diff.con commit push: require force for annotated tags (40eff17)
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "quote.h"
   6#include "diff.h"
   7#include "diffcore.h"
   8#include "delta.h"
   9#include "xdiff-interface.h"
  10#include "color.h"
  11#include "attr.h"
  12#include "run-command.h"
  13#include "utf8.h"
  14#include "userdiff.h"
  15#include "sigchain.h"
  16#include "submodule.h"
  17#include "ll-merge.h"
  18#include "string-list.h"
  19
  20#ifdef NO_FAST_WORKING_DIRECTORY
  21#define FAST_WORKING_DIRECTORY 0
  22#else
  23#define FAST_WORKING_DIRECTORY 1
  24#endif
  25
  26static int diff_detect_rename_default;
  27static int diff_rename_limit_default = 400;
  28static int diff_suppress_blank_empty;
  29static int diff_use_color_default = -1;
  30static int diff_context_default = 3;
  31static const char *diff_word_regex_cfg;
  32static const char *external_diff_cmd_cfg;
  33int diff_auto_refresh_index = 1;
  34static int diff_mnemonic_prefix;
  35static int diff_no_prefix;
  36static int diff_stat_graph_width;
  37static int diff_dirstat_permille_default = 30;
  38static struct diff_options default_diff_options;
  39
  40static char diff_colors[][COLOR_MAXLEN] = {
  41        GIT_COLOR_RESET,
  42        GIT_COLOR_NORMAL,       /* PLAIN */
  43        GIT_COLOR_BOLD,         /* METAINFO */
  44        GIT_COLOR_CYAN,         /* FRAGINFO */
  45        GIT_COLOR_RED,          /* OLD */
  46        GIT_COLOR_GREEN,        /* NEW */
  47        GIT_COLOR_YELLOW,       /* COMMIT */
  48        GIT_COLOR_BG_RED,       /* WHITESPACE */
  49        GIT_COLOR_NORMAL,       /* FUNCINFO */
  50};
  51
  52static int parse_diff_color_slot(const char *var, int ofs)
  53{
  54        if (!strcasecmp(var+ofs, "plain"))
  55                return DIFF_PLAIN;
  56        if (!strcasecmp(var+ofs, "meta"))
  57                return DIFF_METAINFO;
  58        if (!strcasecmp(var+ofs, "frag"))
  59                return DIFF_FRAGINFO;
  60        if (!strcasecmp(var+ofs, "old"))
  61                return DIFF_FILE_OLD;
  62        if (!strcasecmp(var+ofs, "new"))
  63                return DIFF_FILE_NEW;
  64        if (!strcasecmp(var+ofs, "commit"))
  65                return DIFF_COMMIT;
  66        if (!strcasecmp(var+ofs, "whitespace"))
  67                return DIFF_WHITESPACE;
  68        if (!strcasecmp(var+ofs, "func"))
  69                return DIFF_FUNCINFO;
  70        return -1;
  71}
  72
  73static int parse_dirstat_params(struct diff_options *options, const char *params_string,
  74                                struct strbuf *errmsg)
  75{
  76        char *params_copy = xstrdup(params_string);
  77        struct string_list params = STRING_LIST_INIT_NODUP;
  78        int ret = 0;
  79        int i;
  80
  81        if (*params_copy)
  82                string_list_split_in_place(&params, params_copy, ',', -1);
  83        for (i = 0; i < params.nr; i++) {
  84                const char *p = params.items[i].string;
  85                if (!strcmp(p, "changes")) {
  86                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  87                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  88                } else if (!strcmp(p, "lines")) {
  89                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  90                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  91                } else if (!strcmp(p, "files")) {
  92                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  93                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  94                } else if (!strcmp(p, "noncumulative")) {
  95                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
  96                } else if (!strcmp(p, "cumulative")) {
  97                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
  98                } else if (isdigit(*p)) {
  99                        char *end;
 100                        int permille = strtoul(p, &end, 10) * 10;
 101                        if (*end == '.' && isdigit(*++end)) {
 102                                /* only use first digit */
 103                                permille += *end - '0';
 104                                /* .. and ignore any further digits */
 105                                while (isdigit(*++end))
 106                                        ; /* nothing */
 107                        }
 108                        if (!*end)
 109                                options->dirstat_permille = permille;
 110                        else {
 111                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%s'\n"),
 112                                            p);
 113                                ret++;
 114                        }
 115                } else {
 116                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%s'\n"), p);
 117                        ret++;
 118                }
 119
 120        }
 121        string_list_clear(&params, 0);
 122        free(params_copy);
 123        return ret;
 124}
 125
 126static int git_config_rename(const char *var, const char *value)
 127{
 128        if (!value)
 129                return DIFF_DETECT_RENAME;
 130        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 131                return  DIFF_DETECT_COPY;
 132        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 133}
 134
 135/*
 136 * These are to give UI layer defaults.
 137 * The core-level commands such as git-diff-files should
 138 * never be affected by the setting of diff.renames
 139 * the user happens to have in the configuration file.
 140 */
 141int git_diff_ui_config(const char *var, const char *value, void *cb)
 142{
 143        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 144                diff_use_color_default = git_config_colorbool(var, value);
 145                return 0;
 146        }
 147        if (!strcmp(var, "diff.context")) {
 148                diff_context_default = git_config_int(var, value);
 149                if (diff_context_default < 0)
 150                        return -1;
 151                return 0;
 152        }
 153        if (!strcmp(var, "diff.renames")) {
 154                diff_detect_rename_default = git_config_rename(var, value);
 155                return 0;
 156        }
 157        if (!strcmp(var, "diff.autorefreshindex")) {
 158                diff_auto_refresh_index = git_config_bool(var, value);
 159                return 0;
 160        }
 161        if (!strcmp(var, "diff.mnemonicprefix")) {
 162                diff_mnemonic_prefix = git_config_bool(var, value);
 163                return 0;
 164        }
 165        if (!strcmp(var, "diff.noprefix")) {
 166                diff_no_prefix = git_config_bool(var, value);
 167                return 0;
 168        }
 169        if (!strcmp(var, "diff.statgraphwidth")) {
 170                diff_stat_graph_width = git_config_int(var, value);
 171                return 0;
 172        }
 173        if (!strcmp(var, "diff.external"))
 174                return git_config_string(&external_diff_cmd_cfg, var, value);
 175        if (!strcmp(var, "diff.wordregex"))
 176                return git_config_string(&diff_word_regex_cfg, var, value);
 177
 178        if (!strcmp(var, "diff.ignoresubmodules"))
 179                handle_ignore_submodules_arg(&default_diff_options, value);
 180
 181        if (git_color_config(var, value, cb) < 0)
 182                return -1;
 183
 184        return git_diff_basic_config(var, value, cb);
 185}
 186
 187int git_diff_basic_config(const char *var, const char *value, void *cb)
 188{
 189        if (!strcmp(var, "diff.renamelimit")) {
 190                diff_rename_limit_default = git_config_int(var, value);
 191                return 0;
 192        }
 193
 194        if (userdiff_config(var, value) < 0)
 195                return -1;
 196
 197        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 198                int slot = parse_diff_color_slot(var, 11);
 199                if (slot < 0)
 200                        return 0;
 201                if (!value)
 202                        return config_error_nonbool(var);
 203                color_parse(value, var, diff_colors[slot]);
 204                return 0;
 205        }
 206
 207        /* like GNU diff's --suppress-blank-empty option  */
 208        if (!strcmp(var, "diff.suppressblankempty") ||
 209                        /* for backwards compatibility */
 210                        !strcmp(var, "diff.suppress-blank-empty")) {
 211                diff_suppress_blank_empty = git_config_bool(var, value);
 212                return 0;
 213        }
 214
 215        if (!strcmp(var, "diff.dirstat")) {
 216                struct strbuf errmsg = STRBUF_INIT;
 217                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 218                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 219                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 220                                errmsg.buf);
 221                strbuf_release(&errmsg);
 222                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 223                return 0;
 224        }
 225
 226        if (!prefixcmp(var, "submodule."))
 227                return parse_submodule_config_option(var, value);
 228
 229        return git_default_config(var, value, cb);
 230}
 231
 232static char *quote_two(const char *one, const char *two)
 233{
 234        int need_one = quote_c_style(one, NULL, NULL, 1);
 235        int need_two = quote_c_style(two, NULL, NULL, 1);
 236        struct strbuf res = STRBUF_INIT;
 237
 238        if (need_one + need_two) {
 239                strbuf_addch(&res, '"');
 240                quote_c_style(one, &res, NULL, 1);
 241                quote_c_style(two, &res, NULL, 1);
 242                strbuf_addch(&res, '"');
 243        } else {
 244                strbuf_addstr(&res, one);
 245                strbuf_addstr(&res, two);
 246        }
 247        return strbuf_detach(&res, NULL);
 248}
 249
 250static const char *external_diff(void)
 251{
 252        static const char *external_diff_cmd = NULL;
 253        static int done_preparing = 0;
 254
 255        if (done_preparing)
 256                return external_diff_cmd;
 257        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 258        if (!external_diff_cmd)
 259                external_diff_cmd = external_diff_cmd_cfg;
 260        done_preparing = 1;
 261        return external_diff_cmd;
 262}
 263
 264static struct diff_tempfile {
 265        const char *name; /* filename external diff should read from */
 266        char hex[41];
 267        char mode[10];
 268        char tmp_path[PATH_MAX];
 269} diff_temp[2];
 270
 271typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 272
 273struct emit_callback {
 274        int color_diff;
 275        unsigned ws_rule;
 276        int blank_at_eof_in_preimage;
 277        int blank_at_eof_in_postimage;
 278        int lno_in_preimage;
 279        int lno_in_postimage;
 280        sane_truncate_fn truncate;
 281        const char **label_path;
 282        struct diff_words_data *diff_words;
 283        struct diff_options *opt;
 284        int *found_changesp;
 285        struct strbuf *header;
 286};
 287
 288static int count_lines(const char *data, int size)
 289{
 290        int count, ch, completely_empty = 1, nl_just_seen = 0;
 291        count = 0;
 292        while (0 < size--) {
 293                ch = *data++;
 294                if (ch == '\n') {
 295                        count++;
 296                        nl_just_seen = 1;
 297                        completely_empty = 0;
 298                }
 299                else {
 300                        nl_just_seen = 0;
 301                        completely_empty = 0;
 302                }
 303        }
 304        if (completely_empty)
 305                return 0;
 306        if (!nl_just_seen)
 307                count++; /* no trailing newline */
 308        return count;
 309}
 310
 311static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 312{
 313        if (!DIFF_FILE_VALID(one)) {
 314                mf->ptr = (char *)""; /* does not matter */
 315                mf->size = 0;
 316                return 0;
 317        }
 318        else if (diff_populate_filespec(one, 0))
 319                return -1;
 320
 321        mf->ptr = one->data;
 322        mf->size = one->size;
 323        return 0;
 324}
 325
 326/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 327static unsigned long diff_filespec_size(struct diff_filespec *one)
 328{
 329        if (!DIFF_FILE_VALID(one))
 330                return 0;
 331        diff_populate_filespec(one, 1);
 332        return one->size;
 333}
 334
 335static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 336{
 337        char *ptr = mf->ptr;
 338        long size = mf->size;
 339        int cnt = 0;
 340
 341        if (!size)
 342                return cnt;
 343        ptr += size - 1; /* pointing at the very end */
 344        if (*ptr != '\n')
 345                ; /* incomplete line */
 346        else
 347                ptr--; /* skip the last LF */
 348        while (mf->ptr < ptr) {
 349                char *prev_eol;
 350                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 351                        if (*prev_eol == '\n')
 352                                break;
 353                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 354                        break;
 355                cnt++;
 356                ptr = prev_eol - 1;
 357        }
 358        return cnt;
 359}
 360
 361static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 362                               struct emit_callback *ecbdata)
 363{
 364        int l1, l2, at;
 365        unsigned ws_rule = ecbdata->ws_rule;
 366        l1 = count_trailing_blank(mf1, ws_rule);
 367        l2 = count_trailing_blank(mf2, ws_rule);
 368        if (l2 <= l1) {
 369                ecbdata->blank_at_eof_in_preimage = 0;
 370                ecbdata->blank_at_eof_in_postimage = 0;
 371                return;
 372        }
 373        at = count_lines(mf1->ptr, mf1->size);
 374        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 375
 376        at = count_lines(mf2->ptr, mf2->size);
 377        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 378}
 379
 380static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 381                        int first, const char *line, int len)
 382{
 383        int has_trailing_newline, has_trailing_carriage_return;
 384        int nofirst;
 385        FILE *file = o->file;
 386
 387        if (o->output_prefix) {
 388                struct strbuf *msg = NULL;
 389                msg = o->output_prefix(o, o->output_prefix_data);
 390                assert(msg);
 391                fwrite(msg->buf, msg->len, 1, file);
 392        }
 393
 394        if (len == 0) {
 395                has_trailing_newline = (first == '\n');
 396                has_trailing_carriage_return = (!has_trailing_newline &&
 397                                                (first == '\r'));
 398                nofirst = has_trailing_newline || has_trailing_carriage_return;
 399        } else {
 400                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 401                if (has_trailing_newline)
 402                        len--;
 403                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 404                if (has_trailing_carriage_return)
 405                        len--;
 406                nofirst = 0;
 407        }
 408
 409        if (len || !nofirst) {
 410                fputs(set, file);
 411                if (!nofirst)
 412                        fputc(first, file);
 413                fwrite(line, len, 1, file);
 414                fputs(reset, file);
 415        }
 416        if (has_trailing_carriage_return)
 417                fputc('\r', file);
 418        if (has_trailing_newline)
 419                fputc('\n', file);
 420}
 421
 422static void emit_line(struct diff_options *o, const char *set, const char *reset,
 423                      const char *line, int len)
 424{
 425        emit_line_0(o, set, reset, line[0], line+1, len-1);
 426}
 427
 428static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 429{
 430        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 431              ecbdata->blank_at_eof_in_preimage &&
 432              ecbdata->blank_at_eof_in_postimage &&
 433              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 434              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 435                return 0;
 436        return ws_blank_line(line, len, ecbdata->ws_rule);
 437}
 438
 439static void emit_add_line(const char *reset,
 440                          struct emit_callback *ecbdata,
 441                          const char *line, int len)
 442{
 443        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 444        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 445
 446        if (!*ws)
 447                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 448        else if (new_blank_line_at_eof(ecbdata, line, len))
 449                /* Blank line at EOF - paint '+' as well */
 450                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 451        else {
 452                /* Emit just the prefix, then the rest. */
 453                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 454                ws_check_emit(line, len, ecbdata->ws_rule,
 455                              ecbdata->opt->file, set, reset, ws);
 456        }
 457}
 458
 459static void emit_hunk_header(struct emit_callback *ecbdata,
 460                             const char *line, int len)
 461{
 462        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 463        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 464        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 465        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 466        static const char atat[2] = { '@', '@' };
 467        const char *cp, *ep;
 468        struct strbuf msgbuf = STRBUF_INIT;
 469        int org_len = len;
 470        int i = 1;
 471
 472        /*
 473         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 474         * it always is at least 10 bytes long.
 475         */
 476        if (len < 10 ||
 477            memcmp(line, atat, 2) ||
 478            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 479                emit_line(ecbdata->opt, plain, reset, line, len);
 480                return;
 481        }
 482        ep += 2; /* skip over @@ */
 483
 484        /* The hunk header in fraginfo color */
 485        strbuf_add(&msgbuf, frag, strlen(frag));
 486        strbuf_add(&msgbuf, line, ep - line);
 487        strbuf_add(&msgbuf, reset, strlen(reset));
 488
 489        /*
 490         * trailing "\r\n"
 491         */
 492        for ( ; i < 3; i++)
 493                if (line[len - i] == '\r' || line[len - i] == '\n')
 494                        len--;
 495
 496        /* blank before the func header */
 497        for (cp = ep; ep - line < len; ep++)
 498                if (*ep != ' ' && *ep != '\t')
 499                        break;
 500        if (ep != cp) {
 501                strbuf_add(&msgbuf, plain, strlen(plain));
 502                strbuf_add(&msgbuf, cp, ep - cp);
 503                strbuf_add(&msgbuf, reset, strlen(reset));
 504        }
 505
 506        if (ep < line + len) {
 507                strbuf_add(&msgbuf, func, strlen(func));
 508                strbuf_add(&msgbuf, ep, line + len - ep);
 509                strbuf_add(&msgbuf, reset, strlen(reset));
 510        }
 511
 512        strbuf_add(&msgbuf, line + len, org_len - len);
 513        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 514        strbuf_release(&msgbuf);
 515}
 516
 517static struct diff_tempfile *claim_diff_tempfile(void) {
 518        int i;
 519        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 520                if (!diff_temp[i].name)
 521                        return diff_temp + i;
 522        die("BUG: diff is failing to clean up its tempfiles");
 523}
 524
 525static int remove_tempfile_installed;
 526
 527static void remove_tempfile(void)
 528{
 529        int i;
 530        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 531                if (diff_temp[i].name == diff_temp[i].tmp_path)
 532                        unlink_or_warn(diff_temp[i].name);
 533                diff_temp[i].name = NULL;
 534        }
 535}
 536
 537static void remove_tempfile_on_signal(int signo)
 538{
 539        remove_tempfile();
 540        sigchain_pop(signo);
 541        raise(signo);
 542}
 543
 544static void print_line_count(FILE *file, int count)
 545{
 546        switch (count) {
 547        case 0:
 548                fprintf(file, "0,0");
 549                break;
 550        case 1:
 551                fprintf(file, "1");
 552                break;
 553        default:
 554                fprintf(file, "1,%d", count);
 555                break;
 556        }
 557}
 558
 559static void emit_rewrite_lines(struct emit_callback *ecb,
 560                               int prefix, const char *data, int size)
 561{
 562        const char *endp = NULL;
 563        static const char *nneof = " No newline at end of file\n";
 564        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 565        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 566
 567        while (0 < size) {
 568                int len;
 569
 570                endp = memchr(data, '\n', size);
 571                len = endp ? (endp - data + 1) : size;
 572                if (prefix != '+') {
 573                        ecb->lno_in_preimage++;
 574                        emit_line_0(ecb->opt, old, reset, '-',
 575                                    data, len);
 576                } else {
 577                        ecb->lno_in_postimage++;
 578                        emit_add_line(reset, ecb, data, len);
 579                }
 580                size -= len;
 581                data += len;
 582        }
 583        if (!endp) {
 584                const char *plain = diff_get_color(ecb->color_diff,
 585                                                   DIFF_PLAIN);
 586                putc('\n', ecb->opt->file);
 587                emit_line_0(ecb->opt, plain, reset, '\\',
 588                            nneof, strlen(nneof));
 589        }
 590}
 591
 592static void emit_rewrite_diff(const char *name_a,
 593                              const char *name_b,
 594                              struct diff_filespec *one,
 595                              struct diff_filespec *two,
 596                              struct userdiff_driver *textconv_one,
 597                              struct userdiff_driver *textconv_two,
 598                              struct diff_options *o)
 599{
 600        int lc_a, lc_b;
 601        const char *name_a_tab, *name_b_tab;
 602        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 603        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 604        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 605        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 606        const char *a_prefix, *b_prefix;
 607        char *data_one, *data_two;
 608        size_t size_one, size_two;
 609        struct emit_callback ecbdata;
 610        char *line_prefix = "";
 611        struct strbuf *msgbuf;
 612
 613        if (o && o->output_prefix) {
 614                msgbuf = o->output_prefix(o, o->output_prefix_data);
 615                line_prefix = msgbuf->buf;
 616        }
 617
 618        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 619                a_prefix = o->b_prefix;
 620                b_prefix = o->a_prefix;
 621        } else {
 622                a_prefix = o->a_prefix;
 623                b_prefix = o->b_prefix;
 624        }
 625
 626        name_a += (*name_a == '/');
 627        name_b += (*name_b == '/');
 628        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 629        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 630
 631        strbuf_reset(&a_name);
 632        strbuf_reset(&b_name);
 633        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 634        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 635
 636        size_one = fill_textconv(textconv_one, one, &data_one);
 637        size_two = fill_textconv(textconv_two, two, &data_two);
 638
 639        memset(&ecbdata, 0, sizeof(ecbdata));
 640        ecbdata.color_diff = want_color(o->use_color);
 641        ecbdata.found_changesp = &o->found_changes;
 642        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 643        ecbdata.opt = o;
 644        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 645                mmfile_t mf1, mf2;
 646                mf1.ptr = (char *)data_one;
 647                mf2.ptr = (char *)data_two;
 648                mf1.size = size_one;
 649                mf2.size = size_two;
 650                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 651        }
 652        ecbdata.lno_in_preimage = 1;
 653        ecbdata.lno_in_postimage = 1;
 654
 655        lc_a = count_lines(data_one, size_one);
 656        lc_b = count_lines(data_two, size_two);
 657        fprintf(o->file,
 658                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 659                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 660                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 661                line_prefix, fraginfo);
 662        if (!o->irreversible_delete)
 663                print_line_count(o->file, lc_a);
 664        else
 665                fprintf(o->file, "?,?");
 666        fprintf(o->file, " +");
 667        print_line_count(o->file, lc_b);
 668        fprintf(o->file, " @@%s\n", reset);
 669        if (lc_a && !o->irreversible_delete)
 670                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 671        if (lc_b)
 672                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 673        if (textconv_one)
 674                free((char *)data_one);
 675        if (textconv_two)
 676                free((char *)data_two);
 677}
 678
 679struct diff_words_buffer {
 680        mmfile_t text;
 681        long alloc;
 682        struct diff_words_orig {
 683                const char *begin, *end;
 684        } *orig;
 685        int orig_nr, orig_alloc;
 686};
 687
 688static void diff_words_append(char *line, unsigned long len,
 689                struct diff_words_buffer *buffer)
 690{
 691        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 692        line++;
 693        len--;
 694        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 695        buffer->text.size += len;
 696        buffer->text.ptr[buffer->text.size] = '\0';
 697}
 698
 699struct diff_words_style_elem {
 700        const char *prefix;
 701        const char *suffix;
 702        const char *color; /* NULL; filled in by the setup code if
 703                            * color is enabled */
 704};
 705
 706struct diff_words_style {
 707        enum diff_words_type type;
 708        struct diff_words_style_elem new, old, ctx;
 709        const char *newline;
 710};
 711
 712static struct diff_words_style diff_words_styles[] = {
 713        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 714        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 715        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 716};
 717
 718struct diff_words_data {
 719        struct diff_words_buffer minus, plus;
 720        const char *current_plus;
 721        int last_minus;
 722        struct diff_options *opt;
 723        regex_t *word_regex;
 724        enum diff_words_type type;
 725        struct diff_words_style *style;
 726};
 727
 728static int fn_out_diff_words_write_helper(FILE *fp,
 729                                          struct diff_words_style_elem *st_el,
 730                                          const char *newline,
 731                                          size_t count, const char *buf,
 732                                          const char *line_prefix)
 733{
 734        int print = 0;
 735
 736        while (count) {
 737                char *p = memchr(buf, '\n', count);
 738                if (print)
 739                        fputs(line_prefix, fp);
 740                if (p != buf) {
 741                        if (st_el->color && fputs(st_el->color, fp) < 0)
 742                                return -1;
 743                        if (fputs(st_el->prefix, fp) < 0 ||
 744                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 745                            fputs(st_el->suffix, fp) < 0)
 746                                return -1;
 747                        if (st_el->color && *st_el->color
 748                            && fputs(GIT_COLOR_RESET, fp) < 0)
 749                                return -1;
 750                }
 751                if (!p)
 752                        return 0;
 753                if (fputs(newline, fp) < 0)
 754                        return -1;
 755                count -= p + 1 - buf;
 756                buf = p + 1;
 757                print = 1;
 758        }
 759        return 0;
 760}
 761
 762/*
 763 * '--color-words' algorithm can be described as:
 764 *
 765 *   1. collect a the minus/plus lines of a diff hunk, divided into
 766 *      minus-lines and plus-lines;
 767 *
 768 *   2. break both minus-lines and plus-lines into words and
 769 *      place them into two mmfile_t with one word for each line;
 770 *
 771 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 772 *
 773 * And for the common parts of the both file, we output the plus side text.
 774 * diff_words->current_plus is used to trace the current position of the plus file
 775 * which printed. diff_words->last_minus is used to trace the last minus word
 776 * printed.
 777 *
 778 * For '--graph' to work with '--color-words', we need to output the graph prefix
 779 * on each line of color words output. Generally, there are two conditions on
 780 * which we should output the prefix.
 781 *
 782 *   1. diff_words->last_minus == 0 &&
 783 *      diff_words->current_plus == diff_words->plus.text.ptr
 784 *
 785 *      that is: the plus text must start as a new line, and if there is no minus
 786 *      word printed, a graph prefix must be printed.
 787 *
 788 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 789 *      *(diff_words->current_plus - 1) == '\n'
 790 *
 791 *      that is: a graph prefix must be printed following a '\n'
 792 */
 793static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 794{
 795        if ((diff_words->last_minus == 0 &&
 796                diff_words->current_plus == diff_words->plus.text.ptr) ||
 797                (diff_words->current_plus > diff_words->plus.text.ptr &&
 798                *(diff_words->current_plus - 1) == '\n')) {
 799                return 1;
 800        } else {
 801                return 0;
 802        }
 803}
 804
 805static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 806{
 807        struct diff_words_data *diff_words = priv;
 808        struct diff_words_style *style = diff_words->style;
 809        int minus_first, minus_len, plus_first, plus_len;
 810        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 811        struct diff_options *opt = diff_words->opt;
 812        struct strbuf *msgbuf;
 813        char *line_prefix = "";
 814
 815        if (line[0] != '@' || parse_hunk_header(line, len,
 816                        &minus_first, &minus_len, &plus_first, &plus_len))
 817                return;
 818
 819        assert(opt);
 820        if (opt->output_prefix) {
 821                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 822                line_prefix = msgbuf->buf;
 823        }
 824
 825        /* POSIX requires that first be decremented by one if len == 0... */
 826        if (minus_len) {
 827                minus_begin = diff_words->minus.orig[minus_first].begin;
 828                minus_end =
 829                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 830        } else
 831                minus_begin = minus_end =
 832                        diff_words->minus.orig[minus_first].end;
 833
 834        if (plus_len) {
 835                plus_begin = diff_words->plus.orig[plus_first].begin;
 836                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 837        } else
 838                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 839
 840        if (color_words_output_graph_prefix(diff_words)) {
 841                fputs(line_prefix, diff_words->opt->file);
 842        }
 843        if (diff_words->current_plus != plus_begin) {
 844                fn_out_diff_words_write_helper(diff_words->opt->file,
 845                                &style->ctx, style->newline,
 846                                plus_begin - diff_words->current_plus,
 847                                diff_words->current_plus, line_prefix);
 848                if (*(plus_begin - 1) == '\n')
 849                        fputs(line_prefix, diff_words->opt->file);
 850        }
 851        if (minus_begin != minus_end) {
 852                fn_out_diff_words_write_helper(diff_words->opt->file,
 853                                &style->old, style->newline,
 854                                minus_end - minus_begin, minus_begin,
 855                                line_prefix);
 856        }
 857        if (plus_begin != plus_end) {
 858                fn_out_diff_words_write_helper(diff_words->opt->file,
 859                                &style->new, style->newline,
 860                                plus_end - plus_begin, plus_begin,
 861                                line_prefix);
 862        }
 863
 864        diff_words->current_plus = plus_end;
 865        diff_words->last_minus = minus_first;
 866}
 867
 868/* This function starts looking at *begin, and returns 0 iff a word was found. */
 869static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 870                int *begin, int *end)
 871{
 872        if (word_regex && *begin < buffer->size) {
 873                regmatch_t match[1];
 874                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 875                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 876                                        '\n', match[0].rm_eo - match[0].rm_so);
 877                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 878                        *begin += match[0].rm_so;
 879                        return *begin >= *end;
 880                }
 881                return -1;
 882        }
 883
 884        /* find the next word */
 885        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 886                (*begin)++;
 887        if (*begin >= buffer->size)
 888                return -1;
 889
 890        /* find the end of the word */
 891        *end = *begin + 1;
 892        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 893                (*end)++;
 894
 895        return 0;
 896}
 897
 898/*
 899 * This function splits the words in buffer->text, stores the list with
 900 * newline separator into out, and saves the offsets of the original words
 901 * in buffer->orig.
 902 */
 903static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 904                regex_t *word_regex)
 905{
 906        int i, j;
 907        long alloc = 0;
 908
 909        out->size = 0;
 910        out->ptr = NULL;
 911
 912        /* fake an empty "0th" word */
 913        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 914        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 915        buffer->orig_nr = 1;
 916
 917        for (i = 0; i < buffer->text.size; i++) {
 918                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 919                        return;
 920
 921                /* store original boundaries */
 922                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 923                                buffer->orig_alloc);
 924                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 925                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 926                buffer->orig_nr++;
 927
 928                /* store one word */
 929                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 930                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 931                out->ptr[out->size + j - i] = '\n';
 932                out->size += j - i + 1;
 933
 934                i = j - 1;
 935        }
 936}
 937
 938/* this executes the word diff on the accumulated buffers */
 939static void diff_words_show(struct diff_words_data *diff_words)
 940{
 941        xpparam_t xpp;
 942        xdemitconf_t xecfg;
 943        mmfile_t minus, plus;
 944        struct diff_words_style *style = diff_words->style;
 945
 946        struct diff_options *opt = diff_words->opt;
 947        struct strbuf *msgbuf;
 948        char *line_prefix = "";
 949
 950        assert(opt);
 951        if (opt->output_prefix) {
 952                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 953                line_prefix = msgbuf->buf;
 954        }
 955
 956        /* special case: only removal */
 957        if (!diff_words->plus.text.size) {
 958                fputs(line_prefix, diff_words->opt->file);
 959                fn_out_diff_words_write_helper(diff_words->opt->file,
 960                        &style->old, style->newline,
 961                        diff_words->minus.text.size,
 962                        diff_words->minus.text.ptr, line_prefix);
 963                diff_words->minus.text.size = 0;
 964                return;
 965        }
 966
 967        diff_words->current_plus = diff_words->plus.text.ptr;
 968        diff_words->last_minus = 0;
 969
 970        memset(&xpp, 0, sizeof(xpp));
 971        memset(&xecfg, 0, sizeof(xecfg));
 972        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 973        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 974        xpp.flags = 0;
 975        /* as only the hunk header will be parsed, we need a 0-context */
 976        xecfg.ctxlen = 0;
 977        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 978                      &xpp, &xecfg);
 979        free(minus.ptr);
 980        free(plus.ptr);
 981        if (diff_words->current_plus != diff_words->plus.text.ptr +
 982                        diff_words->plus.text.size) {
 983                if (color_words_output_graph_prefix(diff_words))
 984                        fputs(line_prefix, diff_words->opt->file);
 985                fn_out_diff_words_write_helper(diff_words->opt->file,
 986                        &style->ctx, style->newline,
 987                        diff_words->plus.text.ptr + diff_words->plus.text.size
 988                        - diff_words->current_plus, diff_words->current_plus,
 989                        line_prefix);
 990        }
 991        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 992}
 993
 994/* In "color-words" mode, show word-diff of words accumulated in the buffer */
 995static void diff_words_flush(struct emit_callback *ecbdata)
 996{
 997        if (ecbdata->diff_words->minus.text.size ||
 998            ecbdata->diff_words->plus.text.size)
 999                diff_words_show(ecbdata->diff_words);
1000}
1001
1002static void diff_filespec_load_driver(struct diff_filespec *one)
1003{
1004        /* Use already-loaded driver */
1005        if (one->driver)
1006                return;
1007
1008        if (S_ISREG(one->mode))
1009                one->driver = userdiff_find_by_path(one->path);
1010
1011        /* Fallback to default settings */
1012        if (!one->driver)
1013                one->driver = userdiff_find_by_name("default");
1014}
1015
1016static const char *userdiff_word_regex(struct diff_filespec *one)
1017{
1018        diff_filespec_load_driver(one);
1019        return one->driver->word_regex;
1020}
1021
1022static void init_diff_words_data(struct emit_callback *ecbdata,
1023                                 struct diff_options *orig_opts,
1024                                 struct diff_filespec *one,
1025                                 struct diff_filespec *two)
1026{
1027        int i;
1028        struct diff_options *o = xmalloc(sizeof(struct diff_options));
1029        memcpy(o, orig_opts, sizeof(struct diff_options));
1030
1031        ecbdata->diff_words =
1032                xcalloc(1, sizeof(struct diff_words_data));
1033        ecbdata->diff_words->type = o->word_diff;
1034        ecbdata->diff_words->opt = o;
1035        if (!o->word_regex)
1036                o->word_regex = userdiff_word_regex(one);
1037        if (!o->word_regex)
1038                o->word_regex = userdiff_word_regex(two);
1039        if (!o->word_regex)
1040                o->word_regex = diff_word_regex_cfg;
1041        if (o->word_regex) {
1042                ecbdata->diff_words->word_regex = (regex_t *)
1043                        xmalloc(sizeof(regex_t));
1044                if (regcomp(ecbdata->diff_words->word_regex,
1045                            o->word_regex,
1046                            REG_EXTENDED | REG_NEWLINE))
1047                        die ("Invalid regular expression: %s",
1048                             o->word_regex);
1049        }
1050        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1051                if (o->word_diff == diff_words_styles[i].type) {
1052                        ecbdata->diff_words->style =
1053                                &diff_words_styles[i];
1054                        break;
1055                }
1056        }
1057        if (want_color(o->use_color)) {
1058                struct diff_words_style *st = ecbdata->diff_words->style;
1059                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1060                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1061                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1062        }
1063}
1064
1065static void free_diff_words_data(struct emit_callback *ecbdata)
1066{
1067        if (ecbdata->diff_words) {
1068                diff_words_flush(ecbdata);
1069                free (ecbdata->diff_words->opt);
1070                free (ecbdata->diff_words->minus.text.ptr);
1071                free (ecbdata->diff_words->minus.orig);
1072                free (ecbdata->diff_words->plus.text.ptr);
1073                free (ecbdata->diff_words->plus.orig);
1074                if (ecbdata->diff_words->word_regex) {
1075                        regfree(ecbdata->diff_words->word_regex);
1076                        free(ecbdata->diff_words->word_regex);
1077                }
1078                free(ecbdata->diff_words);
1079                ecbdata->diff_words = NULL;
1080        }
1081}
1082
1083const char *diff_get_color(int diff_use_color, enum color_diff ix)
1084{
1085        if (want_color(diff_use_color))
1086                return diff_colors[ix];
1087        return "";
1088}
1089
1090static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1091{
1092        const char *cp;
1093        unsigned long allot;
1094        size_t l = len;
1095
1096        if (ecb->truncate)
1097                return ecb->truncate(line, len);
1098        cp = line;
1099        allot = l;
1100        while (0 < l) {
1101                (void) utf8_width(&cp, &l);
1102                if (!cp)
1103                        break; /* truncated in the middle? */
1104        }
1105        return allot - l;
1106}
1107
1108static void find_lno(const char *line, struct emit_callback *ecbdata)
1109{
1110        const char *p;
1111        ecbdata->lno_in_preimage = 0;
1112        ecbdata->lno_in_postimage = 0;
1113        p = strchr(line, '-');
1114        if (!p)
1115                return; /* cannot happen */
1116        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1117        p = strchr(p, '+');
1118        if (!p)
1119                return; /* cannot happen */
1120        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1121}
1122
1123static void fn_out_consume(void *priv, char *line, unsigned long len)
1124{
1125        struct emit_callback *ecbdata = priv;
1126        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1127        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1128        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1129        struct diff_options *o = ecbdata->opt;
1130        char *line_prefix = "";
1131        struct strbuf *msgbuf;
1132
1133        if (o && o->output_prefix) {
1134                msgbuf = o->output_prefix(o, o->output_prefix_data);
1135                line_prefix = msgbuf->buf;
1136        }
1137
1138        if (ecbdata->header) {
1139                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1140                strbuf_reset(ecbdata->header);
1141                ecbdata->header = NULL;
1142        }
1143        *(ecbdata->found_changesp) = 1;
1144
1145        if (ecbdata->label_path[0]) {
1146                const char *name_a_tab, *name_b_tab;
1147
1148                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1149                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1150
1151                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1152                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1153                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1154                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1155                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1156        }
1157
1158        if (diff_suppress_blank_empty
1159            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1160                line[0] = '\n';
1161                len = 1;
1162        }
1163
1164        if (line[0] == '@') {
1165                if (ecbdata->diff_words)
1166                        diff_words_flush(ecbdata);
1167                len = sane_truncate_line(ecbdata, line, len);
1168                find_lno(line, ecbdata);
1169                emit_hunk_header(ecbdata, line, len);
1170                if (line[len-1] != '\n')
1171                        putc('\n', ecbdata->opt->file);
1172                return;
1173        }
1174
1175        if (len < 1) {
1176                emit_line(ecbdata->opt, reset, reset, line, len);
1177                if (ecbdata->diff_words
1178                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1179                        fputs("~\n", ecbdata->opt->file);
1180                return;
1181        }
1182
1183        if (ecbdata->diff_words) {
1184                if (line[0] == '-') {
1185                        diff_words_append(line, len,
1186                                          &ecbdata->diff_words->minus);
1187                        return;
1188                } else if (line[0] == '+') {
1189                        diff_words_append(line, len,
1190                                          &ecbdata->diff_words->plus);
1191                        return;
1192                } else if (!prefixcmp(line, "\\ ")) {
1193                        /*
1194                         * Eat the "no newline at eof" marker as if we
1195                         * saw a "+" or "-" line with nothing on it,
1196                         * and return without diff_words_flush() to
1197                         * defer processing. If this is the end of
1198                         * preimage, more "+" lines may come after it.
1199                         */
1200                        return;
1201                }
1202                diff_words_flush(ecbdata);
1203                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1204                        emit_line(ecbdata->opt, plain, reset, line, len);
1205                        fputs("~\n", ecbdata->opt->file);
1206                } else {
1207                        /*
1208                         * Skip the prefix character, if any.  With
1209                         * diff_suppress_blank_empty, there may be
1210                         * none.
1211                         */
1212                        if (line[0] != '\n') {
1213                              line++;
1214                              len--;
1215                        }
1216                        emit_line(ecbdata->opt, plain, reset, line, len);
1217                }
1218                return;
1219        }
1220
1221        if (line[0] != '+') {
1222                const char *color =
1223                        diff_get_color(ecbdata->color_diff,
1224                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1225                ecbdata->lno_in_preimage++;
1226                if (line[0] == ' ')
1227                        ecbdata->lno_in_postimage++;
1228                emit_line(ecbdata->opt, color, reset, line, len);
1229        } else {
1230                ecbdata->lno_in_postimage++;
1231                emit_add_line(reset, ecbdata, line + 1, len - 1);
1232        }
1233}
1234
1235static char *pprint_rename(const char *a, const char *b)
1236{
1237        const char *old = a;
1238        const char *new = b;
1239        struct strbuf name = STRBUF_INIT;
1240        int pfx_length, sfx_length;
1241        int len_a = strlen(a);
1242        int len_b = strlen(b);
1243        int a_midlen, b_midlen;
1244        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1245        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1246
1247        if (qlen_a || qlen_b) {
1248                quote_c_style(a, &name, NULL, 0);
1249                strbuf_addstr(&name, " => ");
1250                quote_c_style(b, &name, NULL, 0);
1251                return strbuf_detach(&name, NULL);
1252        }
1253
1254        /* Find common prefix */
1255        pfx_length = 0;
1256        while (*old && *new && *old == *new) {
1257                if (*old == '/')
1258                        pfx_length = old - a + 1;
1259                old++;
1260                new++;
1261        }
1262
1263        /* Find common suffix */
1264        old = a + len_a;
1265        new = b + len_b;
1266        sfx_length = 0;
1267        while (a <= old && b <= new && *old == *new) {
1268                if (*old == '/')
1269                        sfx_length = len_a - (old - a);
1270                old--;
1271                new--;
1272        }
1273
1274        /*
1275         * pfx{mid-a => mid-b}sfx
1276         * {pfx-a => pfx-b}sfx
1277         * pfx{sfx-a => sfx-b}
1278         * name-a => name-b
1279         */
1280        a_midlen = len_a - pfx_length - sfx_length;
1281        b_midlen = len_b - pfx_length - sfx_length;
1282        if (a_midlen < 0)
1283                a_midlen = 0;
1284        if (b_midlen < 0)
1285                b_midlen = 0;
1286
1287        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1288        if (pfx_length + sfx_length) {
1289                strbuf_add(&name, a, pfx_length);
1290                strbuf_addch(&name, '{');
1291        }
1292        strbuf_add(&name, a + pfx_length, a_midlen);
1293        strbuf_addstr(&name, " => ");
1294        strbuf_add(&name, b + pfx_length, b_midlen);
1295        if (pfx_length + sfx_length) {
1296                strbuf_addch(&name, '}');
1297                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1298        }
1299        return strbuf_detach(&name, NULL);
1300}
1301
1302struct diffstat_t {
1303        int nr;
1304        int alloc;
1305        struct diffstat_file {
1306                char *from_name;
1307                char *name;
1308                char *print_name;
1309                unsigned is_unmerged:1;
1310                unsigned is_binary:1;
1311                unsigned is_renamed:1;
1312                uintmax_t added, deleted;
1313        } **files;
1314};
1315
1316static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1317                                          const char *name_a,
1318                                          const char *name_b)
1319{
1320        struct diffstat_file *x;
1321        x = xcalloc(sizeof (*x), 1);
1322        if (diffstat->nr == diffstat->alloc) {
1323                diffstat->alloc = alloc_nr(diffstat->alloc);
1324                diffstat->files = xrealloc(diffstat->files,
1325                                diffstat->alloc * sizeof(x));
1326        }
1327        diffstat->files[diffstat->nr++] = x;
1328        if (name_b) {
1329                x->from_name = xstrdup(name_a);
1330                x->name = xstrdup(name_b);
1331                x->is_renamed = 1;
1332        }
1333        else {
1334                x->from_name = NULL;
1335                x->name = xstrdup(name_a);
1336        }
1337        return x;
1338}
1339
1340static void diffstat_consume(void *priv, char *line, unsigned long len)
1341{
1342        struct diffstat_t *diffstat = priv;
1343        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1344
1345        if (line[0] == '+')
1346                x->added++;
1347        else if (line[0] == '-')
1348                x->deleted++;
1349}
1350
1351const char mime_boundary_leader[] = "------------";
1352
1353static int scale_linear(int it, int width, int max_change)
1354{
1355        if (!it)
1356                return 0;
1357        /*
1358         * make sure that at least one '-' or '+' is printed if
1359         * there is any change to this path. The easiest way is to
1360         * scale linearly as if the alloted width is one column shorter
1361         * than it is, and then add 1 to the result.
1362         */
1363        return 1 + (it * (width - 1) / max_change);
1364}
1365
1366static void show_name(FILE *file,
1367                      const char *prefix, const char *name, int len)
1368{
1369        fprintf(file, " %s%-*s |", prefix, len, name);
1370}
1371
1372static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1373{
1374        if (cnt <= 0)
1375                return;
1376        fprintf(file, "%s", set);
1377        while (cnt--)
1378                putc(ch, file);
1379        fprintf(file, "%s", reset);
1380}
1381
1382static void fill_print_name(struct diffstat_file *file)
1383{
1384        char *pname;
1385
1386        if (file->print_name)
1387                return;
1388
1389        if (!file->is_renamed) {
1390                struct strbuf buf = STRBUF_INIT;
1391                if (quote_c_style(file->name, &buf, NULL, 0)) {
1392                        pname = strbuf_detach(&buf, NULL);
1393                } else {
1394                        pname = file->name;
1395                        strbuf_release(&buf);
1396                }
1397        } else {
1398                pname = pprint_rename(file->from_name, file->name);
1399        }
1400        file->print_name = pname;
1401}
1402
1403int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1404{
1405        struct strbuf sb = STRBUF_INIT;
1406        int ret;
1407
1408        if (!files) {
1409                assert(insertions == 0 && deletions == 0);
1410                return fprintf(fp, "%s\n", " 0 files changed");
1411        }
1412
1413        strbuf_addf(&sb,
1414                    (files == 1) ? " %d file changed" : " %d files changed",
1415                    files);
1416
1417        /*
1418         * For binary diff, the caller may want to print "x files
1419         * changed" with insertions == 0 && deletions == 0.
1420         *
1421         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1422         * is probably less confusing (i.e skip over "2 files changed
1423         * but nothing about added/removed lines? Is this a bug in Git?").
1424         */
1425        if (insertions || deletions == 0) {
1426                /*
1427                 * TRANSLATORS: "+" in (+) is a line addition marker;
1428                 * do not translate it.
1429                 */
1430                strbuf_addf(&sb,
1431                            (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1432                            insertions);
1433        }
1434
1435        if (deletions || insertions == 0) {
1436                /*
1437                 * TRANSLATORS: "-" in (-) is a line removal marker;
1438                 * do not translate it.
1439                 */
1440                strbuf_addf(&sb,
1441                            (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1442                            deletions);
1443        }
1444        strbuf_addch(&sb, '\n');
1445        ret = fputs(sb.buf, fp);
1446        strbuf_release(&sb);
1447        return ret;
1448}
1449
1450static void show_stats(struct diffstat_t *data, struct diff_options *options)
1451{
1452        int i, len, add, del, adds = 0, dels = 0;
1453        uintmax_t max_change = 0, max_len = 0;
1454        int total_files = data->nr, count;
1455        int width, name_width, graph_width, number_width = 0, bin_width = 0;
1456        const char *reset, *add_c, *del_c;
1457        const char *line_prefix = "";
1458        int extra_shown = 0;
1459        struct strbuf *msg = NULL;
1460
1461        if (data->nr == 0)
1462                return;
1463
1464        if (options->output_prefix) {
1465                msg = options->output_prefix(options, options->output_prefix_data);
1466                line_prefix = msg->buf;
1467        }
1468
1469        count = options->stat_count ? options->stat_count : data->nr;
1470
1471        reset = diff_get_color_opt(options, DIFF_RESET);
1472        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1473        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1474
1475        /*
1476         * Find the longest filename and max number of changes
1477         */
1478        for (i = 0; (i < count) && (i < data->nr); i++) {
1479                struct diffstat_file *file = data->files[i];
1480                uintmax_t change = file->added + file->deleted;
1481                if (!data->files[i]->is_renamed &&
1482                         (change == 0)) {
1483                        count++; /* not shown == room for one more */
1484                        continue;
1485                }
1486                fill_print_name(file);
1487                len = strlen(file->print_name);
1488                if (max_len < len)
1489                        max_len = len;
1490
1491                if (file->is_unmerged) {
1492                        /* "Unmerged" is 8 characters */
1493                        bin_width = bin_width < 8 ? 8 : bin_width;
1494                        continue;
1495                }
1496                if (file->is_binary) {
1497                        /* "Bin XXX -> YYY bytes" */
1498                        int w = 14 + decimal_width(file->added)
1499                                + decimal_width(file->deleted);
1500                        bin_width = bin_width < w ? w : bin_width;
1501                        /* Display change counts aligned with "Bin" */
1502                        number_width = 3;
1503                        continue;
1504                }
1505
1506                if (max_change < change)
1507                        max_change = change;
1508        }
1509        count = i; /* min(count, data->nr) */
1510
1511        /*
1512         * We have width = stat_width or term_columns() columns total.
1513         * We want a maximum of min(max_len, stat_name_width) for the name part.
1514         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1515         * We also need 1 for " " and 4 + decimal_width(max_change)
1516         * for " | NNNN " and one the empty column at the end, altogether
1517         * 6 + decimal_width(max_change).
1518         *
1519         * If there's not enough space, we will use the smaller of
1520         * stat_name_width (if set) and 5/8*width for the filename,
1521         * and the rest for constant elements + graph part, but no more
1522         * than stat_graph_width for the graph part.
1523         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1524         * for the standard terminal size).
1525         *
1526         * In other words: stat_width limits the maximum width, and
1527         * stat_name_width fixes the maximum width of the filename,
1528         * and is also used to divide available columns if there
1529         * aren't enough.
1530         *
1531         * Binary files are displayed with "Bin XXX -> YYY bytes"
1532         * instead of the change count and graph. This part is treated
1533         * similarly to the graph part, except that it is not
1534         * "scaled". If total width is too small to accomodate the
1535         * guaranteed minimum width of the filename part and the
1536         * separators and this message, this message will "overflow"
1537         * making the line longer than the maximum width.
1538         */
1539
1540        if (options->stat_width == -1)
1541                width = term_columns() - options->output_prefix_length;
1542        else
1543                width = options->stat_width ? options->stat_width : 80;
1544        number_width = decimal_width(max_change) > number_width ?
1545                decimal_width(max_change) : number_width;
1546
1547        if (options->stat_graph_width == -1)
1548                options->stat_graph_width = diff_stat_graph_width;
1549
1550        /*
1551         * Guarantee 3/8*16==6 for the graph part
1552         * and 5/8*16==10 for the filename part
1553         */
1554        if (width < 16 + 6 + number_width)
1555                width = 16 + 6 + number_width;
1556
1557        /*
1558         * First assign sizes that are wanted, ignoring available width.
1559         * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1560         * starting from "XXX" should fit in graph_width.
1561         */
1562        graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1563        if (options->stat_graph_width &&
1564            options->stat_graph_width < graph_width)
1565                graph_width = options->stat_graph_width;
1566
1567        name_width = (options->stat_name_width > 0 &&
1568                      options->stat_name_width < max_len) ?
1569                options->stat_name_width : max_len;
1570
1571        /*
1572         * Adjust adjustable widths not to exceed maximum width
1573         */
1574        if (name_width + number_width + 6 + graph_width > width) {
1575                if (graph_width > width * 3/8 - number_width - 6) {
1576                        graph_width = width * 3/8 - number_width - 6;
1577                        if (graph_width < 6)
1578                                graph_width = 6;
1579                }
1580
1581                if (options->stat_graph_width &&
1582                    graph_width > options->stat_graph_width)
1583                        graph_width = options->stat_graph_width;
1584                if (name_width > width - number_width - 6 - graph_width)
1585                        name_width = width - number_width - 6 - graph_width;
1586                else
1587                        graph_width = width - number_width - 6 - name_width;
1588        }
1589
1590        /*
1591         * From here name_width is the width of the name area,
1592         * and graph_width is the width of the graph area.
1593         * max_change is used to scale graph properly.
1594         */
1595        for (i = 0; i < count; i++) {
1596                const char *prefix = "";
1597                char *name = data->files[i]->print_name;
1598                uintmax_t added = data->files[i]->added;
1599                uintmax_t deleted = data->files[i]->deleted;
1600                int name_len;
1601
1602                if (!data->files[i]->is_renamed &&
1603                         (added + deleted == 0)) {
1604                        total_files--;
1605                        continue;
1606                }
1607                /*
1608                 * "scale" the filename
1609                 */
1610                len = name_width;
1611                name_len = strlen(name);
1612                if (name_width < name_len) {
1613                        char *slash;
1614                        prefix = "...";
1615                        len -= 3;
1616                        name += name_len - len;
1617                        slash = strchr(name, '/');
1618                        if (slash)
1619                                name = slash;
1620                }
1621
1622                if (data->files[i]->is_binary) {
1623                        fprintf(options->file, "%s", line_prefix);
1624                        show_name(options->file, prefix, name, len);
1625                        fprintf(options->file, " %*s", number_width, "Bin");
1626                        if (!added && !deleted) {
1627                                putc('\n', options->file);
1628                                continue;
1629                        }
1630                        fprintf(options->file, " %s%"PRIuMAX"%s",
1631                                del_c, deleted, reset);
1632                        fprintf(options->file, " -> ");
1633                        fprintf(options->file, "%s%"PRIuMAX"%s",
1634                                add_c, added, reset);
1635                        fprintf(options->file, " bytes");
1636                        fprintf(options->file, "\n");
1637                        continue;
1638                }
1639                else if (data->files[i]->is_unmerged) {
1640                        fprintf(options->file, "%s", line_prefix);
1641                        show_name(options->file, prefix, name, len);
1642                        fprintf(options->file, " Unmerged\n");
1643                        continue;
1644                }
1645
1646                /*
1647                 * scale the add/delete
1648                 */
1649                add = added;
1650                del = deleted;
1651                adds += add;
1652                dels += del;
1653
1654                if (graph_width <= max_change) {
1655                        int total = add + del;
1656
1657                        total = scale_linear(add + del, graph_width, max_change);
1658                        if (total < 2 && add && del)
1659                                /* width >= 2 due to the sanity check */
1660                                total = 2;
1661                        if (add < del) {
1662                                add = scale_linear(add, graph_width, max_change);
1663                                del = total - add;
1664                        } else {
1665                                del = scale_linear(del, graph_width, max_change);
1666                                add = total - del;
1667                        }
1668                }
1669                fprintf(options->file, "%s", line_prefix);
1670                show_name(options->file, prefix, name, len);
1671                fprintf(options->file, " %*"PRIuMAX"%s",
1672                        number_width, added + deleted,
1673                        added + deleted ? " " : "");
1674                show_graph(options->file, '+', add, add_c, reset);
1675                show_graph(options->file, '-', del, del_c, reset);
1676                fprintf(options->file, "\n");
1677        }
1678        for (i = count; i < data->nr; i++) {
1679                uintmax_t added = data->files[i]->added;
1680                uintmax_t deleted = data->files[i]->deleted;
1681                if (!data->files[i]->is_renamed &&
1682                         (added + deleted == 0)) {
1683                        total_files--;
1684                        continue;
1685                }
1686                adds += added;
1687                dels += deleted;
1688                if (!extra_shown)
1689                        fprintf(options->file, "%s ...\n", line_prefix);
1690                extra_shown = 1;
1691        }
1692        fprintf(options->file, "%s", line_prefix);
1693        print_stat_summary(options->file, total_files, adds, dels);
1694}
1695
1696static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1697{
1698        int i, adds = 0, dels = 0, total_files = data->nr;
1699
1700        if (data->nr == 0)
1701                return;
1702
1703        for (i = 0; i < data->nr; i++) {
1704                int added = data->files[i]->added;
1705                int deleted= data->files[i]->deleted;
1706
1707                if (data->files[i]->is_unmerged)
1708                        continue;
1709                if (!data->files[i]->is_renamed && (added + deleted == 0)) {
1710                        total_files--;
1711                } else if (!data->files[i]->is_binary) { /* don't count bytes */
1712                        adds += added;
1713                        dels += deleted;
1714                }
1715        }
1716        if (options->output_prefix) {
1717                struct strbuf *msg = NULL;
1718                msg = options->output_prefix(options,
1719                                options->output_prefix_data);
1720                fprintf(options->file, "%s", msg->buf);
1721        }
1722        print_stat_summary(options->file, total_files, adds, dels);
1723}
1724
1725static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1726{
1727        int i;
1728
1729        if (data->nr == 0)
1730                return;
1731
1732        for (i = 0; i < data->nr; i++) {
1733                struct diffstat_file *file = data->files[i];
1734
1735                if (options->output_prefix) {
1736                        struct strbuf *msg = NULL;
1737                        msg = options->output_prefix(options,
1738                                        options->output_prefix_data);
1739                        fprintf(options->file, "%s", msg->buf);
1740                }
1741
1742                if (file->is_binary)
1743                        fprintf(options->file, "-\t-\t");
1744                else
1745                        fprintf(options->file,
1746                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1747                                file->added, file->deleted);
1748                if (options->line_termination) {
1749                        fill_print_name(file);
1750                        if (!file->is_renamed)
1751                                write_name_quoted(file->name, options->file,
1752                                                  options->line_termination);
1753                        else {
1754                                fputs(file->print_name, options->file);
1755                                putc(options->line_termination, options->file);
1756                        }
1757                } else {
1758                        if (file->is_renamed) {
1759                                putc('\0', options->file);
1760                                write_name_quoted(file->from_name, options->file, '\0');
1761                        }
1762                        write_name_quoted(file->name, options->file, '\0');
1763                }
1764        }
1765}
1766
1767struct dirstat_file {
1768        const char *name;
1769        unsigned long changed;
1770};
1771
1772struct dirstat_dir {
1773        struct dirstat_file *files;
1774        int alloc, nr, permille, cumulative;
1775};
1776
1777static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1778                unsigned long changed, const char *base, int baselen)
1779{
1780        unsigned long this_dir = 0;
1781        unsigned int sources = 0;
1782        const char *line_prefix = "";
1783        struct strbuf *msg = NULL;
1784
1785        if (opt->output_prefix) {
1786                msg = opt->output_prefix(opt, opt->output_prefix_data);
1787                line_prefix = msg->buf;
1788        }
1789
1790        while (dir->nr) {
1791                struct dirstat_file *f = dir->files;
1792                int namelen = strlen(f->name);
1793                unsigned long this;
1794                char *slash;
1795
1796                if (namelen < baselen)
1797                        break;
1798                if (memcmp(f->name, base, baselen))
1799                        break;
1800                slash = strchr(f->name + baselen, '/');
1801                if (slash) {
1802                        int newbaselen = slash + 1 - f->name;
1803                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1804                        sources++;
1805                } else {
1806                        this = f->changed;
1807                        dir->files++;
1808                        dir->nr--;
1809                        sources += 2;
1810                }
1811                this_dir += this;
1812        }
1813
1814        /*
1815         * We don't report dirstat's for
1816         *  - the top level
1817         *  - or cases where everything came from a single directory
1818         *    under this directory (sources == 1).
1819         */
1820        if (baselen && sources != 1) {
1821                if (this_dir) {
1822                        int permille = this_dir * 1000 / changed;
1823                        if (permille >= dir->permille) {
1824                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1825                                        permille / 10, permille % 10, baselen, base);
1826                                if (!dir->cumulative)
1827                                        return 0;
1828                        }
1829                }
1830        }
1831        return this_dir;
1832}
1833
1834static int dirstat_compare(const void *_a, const void *_b)
1835{
1836        const struct dirstat_file *a = _a;
1837        const struct dirstat_file *b = _b;
1838        return strcmp(a->name, b->name);
1839}
1840
1841static void show_dirstat(struct diff_options *options)
1842{
1843        int i;
1844        unsigned long changed;
1845        struct dirstat_dir dir;
1846        struct diff_queue_struct *q = &diff_queued_diff;
1847
1848        dir.files = NULL;
1849        dir.alloc = 0;
1850        dir.nr = 0;
1851        dir.permille = options->dirstat_permille;
1852        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1853
1854        changed = 0;
1855        for (i = 0; i < q->nr; i++) {
1856                struct diff_filepair *p = q->queue[i];
1857                const char *name;
1858                unsigned long copied, added, damage;
1859                int content_changed;
1860
1861                name = p->two->path ? p->two->path : p->one->path;
1862
1863                if (p->one->sha1_valid && p->two->sha1_valid)
1864                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1865                else
1866                        content_changed = 1;
1867
1868                if (!content_changed) {
1869                        /*
1870                         * The SHA1 has not changed, so pre-/post-content is
1871                         * identical. We can therefore skip looking at the
1872                         * file contents altogether.
1873                         */
1874                        damage = 0;
1875                        goto found_damage;
1876                }
1877
1878                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1879                        /*
1880                         * In --dirstat-by-file mode, we don't really need to
1881                         * look at the actual file contents at all.
1882                         * The fact that the SHA1 changed is enough for us to
1883                         * add this file to the list of results
1884                         * (with each file contributing equal damage).
1885                         */
1886                        damage = 1;
1887                        goto found_damage;
1888                }
1889
1890                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1891                        diff_populate_filespec(p->one, 0);
1892                        diff_populate_filespec(p->two, 0);
1893                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1894                                               &copied, &added);
1895                        diff_free_filespec_data(p->one);
1896                        diff_free_filespec_data(p->two);
1897                } else if (DIFF_FILE_VALID(p->one)) {
1898                        diff_populate_filespec(p->one, 1);
1899                        copied = added = 0;
1900                        diff_free_filespec_data(p->one);
1901                } else if (DIFF_FILE_VALID(p->two)) {
1902                        diff_populate_filespec(p->two, 1);
1903                        copied = 0;
1904                        added = p->two->size;
1905                        diff_free_filespec_data(p->two);
1906                } else
1907                        continue;
1908
1909                /*
1910                 * Original minus copied is the removed material,
1911                 * added is the new material.  They are both damages
1912                 * made to the preimage.
1913                 * If the resulting damage is zero, we know that
1914                 * diffcore_count_changes() considers the two entries to
1915                 * be identical, but since content_changed is true, we
1916                 * know that there must have been _some_ kind of change,
1917                 * so we force all entries to have damage > 0.
1918                 */
1919                damage = (p->one->size - copied) + added;
1920                if (!damage)
1921                        damage = 1;
1922
1923found_damage:
1924                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1925                dir.files[dir.nr].name = name;
1926                dir.files[dir.nr].changed = damage;
1927                changed += damage;
1928                dir.nr++;
1929        }
1930
1931        /* This can happen even with many files, if everything was renames */
1932        if (!changed)
1933                return;
1934
1935        /* Show all directories with more than x% of the changes */
1936        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1937        gather_dirstat(options, &dir, changed, "", 0);
1938}
1939
1940static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1941{
1942        int i;
1943        unsigned long changed;
1944        struct dirstat_dir dir;
1945
1946        if (data->nr == 0)
1947                return;
1948
1949        dir.files = NULL;
1950        dir.alloc = 0;
1951        dir.nr = 0;
1952        dir.permille = options->dirstat_permille;
1953        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1954
1955        changed = 0;
1956        for (i = 0; i < data->nr; i++) {
1957                struct diffstat_file *file = data->files[i];
1958                unsigned long damage = file->added + file->deleted;
1959                if (file->is_binary)
1960                        /*
1961                         * binary files counts bytes, not lines. Must find some
1962                         * way to normalize binary bytes vs. textual lines.
1963                         * The following heuristic assumes that there are 64
1964                         * bytes per "line".
1965                         * This is stupid and ugly, but very cheap...
1966                         */
1967                        damage = (damage + 63) / 64;
1968                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1969                dir.files[dir.nr].name = file->name;
1970                dir.files[dir.nr].changed = damage;
1971                changed += damage;
1972                dir.nr++;
1973        }
1974
1975        /* This can happen even with many files, if everything was renames */
1976        if (!changed)
1977                return;
1978
1979        /* Show all directories with more than x% of the changes */
1980        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1981        gather_dirstat(options, &dir, changed, "", 0);
1982}
1983
1984static void free_diffstat_info(struct diffstat_t *diffstat)
1985{
1986        int i;
1987        for (i = 0; i < diffstat->nr; i++) {
1988                struct diffstat_file *f = diffstat->files[i];
1989                if (f->name != f->print_name)
1990                        free(f->print_name);
1991                free(f->name);
1992                free(f->from_name);
1993                free(f);
1994        }
1995        free(diffstat->files);
1996}
1997
1998struct checkdiff_t {
1999        const char *filename;
2000        int lineno;
2001        int conflict_marker_size;
2002        struct diff_options *o;
2003        unsigned ws_rule;
2004        unsigned status;
2005};
2006
2007static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2008{
2009        char firstchar;
2010        int cnt;
2011
2012        if (len < marker_size + 1)
2013                return 0;
2014        firstchar = line[0];
2015        switch (firstchar) {
2016        case '=': case '>': case '<': case '|':
2017                break;
2018        default:
2019                return 0;
2020        }
2021        for (cnt = 1; cnt < marker_size; cnt++)
2022                if (line[cnt] != firstchar)
2023                        return 0;
2024        /* line[1] thru line[marker_size-1] are same as firstchar */
2025        if (len < marker_size + 1 || !isspace(line[marker_size]))
2026                return 0;
2027        return 1;
2028}
2029
2030static void checkdiff_consume(void *priv, char *line, unsigned long len)
2031{
2032        struct checkdiff_t *data = priv;
2033        int marker_size = data->conflict_marker_size;
2034        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2035        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2036        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2037        char *err;
2038        char *line_prefix = "";
2039        struct strbuf *msgbuf;
2040
2041        assert(data->o);
2042        if (data->o->output_prefix) {
2043                msgbuf = data->o->output_prefix(data->o,
2044                        data->o->output_prefix_data);
2045                line_prefix = msgbuf->buf;
2046        }
2047
2048        if (line[0] == '+') {
2049                unsigned bad;
2050                data->lineno++;
2051                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2052                        data->status |= 1;
2053                        fprintf(data->o->file,
2054                                "%s%s:%d: leftover conflict marker\n",
2055                                line_prefix, data->filename, data->lineno);
2056                }
2057                bad = ws_check(line + 1, len - 1, data->ws_rule);
2058                if (!bad)
2059                        return;
2060                data->status |= bad;
2061                err = whitespace_error_string(bad);
2062                fprintf(data->o->file, "%s%s:%d: %s.\n",
2063                        line_prefix, data->filename, data->lineno, err);
2064                free(err);
2065                emit_line(data->o, set, reset, line, 1);
2066                ws_check_emit(line + 1, len - 1, data->ws_rule,
2067                              data->o->file, set, reset, ws);
2068        } else if (line[0] == ' ') {
2069                data->lineno++;
2070        } else if (line[0] == '@') {
2071                char *plus = strchr(line, '+');
2072                if (plus)
2073                        data->lineno = strtol(plus, NULL, 10) - 1;
2074                else
2075                        die("invalid diff");
2076        }
2077}
2078
2079static unsigned char *deflate_it(char *data,
2080                                 unsigned long size,
2081                                 unsigned long *result_size)
2082{
2083        int bound;
2084        unsigned char *deflated;
2085        git_zstream stream;
2086
2087        memset(&stream, 0, sizeof(stream));
2088        git_deflate_init(&stream, zlib_compression_level);
2089        bound = git_deflate_bound(&stream, size);
2090        deflated = xmalloc(bound);
2091        stream.next_out = deflated;
2092        stream.avail_out = bound;
2093
2094        stream.next_in = (unsigned char *)data;
2095        stream.avail_in = size;
2096        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2097                ; /* nothing */
2098        git_deflate_end(&stream);
2099        *result_size = stream.total_out;
2100        return deflated;
2101}
2102
2103static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2104{
2105        void *cp;
2106        void *delta;
2107        void *deflated;
2108        void *data;
2109        unsigned long orig_size;
2110        unsigned long delta_size;
2111        unsigned long deflate_size;
2112        unsigned long data_size;
2113
2114        /* We could do deflated delta, or we could do just deflated two,
2115         * whichever is smaller.
2116         */
2117        delta = NULL;
2118        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2119        if (one->size && two->size) {
2120                delta = diff_delta(one->ptr, one->size,
2121                                   two->ptr, two->size,
2122                                   &delta_size, deflate_size);
2123                if (delta) {
2124                        void *to_free = delta;
2125                        orig_size = delta_size;
2126                        delta = deflate_it(delta, delta_size, &delta_size);
2127                        free(to_free);
2128                }
2129        }
2130
2131        if (delta && delta_size < deflate_size) {
2132                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2133                free(deflated);
2134                data = delta;
2135                data_size = delta_size;
2136        }
2137        else {
2138                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2139                free(delta);
2140                data = deflated;
2141                data_size = deflate_size;
2142        }
2143
2144        /* emit data encoded in base85 */
2145        cp = data;
2146        while (data_size) {
2147                int bytes = (52 < data_size) ? 52 : data_size;
2148                char line[70];
2149                data_size -= bytes;
2150                if (bytes <= 26)
2151                        line[0] = bytes + 'A' - 1;
2152                else
2153                        line[0] = bytes - 26 + 'a' - 1;
2154                encode_85(line + 1, cp, bytes);
2155                cp = (char *) cp + bytes;
2156                fprintf(file, "%s", prefix);
2157                fputs(line, file);
2158                fputc('\n', file);
2159        }
2160        fprintf(file, "%s\n", prefix);
2161        free(data);
2162}
2163
2164static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2165{
2166        fprintf(file, "%sGIT binary patch\n", prefix);
2167        emit_binary_diff_body(file, one, two, prefix);
2168        emit_binary_diff_body(file, two, one, prefix);
2169}
2170
2171int diff_filespec_is_binary(struct diff_filespec *one)
2172{
2173        if (one->is_binary == -1) {
2174                diff_filespec_load_driver(one);
2175                if (one->driver->binary != -1)
2176                        one->is_binary = one->driver->binary;
2177                else {
2178                        if (!one->data && DIFF_FILE_VALID(one))
2179                                diff_populate_filespec(one, 0);
2180                        if (one->data)
2181                                one->is_binary = buffer_is_binary(one->data,
2182                                                one->size);
2183                        if (one->is_binary == -1)
2184                                one->is_binary = 0;
2185                }
2186        }
2187        return one->is_binary;
2188}
2189
2190static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2191{
2192        diff_filespec_load_driver(one);
2193        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2194}
2195
2196void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2197{
2198        if (!options->a_prefix)
2199                options->a_prefix = a;
2200        if (!options->b_prefix)
2201                options->b_prefix = b;
2202}
2203
2204struct userdiff_driver *get_textconv(struct diff_filespec *one)
2205{
2206        if (!DIFF_FILE_VALID(one))
2207                return NULL;
2208
2209        diff_filespec_load_driver(one);
2210        return userdiff_get_textconv(one->driver);
2211}
2212
2213static void builtin_diff(const char *name_a,
2214                         const char *name_b,
2215                         struct diff_filespec *one,
2216                         struct diff_filespec *two,
2217                         const char *xfrm_msg,
2218                         int must_show_header,
2219                         struct diff_options *o,
2220                         int complete_rewrite)
2221{
2222        mmfile_t mf1, mf2;
2223        const char *lbl[2];
2224        char *a_one, *b_two;
2225        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2226        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2227        const char *a_prefix, *b_prefix;
2228        struct userdiff_driver *textconv_one = NULL;
2229        struct userdiff_driver *textconv_two = NULL;
2230        struct strbuf header = STRBUF_INIT;
2231        struct strbuf *msgbuf;
2232        char *line_prefix = "";
2233
2234        if (o->output_prefix) {
2235                msgbuf = o->output_prefix(o, o->output_prefix_data);
2236                line_prefix = msgbuf->buf;
2237        }
2238
2239        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2240                        (!one->mode || S_ISGITLINK(one->mode)) &&
2241                        (!two->mode || S_ISGITLINK(two->mode))) {
2242                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2243                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2244                show_submodule_summary(o->file, one ? one->path : two->path,
2245                                one->sha1, two->sha1, two->dirty_submodule,
2246                                del, add, reset);
2247                return;
2248        }
2249
2250        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2251                textconv_one = get_textconv(one);
2252                textconv_two = get_textconv(two);
2253        }
2254
2255        diff_set_mnemonic_prefix(o, "a/", "b/");
2256        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2257                a_prefix = o->b_prefix;
2258                b_prefix = o->a_prefix;
2259        } else {
2260                a_prefix = o->a_prefix;
2261                b_prefix = o->b_prefix;
2262        }
2263
2264        /* Never use a non-valid filename anywhere if at all possible */
2265        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2266        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2267
2268        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2269        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2270        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2271        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2272        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2273        if (lbl[0][0] == '/') {
2274                /* /dev/null */
2275                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2276                if (xfrm_msg)
2277                        strbuf_addstr(&header, xfrm_msg);
2278                must_show_header = 1;
2279        }
2280        else if (lbl[1][0] == '/') {
2281                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2282                if (xfrm_msg)
2283                        strbuf_addstr(&header, xfrm_msg);
2284                must_show_header = 1;
2285        }
2286        else {
2287                if (one->mode != two->mode) {
2288                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2289                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2290                        must_show_header = 1;
2291                }
2292                if (xfrm_msg)
2293                        strbuf_addstr(&header, xfrm_msg);
2294
2295                /*
2296                 * we do not run diff between different kind
2297                 * of objects.
2298                 */
2299                if ((one->mode ^ two->mode) & S_IFMT)
2300                        goto free_ab_and_return;
2301                if (complete_rewrite &&
2302                    (textconv_one || !diff_filespec_is_binary(one)) &&
2303                    (textconv_two || !diff_filespec_is_binary(two))) {
2304                        fprintf(o->file, "%s", header.buf);
2305                        strbuf_reset(&header);
2306                        emit_rewrite_diff(name_a, name_b, one, two,
2307                                                textconv_one, textconv_two, o);
2308                        o->found_changes = 1;
2309                        goto free_ab_and_return;
2310                }
2311        }
2312
2313        if (o->irreversible_delete && lbl[1][0] == '/') {
2314                fprintf(o->file, "%s", header.buf);
2315                strbuf_reset(&header);
2316                goto free_ab_and_return;
2317        } else if (!DIFF_OPT_TST(o, TEXT) &&
2318            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2319              (!textconv_two && diff_filespec_is_binary(two)) )) {
2320                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2321                        die("unable to read files to diff");
2322                /* Quite common confusing case */
2323                if (mf1.size == mf2.size &&
2324                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2325                        if (must_show_header)
2326                                fprintf(o->file, "%s", header.buf);
2327                        goto free_ab_and_return;
2328                }
2329                fprintf(o->file, "%s", header.buf);
2330                strbuf_reset(&header);
2331                if (DIFF_OPT_TST(o, BINARY))
2332                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2333                else
2334                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2335                                line_prefix, lbl[0], lbl[1]);
2336                o->found_changes = 1;
2337        } else {
2338                /* Crazy xdl interfaces.. */
2339                const char *diffopts = getenv("GIT_DIFF_OPTS");
2340                xpparam_t xpp;
2341                xdemitconf_t xecfg;
2342                struct emit_callback ecbdata;
2343                const struct userdiff_funcname *pe;
2344
2345                if (must_show_header) {
2346                        fprintf(o->file, "%s", header.buf);
2347                        strbuf_reset(&header);
2348                }
2349
2350                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2351                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2352
2353                pe = diff_funcname_pattern(one);
2354                if (!pe)
2355                        pe = diff_funcname_pattern(two);
2356
2357                memset(&xpp, 0, sizeof(xpp));
2358                memset(&xecfg, 0, sizeof(xecfg));
2359                memset(&ecbdata, 0, sizeof(ecbdata));
2360                ecbdata.label_path = lbl;
2361                ecbdata.color_diff = want_color(o->use_color);
2362                ecbdata.found_changesp = &o->found_changes;
2363                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2364                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2365                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2366                ecbdata.opt = o;
2367                ecbdata.header = header.len ? &header : NULL;
2368                xpp.flags = o->xdl_opts;
2369                xecfg.ctxlen = o->context;
2370                xecfg.interhunkctxlen = o->interhunkcontext;
2371                xecfg.flags = XDL_EMIT_FUNCNAMES;
2372                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2373                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2374                if (pe)
2375                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2376                if (!diffopts)
2377                        ;
2378                else if (!prefixcmp(diffopts, "--unified="))
2379                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2380                else if (!prefixcmp(diffopts, "-u"))
2381                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2382                if (o->word_diff)
2383                        init_diff_words_data(&ecbdata, o, one, two);
2384                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2385                              &xpp, &xecfg);
2386                if (o->word_diff)
2387                        free_diff_words_data(&ecbdata);
2388                if (textconv_one)
2389                        free(mf1.ptr);
2390                if (textconv_two)
2391                        free(mf2.ptr);
2392                xdiff_clear_find_func(&xecfg);
2393        }
2394
2395 free_ab_and_return:
2396        strbuf_release(&header);
2397        diff_free_filespec_data(one);
2398        diff_free_filespec_data(two);
2399        free(a_one);
2400        free(b_two);
2401        return;
2402}
2403
2404static void builtin_diffstat(const char *name_a, const char *name_b,
2405                             struct diff_filespec *one,
2406                             struct diff_filespec *two,
2407                             struct diffstat_t *diffstat,
2408                             struct diff_options *o,
2409                             int complete_rewrite)
2410{
2411        mmfile_t mf1, mf2;
2412        struct diffstat_file *data;
2413        int same_contents;
2414
2415        data = diffstat_add(diffstat, name_a, name_b);
2416
2417        if (!one || !two) {
2418                data->is_unmerged = 1;
2419                return;
2420        }
2421
2422        same_contents = !hashcmp(one->sha1, two->sha1);
2423
2424        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2425                data->is_binary = 1;
2426                if (same_contents) {
2427                        data->added = 0;
2428                        data->deleted = 0;
2429                } else {
2430                        data->added = diff_filespec_size(two);
2431                        data->deleted = diff_filespec_size(one);
2432                }
2433        }
2434
2435        else if (complete_rewrite) {
2436                diff_populate_filespec(one, 0);
2437                diff_populate_filespec(two, 0);
2438                data->deleted = count_lines(one->data, one->size);
2439                data->added = count_lines(two->data, two->size);
2440        }
2441
2442        else if (!same_contents) {
2443                /* Crazy xdl interfaces.. */
2444                xpparam_t xpp;
2445                xdemitconf_t xecfg;
2446
2447                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2448                        die("unable to read files to diff");
2449
2450                memset(&xpp, 0, sizeof(xpp));
2451                memset(&xecfg, 0, sizeof(xecfg));
2452                xpp.flags = o->xdl_opts;
2453                xecfg.ctxlen = o->context;
2454                xecfg.interhunkctxlen = o->interhunkcontext;
2455                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2456                              &xpp, &xecfg);
2457        }
2458
2459        diff_free_filespec_data(one);
2460        diff_free_filespec_data(two);
2461}
2462
2463static void builtin_checkdiff(const char *name_a, const char *name_b,
2464                              const char *attr_path,
2465                              struct diff_filespec *one,
2466                              struct diff_filespec *two,
2467                              struct diff_options *o)
2468{
2469        mmfile_t mf1, mf2;
2470        struct checkdiff_t data;
2471
2472        if (!two)
2473                return;
2474
2475        memset(&data, 0, sizeof(data));
2476        data.filename = name_b ? name_b : name_a;
2477        data.lineno = 0;
2478        data.o = o;
2479        data.ws_rule = whitespace_rule(attr_path);
2480        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2481
2482        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2483                die("unable to read files to diff");
2484
2485        /*
2486         * All the other codepaths check both sides, but not checking
2487         * the "old" side here is deliberate.  We are checking the newly
2488         * introduced changes, and as long as the "new" side is text, we
2489         * can and should check what it introduces.
2490         */
2491        if (diff_filespec_is_binary(two))
2492                goto free_and_return;
2493        else {
2494                /* Crazy xdl interfaces.. */
2495                xpparam_t xpp;
2496                xdemitconf_t xecfg;
2497
2498                memset(&xpp, 0, sizeof(xpp));
2499                memset(&xecfg, 0, sizeof(xecfg));
2500                xecfg.ctxlen = 1; /* at least one context line */
2501                xpp.flags = 0;
2502                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2503                              &xpp, &xecfg);
2504
2505                if (data.ws_rule & WS_BLANK_AT_EOF) {
2506                        struct emit_callback ecbdata;
2507                        int blank_at_eof;
2508
2509                        ecbdata.ws_rule = data.ws_rule;
2510                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2511                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2512
2513                        if (blank_at_eof) {
2514                                static char *err;
2515                                if (!err)
2516                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2517                                fprintf(o->file, "%s:%d: %s.\n",
2518                                        data.filename, blank_at_eof, err);
2519                                data.status = 1; /* report errors */
2520                        }
2521                }
2522        }
2523 free_and_return:
2524        diff_free_filespec_data(one);
2525        diff_free_filespec_data(two);
2526        if (data.status)
2527                DIFF_OPT_SET(o, CHECK_FAILED);
2528}
2529
2530struct diff_filespec *alloc_filespec(const char *path)
2531{
2532        int namelen = strlen(path);
2533        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2534
2535        memset(spec, 0, sizeof(*spec));
2536        spec->path = (char *)(spec + 1);
2537        memcpy(spec->path, path, namelen+1);
2538        spec->count = 1;
2539        spec->is_binary = -1;
2540        return spec;
2541}
2542
2543void free_filespec(struct diff_filespec *spec)
2544{
2545        if (!--spec->count) {
2546                diff_free_filespec_data(spec);
2547                free(spec);
2548        }
2549}
2550
2551void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2552                   int sha1_valid, unsigned short mode)
2553{
2554        if (mode) {
2555                spec->mode = canon_mode(mode);
2556                hashcpy(spec->sha1, sha1);
2557                spec->sha1_valid = sha1_valid;
2558        }
2559}
2560
2561/*
2562 * Given a name and sha1 pair, if the index tells us the file in
2563 * the work tree has that object contents, return true, so that
2564 * prepare_temp_file() does not have to inflate and extract.
2565 */
2566static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2567{
2568        struct cache_entry *ce;
2569        struct stat st;
2570        int pos, len;
2571
2572        /*
2573         * We do not read the cache ourselves here, because the
2574         * benchmark with my previous version that always reads cache
2575         * shows that it makes things worse for diff-tree comparing
2576         * two linux-2.6 kernel trees in an already checked out work
2577         * tree.  This is because most diff-tree comparisons deal with
2578         * only a small number of files, while reading the cache is
2579         * expensive for a large project, and its cost outweighs the
2580         * savings we get by not inflating the object to a temporary
2581         * file.  Practically, this code only helps when we are used
2582         * by diff-cache --cached, which does read the cache before
2583         * calling us.
2584         */
2585        if (!active_cache)
2586                return 0;
2587
2588        /* We want to avoid the working directory if our caller
2589         * doesn't need the data in a normal file, this system
2590         * is rather slow with its stat/open/mmap/close syscalls,
2591         * and the object is contained in a pack file.  The pack
2592         * is probably already open and will be faster to obtain
2593         * the data through than the working directory.  Loose
2594         * objects however would tend to be slower as they need
2595         * to be individually opened and inflated.
2596         */
2597        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2598                return 0;
2599
2600        len = strlen(name);
2601        pos = cache_name_pos(name, len);
2602        if (pos < 0)
2603                return 0;
2604        ce = active_cache[pos];
2605
2606        /*
2607         * This is not the sha1 we are looking for, or
2608         * unreusable because it is not a regular file.
2609         */
2610        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2611                return 0;
2612
2613        /*
2614         * If ce is marked as "assume unchanged", there is no
2615         * guarantee that work tree matches what we are looking for.
2616         */
2617        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2618                return 0;
2619
2620        /*
2621         * If ce matches the file in the work tree, we can reuse it.
2622         */
2623        if (ce_uptodate(ce) ||
2624            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2625                return 1;
2626
2627        return 0;
2628}
2629
2630static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2631{
2632        int len;
2633        char *data = xmalloc(100), *dirty = "";
2634
2635        /* Are we looking at the work tree? */
2636        if (s->dirty_submodule)
2637                dirty = "-dirty";
2638
2639        len = snprintf(data, 100,
2640                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2641        s->data = data;
2642        s->size = len;
2643        s->should_free = 1;
2644        if (size_only) {
2645                s->data = NULL;
2646                free(data);
2647        }
2648        return 0;
2649}
2650
2651/*
2652 * While doing rename detection and pickaxe operation, we may need to
2653 * grab the data for the blob (or file) for our own in-core comparison.
2654 * diff_filespec has data and size fields for this purpose.
2655 */
2656int diff_populate_filespec(struct diff_filespec *s, int size_only)
2657{
2658        int err = 0;
2659        if (!DIFF_FILE_VALID(s))
2660                die("internal error: asking to populate invalid file.");
2661        if (S_ISDIR(s->mode))
2662                return -1;
2663
2664        if (s->data)
2665                return 0;
2666
2667        if (size_only && 0 < s->size)
2668                return 0;
2669
2670        if (S_ISGITLINK(s->mode))
2671                return diff_populate_gitlink(s, size_only);
2672
2673        if (!s->sha1_valid ||
2674            reuse_worktree_file(s->path, s->sha1, 0)) {
2675                struct strbuf buf = STRBUF_INIT;
2676                struct stat st;
2677                int fd;
2678
2679                if (lstat(s->path, &st) < 0) {
2680                        if (errno == ENOENT) {
2681                        err_empty:
2682                                err = -1;
2683                        empty:
2684                                s->data = (char *)"";
2685                                s->size = 0;
2686                                return err;
2687                        }
2688                }
2689                s->size = xsize_t(st.st_size);
2690                if (!s->size)
2691                        goto empty;
2692                if (S_ISLNK(st.st_mode)) {
2693                        struct strbuf sb = STRBUF_INIT;
2694
2695                        if (strbuf_readlink(&sb, s->path, s->size))
2696                                goto err_empty;
2697                        s->size = sb.len;
2698                        s->data = strbuf_detach(&sb, NULL);
2699                        s->should_free = 1;
2700                        return 0;
2701                }
2702                if (size_only)
2703                        return 0;
2704                fd = open(s->path, O_RDONLY);
2705                if (fd < 0)
2706                        goto err_empty;
2707                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2708                close(fd);
2709                s->should_munmap = 1;
2710
2711                /*
2712                 * Convert from working tree format to canonical git format
2713                 */
2714                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2715                        size_t size = 0;
2716                        munmap(s->data, s->size);
2717                        s->should_munmap = 0;
2718                        s->data = strbuf_detach(&buf, &size);
2719                        s->size = size;
2720                        s->should_free = 1;
2721                }
2722        }
2723        else {
2724                enum object_type type;
2725                if (size_only) {
2726                        type = sha1_object_info(s->sha1, &s->size);
2727                        if (type < 0)
2728                                die("unable to read %s", sha1_to_hex(s->sha1));
2729                } else {
2730                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2731                        if (!s->data)
2732                                die("unable to read %s", sha1_to_hex(s->sha1));
2733                        s->should_free = 1;
2734                }
2735        }
2736        return 0;
2737}
2738
2739void diff_free_filespec_blob(struct diff_filespec *s)
2740{
2741        if (s->should_free)
2742                free(s->data);
2743        else if (s->should_munmap)
2744                munmap(s->data, s->size);
2745
2746        if (s->should_free || s->should_munmap) {
2747                s->should_free = s->should_munmap = 0;
2748                s->data = NULL;
2749        }
2750}
2751
2752void diff_free_filespec_data(struct diff_filespec *s)
2753{
2754        diff_free_filespec_blob(s);
2755        free(s->cnt_data);
2756        s->cnt_data = NULL;
2757}
2758
2759static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2760                           void *blob,
2761                           unsigned long size,
2762                           const unsigned char *sha1,
2763                           int mode)
2764{
2765        int fd;
2766        struct strbuf buf = STRBUF_INIT;
2767        struct strbuf template = STRBUF_INIT;
2768        char *path_dup = xstrdup(path);
2769        const char *base = basename(path_dup);
2770
2771        /* Generate "XXXXXX_basename.ext" */
2772        strbuf_addstr(&template, "XXXXXX_");
2773        strbuf_addstr(&template, base);
2774
2775        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2776                        strlen(base) + 1);
2777        if (fd < 0)
2778                die_errno("unable to create temp-file");
2779        if (convert_to_working_tree(path,
2780                        (const char *)blob, (size_t)size, &buf)) {
2781                blob = buf.buf;
2782                size = buf.len;
2783        }
2784        if (write_in_full(fd, blob, size) != size)
2785                die_errno("unable to write temp-file");
2786        close(fd);
2787        temp->name = temp->tmp_path;
2788        strcpy(temp->hex, sha1_to_hex(sha1));
2789        temp->hex[40] = 0;
2790        sprintf(temp->mode, "%06o", mode);
2791        strbuf_release(&buf);
2792        strbuf_release(&template);
2793        free(path_dup);
2794}
2795
2796static struct diff_tempfile *prepare_temp_file(const char *name,
2797                struct diff_filespec *one)
2798{
2799        struct diff_tempfile *temp = claim_diff_tempfile();
2800
2801        if (!DIFF_FILE_VALID(one)) {
2802        not_a_valid_file:
2803                /* A '-' entry produces this for file-2, and
2804                 * a '+' entry produces this for file-1.
2805                 */
2806                temp->name = "/dev/null";
2807                strcpy(temp->hex, ".");
2808                strcpy(temp->mode, ".");
2809                return temp;
2810        }
2811
2812        if (!remove_tempfile_installed) {
2813                atexit(remove_tempfile);
2814                sigchain_push_common(remove_tempfile_on_signal);
2815                remove_tempfile_installed = 1;
2816        }
2817
2818        if (!one->sha1_valid ||
2819            reuse_worktree_file(name, one->sha1, 1)) {
2820                struct stat st;
2821                if (lstat(name, &st) < 0) {
2822                        if (errno == ENOENT)
2823                                goto not_a_valid_file;
2824                        die_errno("stat(%s)", name);
2825                }
2826                if (S_ISLNK(st.st_mode)) {
2827                        struct strbuf sb = STRBUF_INIT;
2828                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2829                                die_errno("readlink(%s)", name);
2830                        prep_temp_blob(name, temp, sb.buf, sb.len,
2831                                       (one->sha1_valid ?
2832                                        one->sha1 : null_sha1),
2833                                       (one->sha1_valid ?
2834                                        one->mode : S_IFLNK));
2835                        strbuf_release(&sb);
2836                }
2837                else {
2838                        /* we can borrow from the file in the work tree */
2839                        temp->name = name;
2840                        if (!one->sha1_valid)
2841                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2842                        else
2843                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2844                        /* Even though we may sometimes borrow the
2845                         * contents from the work tree, we always want
2846                         * one->mode.  mode is trustworthy even when
2847                         * !(one->sha1_valid), as long as
2848                         * DIFF_FILE_VALID(one).
2849                         */
2850                        sprintf(temp->mode, "%06o", one->mode);
2851                }
2852                return temp;
2853        }
2854        else {
2855                if (diff_populate_filespec(one, 0))
2856                        die("cannot read data blob for %s", one->path);
2857                prep_temp_blob(name, temp, one->data, one->size,
2858                               one->sha1, one->mode);
2859        }
2860        return temp;
2861}
2862
2863/* An external diff command takes:
2864 *
2865 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2866 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2867 *
2868 */
2869static void run_external_diff(const char *pgm,
2870                              const char *name,
2871                              const char *other,
2872                              struct diff_filespec *one,
2873                              struct diff_filespec *two,
2874                              const char *xfrm_msg,
2875                              int complete_rewrite)
2876{
2877        const char *spawn_arg[10];
2878        int retval;
2879        const char **arg = &spawn_arg[0];
2880
2881        if (one && two) {
2882                struct diff_tempfile *temp_one, *temp_two;
2883                const char *othername = (other ? other : name);
2884                temp_one = prepare_temp_file(name, one);
2885                temp_two = prepare_temp_file(othername, two);
2886                *arg++ = pgm;
2887                *arg++ = name;
2888                *arg++ = temp_one->name;
2889                *arg++ = temp_one->hex;
2890                *arg++ = temp_one->mode;
2891                *arg++ = temp_two->name;
2892                *arg++ = temp_two->hex;
2893                *arg++ = temp_two->mode;
2894                if (other) {
2895                        *arg++ = other;
2896                        *arg++ = xfrm_msg;
2897                }
2898        } else {
2899                *arg++ = pgm;
2900                *arg++ = name;
2901        }
2902        *arg = NULL;
2903        fflush(NULL);
2904        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2905        remove_tempfile();
2906        if (retval) {
2907                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2908                exit(1);
2909        }
2910}
2911
2912static int similarity_index(struct diff_filepair *p)
2913{
2914        return p->score * 100 / MAX_SCORE;
2915}
2916
2917static void fill_metainfo(struct strbuf *msg,
2918                          const char *name,
2919                          const char *other,
2920                          struct diff_filespec *one,
2921                          struct diff_filespec *two,
2922                          struct diff_options *o,
2923                          struct diff_filepair *p,
2924                          int *must_show_header,
2925                          int use_color)
2926{
2927        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2928        const char *reset = diff_get_color(use_color, DIFF_RESET);
2929        struct strbuf *msgbuf;
2930        char *line_prefix = "";
2931
2932        *must_show_header = 1;
2933        if (o->output_prefix) {
2934                msgbuf = o->output_prefix(o, o->output_prefix_data);
2935                line_prefix = msgbuf->buf;
2936        }
2937        strbuf_init(msg, PATH_MAX * 2 + 300);
2938        switch (p->status) {
2939        case DIFF_STATUS_COPIED:
2940                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2941                            line_prefix, set, similarity_index(p));
2942                strbuf_addf(msg, "%s\n%s%scopy from ",
2943                            reset,  line_prefix, set);
2944                quote_c_style(name, msg, NULL, 0);
2945                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2946                quote_c_style(other, msg, NULL, 0);
2947                strbuf_addf(msg, "%s\n", reset);
2948                break;
2949        case DIFF_STATUS_RENAMED:
2950                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2951                            line_prefix, set, similarity_index(p));
2952                strbuf_addf(msg, "%s\n%s%srename from ",
2953                            reset, line_prefix, set);
2954                quote_c_style(name, msg, NULL, 0);
2955                strbuf_addf(msg, "%s\n%s%srename to ",
2956                            reset, line_prefix, set);
2957                quote_c_style(other, msg, NULL, 0);
2958                strbuf_addf(msg, "%s\n", reset);
2959                break;
2960        case DIFF_STATUS_MODIFIED:
2961                if (p->score) {
2962                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2963                                    line_prefix,
2964                                    set, similarity_index(p), reset);
2965                        break;
2966                }
2967                /* fallthru */
2968        default:
2969                *must_show_header = 0;
2970        }
2971        if (one && two && hashcmp(one->sha1, two->sha1)) {
2972                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2973
2974                if (DIFF_OPT_TST(o, BINARY)) {
2975                        mmfile_t mf;
2976                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2977                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2978                                abbrev = 40;
2979                }
2980                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2981                            find_unique_abbrev(one->sha1, abbrev));
2982                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2983                if (one->mode == two->mode)
2984                        strbuf_addf(msg, " %06o", one->mode);
2985                strbuf_addf(msg, "%s\n", reset);
2986        }
2987}
2988
2989static void run_diff_cmd(const char *pgm,
2990                         const char *name,
2991                         const char *other,
2992                         const char *attr_path,
2993                         struct diff_filespec *one,
2994                         struct diff_filespec *two,
2995                         struct strbuf *msg,
2996                         struct diff_options *o,
2997                         struct diff_filepair *p)
2998{
2999        const char *xfrm_msg = NULL;
3000        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3001        int must_show_header = 0;
3002
3003
3004        if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3005                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3006                if (drv && drv->external)
3007                        pgm = drv->external;
3008        }
3009
3010        if (msg) {
3011                /*
3012                 * don't use colors when the header is intended for an
3013                 * external diff driver
3014                 */
3015                fill_metainfo(msg, name, other, one, two, o, p,
3016                              &must_show_header,
3017                              want_color(o->use_color) && !pgm);
3018                xfrm_msg = msg->len ? msg->buf : NULL;
3019        }
3020
3021        if (pgm) {
3022                run_external_diff(pgm, name, other, one, two, xfrm_msg,
3023                                  complete_rewrite);
3024                return;
3025        }
3026        if (one && two)
3027                builtin_diff(name, other ? other : name,
3028                             one, two, xfrm_msg, must_show_header,
3029                             o, complete_rewrite);
3030        else
3031                fprintf(o->file, "* Unmerged path %s\n", name);
3032}
3033
3034static void diff_fill_sha1_info(struct diff_filespec *one)
3035{
3036        if (DIFF_FILE_VALID(one)) {
3037                if (!one->sha1_valid) {
3038                        struct stat st;
3039                        if (one->is_stdin) {
3040                                hashcpy(one->sha1, null_sha1);
3041                                return;
3042                        }
3043                        if (lstat(one->path, &st) < 0)
3044                                die_errno("stat '%s'", one->path);
3045                        if (index_path(one->sha1, one->path, &st, 0))
3046                                die("cannot hash %s", one->path);
3047                }
3048        }
3049        else
3050                hashclr(one->sha1);
3051}
3052
3053static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3054{
3055        /* Strip the prefix but do not molest /dev/null and absolute paths */
3056        if (*namep && **namep != '/') {
3057                *namep += prefix_length;
3058                if (**namep == '/')
3059                        ++*namep;
3060        }
3061        if (*otherp && **otherp != '/') {
3062                *otherp += prefix_length;
3063                if (**otherp == '/')
3064                        ++*otherp;
3065        }
3066}
3067
3068static void run_diff(struct diff_filepair *p, struct diff_options *o)
3069{
3070        const char *pgm = external_diff();
3071        struct strbuf msg;
3072        struct diff_filespec *one = p->one;
3073        struct diff_filespec *two = p->two;
3074        const char *name;
3075        const char *other;
3076        const char *attr_path;
3077
3078        name  = p->one->path;
3079        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3080        attr_path = name;
3081        if (o->prefix_length)
3082                strip_prefix(o->prefix_length, &name, &other);
3083
3084        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3085                pgm = NULL;
3086
3087        if (DIFF_PAIR_UNMERGED(p)) {
3088                run_diff_cmd(pgm, name, NULL, attr_path,
3089                             NULL, NULL, NULL, o, p);
3090                return;
3091        }
3092
3093        diff_fill_sha1_info(one);
3094        diff_fill_sha1_info(two);
3095
3096        if (!pgm &&
3097            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3098            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3099                /*
3100                 * a filepair that changes between file and symlink
3101                 * needs to be split into deletion and creation.
3102                 */
3103                struct diff_filespec *null = alloc_filespec(two->path);
3104                run_diff_cmd(NULL, name, other, attr_path,
3105                             one, null, &msg, o, p);
3106                free(null);
3107                strbuf_release(&msg);
3108
3109                null = alloc_filespec(one->path);
3110                run_diff_cmd(NULL, name, other, attr_path,
3111                             null, two, &msg, o, p);
3112                free(null);
3113        }
3114        else
3115                run_diff_cmd(pgm, name, other, attr_path,
3116                             one, two, &msg, o, p);
3117
3118        strbuf_release(&msg);
3119}
3120
3121static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3122                         struct diffstat_t *diffstat)
3123{
3124        const char *name;
3125        const char *other;
3126        int complete_rewrite = 0;
3127
3128        if (DIFF_PAIR_UNMERGED(p)) {
3129                /* unmerged */
3130                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3131                return;
3132        }
3133
3134        name = p->one->path;
3135        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3136
3137        if (o->prefix_length)
3138                strip_prefix(o->prefix_length, &name, &other);
3139
3140        diff_fill_sha1_info(p->one);
3141        diff_fill_sha1_info(p->two);
3142
3143        if (p->status == DIFF_STATUS_MODIFIED && p->score)
3144                complete_rewrite = 1;
3145        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3146}
3147
3148static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3149{
3150        const char *name;
3151        const char *other;
3152        const char *attr_path;
3153
3154        if (DIFF_PAIR_UNMERGED(p)) {
3155                /* unmerged */
3156                return;
3157        }
3158
3159        name = p->one->path;
3160        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3161        attr_path = other ? other : name;
3162
3163        if (o->prefix_length)
3164                strip_prefix(o->prefix_length, &name, &other);
3165
3166        diff_fill_sha1_info(p->one);
3167        diff_fill_sha1_info(p->two);
3168
3169        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3170}
3171
3172void diff_setup(struct diff_options *options)
3173{
3174        memcpy(options, &default_diff_options, sizeof(*options));
3175
3176        options->file = stdout;
3177
3178        options->line_termination = '\n';
3179        options->break_opt = -1;
3180        options->rename_limit = -1;
3181        options->dirstat_permille = diff_dirstat_permille_default;
3182        options->context = diff_context_default;
3183        DIFF_OPT_SET(options, RENAME_EMPTY);
3184
3185        options->change = diff_change;
3186        options->add_remove = diff_addremove;
3187        options->use_color = diff_use_color_default;
3188        options->detect_rename = diff_detect_rename_default;
3189
3190        if (diff_no_prefix) {
3191                options->a_prefix = options->b_prefix = "";
3192        } else if (!diff_mnemonic_prefix) {
3193                options->a_prefix = "a/";
3194                options->b_prefix = "b/";
3195        }
3196}
3197
3198void diff_setup_done(struct diff_options *options)
3199{
3200        int count = 0;
3201
3202        if (options->output_format & DIFF_FORMAT_NAME)
3203                count++;
3204        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3205                count++;
3206        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3207                count++;
3208        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3209                count++;
3210        if (count > 1)
3211                die("--name-only, --name-status, --check and -s are mutually exclusive");
3212
3213        /*
3214         * Most of the time we can say "there are changes"
3215         * only by checking if there are changed paths, but
3216         * --ignore-whitespace* options force us to look
3217         * inside contents.
3218         */
3219
3220        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3221            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3222            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3223                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3224        else
3225                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3226
3227        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3228                options->detect_rename = DIFF_DETECT_COPY;
3229
3230        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3231                options->prefix = NULL;
3232        if (options->prefix)
3233                options->prefix_length = strlen(options->prefix);
3234        else
3235                options->prefix_length = 0;
3236
3237        if (options->output_format & (DIFF_FORMAT_NAME |
3238                                      DIFF_FORMAT_NAME_STATUS |
3239                                      DIFF_FORMAT_CHECKDIFF |
3240                                      DIFF_FORMAT_NO_OUTPUT))
3241                options->output_format &= ~(DIFF_FORMAT_RAW |
3242                                            DIFF_FORMAT_NUMSTAT |
3243                                            DIFF_FORMAT_DIFFSTAT |
3244                                            DIFF_FORMAT_SHORTSTAT |
3245                                            DIFF_FORMAT_DIRSTAT |
3246                                            DIFF_FORMAT_SUMMARY |
3247                                            DIFF_FORMAT_PATCH);
3248
3249        /*
3250         * These cases always need recursive; we do not drop caller-supplied
3251         * recursive bits for other formats here.
3252         */
3253        if (options->output_format & (DIFF_FORMAT_PATCH |
3254                                      DIFF_FORMAT_NUMSTAT |
3255                                      DIFF_FORMAT_DIFFSTAT |
3256                                      DIFF_FORMAT_SHORTSTAT |
3257                                      DIFF_FORMAT_DIRSTAT |
3258                                      DIFF_FORMAT_SUMMARY |
3259                                      DIFF_FORMAT_CHECKDIFF))
3260                DIFF_OPT_SET(options, RECURSIVE);
3261        /*
3262         * Also pickaxe would not work very well if you do not say recursive
3263         */
3264        if (options->pickaxe)
3265                DIFF_OPT_SET(options, RECURSIVE);
3266        /*
3267         * When patches are generated, submodules diffed against the work tree
3268         * must be checked for dirtiness too so it can be shown in the output
3269         */
3270        if (options->output_format & DIFF_FORMAT_PATCH)
3271                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3272
3273        if (options->detect_rename && options->rename_limit < 0)
3274                options->rename_limit = diff_rename_limit_default;
3275        if (options->setup & DIFF_SETUP_USE_CACHE) {
3276                if (!active_cache)
3277                        /* read-cache does not die even when it fails
3278                         * so it is safe for us to do this here.  Also
3279                         * it does not smudge active_cache or active_nr
3280                         * when it fails, so we do not have to worry about
3281                         * cleaning it up ourselves either.
3282                         */
3283                        read_cache();
3284        }
3285        if (options->abbrev <= 0 || 40 < options->abbrev)
3286                options->abbrev = 40; /* full */
3287
3288        /*
3289         * It does not make sense to show the first hit we happened
3290         * to have found.  It does not make sense not to return with
3291         * exit code in such a case either.
3292         */
3293        if (DIFF_OPT_TST(options, QUICK)) {
3294                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3295                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3296        }
3297}
3298
3299static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3300{
3301        char c, *eq;
3302        int len;
3303
3304        if (*arg != '-')
3305                return 0;
3306        c = *++arg;
3307        if (!c)
3308                return 0;
3309        if (c == arg_short) {
3310                c = *++arg;
3311                if (!c)
3312                        return 1;
3313                if (val && isdigit(c)) {
3314                        char *end;
3315                        int n = strtoul(arg, &end, 10);
3316                        if (*end)
3317                                return 0;
3318                        *val = n;
3319                        return 1;
3320                }
3321                return 0;
3322        }
3323        if (c != '-')
3324                return 0;
3325        arg++;
3326        eq = strchr(arg, '=');
3327        if (eq)
3328                len = eq - arg;
3329        else
3330                len = strlen(arg);
3331        if (!len || strncmp(arg, arg_long, len))
3332                return 0;
3333        if (eq) {
3334                int n;
3335                char *end;
3336                if (!isdigit(*++eq))
3337                        return 0;
3338                n = strtoul(eq, &end, 10);
3339                if (*end)
3340                        return 0;
3341                *val = n;
3342        }
3343        return 1;
3344}
3345
3346static int diff_scoreopt_parse(const char *opt);
3347
3348static inline int short_opt(char opt, const char **argv,
3349                            const char **optarg)
3350{
3351        const char *arg = argv[0];
3352        if (arg[0] != '-' || arg[1] != opt)
3353                return 0;
3354        if (arg[2] != '\0') {
3355                *optarg = arg + 2;
3356                return 1;
3357        }
3358        if (!argv[1])
3359                die("Option '%c' requires a value", opt);
3360        *optarg = argv[1];
3361        return 2;
3362}
3363
3364int parse_long_opt(const char *opt, const char **argv,
3365                   const char **optarg)
3366{
3367        const char *arg = argv[0];
3368        if (arg[0] != '-' || arg[1] != '-')
3369                return 0;
3370        arg += strlen("--");
3371        if (prefixcmp(arg, opt))
3372                return 0;
3373        arg += strlen(opt);
3374        if (*arg == '=') { /* sticked form: --option=value */
3375                *optarg = arg + 1;
3376                return 1;
3377        }
3378        if (*arg != '\0')
3379                return 0;
3380        /* separate form: --option value */
3381        if (!argv[1])
3382                die("Option '--%s' requires a value", opt);
3383        *optarg = argv[1];
3384        return 2;
3385}
3386
3387static int stat_opt(struct diff_options *options, const char **av)
3388{
3389        const char *arg = av[0];
3390        char *end;
3391        int width = options->stat_width;
3392        int name_width = options->stat_name_width;
3393        int graph_width = options->stat_graph_width;
3394        int count = options->stat_count;
3395        int argcount = 1;
3396
3397        arg += strlen("--stat");
3398        end = (char *)arg;
3399
3400        switch (*arg) {
3401        case '-':
3402                if (!prefixcmp(arg, "-width")) {
3403                        arg += strlen("-width");
3404                        if (*arg == '=')
3405                                width = strtoul(arg + 1, &end, 10);
3406                        else if (!*arg && !av[1])
3407                                die("Option '--stat-width' requires a value");
3408                        else if (!*arg) {
3409                                width = strtoul(av[1], &end, 10);
3410                                argcount = 2;
3411                        }
3412                } else if (!prefixcmp(arg, "-name-width")) {
3413                        arg += strlen("-name-width");
3414                        if (*arg == '=')
3415                                name_width = strtoul(arg + 1, &end, 10);
3416                        else if (!*arg && !av[1])
3417                                die("Option '--stat-name-width' requires a value");
3418                        else if (!*arg) {
3419                                name_width = strtoul(av[1], &end, 10);
3420                                argcount = 2;
3421                        }
3422                } else if (!prefixcmp(arg, "-graph-width")) {
3423                        arg += strlen("-graph-width");
3424                        if (*arg == '=')
3425                                graph_width = strtoul(arg + 1, &end, 10);
3426                        else if (!*arg && !av[1])
3427                                die("Option '--stat-graph-width' requires a value");
3428                        else if (!*arg) {
3429                                graph_width = strtoul(av[1], &end, 10);
3430                                argcount = 2;
3431                        }
3432                } else if (!prefixcmp(arg, "-count")) {
3433                        arg += strlen("-count");
3434                        if (*arg == '=')
3435                                count = strtoul(arg + 1, &end, 10);
3436                        else if (!*arg && !av[1])
3437                                die("Option '--stat-count' requires a value");
3438                        else if (!*arg) {
3439                                count = strtoul(av[1], &end, 10);
3440                                argcount = 2;
3441                        }
3442                }
3443                break;
3444        case '=':
3445                width = strtoul(arg+1, &end, 10);
3446                if (*end == ',')
3447                        name_width = strtoul(end+1, &end, 10);
3448                if (*end == ',')
3449                        count = strtoul(end+1, &end, 10);
3450        }
3451
3452        /* Important! This checks all the error cases! */
3453        if (*end)
3454                return 0;
3455        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3456        options->stat_name_width = name_width;
3457        options->stat_graph_width = graph_width;
3458        options->stat_width = width;
3459        options->stat_count = count;
3460        return argcount;
3461}
3462
3463static int parse_dirstat_opt(struct diff_options *options, const char *params)
3464{
3465        struct strbuf errmsg = STRBUF_INIT;
3466        if (parse_dirstat_params(options, params, &errmsg))
3467                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3468                    errmsg.buf);
3469        strbuf_release(&errmsg);
3470        /*
3471         * The caller knows a dirstat-related option is given from the command
3472         * line; allow it to say "return this_function();"
3473         */
3474        options->output_format |= DIFF_FORMAT_DIRSTAT;
3475        return 1;
3476}
3477
3478int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3479{
3480        const char *arg = av[0];
3481        const char *optarg;
3482        int argcount;
3483
3484        /* Output format options */
3485        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3486                options->output_format |= DIFF_FORMAT_PATCH;
3487        else if (opt_arg(arg, 'U', "unified", &options->context))
3488                options->output_format |= DIFF_FORMAT_PATCH;
3489        else if (!strcmp(arg, "--raw"))
3490                options->output_format |= DIFF_FORMAT_RAW;
3491        else if (!strcmp(arg, "--patch-with-raw"))
3492                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3493        else if (!strcmp(arg, "--numstat"))
3494                options->output_format |= DIFF_FORMAT_NUMSTAT;
3495        else if (!strcmp(arg, "--shortstat"))
3496                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3497        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3498                return parse_dirstat_opt(options, "");
3499        else if (!prefixcmp(arg, "-X"))
3500                return parse_dirstat_opt(options, arg + 2);
3501        else if (!prefixcmp(arg, "--dirstat="))
3502                return parse_dirstat_opt(options, arg + 10);
3503        else if (!strcmp(arg, "--cumulative"))
3504                return parse_dirstat_opt(options, "cumulative");
3505        else if (!strcmp(arg, "--dirstat-by-file"))
3506                return parse_dirstat_opt(options, "files");
3507        else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3508                parse_dirstat_opt(options, "files");
3509                return parse_dirstat_opt(options, arg + 18);
3510        }
3511        else if (!strcmp(arg, "--check"))
3512                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3513        else if (!strcmp(arg, "--summary"))
3514                options->output_format |= DIFF_FORMAT_SUMMARY;
3515        else if (!strcmp(arg, "--patch-with-stat"))
3516                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3517        else if (!strcmp(arg, "--name-only"))
3518                options->output_format |= DIFF_FORMAT_NAME;
3519        else if (!strcmp(arg, "--name-status"))
3520                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3521        else if (!strcmp(arg, "-s"))
3522                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3523        else if (!prefixcmp(arg, "--stat"))
3524                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3525                return stat_opt(options, av);
3526
3527        /* renames options */
3528        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3529                 !strcmp(arg, "--break-rewrites")) {
3530                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3531                        return error("invalid argument to -B: %s", arg+2);
3532        }
3533        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3534                 !strcmp(arg, "--find-renames")) {
3535                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3536                        return error("invalid argument to -M: %s", arg+2);
3537                options->detect_rename = DIFF_DETECT_RENAME;
3538        }
3539        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3540                options->irreversible_delete = 1;
3541        }
3542        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3543                 !strcmp(arg, "--find-copies")) {
3544                if (options->detect_rename == DIFF_DETECT_COPY)
3545                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3546                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3547                        return error("invalid argument to -C: %s", arg+2);
3548                options->detect_rename = DIFF_DETECT_COPY;
3549        }
3550        else if (!strcmp(arg, "--no-renames"))
3551                options->detect_rename = 0;
3552        else if (!strcmp(arg, "--rename-empty"))
3553                DIFF_OPT_SET(options, RENAME_EMPTY);
3554        else if (!strcmp(arg, "--no-rename-empty"))
3555                DIFF_OPT_CLR(options, RENAME_EMPTY);
3556        else if (!strcmp(arg, "--relative"))
3557                DIFF_OPT_SET(options, RELATIVE_NAME);
3558        else if (!prefixcmp(arg, "--relative=")) {
3559                DIFF_OPT_SET(options, RELATIVE_NAME);
3560                options->prefix = arg + 11;
3561        }
3562
3563        /* xdiff options */
3564        else if (!strcmp(arg, "--minimal"))
3565                DIFF_XDL_SET(options, NEED_MINIMAL);
3566        else if (!strcmp(arg, "--no-minimal"))
3567                DIFF_XDL_CLR(options, NEED_MINIMAL);
3568        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3569                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3570        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3571                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3572        else if (!strcmp(arg, "--ignore-space-at-eol"))
3573                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3574        else if (!strcmp(arg, "--patience"))
3575                options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3576        else if (!strcmp(arg, "--histogram"))
3577                options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3578
3579        /* flags options */
3580        else if (!strcmp(arg, "--binary")) {
3581                options->output_format |= DIFF_FORMAT_PATCH;
3582                DIFF_OPT_SET(options, BINARY);
3583        }
3584        else if (!strcmp(arg, "--full-index"))
3585                DIFF_OPT_SET(options, FULL_INDEX);
3586        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3587                DIFF_OPT_SET(options, TEXT);
3588        else if (!strcmp(arg, "-R"))
3589                DIFF_OPT_SET(options, REVERSE_DIFF);
3590        else if (!strcmp(arg, "--find-copies-harder"))
3591                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3592        else if (!strcmp(arg, "--follow"))
3593                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3594        else if (!strcmp(arg, "--color"))
3595                options->use_color = 1;
3596        else if (!prefixcmp(arg, "--color=")) {
3597                int value = git_config_colorbool(NULL, arg+8);
3598                if (value < 0)
3599                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3600                options->use_color = value;
3601        }
3602        else if (!strcmp(arg, "--no-color"))
3603                options->use_color = 0;
3604        else if (!strcmp(arg, "--color-words")) {
3605                options->use_color = 1;
3606                options->word_diff = DIFF_WORDS_COLOR;
3607        }
3608        else if (!prefixcmp(arg, "--color-words=")) {
3609                options->use_color = 1;
3610                options->word_diff = DIFF_WORDS_COLOR;
3611                options->word_regex = arg + 14;
3612        }
3613        else if (!strcmp(arg, "--word-diff")) {
3614                if (options->word_diff == DIFF_WORDS_NONE)
3615                        options->word_diff = DIFF_WORDS_PLAIN;
3616        }
3617        else if (!prefixcmp(arg, "--word-diff=")) {
3618                const char *type = arg + 12;
3619                if (!strcmp(type, "plain"))
3620                        options->word_diff = DIFF_WORDS_PLAIN;
3621                else if (!strcmp(type, "color")) {
3622                        options->use_color = 1;
3623                        options->word_diff = DIFF_WORDS_COLOR;
3624                }
3625                else if (!strcmp(type, "porcelain"))
3626                        options->word_diff = DIFF_WORDS_PORCELAIN;
3627                else if (!strcmp(type, "none"))
3628                        options->word_diff = DIFF_WORDS_NONE;
3629                else
3630                        die("bad --word-diff argument: %s", type);
3631        }
3632        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3633                if (options->word_diff == DIFF_WORDS_NONE)
3634                        options->word_diff = DIFF_WORDS_PLAIN;
3635                options->word_regex = optarg;
3636                return argcount;
3637        }
3638        else if (!strcmp(arg, "--exit-code"))
3639                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3640        else if (!strcmp(arg, "--quiet"))
3641                DIFF_OPT_SET(options, QUICK);
3642        else if (!strcmp(arg, "--ext-diff"))
3643                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3644        else if (!strcmp(arg, "--no-ext-diff"))
3645                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3646        else if (!strcmp(arg, "--textconv"))
3647                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3648        else if (!strcmp(arg, "--no-textconv"))
3649                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3650        else if (!strcmp(arg, "--ignore-submodules")) {
3651                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3652                handle_ignore_submodules_arg(options, "all");
3653        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3654                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3655                handle_ignore_submodules_arg(options, arg + 20);
3656        } else if (!strcmp(arg, "--submodule"))
3657                DIFF_OPT_SET(options, SUBMODULE_LOG);
3658        else if (!prefixcmp(arg, "--submodule=")) {
3659                if (!strcmp(arg + 12, "log"))
3660                        DIFF_OPT_SET(options, SUBMODULE_LOG);
3661        }
3662
3663        /* misc options */
3664        else if (!strcmp(arg, "-z"))
3665                options->line_termination = 0;
3666        else if ((argcount = short_opt('l', av, &optarg))) {
3667                options->rename_limit = strtoul(optarg, NULL, 10);
3668                return argcount;
3669        }
3670        else if ((argcount = short_opt('S', av, &optarg))) {
3671                options->pickaxe = optarg;
3672                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3673                return argcount;
3674        } else if ((argcount = short_opt('G', av, &optarg))) {
3675                options->pickaxe = optarg;
3676                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3677                return argcount;
3678        }
3679        else if (!strcmp(arg, "--pickaxe-all"))
3680                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3681        else if (!strcmp(arg, "--pickaxe-regex"))
3682                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3683        else if ((argcount = short_opt('O', av, &optarg))) {
3684                options->orderfile = optarg;
3685                return argcount;
3686        }
3687        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3688                options->filter = optarg;
3689                return argcount;
3690        }
3691        else if (!strcmp(arg, "--abbrev"))
3692                options->abbrev = DEFAULT_ABBREV;
3693        else if (!prefixcmp(arg, "--abbrev=")) {
3694                options->abbrev = strtoul(arg + 9, NULL, 10);
3695                if (options->abbrev < MINIMUM_ABBREV)
3696                        options->abbrev = MINIMUM_ABBREV;
3697                else if (40 < options->abbrev)
3698                        options->abbrev = 40;
3699        }
3700        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3701                options->a_prefix = optarg;
3702                return argcount;
3703        }
3704        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3705                options->b_prefix = optarg;
3706                return argcount;
3707        }
3708        else if (!strcmp(arg, "--no-prefix"))
3709                options->a_prefix = options->b_prefix = "";
3710        else if (opt_arg(arg, '\0', "inter-hunk-context",
3711                         &options->interhunkcontext))
3712                ;
3713        else if (!strcmp(arg, "-W"))
3714                DIFF_OPT_SET(options, FUNCCONTEXT);
3715        else if (!strcmp(arg, "--function-context"))
3716                DIFF_OPT_SET(options, FUNCCONTEXT);
3717        else if (!strcmp(arg, "--no-function-context"))
3718                DIFF_OPT_CLR(options, FUNCCONTEXT);
3719        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3720                options->file = fopen(optarg, "w");
3721                if (!options->file)
3722                        die_errno("Could not open '%s'", optarg);
3723                options->close_file = 1;
3724                return argcount;
3725        } else
3726                return 0;
3727        return 1;
3728}
3729
3730int parse_rename_score(const char **cp_p)
3731{
3732        unsigned long num, scale;
3733        int ch, dot;
3734        const char *cp = *cp_p;
3735
3736        num = 0;
3737        scale = 1;
3738        dot = 0;
3739        for (;;) {
3740                ch = *cp;
3741                if ( !dot && ch == '.' ) {
3742                        scale = 1;
3743                        dot = 1;
3744                } else if ( ch == '%' ) {
3745                        scale = dot ? scale*100 : 100;
3746                        cp++;   /* % is always at the end */
3747                        break;
3748                } else if ( ch >= '0' && ch <= '9' ) {
3749                        if ( scale < 100000 ) {
3750                                scale *= 10;
3751                                num = (num*10) + (ch-'0');
3752                        }
3753                } else {
3754                        break;
3755                }
3756                cp++;
3757        }
3758        *cp_p = cp;
3759
3760        /* user says num divided by scale and we say internally that
3761         * is MAX_SCORE * num / scale.
3762         */
3763        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3764}
3765
3766static int diff_scoreopt_parse(const char *opt)
3767{
3768        int opt1, opt2, cmd;
3769
3770        if (*opt++ != '-')
3771                return -1;
3772        cmd = *opt++;
3773        if (cmd == '-') {
3774                /* convert the long-form arguments into short-form versions */
3775                if (!prefixcmp(opt, "break-rewrites")) {
3776                        opt += strlen("break-rewrites");
3777                        if (*opt == 0 || *opt++ == '=')
3778                                cmd = 'B';
3779                } else if (!prefixcmp(opt, "find-copies")) {
3780                        opt += strlen("find-copies");
3781                        if (*opt == 0 || *opt++ == '=')
3782                                cmd = 'C';
3783                } else if (!prefixcmp(opt, "find-renames")) {
3784                        opt += strlen("find-renames");
3785                        if (*opt == 0 || *opt++ == '=')
3786                                cmd = 'M';
3787                }
3788        }
3789        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3790                return -1; /* that is not a -M, -C nor -B option */
3791
3792        opt1 = parse_rename_score(&opt);
3793        if (cmd != 'B')
3794                opt2 = 0;
3795        else {
3796                if (*opt == 0)
3797                        opt2 = 0;
3798                else if (*opt != '/')
3799                        return -1; /* we expect -B80/99 or -B80 */
3800                else {
3801                        opt++;
3802                        opt2 = parse_rename_score(&opt);
3803                }
3804        }
3805        if (*opt != 0)
3806                return -1;
3807        return opt1 | (opt2 << 16);
3808}
3809
3810struct diff_queue_struct diff_queued_diff;
3811
3812void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3813{
3814        if (queue->alloc <= queue->nr) {
3815                queue->alloc = alloc_nr(queue->alloc);
3816                queue->queue = xrealloc(queue->queue,
3817                                        sizeof(dp) * queue->alloc);
3818        }
3819        queue->queue[queue->nr++] = dp;
3820}
3821
3822struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3823                                 struct diff_filespec *one,
3824                                 struct diff_filespec *two)
3825{
3826        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3827        dp->one = one;
3828        dp->two = two;
3829        if (queue)
3830                diff_q(queue, dp);
3831        return dp;
3832}
3833
3834void diff_free_filepair(struct diff_filepair *p)
3835{
3836        free_filespec(p->one);
3837        free_filespec(p->two);
3838        free(p);
3839}
3840
3841/* This is different from find_unique_abbrev() in that
3842 * it stuffs the result with dots for alignment.
3843 */
3844const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3845{
3846        int abblen;
3847        const char *abbrev;
3848        if (len == 40)
3849                return sha1_to_hex(sha1);
3850
3851        abbrev = find_unique_abbrev(sha1, len);
3852        abblen = strlen(abbrev);
3853        if (abblen < 37) {
3854                static char hex[41];
3855                if (len < abblen && abblen <= len + 2)
3856                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3857                else
3858                        sprintf(hex, "%s...", abbrev);
3859                return hex;
3860        }
3861        return sha1_to_hex(sha1);
3862}
3863
3864static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3865{
3866        int line_termination = opt->line_termination;
3867        int inter_name_termination = line_termination ? '\t' : '\0';
3868        if (opt->output_prefix) {
3869                struct strbuf *msg = NULL;
3870                msg = opt->output_prefix(opt, opt->output_prefix_data);
3871                fprintf(opt->file, "%s", msg->buf);
3872        }
3873
3874        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3875                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3876                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3877                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3878        }
3879        if (p->score) {
3880                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3881                        inter_name_termination);
3882        } else {
3883                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3884        }
3885
3886        if (p->status == DIFF_STATUS_COPIED ||
3887            p->status == DIFF_STATUS_RENAMED) {
3888                const char *name_a, *name_b;
3889                name_a = p->one->path;
3890                name_b = p->two->path;
3891                strip_prefix(opt->prefix_length, &name_a, &name_b);
3892                write_name_quoted(name_a, opt->file, inter_name_termination);
3893                write_name_quoted(name_b, opt->file, line_termination);
3894        } else {
3895                const char *name_a, *name_b;
3896                name_a = p->one->mode ? p->one->path : p->two->path;
3897                name_b = NULL;
3898                strip_prefix(opt->prefix_length, &name_a, &name_b);
3899                write_name_quoted(name_a, opt->file, line_termination);
3900        }
3901}
3902
3903int diff_unmodified_pair(struct diff_filepair *p)
3904{
3905        /* This function is written stricter than necessary to support
3906         * the currently implemented transformers, but the idea is to
3907         * let transformers to produce diff_filepairs any way they want,
3908         * and filter and clean them up here before producing the output.
3909         */
3910        struct diff_filespec *one = p->one, *two = p->two;
3911
3912        if (DIFF_PAIR_UNMERGED(p))
3913                return 0; /* unmerged is interesting */
3914
3915        /* deletion, addition, mode or type change
3916         * and rename are all interesting.
3917         */
3918        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3919            DIFF_PAIR_MODE_CHANGED(p) ||
3920            strcmp(one->path, two->path))
3921                return 0;
3922
3923        /* both are valid and point at the same path.  that is, we are
3924         * dealing with a change.
3925         */
3926        if (one->sha1_valid && two->sha1_valid &&
3927            !hashcmp(one->sha1, two->sha1) &&
3928            !one->dirty_submodule && !two->dirty_submodule)
3929                return 1; /* no change */
3930        if (!one->sha1_valid && !two->sha1_valid)
3931                return 1; /* both look at the same file on the filesystem. */
3932        return 0;
3933}
3934
3935static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3936{
3937        if (diff_unmodified_pair(p))
3938                return;
3939
3940        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3941            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3942                return; /* no tree diffs in patch format */
3943
3944        run_diff(p, o);
3945}
3946
3947static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3948                            struct diffstat_t *diffstat)
3949{
3950        if (diff_unmodified_pair(p))
3951                return;
3952
3953        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3954            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3955                return; /* no useful stat for tree diffs */
3956
3957        run_diffstat(p, o, diffstat);
3958}
3959
3960static void diff_flush_checkdiff(struct diff_filepair *p,
3961                struct diff_options *o)
3962{
3963        if (diff_unmodified_pair(p))
3964                return;
3965
3966        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3967            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3968                return; /* nothing to check in tree diffs */
3969
3970        run_checkdiff(p, o);
3971}
3972
3973int diff_queue_is_empty(void)
3974{
3975        struct diff_queue_struct *q = &diff_queued_diff;
3976        int i;
3977        for (i = 0; i < q->nr; i++)
3978                if (!diff_unmodified_pair(q->queue[i]))
3979                        return 0;
3980        return 1;
3981}
3982
3983#if DIFF_DEBUG
3984void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3985{
3986        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3987                x, one ? one : "",
3988                s->path,
3989                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3990                s->mode,
3991                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3992        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3993                x, one ? one : "",
3994                s->size, s->xfrm_flags);
3995}
3996
3997void diff_debug_filepair(const struct diff_filepair *p, int i)
3998{
3999        diff_debug_filespec(p->one, i, "one");
4000        diff_debug_filespec(p->two, i, "two");
4001        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4002                p->score, p->status ? p->status : '?',
4003                p->one->rename_used, p->broken_pair);
4004}
4005
4006void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4007{
4008        int i;
4009        if (msg)
4010                fprintf(stderr, "%s\n", msg);
4011        fprintf(stderr, "q->nr = %d\n", q->nr);
4012        for (i = 0; i < q->nr; i++) {
4013                struct diff_filepair *p = q->queue[i];
4014                diff_debug_filepair(p, i);
4015        }
4016}
4017#endif
4018
4019static void diff_resolve_rename_copy(void)
4020{
4021        int i;
4022        struct diff_filepair *p;
4023        struct diff_queue_struct *q = &diff_queued_diff;
4024
4025        diff_debug_queue("resolve-rename-copy", q);
4026
4027        for (i = 0; i < q->nr; i++) {
4028                p = q->queue[i];
4029                p->status = 0; /* undecided */
4030                if (DIFF_PAIR_UNMERGED(p))
4031                        p->status = DIFF_STATUS_UNMERGED;
4032                else if (!DIFF_FILE_VALID(p->one))
4033                        p->status = DIFF_STATUS_ADDED;
4034                else if (!DIFF_FILE_VALID(p->two))
4035                        p->status = DIFF_STATUS_DELETED;
4036                else if (DIFF_PAIR_TYPE_CHANGED(p))
4037                        p->status = DIFF_STATUS_TYPE_CHANGED;
4038
4039                /* from this point on, we are dealing with a pair
4040                 * whose both sides are valid and of the same type, i.e.
4041                 * either in-place edit or rename/copy edit.
4042                 */
4043                else if (DIFF_PAIR_RENAME(p)) {
4044                        /*
4045                         * A rename might have re-connected a broken
4046                         * pair up, causing the pathnames to be the
4047                         * same again. If so, that's not a rename at
4048                         * all, just a modification..
4049                         *
4050                         * Otherwise, see if this source was used for
4051                         * multiple renames, in which case we decrement
4052                         * the count, and call it a copy.
4053                         */
4054                        if (!strcmp(p->one->path, p->two->path))
4055                                p->status = DIFF_STATUS_MODIFIED;
4056                        else if (--p->one->rename_used > 0)
4057                                p->status = DIFF_STATUS_COPIED;
4058                        else
4059                                p->status = DIFF_STATUS_RENAMED;
4060                }
4061                else if (hashcmp(p->one->sha1, p->two->sha1) ||
4062                         p->one->mode != p->two->mode ||
4063                         p->one->dirty_submodule ||
4064                         p->two->dirty_submodule ||
4065                         is_null_sha1(p->one->sha1))
4066                        p->status = DIFF_STATUS_MODIFIED;
4067                else {
4068                        /* This is a "no-change" entry and should not
4069                         * happen anymore, but prepare for broken callers.
4070                         */
4071                        error("feeding unmodified %s to diffcore",
4072                              p->one->path);
4073                        p->status = DIFF_STATUS_UNKNOWN;
4074                }
4075        }
4076        diff_debug_queue("resolve-rename-copy done", q);
4077}
4078
4079static int check_pair_status(struct diff_filepair *p)
4080{
4081        switch (p->status) {
4082        case DIFF_STATUS_UNKNOWN:
4083                return 0;
4084        case 0:
4085                die("internal error in diff-resolve-rename-copy");
4086        default:
4087                return 1;
4088        }
4089}
4090
4091static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4092{
4093        int fmt = opt->output_format;
4094
4095        if (fmt & DIFF_FORMAT_CHECKDIFF)
4096                diff_flush_checkdiff(p, opt);
4097        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4098                diff_flush_raw(p, opt);
4099        else if (fmt & DIFF_FORMAT_NAME) {
4100                const char *name_a, *name_b;
4101                name_a = p->two->path;
4102                name_b = NULL;
4103                strip_prefix(opt->prefix_length, &name_a, &name_b);
4104                write_name_quoted(name_a, opt->file, opt->line_termination);
4105        }
4106}
4107
4108static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4109{
4110        if (fs->mode)
4111                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4112        else
4113                fprintf(file, " %s ", newdelete);
4114        write_name_quoted(fs->path, file, '\n');
4115}
4116
4117
4118static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4119                const char *line_prefix)
4120{
4121        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4122                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4123                        p->two->mode, show_name ? ' ' : '\n');
4124                if (show_name) {
4125                        write_name_quoted(p->two->path, file, '\n');
4126                }
4127        }
4128}
4129
4130static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4131                        const char *line_prefix)
4132{
4133        char *names = pprint_rename(p->one->path, p->two->path);
4134
4135        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4136        free(names);
4137        show_mode_change(file, p, 0, line_prefix);
4138}
4139
4140static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4141{
4142        FILE *file = opt->file;
4143        char *line_prefix = "";
4144
4145        if (opt->output_prefix) {
4146                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4147                line_prefix = buf->buf;
4148        }
4149
4150        switch(p->status) {
4151        case DIFF_STATUS_DELETED:
4152                fputs(line_prefix, file);
4153                show_file_mode_name(file, "delete", p->one);
4154                break;
4155        case DIFF_STATUS_ADDED:
4156                fputs(line_prefix, file);
4157                show_file_mode_name(file, "create", p->two);
4158                break;
4159        case DIFF_STATUS_COPIED:
4160                fputs(line_prefix, file);
4161                show_rename_copy(file, "copy", p, line_prefix);
4162                break;
4163        case DIFF_STATUS_RENAMED:
4164                fputs(line_prefix, file);
4165                show_rename_copy(file, "rename", p, line_prefix);
4166                break;
4167        default:
4168                if (p->score) {
4169                        fprintf(file, "%s rewrite ", line_prefix);
4170                        write_name_quoted(p->two->path, file, ' ');
4171                        fprintf(file, "(%d%%)\n", similarity_index(p));
4172                }
4173                show_mode_change(file, p, !p->score, line_prefix);
4174                break;
4175        }
4176}
4177
4178struct patch_id_t {
4179        git_SHA_CTX *ctx;
4180        int patchlen;
4181};
4182
4183static int remove_space(char *line, int len)
4184{
4185        int i;
4186        char *dst = line;
4187        unsigned char c;
4188
4189        for (i = 0; i < len; i++)
4190                if (!isspace((c = line[i])))
4191                        *dst++ = c;
4192
4193        return dst - line;
4194}
4195
4196static void patch_id_consume(void *priv, char *line, unsigned long len)
4197{
4198        struct patch_id_t *data = priv;
4199        int new_len;
4200
4201        /* Ignore line numbers when computing the SHA1 of the patch */
4202        if (!prefixcmp(line, "@@ -"))
4203                return;
4204
4205        new_len = remove_space(line, len);
4206
4207        git_SHA1_Update(data->ctx, line, new_len);
4208        data->patchlen += new_len;
4209}
4210
4211/* returns 0 upon success, and writes result into sha1 */
4212static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4213{
4214        struct diff_queue_struct *q = &diff_queued_diff;
4215        int i;
4216        git_SHA_CTX ctx;
4217        struct patch_id_t data;
4218        char buffer[PATH_MAX * 4 + 20];
4219
4220        git_SHA1_Init(&ctx);
4221        memset(&data, 0, sizeof(struct patch_id_t));
4222        data.ctx = &ctx;
4223
4224        for (i = 0; i < q->nr; i++) {
4225                xpparam_t xpp;
4226                xdemitconf_t xecfg;
4227                mmfile_t mf1, mf2;
4228                struct diff_filepair *p = q->queue[i];
4229                int len1, len2;
4230
4231                memset(&xpp, 0, sizeof(xpp));
4232                memset(&xecfg, 0, sizeof(xecfg));
4233                if (p->status == 0)
4234                        return error("internal diff status error");
4235                if (p->status == DIFF_STATUS_UNKNOWN)
4236                        continue;
4237                if (diff_unmodified_pair(p))
4238                        continue;
4239                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4240                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4241                        continue;
4242                if (DIFF_PAIR_UNMERGED(p))
4243                        continue;
4244
4245                diff_fill_sha1_info(p->one);
4246                diff_fill_sha1_info(p->two);
4247                if (fill_mmfile(&mf1, p->one) < 0 ||
4248                                fill_mmfile(&mf2, p->two) < 0)
4249                        return error("unable to read files to diff");
4250
4251                len1 = remove_space(p->one->path, strlen(p->one->path));
4252                len2 = remove_space(p->two->path, strlen(p->two->path));
4253                if (p->one->mode == 0)
4254                        len1 = snprintf(buffer, sizeof(buffer),
4255                                        "diff--gita/%.*sb/%.*s"
4256                                        "newfilemode%06o"
4257                                        "---/dev/null"
4258                                        "+++b/%.*s",
4259                                        len1, p->one->path,
4260                                        len2, p->two->path,
4261                                        p->two->mode,
4262                                        len2, p->two->path);
4263                else if (p->two->mode == 0)
4264                        len1 = snprintf(buffer, sizeof(buffer),
4265                                        "diff--gita/%.*sb/%.*s"
4266                                        "deletedfilemode%06o"
4267                                        "---a/%.*s"
4268                                        "+++/dev/null",
4269                                        len1, p->one->path,
4270                                        len2, p->two->path,
4271                                        p->one->mode,
4272                                        len1, p->one->path);
4273                else
4274                        len1 = snprintf(buffer, sizeof(buffer),
4275                                        "diff--gita/%.*sb/%.*s"
4276                                        "---a/%.*s"
4277                                        "+++b/%.*s",
4278                                        len1, p->one->path,
4279                                        len2, p->two->path,
4280                                        len1, p->one->path,
4281                                        len2, p->two->path);
4282                git_SHA1_Update(&ctx, buffer, len1);
4283
4284                if (diff_filespec_is_binary(p->one) ||
4285                    diff_filespec_is_binary(p->two)) {
4286                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4287                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4288                        continue;
4289                }
4290
4291                xpp.flags = 0;
4292                xecfg.ctxlen = 3;
4293                xecfg.flags = 0;
4294                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4295                              &xpp, &xecfg);
4296        }
4297
4298        git_SHA1_Final(sha1, &ctx);
4299        return 0;
4300}
4301
4302int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4303{
4304        struct diff_queue_struct *q = &diff_queued_diff;
4305        int i;
4306        int result = diff_get_patch_id(options, sha1);
4307
4308        for (i = 0; i < q->nr; i++)
4309                diff_free_filepair(q->queue[i]);
4310
4311        free(q->queue);
4312        DIFF_QUEUE_CLEAR(q);
4313
4314        return result;
4315}
4316
4317static int is_summary_empty(const struct diff_queue_struct *q)
4318{
4319        int i;
4320
4321        for (i = 0; i < q->nr; i++) {
4322                const struct diff_filepair *p = q->queue[i];
4323
4324                switch (p->status) {
4325                case DIFF_STATUS_DELETED:
4326                case DIFF_STATUS_ADDED:
4327                case DIFF_STATUS_COPIED:
4328                case DIFF_STATUS_RENAMED:
4329                        return 0;
4330                default:
4331                        if (p->score)
4332                                return 0;
4333                        if (p->one->mode && p->two->mode &&
4334                            p->one->mode != p->two->mode)
4335                                return 0;
4336                        break;
4337                }
4338        }
4339        return 1;
4340}
4341
4342static const char rename_limit_warning[] =
4343"inexact rename detection was skipped due to too many files.";
4344
4345static const char degrade_cc_to_c_warning[] =
4346"only found copies from modified paths due to too many files.";
4347
4348static const char rename_limit_advice[] =
4349"you may want to set your %s variable to at least "
4350"%d and retry the command.";
4351
4352void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4353{
4354        if (degraded_cc)
4355                warning(degrade_cc_to_c_warning);
4356        else if (needed)
4357                warning(rename_limit_warning);
4358        else
4359                return;
4360        if (0 < needed && needed < 32767)
4361                warning(rename_limit_advice, varname, needed);
4362}
4363
4364void diff_flush(struct diff_options *options)
4365{
4366        struct diff_queue_struct *q = &diff_queued_diff;
4367        int i, output_format = options->output_format;
4368        int separator = 0;
4369        int dirstat_by_line = 0;
4370
4371        /*
4372         * Order: raw, stat, summary, patch
4373         * or:    name/name-status/checkdiff (other bits clear)
4374         */
4375        if (!q->nr)
4376                goto free_queue;
4377
4378        if (output_format & (DIFF_FORMAT_RAW |
4379                             DIFF_FORMAT_NAME |
4380                             DIFF_FORMAT_NAME_STATUS |
4381                             DIFF_FORMAT_CHECKDIFF)) {
4382                for (i = 0; i < q->nr; i++) {
4383                        struct diff_filepair *p = q->queue[i];
4384                        if (check_pair_status(p))
4385                                flush_one_pair(p, options);
4386                }
4387                separator++;
4388        }
4389
4390        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4391                dirstat_by_line = 1;
4392
4393        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4394            dirstat_by_line) {
4395                struct diffstat_t diffstat;
4396
4397                memset(&diffstat, 0, sizeof(struct diffstat_t));
4398                for (i = 0; i < q->nr; i++) {
4399                        struct diff_filepair *p = q->queue[i];
4400                        if (check_pair_status(p))
4401                                diff_flush_stat(p, options, &diffstat);
4402                }
4403                if (output_format & DIFF_FORMAT_NUMSTAT)
4404                        show_numstat(&diffstat, options);
4405                if (output_format & DIFF_FORMAT_DIFFSTAT)
4406                        show_stats(&diffstat, options);
4407                if (output_format & DIFF_FORMAT_SHORTSTAT)
4408                        show_shortstats(&diffstat, options);
4409                if (output_format & DIFF_FORMAT_DIRSTAT)
4410                        show_dirstat_by_line(&diffstat, options);
4411                free_diffstat_info(&diffstat);
4412                separator++;
4413        }
4414        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4415                show_dirstat(options);
4416
4417        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4418                for (i = 0; i < q->nr; i++) {
4419                        diff_summary(options, q->queue[i]);
4420                }
4421                separator++;
4422        }
4423
4424        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4425            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4426            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4427                /*
4428                 * run diff_flush_patch for the exit status. setting
4429                 * options->file to /dev/null should be safe, becaue we
4430                 * aren't supposed to produce any output anyway.
4431                 */
4432                if (options->close_file)
4433                        fclose(options->file);
4434                options->file = fopen("/dev/null", "w");
4435                if (!options->file)
4436                        die_errno("Could not open /dev/null");
4437                options->close_file = 1;
4438                for (i = 0; i < q->nr; i++) {
4439                        struct diff_filepair *p = q->queue[i];
4440                        if (check_pair_status(p))
4441                                diff_flush_patch(p, options);
4442                        if (options->found_changes)
4443                                break;
4444                }
4445        }
4446
4447        if (output_format & DIFF_FORMAT_PATCH) {
4448                if (separator) {
4449                        if (options->output_prefix) {
4450                                struct strbuf *msg = NULL;
4451                                msg = options->output_prefix(options,
4452                                        options->output_prefix_data);
4453                                fwrite(msg->buf, msg->len, 1, stdout);
4454                        }
4455                        putc(options->line_termination, options->file);
4456                        if (options->stat_sep) {
4457                                /* attach patch instead of inline */
4458                                fputs(options->stat_sep, options->file);
4459                        }
4460                }
4461
4462                for (i = 0; i < q->nr; i++) {
4463                        struct diff_filepair *p = q->queue[i];
4464                        if (check_pair_status(p))
4465                                diff_flush_patch(p, options);
4466                }
4467        }
4468
4469        if (output_format & DIFF_FORMAT_CALLBACK)
4470                options->format_callback(q, options, options->format_callback_data);
4471
4472        for (i = 0; i < q->nr; i++)
4473                diff_free_filepair(q->queue[i]);
4474free_queue:
4475        free(q->queue);
4476        DIFF_QUEUE_CLEAR(q);
4477        if (options->close_file)
4478                fclose(options->file);
4479
4480        /*
4481         * Report the content-level differences with HAS_CHANGES;
4482         * diff_addremove/diff_change does not set the bit when
4483         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4484         */
4485        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4486                if (options->found_changes)
4487                        DIFF_OPT_SET(options, HAS_CHANGES);
4488                else
4489                        DIFF_OPT_CLR(options, HAS_CHANGES);
4490        }
4491}
4492
4493static void diffcore_apply_filter(const char *filter)
4494{
4495        int i;
4496        struct diff_queue_struct *q = &diff_queued_diff;
4497        struct diff_queue_struct outq;
4498        DIFF_QUEUE_CLEAR(&outq);
4499
4500        if (!filter)
4501                return;
4502
4503        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4504                int found;
4505                for (i = found = 0; !found && i < q->nr; i++) {
4506                        struct diff_filepair *p = q->queue[i];
4507                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4508                             ((p->score &&
4509                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4510                              (!p->score &&
4511                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4512                            ((p->status != DIFF_STATUS_MODIFIED) &&
4513                             strchr(filter, p->status)))
4514                                found++;
4515                }
4516                if (found)
4517                        return;
4518
4519                /* otherwise we will clear the whole queue
4520                 * by copying the empty outq at the end of this
4521                 * function, but first clear the current entries
4522                 * in the queue.
4523                 */
4524                for (i = 0; i < q->nr; i++)
4525                        diff_free_filepair(q->queue[i]);
4526        }
4527        else {
4528                /* Only the matching ones */
4529                for (i = 0; i < q->nr; i++) {
4530                        struct diff_filepair *p = q->queue[i];
4531
4532                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4533                             ((p->score &&
4534                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4535                              (!p->score &&
4536                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4537                            ((p->status != DIFF_STATUS_MODIFIED) &&
4538                             strchr(filter, p->status)))
4539                                diff_q(&outq, p);
4540                        else
4541                                diff_free_filepair(p);
4542                }
4543        }
4544        free(q->queue);
4545        *q = outq;
4546}
4547
4548/* Check whether two filespecs with the same mode and size are identical */
4549static int diff_filespec_is_identical(struct diff_filespec *one,
4550                                      struct diff_filespec *two)
4551{
4552        if (S_ISGITLINK(one->mode))
4553                return 0;
4554        if (diff_populate_filespec(one, 0))
4555                return 0;
4556        if (diff_populate_filespec(two, 0))
4557                return 0;
4558        return !memcmp(one->data, two->data, one->size);
4559}
4560
4561static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4562{
4563        int i;
4564        struct diff_queue_struct *q = &diff_queued_diff;
4565        struct diff_queue_struct outq;
4566        DIFF_QUEUE_CLEAR(&outq);
4567
4568        for (i = 0; i < q->nr; i++) {
4569                struct diff_filepair *p = q->queue[i];
4570
4571                /*
4572                 * 1. Entries that come from stat info dirtiness
4573                 *    always have both sides (iow, not create/delete),
4574                 *    one side of the object name is unknown, with
4575                 *    the same mode and size.  Keep the ones that
4576                 *    do not match these criteria.  They have real
4577                 *    differences.
4578                 *
4579                 * 2. At this point, the file is known to be modified,
4580                 *    with the same mode and size, and the object
4581                 *    name of one side is unknown.  Need to inspect
4582                 *    the identical contents.
4583                 */
4584                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4585                    !DIFF_FILE_VALID(p->two) ||
4586                    (p->one->sha1_valid && p->two->sha1_valid) ||
4587                    (p->one->mode != p->two->mode) ||
4588                    diff_populate_filespec(p->one, 1) ||
4589                    diff_populate_filespec(p->two, 1) ||
4590                    (p->one->size != p->two->size) ||
4591                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4592                        diff_q(&outq, p);
4593                else {
4594                        /*
4595                         * The caller can subtract 1 from skip_stat_unmatch
4596                         * to determine how many paths were dirty only
4597                         * due to stat info mismatch.
4598                         */
4599                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4600                                diffopt->skip_stat_unmatch++;
4601                        diff_free_filepair(p);
4602                }
4603        }
4604        free(q->queue);
4605        *q = outq;
4606}
4607
4608static int diffnamecmp(const void *a_, const void *b_)
4609{
4610        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4611        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4612        const char *name_a, *name_b;
4613
4614        name_a = a->one ? a->one->path : a->two->path;
4615        name_b = b->one ? b->one->path : b->two->path;
4616        return strcmp(name_a, name_b);
4617}
4618
4619void diffcore_fix_diff_index(struct diff_options *options)
4620{
4621        struct diff_queue_struct *q = &diff_queued_diff;
4622        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4623}
4624
4625void diffcore_std(struct diff_options *options)
4626{
4627        if (options->skip_stat_unmatch)
4628                diffcore_skip_stat_unmatch(options);
4629        if (!options->found_follow) {
4630                /* See try_to_follow_renames() in tree-diff.c */
4631                if (options->break_opt != -1)
4632                        diffcore_break(options->break_opt);
4633                if (options->detect_rename)
4634                        diffcore_rename(options);
4635                if (options->break_opt != -1)
4636                        diffcore_merge_broken();
4637        }
4638        if (options->pickaxe)
4639                diffcore_pickaxe(options);
4640        if (options->orderfile)
4641                diffcore_order(options->orderfile);
4642        if (!options->found_follow)
4643                /* See try_to_follow_renames() in tree-diff.c */
4644                diff_resolve_rename_copy();
4645        diffcore_apply_filter(options->filter);
4646
4647        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4648                DIFF_OPT_SET(options, HAS_CHANGES);
4649        else
4650                DIFF_OPT_CLR(options, HAS_CHANGES);
4651
4652        options->found_follow = 0;
4653}
4654
4655int diff_result_code(struct diff_options *opt, int status)
4656{
4657        int result = 0;
4658
4659        diff_warn_rename_limit("diff.renamelimit",
4660                               opt->needed_rename_limit,
4661                               opt->degraded_cc_to_c);
4662        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4663            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4664                return status;
4665        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4666            DIFF_OPT_TST(opt, HAS_CHANGES))
4667                result |= 01;
4668        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4669            DIFF_OPT_TST(opt, CHECK_FAILED))
4670                result |= 02;
4671        return result;
4672}
4673
4674int diff_can_quit_early(struct diff_options *opt)
4675{
4676        return (DIFF_OPT_TST(opt, QUICK) &&
4677                !opt->filter &&
4678                DIFF_OPT_TST(opt, HAS_CHANGES));
4679}
4680
4681/*
4682 * Shall changes to this submodule be ignored?
4683 *
4684 * Submodule changes can be configured to be ignored separately for each path,
4685 * but that configuration can be overridden from the command line.
4686 */
4687static int is_submodule_ignored(const char *path, struct diff_options *options)
4688{
4689        int ignored = 0;
4690        unsigned orig_flags = options->flags;
4691        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4692                set_diffopt_flags_from_submodule_config(options, path);
4693        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4694                ignored = 1;
4695        options->flags = orig_flags;
4696        return ignored;
4697}
4698
4699void diff_addremove(struct diff_options *options,
4700                    int addremove, unsigned mode,
4701                    const unsigned char *sha1,
4702                    int sha1_valid,
4703                    const char *concatpath, unsigned dirty_submodule)
4704{
4705        struct diff_filespec *one, *two;
4706
4707        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4708                return;
4709
4710        /* This may look odd, but it is a preparation for
4711         * feeding "there are unchanged files which should
4712         * not produce diffs, but when you are doing copy
4713         * detection you would need them, so here they are"
4714         * entries to the diff-core.  They will be prefixed
4715         * with something like '=' or '*' (I haven't decided
4716         * which but should not make any difference).
4717         * Feeding the same new and old to diff_change()
4718         * also has the same effect.
4719         * Before the final output happens, they are pruned after
4720         * merged into rename/copy pairs as appropriate.
4721         */
4722        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4723                addremove = (addremove == '+' ? '-' :
4724                             addremove == '-' ? '+' : addremove);
4725
4726        if (options->prefix &&
4727            strncmp(concatpath, options->prefix, options->prefix_length))
4728                return;
4729
4730        one = alloc_filespec(concatpath);
4731        two = alloc_filespec(concatpath);
4732
4733        if (addremove != '+')
4734                fill_filespec(one, sha1, sha1_valid, mode);
4735        if (addremove != '-') {
4736                fill_filespec(two, sha1, sha1_valid, mode);
4737                two->dirty_submodule = dirty_submodule;
4738        }
4739
4740        diff_queue(&diff_queued_diff, one, two);
4741        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4742                DIFF_OPT_SET(options, HAS_CHANGES);
4743}
4744
4745void diff_change(struct diff_options *options,
4746                 unsigned old_mode, unsigned new_mode,
4747                 const unsigned char *old_sha1,
4748                 const unsigned char *new_sha1,
4749                 int old_sha1_valid, int new_sha1_valid,
4750                 const char *concatpath,
4751                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4752{
4753        struct diff_filespec *one, *two;
4754
4755        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4756            is_submodule_ignored(concatpath, options))
4757                return;
4758
4759        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4760                unsigned tmp;
4761                const unsigned char *tmp_c;
4762                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4763                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4764                tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4765                        new_sha1_valid = tmp;
4766                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4767                        new_dirty_submodule = tmp;
4768        }
4769
4770        if (options->prefix &&
4771            strncmp(concatpath, options->prefix, options->prefix_length))
4772                return;
4773
4774        one = alloc_filespec(concatpath);
4775        two = alloc_filespec(concatpath);
4776        fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4777        fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4778        one->dirty_submodule = old_dirty_submodule;
4779        two->dirty_submodule = new_dirty_submodule;
4780
4781        diff_queue(&diff_queued_diff, one, two);
4782        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4783                DIFF_OPT_SET(options, HAS_CHANGES);
4784}
4785
4786struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4787{
4788        struct diff_filepair *pair;
4789        struct diff_filespec *one, *two;
4790
4791        if (options->prefix &&
4792            strncmp(path, options->prefix, options->prefix_length))
4793                return NULL;
4794
4795        one = alloc_filespec(path);
4796        two = alloc_filespec(path);
4797        pair = diff_queue(&diff_queued_diff, one, two);
4798        pair->is_unmerged = 1;
4799        return pair;
4800}
4801
4802static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4803                size_t *outsize)
4804{
4805        struct diff_tempfile *temp;
4806        const char *argv[3];
4807        const char **arg = argv;
4808        struct child_process child;
4809        struct strbuf buf = STRBUF_INIT;
4810        int err = 0;
4811
4812        temp = prepare_temp_file(spec->path, spec);
4813        *arg++ = pgm;
4814        *arg++ = temp->name;
4815        *arg = NULL;
4816
4817        memset(&child, 0, sizeof(child));
4818        child.use_shell = 1;
4819        child.argv = argv;
4820        child.out = -1;
4821        if (start_command(&child)) {
4822                remove_tempfile();
4823                return NULL;
4824        }
4825
4826        if (strbuf_read(&buf, child.out, 0) < 0)
4827                err = error("error reading from textconv command '%s'", pgm);
4828        close(child.out);
4829
4830        if (finish_command(&child) || err) {
4831                strbuf_release(&buf);
4832                remove_tempfile();
4833                return NULL;
4834        }
4835        remove_tempfile();
4836
4837        return strbuf_detach(&buf, outsize);
4838}
4839
4840size_t fill_textconv(struct userdiff_driver *driver,
4841                     struct diff_filespec *df,
4842                     char **outbuf)
4843{
4844        size_t size;
4845
4846        if (!driver || !driver->textconv) {
4847                if (!DIFF_FILE_VALID(df)) {
4848                        *outbuf = "";
4849                        return 0;
4850                }
4851                if (diff_populate_filespec(df, 0))
4852                        die("unable to read files to diff");
4853                *outbuf = df->data;
4854                return df->size;
4855        }
4856
4857        if (driver->textconv_cache && df->sha1_valid) {
4858                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4859                                          &size);
4860                if (*outbuf)
4861                        return size;
4862        }
4863
4864        *outbuf = run_textconv(driver->textconv, df, &size);
4865        if (!*outbuf)
4866                die("unable to read files to diff");
4867
4868        if (driver->textconv_cache && df->sha1_valid) {
4869                /* ignore errors, as we might be in a readonly repository */
4870                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4871                                size);
4872                /*
4873                 * we could save up changes and flush them all at the end,
4874                 * but we would need an extra call after all diffing is done.
4875                 * Since generating a cache entry is the slow path anyway,
4876                 * this extra overhead probably isn't a big deal.
4877                 */
4878                notes_cache_write(driver->textconv_cache);
4879        }
4880
4881        return size;
4882}
4883
4884void setup_diff_pager(struct diff_options *opt)
4885{
4886        /*
4887         * If the user asked for our exit code, then either they want --quiet
4888         * or --exit-code. We should definitely not bother with a pager in the
4889         * former case, as we will generate no output. Since we still properly
4890         * report our exit code even when a pager is run, we _could_ run a
4891         * pager with --exit-code. But since we have not done so historically,
4892         * and because it is easy to find people oneline advising "git diff
4893         * --exit-code" in hooks and other scripts, we do not do so.
4894         */
4895        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4896            check_pager_config("diff") != 0)
4897                setup_pager();
4898}