pretty.con commit decode file:// and ssh:// URLs (9d2e942)
   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;
  14
  15static void save_user_format(struct rev_info *rev, const char *cp, int is_tformat)
  16{
  17        free(user_format);
  18        user_format = xstrdup(cp);
  19        if (is_tformat)
  20                rev->use_terminator = 1;
  21        rev->commit_format = CMIT_FMT_USERFORMAT;
  22}
  23
  24void get_commit_format(const char *arg, struct rev_info *rev)
  25{
  26        int i;
  27        static struct cmt_fmt_map {
  28                const char *n;
  29                size_t cmp_len;
  30                enum cmit_fmt v;
  31        } cmt_fmts[] = {
  32                { "raw",        1,      CMIT_FMT_RAW },
  33                { "medium",     1,      CMIT_FMT_MEDIUM },
  34                { "short",      1,      CMIT_FMT_SHORT },
  35                { "email",      1,      CMIT_FMT_EMAIL },
  36                { "full",       5,      CMIT_FMT_FULL },
  37                { "fuller",     5,      CMIT_FMT_FULLER },
  38                { "oneline",    1,      CMIT_FMT_ONELINE },
  39        };
  40
  41        rev->use_terminator = 0;
  42        if (!arg || !*arg) {
  43                rev->commit_format = CMIT_FMT_DEFAULT;
  44                return;
  45        }
  46        if (!prefixcmp(arg, "format:") || !prefixcmp(arg, "tformat:")) {
  47                save_user_format(rev, strchr(arg, ':') + 1, arg[0] == 't');
  48                return;
  49        }
  50        for (i = 0; i < ARRAY_SIZE(cmt_fmts); i++) {
  51                if (!strncmp(arg, cmt_fmts[i].n, cmt_fmts[i].cmp_len) &&
  52                    !strncmp(arg, cmt_fmts[i].n, strlen(arg))) {
  53                        if (cmt_fmts[i].v == CMIT_FMT_ONELINE)
  54                                rev->use_terminator = 1;
  55                        rev->commit_format = cmt_fmts[i].v;
  56                        return;
  57                }
  58        }
  59        if (strchr(arg, '%')) {
  60                save_user_format(rev, arg, 1);
  61                return;
  62        }
  63
  64        die("invalid --pretty format: %s", arg);
  65}
  66
  67/*
  68 * Generic support for pretty-printing the header
  69 */
  70static int get_one_line(const char *msg)
  71{
  72        int ret = 0;
  73
  74        for (;;) {
  75                char c = *msg++;
  76                if (!c)
  77                        break;
  78                ret++;
  79                if (c == '\n')
  80                        break;
  81        }
  82        return ret;
  83}
  84
  85/* High bit set, or ISO-2022-INT */
  86static int non_ascii(int ch)
  87{
  88        return !isascii(ch) || ch == '\033';
  89}
  90
  91int has_non_ascii(const char *s)
  92{
  93        int ch;
  94        if (!s)
  95                return 0;
  96        while ((ch = *s++) != '\0') {
  97                if (non_ascii(ch))
  98                        return 1;
  99        }
 100        return 0;
 101}
 102
 103static int is_rfc2047_special(char ch)
 104{
 105        return (non_ascii(ch) || (ch == '=') || (ch == '?') || (ch == '_'));
 106}
 107
 108static void add_rfc2047(struct strbuf *sb, const char *line, int len,
 109                       const char *encoding)
 110{
 111        int i, last;
 112
 113        for (i = 0; i < len; i++) {
 114                int ch = line[i];
 115                if (non_ascii(ch))
 116                        goto needquote;
 117                if ((i + 1 < len) && (ch == '=' && line[i+1] == '?'))
 118                        goto needquote;
 119        }
 120        strbuf_add(sb, line, len);
 121        return;
 122
 123needquote:
 124        strbuf_grow(sb, len * 3 + strlen(encoding) + 100);
 125        strbuf_addf(sb, "=?%s?q?", encoding);
 126        for (i = last = 0; i < len; i++) {
 127                unsigned ch = line[i] & 0xFF;
 128                /*
 129                 * We encode ' ' using '=20' even though rfc2047
 130                 * allows using '_' for readability.  Unfortunately,
 131                 * many programs do not understand this and just
 132                 * leave the underscore in place.
 133                 */
 134                if (is_rfc2047_special(ch) || ch == ' ') {
 135                        strbuf_add(sb, line + last, i - last);
 136                        strbuf_addf(sb, "=%02X", ch);
 137                        last = i + 1;
 138                }
 139        }
 140        strbuf_add(sb, line + last, len - last);
 141        strbuf_addstr(sb, "?=");
 142}
 143
 144void pp_user_info(const char *what, enum cmit_fmt fmt, struct strbuf *sb,
 145                  const char *line, enum date_mode dmode,
 146                  const char *encoding)
 147{
 148        char *date;
 149        int namelen;
 150        unsigned long time;
 151        int tz;
 152
 153        if (fmt == CMIT_FMT_ONELINE)
 154                return;
 155        date = strchr(line, '>');
 156        if (!date)
 157                return;
 158        namelen = ++date - line;
 159        time = strtoul(date, &date, 10);
 160        tz = strtol(date, NULL, 10);
 161
 162        if (fmt == CMIT_FMT_EMAIL) {
 163                char *name_tail = strchr(line, '<');
 164                int display_name_length;
 165                if (!name_tail)
 166                        return;
 167                while (line < name_tail && isspace(name_tail[-1]))
 168                        name_tail--;
 169                display_name_length = name_tail - line;
 170                strbuf_addstr(sb, "From: ");
 171                add_rfc2047(sb, line, display_name_length, encoding);
 172                strbuf_add(sb, name_tail, namelen - display_name_length);
 173                strbuf_addch(sb, '\n');
 174        } else {
 175                strbuf_addf(sb, "%s: %.*s%.*s\n", what,
 176                              (fmt == CMIT_FMT_FULLER) ? 4 : 0,
 177                              "    ", namelen, line);
 178        }
 179        switch (fmt) {
 180        case CMIT_FMT_MEDIUM:
 181                strbuf_addf(sb, "Date:   %s\n", show_date(time, tz, dmode));
 182                break;
 183        case CMIT_FMT_EMAIL:
 184                strbuf_addf(sb, "Date: %s\n", show_date(time, tz, DATE_RFC2822));
 185                break;
 186        case CMIT_FMT_FULLER:
 187                strbuf_addf(sb, "%sDate: %s\n", what, show_date(time, tz, dmode));
 188                break;
 189        default:
 190                /* notin' */
 191                break;
 192        }
 193}
 194
 195static int is_empty_line(const char *line, int *len_p)
 196{
 197        int len = *len_p;
 198        while (len && isspace(line[len-1]))
 199                len--;
 200        *len_p = len;
 201        return !len;
 202}
 203
 204static const char *skip_empty_lines(const char *msg)
 205{
 206        for (;;) {
 207                int linelen = get_one_line(msg);
 208                int ll = linelen;
 209                if (!linelen)
 210                        break;
 211                if (!is_empty_line(msg, &ll))
 212                        break;
 213                msg += linelen;
 214        }
 215        return msg;
 216}
 217
 218static void add_merge_info(enum cmit_fmt fmt, struct strbuf *sb,
 219                        const struct commit *commit, int abbrev)
 220{
 221        struct commit_list *parent = commit->parents;
 222
 223        if ((fmt == CMIT_FMT_ONELINE) || (fmt == CMIT_FMT_EMAIL) ||
 224            !parent || !parent->next)
 225                return;
 226
 227        strbuf_addstr(sb, "Merge:");
 228
 229        while (parent) {
 230                struct commit *p = parent->item;
 231                const char *hex = NULL;
 232                if (abbrev)
 233                        hex = find_unique_abbrev(p->object.sha1, abbrev);
 234                if (!hex)
 235                        hex = sha1_to_hex(p->object.sha1);
 236                parent = parent->next;
 237
 238                strbuf_addf(sb, " %s", hex);
 239        }
 240        strbuf_addch(sb, '\n');
 241}
 242
 243static char *get_header(const struct commit *commit, const char *key)
 244{
 245        int key_len = strlen(key);
 246        const char *line = commit->buffer;
 247
 248        for (;;) {
 249                const char *eol = strchr(line, '\n'), *next;
 250
 251                if (line == eol)
 252                        return NULL;
 253                if (!eol) {
 254                        eol = line + strlen(line);
 255                        next = NULL;
 256                } else
 257                        next = eol + 1;
 258                if (eol - line > key_len &&
 259                    !strncmp(line, key, key_len) &&
 260                    line[key_len] == ' ') {
 261                        return xmemdupz(line + key_len + 1, eol - line - key_len - 1);
 262                }
 263                line = next;
 264        }
 265}
 266
 267static char *replace_encoding_header(char *buf, const char *encoding)
 268{
 269        struct strbuf tmp = STRBUF_INIT;
 270        size_t start, len;
 271        char *cp = buf;
 272
 273        /* guess if there is an encoding header before a \n\n */
 274        while (strncmp(cp, "encoding ", strlen("encoding "))) {
 275                cp = strchr(cp, '\n');
 276                if (!cp || *++cp == '\n')
 277                        return buf;
 278        }
 279        start = cp - buf;
 280        cp = strchr(cp, '\n');
 281        if (!cp)
 282                return buf; /* should not happen but be defensive */
 283        len = cp + 1 - (buf + start);
 284
 285        strbuf_attach(&tmp, buf, strlen(buf), strlen(buf) + 1);
 286        if (is_encoding_utf8(encoding)) {
 287                /* we have re-coded to UTF-8; drop the header */
 288                strbuf_remove(&tmp, start, len);
 289        } else {
 290                /* just replaces XXXX in 'encoding XXXX\n' */
 291                strbuf_splice(&tmp, start + strlen("encoding "),
 292                                          len - strlen("encoding \n"),
 293                                          encoding, strlen(encoding));
 294        }
 295        return strbuf_detach(&tmp, NULL);
 296}
 297
 298static char *logmsg_reencode(const struct commit *commit,
 299                             const char *output_encoding)
 300{
 301        static const char *utf8 = "UTF-8";
 302        const char *use_encoding;
 303        char *encoding;
 304        char *out;
 305
 306        if (!*output_encoding)
 307                return NULL;
 308        encoding = get_header(commit, "encoding");
 309        use_encoding = encoding ? encoding : utf8;
 310        if (!strcmp(use_encoding, output_encoding))
 311                if (encoding) /* we'll strip encoding header later */
 312                        out = xstrdup(commit->buffer);
 313                else
 314                        return NULL; /* nothing to do */
 315        else
 316                out = reencode_string(commit->buffer,
 317                                      output_encoding, use_encoding);
 318        if (out)
 319                out = replace_encoding_header(out, output_encoding);
 320
 321        free(encoding);
 322        return out;
 323}
 324
 325static int mailmap_name(char *email, int email_len, char *name, int name_len)
 326{
 327        static struct string_list *mail_map;
 328        if (!mail_map) {
 329                mail_map = xcalloc(1, sizeof(*mail_map));
 330                read_mailmap(mail_map, NULL);
 331        }
 332        return mail_map->nr && map_user(mail_map, email, email_len, name, name_len);
 333}
 334
 335static size_t format_person_part(struct strbuf *sb, char part,
 336                                 const char *msg, int len, enum date_mode dmode)
 337{
 338        /* currently all placeholders have same length */
 339        const int placeholder_len = 2;
 340        int start, end, tz = 0;
 341        unsigned long date = 0;
 342        char *ep;
 343        const char *name_start, *name_end, *mail_start, *mail_end, *msg_end = msg+len;
 344        char person_name[1024];
 345        char person_mail[1024];
 346
 347        /* advance 'end' to point to email start delimiter */
 348        for (end = 0; end < len && msg[end] != '<'; end++)
 349                ; /* do nothing */
 350
 351        /*
 352         * When end points at the '<' that we found, it should have
 353         * matching '>' later, which means 'end' must be strictly
 354         * below len - 1.
 355         */
 356        if (end >= len - 2)
 357                goto skip;
 358
 359        /* Seek for both name and email part */
 360        name_start = msg;
 361        name_end = msg+end;
 362        while (name_end > name_start && isspace(*(name_end-1)))
 363                name_end--;
 364        mail_start = msg+end+1;
 365        mail_end = mail_start;
 366        while (mail_end < msg_end && *mail_end != '>')
 367                mail_end++;
 368        if (mail_end == msg_end)
 369                goto skip;
 370        end = mail_end-msg;
 371
 372        if (part == 'N' || part == 'E') { /* mailmap lookup */
 373                strlcpy(person_name, name_start, name_end-name_start+1);
 374                strlcpy(person_mail, mail_start, mail_end-mail_start+1);
 375                mailmap_name(person_mail, sizeof(person_mail), person_name, sizeof(person_name));
 376                name_start = person_name;
 377                name_end = name_start + strlen(person_name);
 378                mail_start = person_mail;
 379                mail_end = mail_start +  strlen(person_mail);
 380        }
 381        if (part == 'n' || part == 'N') {       /* name */
 382                strbuf_add(sb, name_start, name_end-name_start);
 383                return placeholder_len;
 384        }
 385        if (part == 'e' || part == 'E') {       /* email */
 386                strbuf_add(sb, mail_start, mail_end-mail_start);
 387                return placeholder_len;
 388        }
 389
 390        /* advance 'start' to point to date start delimiter */
 391        for (start = end + 1; start < len && isspace(msg[start]); start++)
 392                ; /* do nothing */
 393        if (start >= len)
 394                goto skip;
 395        date = strtoul(msg + start, &ep, 10);
 396        if (msg + start == ep)
 397                goto skip;
 398
 399        if (part == 't') {      /* date, UNIX timestamp */
 400                strbuf_add(sb, msg + start, ep - (msg + start));
 401                return placeholder_len;
 402        }
 403
 404        /* parse tz */
 405        for (start = ep - msg + 1; start < len && isspace(msg[start]); start++)
 406                ; /* do nothing */
 407        if (start + 1 < len) {
 408                tz = strtoul(msg + start + 1, NULL, 10);
 409                if (msg[start] == '-')
 410                        tz = -tz;
 411        }
 412
 413        switch (part) {
 414        case 'd':       /* date */
 415                strbuf_addstr(sb, show_date(date, tz, dmode));
 416                return placeholder_len;
 417        case 'D':       /* date, RFC2822 style */
 418                strbuf_addstr(sb, show_date(date, tz, DATE_RFC2822));
 419                return placeholder_len;
 420        case 'r':       /* date, relative */
 421                strbuf_addstr(sb, show_date(date, tz, DATE_RELATIVE));
 422                return placeholder_len;
 423        case 'i':       /* date, ISO 8601 */
 424                strbuf_addstr(sb, show_date(date, tz, DATE_ISO8601));
 425                return placeholder_len;
 426        }
 427
 428skip:
 429        /*
 430         * bogus commit, 'sb' cannot be updated, but we still need to
 431         * compute a valid return value.
 432         */
 433        if (part == 'n' || part == 'e' || part == 't' || part == 'd'
 434            || part == 'D' || part == 'r' || part == 'i')
 435                return placeholder_len;
 436
 437        return 0; /* unknown placeholder */
 438}
 439
 440struct chunk {
 441        size_t off;
 442        size_t len;
 443};
 444
 445struct format_commit_context {
 446        const struct commit *commit;
 447        const struct pretty_print_context *pretty_ctx;
 448        unsigned commit_header_parsed:1;
 449        unsigned commit_message_parsed:1;
 450        size_t width, indent1, indent2;
 451
 452        /* These offsets are relative to the start of the commit message. */
 453        struct chunk author;
 454        struct chunk committer;
 455        struct chunk encoding;
 456        size_t message_off;
 457        size_t subject_off;
 458        size_t body_off;
 459
 460        /* The following ones are relative to the result struct strbuf. */
 461        struct chunk abbrev_commit_hash;
 462        struct chunk abbrev_tree_hash;
 463        struct chunk abbrev_parent_hashes;
 464        size_t wrap_start;
 465};
 466
 467static int add_again(struct strbuf *sb, struct chunk *chunk)
 468{
 469        if (chunk->len) {
 470                strbuf_adddup(sb, chunk->off, chunk->len);
 471                return 1;
 472        }
 473
 474        /*
 475         * We haven't seen this chunk before.  Our caller is surely
 476         * going to add it the hard way now.  Remember the most likely
 477         * start of the to-be-added chunk: the current end of the
 478         * struct strbuf.
 479         */
 480        chunk->off = sb->len;
 481        return 0;
 482}
 483
 484static void parse_commit_header(struct format_commit_context *context)
 485{
 486        const char *msg = context->commit->buffer;
 487        int i;
 488
 489        for (i = 0; msg[i]; i++) {
 490                int eol;
 491                for (eol = i; msg[eol] && msg[eol] != '\n'; eol++)
 492                        ; /* do nothing */
 493
 494                if (i == eol) {
 495                        break;
 496                } else if (!prefixcmp(msg + i, "author ")) {
 497                        context->author.off = i + 7;
 498                        context->author.len = eol - i - 7;
 499                } else if (!prefixcmp(msg + i, "committer ")) {
 500                        context->committer.off = i + 10;
 501                        context->committer.len = eol - i - 10;
 502                } else if (!prefixcmp(msg + i, "encoding ")) {
 503                        context->encoding.off = i + 9;
 504                        context->encoding.len = eol - i - 9;
 505                }
 506                i = eol;
 507        }
 508        context->message_off = i;
 509        context->commit_header_parsed = 1;
 510}
 511
 512static int istitlechar(char c)
 513{
 514        return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
 515                (c >= '0' && c <= '9') || c == '.' || c == '_';
 516}
 517
 518static void format_sanitized_subject(struct strbuf *sb, const char *msg)
 519{
 520        size_t trimlen;
 521        size_t start_len = sb->len;
 522        int space = 2;
 523
 524        for (; *msg && *msg != '\n'; msg++) {
 525                if (istitlechar(*msg)) {
 526                        if (space == 1)
 527                                strbuf_addch(sb, '-');
 528                        space = 0;
 529                        strbuf_addch(sb, *msg);
 530                        if (*msg == '.')
 531                                while (*(msg+1) == '.')
 532                                        msg++;
 533                } else
 534                        space |= 1;
 535        }
 536
 537        /* trim any trailing '.' or '-' characters */
 538        trimlen = 0;
 539        while (sb->len - trimlen > start_len &&
 540                (sb->buf[sb->len - 1 - trimlen] == '.'
 541                || sb->buf[sb->len - 1 - trimlen] == '-'))
 542                trimlen++;
 543        strbuf_remove(sb, sb->len - trimlen, trimlen);
 544}
 545
 546const char *format_subject(struct strbuf *sb, const char *msg,
 547                           const char *line_separator)
 548{
 549        int first = 1;
 550
 551        for (;;) {
 552                const char *line = msg;
 553                int linelen = get_one_line(line);
 554
 555                msg += linelen;
 556                if (!linelen || is_empty_line(line, &linelen))
 557                        break;
 558
 559                if (!sb)
 560                        continue;
 561                strbuf_grow(sb, linelen + 2);
 562                if (!first)
 563                        strbuf_addstr(sb, line_separator);
 564                strbuf_add(sb, line, linelen);
 565                first = 0;
 566        }
 567        return msg;
 568}
 569
 570static void parse_commit_message(struct format_commit_context *c)
 571{
 572        const char *msg = c->commit->buffer + c->message_off;
 573        const char *start = c->commit->buffer;
 574
 575        msg = skip_empty_lines(msg);
 576        c->subject_off = msg - start;
 577
 578        msg = format_subject(NULL, msg, NULL);
 579        msg = skip_empty_lines(msg);
 580        c->body_off = msg - start;
 581
 582        c->commit_message_parsed = 1;
 583}
 584
 585static void format_decoration(struct strbuf *sb, const struct commit *commit)
 586{
 587        struct name_decoration *d;
 588        const char *prefix = " (";
 589
 590        load_ref_decorations(DECORATE_SHORT_REFS);
 591        d = lookup_decoration(&name_decoration, &commit->object);
 592        while (d) {
 593                strbuf_addstr(sb, prefix);
 594                prefix = ", ";
 595                strbuf_addstr(sb, d->name);
 596                d = d->next;
 597        }
 598        if (prefix[0] == ',')
 599                strbuf_addch(sb, ')');
 600}
 601
 602static void strbuf_wrap(struct strbuf *sb, size_t pos,
 603                        size_t width, size_t indent1, size_t indent2)
 604{
 605        struct strbuf tmp = STRBUF_INIT;
 606
 607        if (pos)
 608                strbuf_add(&tmp, sb->buf, pos);
 609        strbuf_add_wrapped_text(&tmp, sb->buf + pos,
 610                                (int) indent1, (int) indent2, (int) width);
 611        strbuf_swap(&tmp, sb);
 612        strbuf_release(&tmp);
 613}
 614
 615static void rewrap_message_tail(struct strbuf *sb,
 616                                struct format_commit_context *c,
 617                                size_t new_width, size_t new_indent1,
 618                                size_t new_indent2)
 619{
 620        if (c->width == new_width && c->indent1 == new_indent1 &&
 621            c->indent2 == new_indent2)
 622                return;
 623        if (c->wrap_start < sb->len)
 624                strbuf_wrap(sb, c->wrap_start, c->width, c->indent1, c->indent2);
 625        c->wrap_start = sb->len;
 626        c->width = new_width;
 627        c->indent1 = new_indent1;
 628        c->indent2 = new_indent2;
 629}
 630
 631static size_t format_commit_one(struct strbuf *sb, const char *placeholder,
 632                                void *context)
 633{
 634        struct format_commit_context *c = context;
 635        const struct commit *commit = c->commit;
 636        const char *msg = commit->buffer;
 637        struct commit_list *p;
 638        int h1, h2;
 639
 640        /* these are independent of the commit */
 641        switch (placeholder[0]) {
 642        case 'C':
 643                if (placeholder[1] == '(') {
 644                        const char *end = strchr(placeholder + 2, ')');
 645                        char color[COLOR_MAXLEN];
 646                        if (!end)
 647                                return 0;
 648                        color_parse_mem(placeholder + 2,
 649                                        end - (placeholder + 2),
 650                                        "--pretty format", color);
 651                        strbuf_addstr(sb, color);
 652                        return end - placeholder + 1;
 653                }
 654                if (!prefixcmp(placeholder + 1, "red")) {
 655                        strbuf_addstr(sb, GIT_COLOR_RED);
 656                        return 4;
 657                } else if (!prefixcmp(placeholder + 1, "green")) {
 658                        strbuf_addstr(sb, GIT_COLOR_GREEN);
 659                        return 6;
 660                } else if (!prefixcmp(placeholder + 1, "blue")) {
 661                        strbuf_addstr(sb, GIT_COLOR_BLUE);
 662                        return 5;
 663                } else if (!prefixcmp(placeholder + 1, "reset")) {
 664                        strbuf_addstr(sb, GIT_COLOR_RESET);
 665                        return 6;
 666                } else
 667                        return 0;
 668        case 'n':               /* newline */
 669                strbuf_addch(sb, '\n');
 670                return 1;
 671        case 'x':
 672                /* %x00 == NUL, %x0a == LF, etc. */
 673                if (0 <= (h1 = hexval_table[0xff & placeholder[1]]) &&
 674                    h1 <= 16 &&
 675                    0 <= (h2 = hexval_table[0xff & placeholder[2]]) &&
 676                    h2 <= 16) {
 677                        strbuf_addch(sb, (h1<<4)|h2);
 678                        return 3;
 679                } else
 680                        return 0;
 681        case 'w':
 682                if (placeholder[1] == '(') {
 683                        unsigned long width = 0, indent1 = 0, indent2 = 0;
 684                        char *next;
 685                        const char *start = placeholder + 2;
 686                        const char *end = strchr(start, ')');
 687                        if (!end)
 688                                return 0;
 689                        if (end > start) {
 690                                width = strtoul(start, &next, 10);
 691                                if (*next == ',') {
 692                                        indent1 = strtoul(next + 1, &next, 10);
 693                                        if (*next == ',') {
 694                                                indent2 = strtoul(next + 1,
 695                                                                 &next, 10);
 696                                        }
 697                                }
 698                                if (*next != ')')
 699                                        return 0;
 700                        }
 701                        rewrap_message_tail(sb, c, width, indent1, indent2);
 702                        return end - placeholder + 1;
 703                } else
 704                        return 0;
 705        }
 706
 707        /* these depend on the commit */
 708        if (!commit->object.parsed)
 709                parse_object(commit->object.sha1);
 710
 711        switch (placeholder[0]) {
 712        case 'H':               /* commit hash */
 713                strbuf_addstr(sb, sha1_to_hex(commit->object.sha1));
 714                return 1;
 715        case 'h':               /* abbreviated commit hash */
 716                if (add_again(sb, &c->abbrev_commit_hash))
 717                        return 1;
 718                strbuf_addstr(sb, find_unique_abbrev(commit->object.sha1,
 719                                                     c->pretty_ctx->abbrev));
 720                c->abbrev_commit_hash.len = sb->len - c->abbrev_commit_hash.off;
 721                return 1;
 722        case 'T':               /* tree hash */
 723                strbuf_addstr(sb, sha1_to_hex(commit->tree->object.sha1));
 724                return 1;
 725        case 't':               /* abbreviated tree hash */
 726                if (add_again(sb, &c->abbrev_tree_hash))
 727                        return 1;
 728                strbuf_addstr(sb, find_unique_abbrev(commit->tree->object.sha1,
 729                                                     c->pretty_ctx->abbrev));
 730                c->abbrev_tree_hash.len = sb->len - c->abbrev_tree_hash.off;
 731                return 1;
 732        case 'P':               /* parent hashes */
 733                for (p = commit->parents; p; p = p->next) {
 734                        if (p != commit->parents)
 735                                strbuf_addch(sb, ' ');
 736                        strbuf_addstr(sb, sha1_to_hex(p->item->object.sha1));
 737                }
 738                return 1;
 739        case 'p':               /* abbreviated parent hashes */
 740                if (add_again(sb, &c->abbrev_parent_hashes))
 741                        return 1;
 742                for (p = commit->parents; p; p = p->next) {
 743                        if (p != commit->parents)
 744                                strbuf_addch(sb, ' ');
 745                        strbuf_addstr(sb, find_unique_abbrev(
 746                                        p->item->object.sha1,
 747                                        c->pretty_ctx->abbrev));
 748                }
 749                c->abbrev_parent_hashes.len = sb->len -
 750                                              c->abbrev_parent_hashes.off;
 751                return 1;
 752        case 'm':               /* left/right/bottom */
 753                strbuf_addch(sb, (commit->object.flags & BOUNDARY)
 754                                 ? '-'
 755                                 : (commit->object.flags & SYMMETRIC_LEFT)
 756                                 ? '<'
 757                                 : '>');
 758                return 1;
 759        case 'd':
 760                format_decoration(sb, commit);
 761                return 1;
 762        case 'g':               /* reflog info */
 763                switch(placeholder[1]) {
 764                case 'd':       /* reflog selector */
 765                case 'D':
 766                        if (c->pretty_ctx->reflog_info)
 767                                get_reflog_selector(sb,
 768                                                    c->pretty_ctx->reflog_info,
 769                                                    c->pretty_ctx->date_mode,
 770                                                    (placeholder[1] == 'd'));
 771                        return 2;
 772                case 's':       /* reflog message */
 773                        if (c->pretty_ctx->reflog_info)
 774                                get_reflog_message(sb, c->pretty_ctx->reflog_info);
 775                        return 2;
 776                }
 777                return 0;       /* unknown %g placeholder */
 778        case 'N':
 779                if (c->pretty_ctx->show_notes) {
 780                        format_display_notes(commit->object.sha1, sb,
 781                                    git_log_output_encoding ? git_log_output_encoding
 782                                                            : git_commit_encoding, 0);
 783                        return 1;
 784                }
 785                return 0;
 786        }
 787
 788        /* For the rest we have to parse the commit header. */
 789        if (!c->commit_header_parsed)
 790                parse_commit_header(c);
 791
 792        switch (placeholder[0]) {
 793        case 'a':       /* author ... */
 794                return format_person_part(sb, placeholder[1],
 795                                   msg + c->author.off, c->author.len,
 796                                   c->pretty_ctx->date_mode);
 797        case 'c':       /* committer ... */
 798                return format_person_part(sb, placeholder[1],
 799                                   msg + c->committer.off, c->committer.len,
 800                                   c->pretty_ctx->date_mode);
 801        case 'e':       /* encoding */
 802                strbuf_add(sb, msg + c->encoding.off, c->encoding.len);
 803                return 1;
 804        case 'B':       /* raw body */
 805                /* message_off is always left at the initial newline */
 806                strbuf_addstr(sb, msg + c->message_off + 1);
 807                return 1;
 808        }
 809
 810        /* Now we need to parse the commit message. */
 811        if (!c->commit_message_parsed)
 812                parse_commit_message(c);
 813
 814        switch (placeholder[0]) {
 815        case 's':       /* subject */
 816                format_subject(sb, msg + c->subject_off, " ");
 817                return 1;
 818        case 'f':       /* sanitized subject */
 819                format_sanitized_subject(sb, msg + c->subject_off);
 820                return 1;
 821        case 'b':       /* body */
 822                strbuf_addstr(sb, msg + c->body_off);
 823                return 1;
 824        }
 825        return 0;       /* unknown placeholder */
 826}
 827
 828static size_t format_commit_item(struct strbuf *sb, const char *placeholder,
 829                                 void *context)
 830{
 831        int consumed;
 832        size_t orig_len;
 833        enum {
 834                NO_MAGIC,
 835                ADD_LF_BEFORE_NON_EMPTY,
 836                DEL_LF_BEFORE_EMPTY,
 837        } magic = NO_MAGIC;
 838
 839        switch (placeholder[0]) {
 840        case '-':
 841                magic = DEL_LF_BEFORE_EMPTY;
 842                break;
 843        case '+':
 844                magic = ADD_LF_BEFORE_NON_EMPTY;
 845                break;
 846        default:
 847                break;
 848        }
 849        if (magic != NO_MAGIC)
 850                placeholder++;
 851
 852        orig_len = sb->len;
 853        consumed = format_commit_one(sb, placeholder, context);
 854        if (magic == NO_MAGIC)
 855                return consumed;
 856
 857        if ((orig_len == sb->len) && magic == DEL_LF_BEFORE_EMPTY) {
 858                while (sb->len && sb->buf[sb->len - 1] == '\n')
 859                        strbuf_setlen(sb, sb->len - 1);
 860        } else if ((orig_len != sb->len) && magic == ADD_LF_BEFORE_NON_EMPTY) {
 861                strbuf_insert(sb, orig_len, "\n", 1);
 862        }
 863        return consumed + 1;
 864}
 865
 866static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
 867                                   void *context)
 868{
 869        struct userformat_want *w = context;
 870
 871        if (*placeholder == '+' || *placeholder == '-')
 872                placeholder++;
 873
 874        switch (*placeholder) {
 875        case 'N':
 876                w->notes = 1;
 877                break;
 878        }
 879        return 0;
 880}
 881
 882void userformat_find_requirements(const char *fmt, struct userformat_want *w)
 883{
 884        struct strbuf dummy = STRBUF_INIT;
 885
 886        if (!fmt) {
 887                if (!user_format)
 888                        return;
 889                fmt = user_format;
 890        }
 891        strbuf_expand(&dummy, user_format, userformat_want_item, w);
 892        strbuf_release(&dummy);
 893}
 894
 895void format_commit_message(const struct commit *commit,
 896                           const char *format, struct strbuf *sb,
 897                           const struct pretty_print_context *pretty_ctx)
 898{
 899        struct format_commit_context context;
 900
 901        memset(&context, 0, sizeof(context));
 902        context.commit = commit;
 903        context.pretty_ctx = pretty_ctx;
 904        context.wrap_start = sb->len;
 905        strbuf_expand(sb, format, format_commit_item, &context);
 906        rewrap_message_tail(sb, &context, 0, 0, 0);
 907}
 908
 909static void pp_header(enum cmit_fmt fmt,
 910                      int abbrev,
 911                      enum date_mode dmode,
 912                      const char *encoding,
 913                      const struct commit *commit,
 914                      const char **msg_p,
 915                      struct strbuf *sb)
 916{
 917        int parents_shown = 0;
 918
 919        for (;;) {
 920                const char *line = *msg_p;
 921                int linelen = get_one_line(*msg_p);
 922
 923                if (!linelen)
 924                        return;
 925                *msg_p += linelen;
 926
 927                if (linelen == 1)
 928                        /* End of header */
 929                        return;
 930
 931                if (fmt == CMIT_FMT_RAW) {
 932                        strbuf_add(sb, line, linelen);
 933                        continue;
 934                }
 935
 936                if (!memcmp(line, "parent ", 7)) {
 937                        if (linelen != 48)
 938                                die("bad parent line in commit");
 939                        continue;
 940                }
 941
 942                if (!parents_shown) {
 943                        struct commit_list *parent;
 944                        int num;
 945                        for (parent = commit->parents, num = 0;
 946                             parent;
 947                             parent = parent->next, num++)
 948                                ;
 949                        /* with enough slop */
 950                        strbuf_grow(sb, num * 50 + 20);
 951                        add_merge_info(fmt, sb, commit, abbrev);
 952                        parents_shown = 1;
 953                }
 954
 955                /*
 956                 * MEDIUM == DEFAULT shows only author with dates.
 957                 * FULL shows both authors but not dates.
 958                 * FULLER shows both authors and dates.
 959                 */
 960                if (!memcmp(line, "author ", 7)) {
 961                        strbuf_grow(sb, linelen + 80);
 962                        pp_user_info("Author", fmt, sb, line + 7, dmode, encoding);
 963                }
 964                if (!memcmp(line, "committer ", 10) &&
 965                    (fmt == CMIT_FMT_FULL || fmt == CMIT_FMT_FULLER)) {
 966                        strbuf_grow(sb, linelen + 80);
 967                        pp_user_info("Commit", fmt, sb, line + 10, dmode, encoding);
 968                }
 969        }
 970}
 971
 972void pp_title_line(enum cmit_fmt fmt,
 973                   const char **msg_p,
 974                   struct strbuf *sb,
 975                   const char *subject,
 976                   const char *after_subject,
 977                   const char *encoding,
 978                   int need_8bit_cte)
 979{
 980        const char *line_separator = (fmt == CMIT_FMT_EMAIL) ? "\n " : " ";
 981        struct strbuf title;
 982
 983        strbuf_init(&title, 80);
 984        *msg_p = format_subject(&title, *msg_p, line_separator);
 985
 986        strbuf_grow(sb, title.len + 1024);
 987        if (subject) {
 988                strbuf_addstr(sb, subject);
 989                add_rfc2047(sb, title.buf, title.len, encoding);
 990        } else {
 991                strbuf_addbuf(sb, &title);
 992        }
 993        strbuf_addch(sb, '\n');
 994
 995        if (need_8bit_cte > 0) {
 996                const char *header_fmt =
 997                        "MIME-Version: 1.0\n"
 998                        "Content-Type: text/plain; charset=%s\n"
 999                        "Content-Transfer-Encoding: 8bit\n";
1000                strbuf_addf(sb, header_fmt, encoding);
1001        }
1002        if (after_subject) {
1003                strbuf_addstr(sb, after_subject);
1004        }
1005        if (fmt == CMIT_FMT_EMAIL) {
1006                strbuf_addch(sb, '\n');
1007        }
1008        strbuf_release(&title);
1009}
1010
1011void pp_remainder(enum cmit_fmt fmt,
1012                  const char **msg_p,
1013                  struct strbuf *sb,
1014                  int indent)
1015{
1016        int first = 1;
1017        for (;;) {
1018                const char *line = *msg_p;
1019                int linelen = get_one_line(line);
1020                *msg_p += linelen;
1021
1022                if (!linelen)
1023                        break;
1024
1025                if (is_empty_line(line, &linelen)) {
1026                        if (first)
1027                                continue;
1028                        if (fmt == CMIT_FMT_SHORT)
1029                                break;
1030                }
1031                first = 0;
1032
1033                strbuf_grow(sb, linelen + indent + 20);
1034                if (indent) {
1035                        memset(sb->buf + sb->len, ' ', indent);
1036                        strbuf_setlen(sb, sb->len + indent);
1037                }
1038                strbuf_add(sb, line, linelen);
1039                strbuf_addch(sb, '\n');
1040        }
1041}
1042
1043char *reencode_commit_message(const struct commit *commit, const char **encoding_p)
1044{
1045        const char *encoding;
1046
1047        encoding = (git_log_output_encoding
1048                    ? git_log_output_encoding
1049                    : git_commit_encoding);
1050        if (!encoding)
1051                encoding = "UTF-8";
1052        if (encoding_p)
1053                *encoding_p = encoding;
1054        return logmsg_reencode(commit, encoding);
1055}
1056
1057void pretty_print_commit(enum cmit_fmt fmt, const struct commit *commit,
1058                         struct strbuf *sb,
1059                         const struct pretty_print_context *context)
1060{
1061        unsigned long beginning_of_body;
1062        int indent = 4;
1063        const char *msg = commit->buffer;
1064        char *reencoded;
1065        const char *encoding;
1066        int need_8bit_cte = context->need_8bit_cte;
1067
1068        if (fmt == CMIT_FMT_USERFORMAT) {
1069                format_commit_message(commit, user_format, sb, context);
1070                return;
1071        }
1072
1073        reencoded = reencode_commit_message(commit, &encoding);
1074        if (reencoded) {
1075                msg = reencoded;
1076        }
1077
1078        if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1079                indent = 0;
1080
1081        /*
1082         * We need to check and emit Content-type: to mark it
1083         * as 8-bit if we haven't done so.
1084         */
1085        if (fmt == CMIT_FMT_EMAIL && need_8bit_cte == 0) {
1086                int i, ch, in_body;
1087
1088                for (in_body = i = 0; (ch = msg[i]); i++) {
1089                        if (!in_body) {
1090                                /* author could be non 7-bit ASCII but
1091                                 * the log may be so; skip over the
1092                                 * header part first.
1093                                 */
1094                                if (ch == '\n' && msg[i+1] == '\n')
1095                                        in_body = 1;
1096                        }
1097                        else if (non_ascii(ch)) {
1098                                need_8bit_cte = 1;
1099                                break;
1100                        }
1101                }
1102        }
1103
1104        pp_header(fmt, context->abbrev, context->date_mode, encoding,
1105                  commit, &msg, sb);
1106        if (fmt != CMIT_FMT_ONELINE && !context->subject) {
1107                strbuf_addch(sb, '\n');
1108        }
1109
1110        /* Skip excess blank lines at the beginning of body, if any... */
1111        msg = skip_empty_lines(msg);
1112
1113        /* These formats treat the title line specially. */
1114        if (fmt == CMIT_FMT_ONELINE || fmt == CMIT_FMT_EMAIL)
1115                pp_title_line(fmt, &msg, sb, context->subject,
1116                              context->after_subject, encoding, need_8bit_cte);
1117
1118        beginning_of_body = sb->len;
1119        if (fmt != CMIT_FMT_ONELINE)
1120                pp_remainder(fmt, &msg, sb, indent);
1121        strbuf_rtrim(sb);
1122
1123        /* Make sure there is an EOLN for the non-oneline case */
1124        if (fmt != CMIT_FMT_ONELINE)
1125                strbuf_addch(sb, '\n');
1126
1127        /*
1128         * The caller may append additional body text in e-mail
1129         * format.  Make sure we did not strip the blank line
1130         * between the header and the body.
1131         */
1132        if (fmt == CMIT_FMT_EMAIL && sb->len <= beginning_of_body)
1133                strbuf_addch(sb, '\n');
1134
1135        if (context->show_notes)
1136                format_display_notes(commit->object.sha1, sb, encoding,
1137                                     NOTES_SHOW_HEADER | NOTES_INDENT);
1138
1139        free(reencoded);
1140}