        int key_len = strlen(key);
 300        const char *line = commit->buffer;
 301
 302        for (;;) {
 303                const char *eol = strchr(line, '\n'), *next;
 304
 305                if (line == eol)
 306                        return NULL;
 307                if (!eol) {
 308                        eol = line + strlen(line);
 309                        next = NULL;
 310                } else
 311                        next = eol + 1;
 312                if (eol - line > key_len &&
 313                    !strncmp(line, key, key_len) &&
 314                    line[key_len] == ' ') {
 315                        return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 316                }
 317                line = next;
 318        }
 319}
 320
 321static char *replace_encoding_header(char *buf, const char *encoding)
 322{
 323        struct strbuf tmp = STRBUF_INIT;
 324        size_t start, len;
 325        char *cp = buf;
 326
 327        /* guess if there is an encoding header before a \n\n */
 328        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 329                cp = strchr(cp, '\n');
 330                if (!cp || *++cp == '\n')
 331                        return buf;
 332        }
 333        start = cp - buf;
 334        cp = strchr(cp, '\n');
 335        if (!cp)
 336                return buf; /* should not happen but be defensive */
 337        len = cp + 1 - (buf + start);
 338
 339        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 340        if (is_encoding_utf8(encoding)) {
 341                /* we have re-coded to UTF-8; drop the header */
 342                strbuf_remove(&tmp, start, len);
 343        } else {
 344                /* just replaces XXXX in 'encoding XXXX\n' */
 345                strbuf_splice(&tmp, start + strlen("encoding "),
 346                                          len - strlen("encoding \n"),
 347                                          encoding, strlen(encoding));
 348        }
 349        return strbuf_detach(&tmp, NULL);
 350}
 351
 352static char *logmsg_reencode(const struct commit *commit,
 353                             const char *output_encoding)
 354{
 355        static const char *utf8 = "UTF-8";
 356        const char *use_encoding;
 357        char *encoding;
 358        char *out;
 359
 360        if (!*output_encoding)
 361                return NULL;
 362        encoding = get_header(commit, "encoding");
 363        use_encoding = encoding ? encoding : utf8;
 364        if (!strcmp(use_encoding, output_encoding))
 365                if (encoding) /* we'll strip encoding header later */
 366                        out = xstrdup(commit->buffer);
 367                else
 368                        return NULL; /* nothing to do */
 369        else
 370                out = reencode_string(commit->buffer,
 371                                      output_encoding, use_encoding);
 372        if (out)
 373                out = replace_encoding_header(out, output_encoding);
 374
 375        free(encoding);
 376        return out;
 377}
 378
 379static int mailmap_name(char *email, int email_len, char *name, int name_len)
 380{
 381        static struct string_list *mail_map;
 382        if (!mail_map) {
 383                mail_map = xcalloc(1, sizeof(*mail_map));
 384                read_mailmap(mail_map, NULL);
 385        }
 386        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 387}
 388
 389static size_t format_person_part(struct strbuf *sb, char part,
 390                                 const char *msg, int len, enum date_mode dmode)
 391{
 392        /* currently all placeholders have same length */
 393        const int placeholder_len = 2;
 394        int start, end, tz = 0;
 395        unsigned long date = 0;
 396        char *ep;
 397        const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 398        char person_name[1024];
 399        char person_mail[1024];
 400
 401        /* advance 'end' to point to email start delimiter */
 402        for (end = 0; end < len && msg[end] != '<'; end++)
 403                ; /* do nothing */
 404
 405        /*
 406         * When end points at the '<' that we found, it should have
 407         * matching '>' later, which means 'end' must be strictly
 408         * below len - 1.
 409         */
 410        if (end >= len - 2)
 411                goto skip;
 412
 413        /* Seek for both name and email part */
 414        name_start = msg;
 415        name_end = msg+end;
 416        while (name_end > name_start && isspace(*(name_end-1)))
 417                name_end--;
 418        mail_start = msg+end+1;
 419        mail_end = mail_start;
 420        while (mail_end < msg_end && *mail_end != '>')
 421                mail_end++;
 422        if (mail_end == msg_end)
 423                goto skip;
 424        end = mail_end-msg;
 425
 426        if (part == 'N' || part == 'E') { /* mailmap lookup */
 427                strlcpy(person_name, name_start, name_end-name_start+1);
 428                strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 429                mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 430                name_start = person_name;
 431                name_end = name_start + strlen(person_name);
 432                mail_start = person_mail;
 433                mail_end = mail_start +  strlen(person_mail);
 434        }
 435        if (part == 'n' || part == 'N') {       /* name */
 436                strbuf_add(sb, name_start, name_end-name_start);
 437                return placeholder_len;
 438        }
 439        if (part == 'e' || part == 'E') {       /* email */
 440                strbuf_add(sb, mail_start, mail_end-mail_start);
 441                return placeholder_len;
 442        }
 443
 444        /* advance 'start' to point to date start delimiter */
 445        for (start = end + 1; start < len && isspace(msg[start]); start++)
 446                ; /* do nothing */
 447        if (start >= len)
 448                goto skip;
 449        date = strtoul(msg + start, &ep, 10);
 450        if (msg + start == ep)
 451                goto skip;
 452
 453        if (part == 't') {      /* date, UNIX timestamp */
 454                strbuf_add(sb, msg + start, ep - (msg + start));
 455                return placeholder_len;
 456        }
 457
 458        /* parse tz */
 459        for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 460                ; /* do nothing */
 461        if (start + 1 < len) {
 462                tz = strtoul(msg + start + 1, NULL, 10);
 463                if (msg[start] == '-')
 464                        tz = -tz;
 465        }
 466
 467        switch (part) {
 468        case 'd':       /* date */
 469                strbuf_addstr(sb, show_date(date, tz, dmode));
 470                return placeholder_len;
 471        case 'D':       /* date, RFC2822 style */
 472                strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 473                return placeholder_len;
 474        case 'r':       /* date, relative */
 475                strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 476                return placeholder_len;
 477        case 'i':       /* date, ISO 8601 */
 478                strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 479                return placeholder_len;
 480        }
 481
 482skip:
 483        /*
 484         * bogus commit, 'sb' cannot be updated, but we still need to
 485         * compute a valid return value.
 486         */
 487        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 488            || part == 'D' || part == 'r' || part == 'i')
 489                return placeholder_len;
 490
 491        return 0; /* unknown placeholder */
 492}
 493
 494struct chunk {
 495        size_t off;
 496        size_t len;
 497};
 498
 499struct format_commit_context {
 500        const struct commit *commit;
 501        const struct pretty_print_context *pretty_ctx;
 502        unsigned commit_header_parsed:1;
 503        unsigned commit_message_parsed:1;
 504        size_t width, indent1, indent2;
 505
 506        /* These offsets are relative to the start of the commit message. */
 507        struct chunk author;
 508        struct chunk committer;
 509        struct chunk encoding;
 510        size_t message_off;
 511        size_t subject_off;
 512        size_t body_off;
 513
 514        /* The following ones are relative to the result struct strbuf. */
 515        struct chunk abbrev_commit_hash;
 516        struct chunk abbrev_tree_hash;
 517        struct chunk abbrev_parent_hashes;
 518        size_t wrap_start;
 519};
 520
 521static int add_again(struct strbuf *sb, struct chunk *chunk)
 522{
 523        if (chunk->len) {
 524                strbuf_adddup(sb, chunk->off, chunk->len);
 525                return 1;
 526        }
 527
 528        /*
 529         * We haven't seen this chunk before.  Our caller is surely
 530         * going to add it the hard way now.  Remember the most likely
 531         * start of the to-be-added chunk: the current end of the
 532         * struct strbuf.
 533         */
 534        chunk->off = sb->len;
 535        return 0;
 536}
 537
 538static void parse_commit_header(struct format_commit_context *context)
 539{
 540        const char *msg = context->commit->buffer;
 541        int i;
 542
 543        for (i = 0; msg[i]; i++) {
 544                int eol;
 545                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 546                        ; /* do nothing */
 547
 548                if (i == eol) {
 549                        break;
 550                } else if (!prefixcmp(msg + i, "author ")) {
 551                        context->author.off = i + 7;
 552                        context->author.len = eol - i - 7;
 553                } else if (!prefixcmp(msg + i, "committer ")) {
 554                        context->committer.off = i + 10;
 555                        context->committer.len = eol - i - 10;
 556                } else if (!prefixcmp(msg + i, "encoding ")) {
 557                        context->encoding.off = i + 9;
 558                        context->encoding.len = eol - i - 9;
 559                }
 560                i = eol;
 561        }
 562        context->message_off = i;
 563        context->commit_header_parsed = 1;
 564}
 565
 566static int istitlechar(char c)
 567{
 568        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 569                (c >= '0' && c <= '9') || c == '.' || c == '_';
 570}
 571
 572static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 573{
 574        size_t trimlen;
 575        size_t start_len = sb->len;
 576        int space = 2;
 577
 578        for (; *msg && *msg != '\n'; msg++) {
 579                if (istitlechar(*msg)) {
 580                        if (space == 1)
 581                                strbuf_addch(sb, '-');
 582                        space = 0;
 583                        strbuf_addch(sb, *msg);
 584                        if (*msg == '.')
 585                                while (*(msg+1) == '.')
 586                                        msg++;
 587                } else
 588                        space |= 1;
 589        }
 590
 591        /* trim any trailing '.' or '-' characters */
 592        trimlen = 0;
 593        while (sb->len - trimlen > start_len &&
 594                (sb->buf[sb->len - 1 - trimlen] == '.'
 595                || sb->buf[sb->len - 1 - trimlen] == '-'))
 596                trimlen++;
 597        strbuf_remove(sb, sb->len - trimlen, trimlen);
 598}
 599
 600const char *format_subject(struct strbuf *sb, const char *msg,
 601                           const char *line_separator)
 602{
 603        int first = 1;
 604
 605        for (;;) {
 606                const char *line = msg;
 607                int linelen = get_one_line(line);
 608
 609                msg += linelen;
 610                if (!linelen || is_empty_line(line, &linelen))
 611                        break;
 612
 613                if (!sb)
 614                        continue;
 615                strbuf_grow(sb, linelen + 2);
 616                if (!first)
 617                        strbuf_addstr(sb, line_separator);
 618                strbuf_add(sb, line, linelen);
 619                first = 0;
 620        }
 621        return msg;
 622}
 623
 624static void parse_commit_message(struct format_commit_context *c)
 625{
 626        const char *msg = c->commit->buffer + c->message_off;
 627        const char *start = c->commit->buffer;
 628
 629        msg = skip_empty_lines(msg);
 630        c->subject_off = msg - start;
 631
 632        msg = format_subject(NULL, msg, NULL);
 633        msg = skip_empty_lines(msg);
 634        c->body_off = msg - start;
 635
 636        c->commit_message_parsed = 1;
 637}
 638
 639static void format_decoration(struct strbuf *sb, const struct commit *commit)
 640{
 641        struct name_decoration *d;
 642        const char *prefix = " (";
 643
 644        load_ref_decorations(DECORATE_SHORT_REFS);
 645        d = lookup_decoration(&name_decoration, &commit->object);
 646        while (d) {
 647                strbuf_addstr(sb, prefix);
 648                prefix = ", ";
 649                strbuf_addstr(sb, d->name);
 650                d = d->next;
 651        }
 652        if (prefix[0] == ',')
 653                strbuf_addch(sb, ')');
 654}
 655
 656static void strbuf_wrap(struct strbuf *sb, size_t pos,
 657                        size_t width, size_t indent1, size_t indent2)
 658{
 659        struct strbuf tmp = STRBUF_INIT;
 660
 661        if (pos)
 662                strbuf_add(&tmp, sb->buf, pos);
 663        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 664                                (int) indent1, (int) indent2, (int) width);
 665        strbuf_swap(&tmp, sb);
 666        strbuf_release(&tmp);
 667}
 668
 669static void rewrap_message_tail(struct strbuf *sb,
 670                                struct format_commit_context *c,
 671                                size_t new_width, size_t new_indent1,
 672                                size_t new_indent2)
 673{
 674        if (c->width == new_width && c->indent1 == new_indent1 &&
 675            c->indent2 == new_indent2)
 676                return;
 677        if (c->wrap_start < sb->len)
 678                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 679        c->wrap_start = sb->len;
 680        c->width = new_width;
 681        c->indent1 = new_indent1;
 682        c->indent2 = new_indent2;
 683}
 684
 685static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 686                                void *context)
 687{
 688        struct format_commit_context *c = context;
 689        const struct commit *commit = c->commit;
 690        const char *msg = commit->buffer;
 691        struct commit_list *p;
 692        int h1, h2;
 693
 694        /* these are independent of the commit */
 695        switch (placeholder[0]) {
 696        case 'C':
 697                if (placeholder[1] == '(') {
 698                        const char *end = strchr(placeholder + 2, ')');
 699                        char color[COLOR_MAXLEN];
 700                        if (!end)
 701                                return 0;
 702                        color_parse_mem(placeholder + 2,
 703                                        end - (placeholder + 2),
 704                                        "--pretty format", color);
 705                        strbuf_addstr(sb, color);
 706                        return end - placeholder + 1;
 707                }
 708                if (!prefixcmp(placeholder + 1, "red")) {
 709                        strbuf_addstr(sb, GIT_COLOR_RED);
 710                        return 4;
 711                } else if (!prefixcmp(placeholder + 1, "green")) {
 712                        strbuf_addstr(sb, GIT_COLOR_GREEN);
 713                        return 6;
 714                } else if (!prefixcmp(placeholder + 1, "blue")) {
 715                        strbuf_addstr(sb, GIT_COLOR_BLUE);
 716                        return 5;
 717                } else if (!prefixcmp(placeholder + 1, "reset")) {
 718                        strbuf_addstr(sb, GIT_COLOR_RESET);
 719                        return 6;
 720                } else
 721                        return 0;
 722        case 'n':               /* newline */
 723                strbuf_addch(sb, '\n');
 724                return 1;
 725        case 'x':
 726                /* %x00 == NUL, %x0a == LF, etc. */
 727                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 728                    h1 <= 16 &&
 729                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 730                    h2 <= 16) {
 731                        strbuf_addch(sb, (h1<<4)|h2);
 732                        return 3;
 733                } else
 734                        return 0;
 735        case 'w':
 736                if (placeholder[1] == '(') {
 737                        unsigned long width = 0, indent1 = 0, indent2 = 0;
 738                        char *next;
 739                        const char *start = placeholder + 2;
 740                        const char *end = strchr(start, ')');
 741                        if (!end)
 742                                return 0;
 743                        if (end > start) {
 744                                width = strtoul(start, &next, 10);
 745                                if (*next == ',') {
 746                                        indent1 = strtoul(next + 1, &next, 10);
 747                                        if (*next == ',') {
 748                                                indent2 = strtoul(next + 1,
 749                                                                 &next, 10);
 750                                        }
 751                                }
 752                                if (*next != ')')
 753                                        return 0;
 754                        }
 755                        rewrap_message_tail(sb, c, width, indent1, indent2);
 756                        return end - placeholder + 1;
 757                } else
 758                        return 0;
 759        }
 760
 761        /* these depend on the commit */
 762        if (!commit->object.parsed)
 763                parse_object(commit->object.sha1);
 764
 765        switch (placeholder[0]) {
 766        case 'H':               /* commit hash */
 767                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 768                return 1;
 769        case 'h':               /* abbreviated commit hash */
 770                if (add_again(sb, &c->abbrev_commit_hash))
 771                        return 1;
 772                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 773                                                     DEFAULT_ABBREV));
 774                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 775                return 1;
 776        case 'T':               /* tree hash */
 777                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 778                return 1;
 779        case 't':               /* abbreviated tree hash */
 780                if (add_again(sb, &c->abbrev_tree_hash))
 781                        return 1;
 782                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 783                                                     DEFAULT_ABBREV));
 784                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 785                return 1;
 786        case 'P':               /* parent hashes */
 787                for (p = commit->parents; p; p = p->next) {
 788                        if (p != commit->parents)
 789                                strbuf_addch(sb, ' ');
 790                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 791                }
 792                return 1;
 793        case 'p':               /* abbreviated parent hashes */
 794                if (add_again(sb, &c->abbrev_parent_hashes))
 795                        return 1;
 796                for (p = commit->parents; p; p = p->next) {
 797                        if (p != commit->parents)
 798                                strbuf_addch(sb, ' ');
 799                        strbuf_addstr(sb, find_unique_abbrev(
 800                                        p->item->object.sha1, DEFAULT_ABBREV));
 801                }
 802                c->abbrev_parent_hashes.len = sb->len -
 803                                              c->abbrev_parent_hashes.off;
 804                return 1;
 805        case 'm':               /* left/right/bottom */
 806                strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 807                                 ? '-'
 808                                 : (commit->object.flags & SYMMETRIC_LEFT)
 809                                 ? '<'
 810                                 : '>');
 811                return 1;
 812        case 'd':
 813                format_decoration(sb, commit);
 814                return 1;
 815        case 'g':               /* reflog info */
 816                switch(placeholder[1]) {
 817                case 'd':       /* reflog selector */
 818                case 'D':
 819                        if (c->pretty_ctx->reflog_info)
 820                                get_reflog_selector(sb,
 821                                                    c->pretty_ctx->reflog_info,
 822                                                    c->pretty_ctx->date_mode,
 823                                                    (placeholder[1] == 'd'));
 824                        return 2;
 825                case 's':       /* reflog message */
 826                        if (c->pretty_ctx->reflog_info)
 827                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
 828                        return 2;
 829                }
 830                return 0;       /* unknown %g placeholder */
 831        case 'N':
 832                if (c->pretty_ctx->show_notes) {
 833                        format_display_notes(commit->object.sha1, sb,
 834                                    git_log_output_encoding ? git_log_output_encoding
 835                                                            : git_commit_encoding, 0);
 836                        return 1;
 837                }
 838                return 0;
 839        }
 840
 841        /* For the rest we have to parse the commit header. */
 842        if (!c->commit_header_parsed)
 843                parse_commit_header(c);
 844
 845        switch (placeholder[0]) {
 846        case 'a':       /* author ... */
 847                return format_person_part(sb, placeholder[1],
 848                                   msg + c->author.off, c->author.len,
 849                                   c->pretty_ctx->date_mode);
 850        case 'c':       /* committer ... */
 851                return format_person_part(sb, placeholder[1],
 852                                   msg + c->committer.off, c->committer.len,
 853                                   c->pretty_ctx->date_mode);
 854        case 'e':       /* encoding */
 855                strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 856                return 1;
 857        }
 858
 859        /* Now we need to parse the commit message. */
 860        if (!c->commit_message_parsed)
 861                parse_commit_message(c);
 862
 863        switch (placeholder[0]) {
 864        case 's':       /* subject */
 865                format_subject(sb, msg + c->subject_off, " ");
 866                return 1;
 867        case 'f':       /* sanitized subject */
 868                format_sanitized_subject(sb, msg + c->subject_off);
 869                return 1;
 870        case 'b':       /* body */
 871                strbuf_addstr(sb, msg + c->body_off);
 872                return 1;
 873        }
 874        return 0;       /* unknown placeholder */
 875}
 876
 877static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 878                                 void *context)
 879{
 880        int consumed;
 881        size_t orig_len;
 882        enum {
 883                NO_MAGIC,
 884                ADD_LF_BEFORE_NON_EMPTY,
 885                DEL_LF_BEFORE_EMPTY,
 886        } magic = NO_MAGIC;
 887
 888        switch (placeholder[0]) {
 889        case '-':
 890                magic = DEL_LF_BEFORE_EMPTY;
 891                break;
 892        case '+':
 893                magic = ADD_LF_BEFORE_NON_EMPTY;
 894                break;
 895        default:
 896                break;
 897        }
 898        if (magic != NO_MAGIC)
 899                placeholder++;
 900
 901        orig_len = sb->len;
 902        consumed = format_commit_one(sb, placeholder, context);
 903        if (magic == NO_MAGIC)
 904                return consumed;
 905
 906        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
 907                while (sb->len && sb->buf[sb->len - 1] == '\n')
 908                        strbuf_setlen(sb, sb->len - 1);
 909        } else if ((orig_len != sb->len) && magic == ADD_LF_BEFORE_NON_EMPTY) {
 910                strbuf_insert(sb, orig_len, "\n", 1);
 911        }
 912        return consumed + 1;
 913}
 914
 915static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
 916                                   void *context)
 917{
 918        struct userformat_want *w = context;
 919
 920        if (*placeholder == '+' || *placeholder == '-')
 921                placeholder++;
 922
 923        switch (*placeholder) {
 924        case 'N':
 925                w->notes = 1;
 926                break;
 927        }
 928        return 0;
 929}
 930
 931void userformat_find_requirements(const char *fmt, struct userformat_want *w)
 932{
 933        struct strbuf dummy = STRBUF_INIT;
 934
 935        if (!fmt) {
 936                if (!user_format)
 937                        return;
 938                fmt = user_format;
 939        }
 940        strbuf_expand(&dummy, user_format, userformat_want_item, w);
 941        strbuf_release(&dummy);
 942}
 943
 944void format_commit_message(const struct commit *commit,
 945                           const char *format, struct strbuf *sb,
 946                           const struct pretty_print_context *pretty_ctx)
 947{
 948        struct format_commit_context context;
 949
 950        memset(&context, 0, sizeof(context));
 951        context.commit = commit;
 952        context.pretty_ctx = pretty_ctx;
 953        context.wrap_start = sb->len;
 954        strbuf_expand(sb, format, format_commit_item, &context);
 955        rewrap_message_tail(sb, &context, 0, 0, 0);
 956}
 957
 958static void pp_header(enum cmit_fmt fmt,
 959                      int abbrev,
 960                      enum date_mode dmode,
 961                      const char *encoding,
 962                      const struct commit *commit,
 963                      const char **msg_p,
 964                      struct strbuf *sb)
 965{
 966        int parents_shown = 0;
 967
 968        for (;;) {
 969                const char *line = *msg_p;
 970                int linelen = get_one_line(*msg_p);
 971
 972                if (!linelen)
 973                        return;
 974                *msg_p += linelen;
 975
 976                if (linelen == 1)
 977                        /* End of header */
 978                        return;
 979
 980                if (fmt == CMIT_FMT_RAW) {
 981                        strbuf_add(sb, line, linelen);
 982                        continue;
 983                }
 984
 985                if (!memcmp(line, "parent ", 7)) {
 986                        if (linelen != 48)
 987                                die("bad parent line in commit");
 988                        continue;
 989                }
 990
 991                if (!parents_shown) {
 992                        struct commit_list *parent;
 993                        int num;
 994                        for (parent = commit->parents, num = 0;
 995                             parent;
 996                             parent = parent->next, num++)
 997                                ;
 998                        /* with enough slop */
 999                        strbuf_grow(sb, num * 50 + 20);
1000                        add_merge_info(fmt, sb, commit, abbrev);
1001                        parents_shown = 1;
1002                }
1003
1004                /*
1005                 * MEDIUM == DEFAULT shows only author with dates.
1006                 * FULL shows both authors but not dates.
1007                 * FULLER shows both authors and dates.
1008                 */
1009                if (!memcmp(line, "author ", 7)) {
1010                        strbuf_grow(sb, linelen + 80);
1011                        pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
1012                }
1013                if (!memcmp(line, "committer ", 10) &&
1014                    (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
1015                        strbuf_grow(sb, linelen + 80);
1016                        pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
1017                }
1018        }
1019}
1020
1021void pp_title_line(enum cmit_fmt fmt,
1022                   const char **msg_p,
1023                   struct strbuf *sb,
1024                   const char *subject,
1025                   const char *after_subject,
1026                   const char *encoding,
1027                   int need_8bit_cte)
1028{
1029        const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
1030        struct strbuf title;
1031
1032        strbuf_init(&title, 80);
1033        *msg_p = format_subject(&title, *msg_p, line_separator);
1034
1035        strbuf_grow(sb, title.len + 1024);
1036        if (subject) {
1037                strbuf_addstr(sb, subject);
1038                add_rfc2047(sb, title.buf, title.len, encoding);
1039        } else {
1040                strbuf_addbuf(sb, &title);
1041        }
1042        strbuf_addch(sb, '\n');
1043
1044        if (need_8bit_cte > 0) {
1045                const char *header_fmt =
1046                        "MIME-Version: 1.0\n"
1047                        "Content-Type: text/plain; charset=%s\n"
1048                        "Content-Transfer-Encoding: 8bit\n";
1049                strbuf_addf(sb, header_fmt, encoding);
1050        }
1051        if (after_subject) {
1052                strbuf_addstr(sb, after_subject);
1053        }
1054        if (fmt == CMIT_FMT_EMAIL) {
1055                strbuf_addch(sb, '\n');
1056        }
1057        strbuf_release(&title);
1058}
1059
1060void pp_remainder(enum cmit_fmt fmt,
1061                  const char **msg_p,
1062                  struct strbuf *sb,
1063                  int indent)
1064{
1065        int first = 1;
1066        for (;;) {
1067                const char *line = *msg_p;
1068                int linelen = get_one_line(line);
1069                *msg_p += linelen;
1070
1071                if (!linelen)
1072                        break;
1073
1074                if (is_empty_line(line, &linelen)) {
1075                        if (first)
1076                                continue;
1077                        if (fmt == CMIT_FMT_SHORT)
1078                                break;
1079                }
1080                first = 0;
1081
1082                strbuf_grow(sb, linelen + indent + 20);
1083                if (indent) {
1084                        memset(sb->buf + sb->len, ' ', indent);
1085                        strbuf_setlen(sb, sb->len + indent);
1086                }
1087                strbuf_add(sb, line, linelen);
1088                strbuf_addch(sb, '\n');
1089        }
1090}
1091
1092char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1093{
1094        const char *encoding;
1095
1096        encoding = (git_log_output_encoding
1097                    ? git_log_output_encoding
1098                    : git_commit_encoding);
1099        if (!encoding)
1100                encoding = "UTF-8";
1101        if (encoding_p)
1102                *encoding_p = encoding;
1103        return logmsg_reencode(commit, encoding);
1104}
1105
1106void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
1107                         struct strbuf *sb,
1108                         const struct pretty_print_context *context)
1109{
1110        unsigned long beginning_of_body;
1111        int indent = 4;
1112        const char *msg = commit->buffer;
1113        char *reencoded;
1114        const char *encoding;
1115        int need_8bit_cte = context->need_8bit_cte;
1116
1117        if (fmt == CMIT_FMT_USERFORMAT) {
1118                format_commit_message(commit, user_format, sb, context);
1119                return;
1120        }
1121
1122        reencoded = reencode_commit_message(commit, &encoding);
1123        if (reencoded) {
1124                msg = reencoded;
1125        }
1126
1127        if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1128                indent = 0;
1129
1130        /*
1131         * We need to check and emit Content-type: to mark it
1132         * as 8-bit if we haven't done so.
1133         */
1134        if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1135                int i, ch, in_body;
1136
1137                for (in_body = i = 0; (ch = msg[i]); i++) {
1138                        if (!in_body) {
1139                                /* author could be non 7-bit ASCII but
1140                                 * the log may be so; skip over the
1141                                 * header part first.
1142                                 */
1143                                if (ch == '\n' && msg[i+1] == '\n')
1144                                        in_body = 1;
1145                        }
1146                        else if (non_ascii(ch)) {
1147                                need_8bit_cte = 1;
1148                                break;
1149                        }
1150                }
1151        }
1152
1153        pp_header(fmt, context->abbrev, context->date_mode, encoding,
1154                  commit, &msg, sb);
1155        if (fmt != CMIT_FMT_ONELINE && !context->subject) {
1156                strbuf_addch(sb, '\n');
1157        }
1158
1159        /* Skip excess blank lines at the beginning of body, if any... */
1160        msg = skip_empty_lines(msg);
1161
1162        /* These formats treat the title line specially. */
1163        if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1164                pp_title_line(fmt, &msg, sb, context->subject,
1165                              context->after_subject, encoding, need_8bit_cte);
1166
1167        beginning_of_body = sb->len;
1168        if (fmt != CMIT_FMT_ONELINE)
1169                pp_remainder(fmt, &msg, sb, indent);
1170        strbuf_rtrim(sb);
1171
1172        /* Make sure there is an EOLN for the non-oneline case */
1173        if (fmt != CMIT_FMT_ONELINE)
1174                strbuf_addch(sb, '\n');
1175
1176        /*
1177         * The caller may append additional body text in e-mail
1178         * format.  Make sure we did not strip the blank line
1179         * between the header and the body.
1180         */
1181        if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1182                strbuf_addch(sb, '\n');
1183
1184        if (context->show_notes)
1185                format_display_notes(commit->object.sha1, sb, encoding,
1186                                     NOTES_SHOW_HEADER | NOTES_INDENT);
1187
1188        free(reencoded);
1189}