diff.con commit Merge branch 'nd/maint-compat-fnmatch-fix' (5ab539b)
   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                unsigned is_interesting:1;
1313                uintmax_t added, deleted;
1314        } **files;
1315};
1316
1317static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1318                                          const char *name_a,
1319                                          const char *name_b)
1320{
1321        struct diffstat_file *x;
1322        x = xcalloc(sizeof (*x), 1);
1323        if (diffstat->nr == diffstat->alloc) {
1324                diffstat->alloc = alloc_nr(diffstat->alloc);
1325                diffstat->files = xrealloc(diffstat->files,
1326                                diffstat->alloc * sizeof(x));
1327        }
1328        diffstat->files[diffstat->nr++] = x;
1329        if (name_b) {
1330                x->from_name = xstrdup(name_a);
1331                x->name = xstrdup(name_b);
1332                x->is_renamed = 1;
1333        }
1334        else {
1335                x->from_name = NULL;
1336                x->name = xstrdup(name_a);
1337        }
1338        return x;
1339}
1340
1341static void diffstat_consume(void *priv, char *line, unsigned long len)
1342{
1343        struct diffstat_t *diffstat = priv;
1344        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1345
1346        if (line[0] == '+')
1347                x->added++;
1348        else if (line[0] == '-')
1349                x->deleted++;
1350}
1351
1352const char mime_boundary_leader[] = "------------";
1353
1354static int scale_linear(int it, int width, int max_change)
1355{
1356        if (!it)
1357                return 0;
1358        /*
1359         * make sure that at least one '-' or '+' is printed if
1360         * there is any change to this path. The easiest way is to
1361         * scale linearly as if the alloted width is one column shorter
1362         * than it is, and then add 1 to the result.
1363         */
1364        return 1 + (it * (width - 1) / max_change);
1365}
1366
1367static void show_name(FILE *file,
1368                      const char *prefix, const char *name, int len)
1369{
1370        fprintf(file, " %s%-*s |", prefix, len, name);
1371}
1372
1373static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1374{
1375        if (cnt <= 0)
1376                return;
1377        fprintf(file, "%s", set);
1378        while (cnt--)
1379                putc(ch, file);
1380        fprintf(file, "%s", reset);
1381}
1382
1383static void fill_print_name(struct diffstat_file *file)
1384{
1385        char *pname;
1386
1387        if (file->print_name)
1388                return;
1389
1390        if (!file->is_renamed) {
1391                struct strbuf buf = STRBUF_INIT;
1392                if (quote_c_style(file->name, &buf, NULL, 0)) {
1393                        pname = strbuf_detach(&buf, NULL);
1394                } else {
1395                        pname = file->name;
1396                        strbuf_release(&buf);
1397                }
1398        } else {
1399                pname = pprint_rename(file->from_name, file->name);
1400        }
1401        file->print_name = pname;
1402}
1403
1404int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1405{
1406        struct strbuf sb = STRBUF_INIT;
1407        int ret;
1408
1409        if (!files) {
1410                assert(insertions == 0 && deletions == 0);
1411                return fprintf(fp, "%s\n", " 0 files changed");
1412        }
1413
1414        strbuf_addf(&sb,
1415                    (files == 1) ? " %d file changed" : " %d files changed",
1416                    files);
1417
1418        /*
1419         * For binary diff, the caller may want to print "x files
1420         * changed" with insertions == 0 && deletions == 0.
1421         *
1422         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1423         * is probably less confusing (i.e skip over "2 files changed
1424         * but nothing about added/removed lines? Is this a bug in Git?").
1425         */
1426        if (insertions || deletions == 0) {
1427                /*
1428                 * TRANSLATORS: "+" in (+) is a line addition marker;
1429                 * do not translate it.
1430                 */
1431                strbuf_addf(&sb,
1432                            (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1433                            insertions);
1434        }
1435
1436        if (deletions || insertions == 0) {
1437                /*
1438                 * TRANSLATORS: "-" in (-) is a line removal marker;
1439                 * do not translate it.
1440                 */
1441                strbuf_addf(&sb,
1442                            (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1443                            deletions);
1444        }
1445        strbuf_addch(&sb, '\n');
1446        ret = fputs(sb.buf, fp);
1447        strbuf_release(&sb);
1448        return ret;
1449}
1450
1451static void show_stats(struct diffstat_t *data, struct diff_options *options)
1452{
1453        int i, len, add, del, adds = 0, dels = 0;
1454        uintmax_t max_change = 0, max_len = 0;
1455        int total_files = data->nr, count;
1456        int width, name_width, graph_width, number_width = 0, bin_width = 0;
1457        const char *reset, *add_c, *del_c;
1458        const char *line_prefix = "";
1459        int extra_shown = 0;
1460        struct strbuf *msg = NULL;
1461
1462        if (data->nr == 0)
1463                return;
1464
1465        if (options->output_prefix) {
1466                msg = options->output_prefix(options, options->output_prefix_data);
1467                line_prefix = msg->buf;
1468        }
1469
1470        count = options->stat_count ? options->stat_count : data->nr;
1471
1472        reset = diff_get_color_opt(options, DIFF_RESET);
1473        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1474        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1475
1476        /*
1477         * Find the longest filename and max number of changes
1478         */
1479        for (i = 0; (i < count) && (i < data->nr); i++) {
1480                struct diffstat_file *file = data->files[i];
1481                uintmax_t change = file->added + file->deleted;
1482                if (!data->files[i]->is_interesting &&
1483                         (change == 0)) {
1484                        count++; /* not shown == room for one more */
1485                        continue;
1486                }
1487                fill_print_name(file);
1488                len = strlen(file->print_name);
1489                if (max_len < len)
1490                        max_len = len;
1491
1492                if (file->is_unmerged) {
1493                        /* "Unmerged" is 8 characters */
1494                        bin_width = bin_width < 8 ? 8 : bin_width;
1495                        continue;
1496                }
1497                if (file->is_binary) {
1498                        /* "Bin XXX -> YYY bytes" */
1499                        int w = 14 + decimal_width(file->added)
1500                                + decimal_width(file->deleted);
1501                        bin_width = bin_width < w ? w : bin_width;
1502                        /* Display change counts aligned with "Bin" */
1503                        number_width = 3;
1504                        continue;
1505                }
1506
1507                if (max_change < change)
1508                        max_change = change;
1509        }
1510        count = i; /* min(count, data->nr) */
1511
1512        /*
1513         * We have width = stat_width or term_columns() columns total.
1514         * We want a maximum of min(max_len, stat_name_width) for the name part.
1515         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1516         * We also need 1 for " " and 4 + decimal_width(max_change)
1517         * for " | NNNN " and one the empty column at the end, altogether
1518         * 6 + decimal_width(max_change).
1519         *
1520         * If there's not enough space, we will use the smaller of
1521         * stat_name_width (if set) and 5/8*width for the filename,
1522         * and the rest for constant elements + graph part, but no more
1523         * than stat_graph_width for the graph part.
1524         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1525         * for the standard terminal size).
1526         *
1527         * In other words: stat_width limits the maximum width, and
1528         * stat_name_width fixes the maximum width of the filename,
1529         * and is also used to divide available columns if there
1530         * aren't enough.
1531         *
1532         * Binary files are displayed with "Bin XXX -> YYY bytes"
1533         * instead of the change count and graph. This part is treated
1534         * similarly to the graph part, except that it is not
1535         * "scaled". If total width is too small to accomodate the
1536         * guaranteed minimum width of the filename part and the
1537         * separators and this message, this message will "overflow"
1538         * making the line longer than the maximum width.
1539         */
1540
1541        if (options->stat_width == -1)
1542                width = term_columns() - options->output_prefix_length;
1543        else
1544                width = options->stat_width ? options->stat_width : 80;
1545        number_width = decimal_width(max_change) > number_width ?
1546                decimal_width(max_change) : number_width;
1547
1548        if (options->stat_graph_width == -1)
1549                options->stat_graph_width = diff_stat_graph_width;
1550
1551        /*
1552         * Guarantee 3/8*16==6 for the graph part
1553         * and 5/8*16==10 for the filename part
1554         */
1555        if (width < 16 + 6 + number_width)
1556                width = 16 + 6 + number_width;
1557
1558        /*
1559         * First assign sizes that are wanted, ignoring available width.
1560         * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1561         * starting from "XXX" should fit in graph_width.
1562         */
1563        graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1564        if (options->stat_graph_width &&
1565            options->stat_graph_width < graph_width)
1566                graph_width = options->stat_graph_width;
1567
1568        name_width = (options->stat_name_width > 0 &&
1569                      options->stat_name_width < max_len) ?
1570                options->stat_name_width : max_len;
1571
1572        /*
1573         * Adjust adjustable widths not to exceed maximum width
1574         */
1575        if (name_width + number_width + 6 + graph_width > width) {
1576                if (graph_width > width * 3/8 - number_width - 6) {
1577                        graph_width = width * 3/8 - number_width - 6;
1578                        if (graph_width < 6)
1579                                graph_width = 6;
1580                }
1581
1582                if (options->stat_graph_width &&
1583                    graph_width > options->stat_graph_width)
1584                        graph_width = options->stat_graph_width;
1585                if (name_width > width - number_width - 6 - graph_width)
1586                        name_width = width - number_width - 6 - graph_width;
1587                else
1588                        graph_width = width - number_width - 6 - name_width;
1589        }
1590
1591        /*
1592         * From here name_width is the width of the name area,
1593         * and graph_width is the width of the graph area.
1594         * max_change is used to scale graph properly.
1595         */
1596        for (i = 0; i < count; i++) {
1597                const char *prefix = "";
1598                char *name = data->files[i]->print_name;
1599                uintmax_t added = data->files[i]->added;
1600                uintmax_t deleted = data->files[i]->deleted;
1601                int name_len;
1602
1603                if (!data->files[i]->is_interesting &&
1604                         (added + deleted == 0)) {
1605                        total_files--;
1606                        continue;
1607                }
1608                /*
1609                 * "scale" the filename
1610                 */
1611                len = name_width;
1612                name_len = strlen(name);
1613                if (name_width < name_len) {
1614                        char *slash;
1615                        prefix = "...";
1616                        len -= 3;
1617                        name += name_len - len;
1618                        slash = strchr(name, '/');
1619                        if (slash)
1620                                name = slash;
1621                }
1622
1623                if (data->files[i]->is_binary) {
1624                        fprintf(options->file, "%s", line_prefix);
1625                        show_name(options->file, prefix, name, len);
1626                        fprintf(options->file, " %*s", number_width, "Bin");
1627                        if (!added && !deleted) {
1628                                putc('\n', options->file);
1629                                continue;
1630                        }
1631                        fprintf(options->file, " %s%"PRIuMAX"%s",
1632                                del_c, deleted, reset);
1633                        fprintf(options->file, " -> ");
1634                        fprintf(options->file, "%s%"PRIuMAX"%s",
1635                                add_c, added, reset);
1636                        fprintf(options->file, " bytes");
1637                        fprintf(options->file, "\n");
1638                        continue;
1639                }
1640                else if (data->files[i]->is_unmerged) {
1641                        fprintf(options->file, "%s", line_prefix);
1642                        show_name(options->file, prefix, name, len);
1643                        fprintf(options->file, " Unmerged\n");
1644                        continue;
1645                }
1646
1647                /*
1648                 * scale the add/delete
1649                 */
1650                add = added;
1651                del = deleted;
1652                adds += add;
1653                dels += del;
1654
1655                if (graph_width <= max_change) {
1656                        int total = add + del;
1657
1658                        total = scale_linear(add + del, graph_width, max_change);
1659                        if (total < 2 && add && del)
1660                                /* width >= 2 due to the sanity check */
1661                                total = 2;
1662                        if (add < del) {
1663                                add = scale_linear(add, graph_width, max_change);
1664                                del = total - add;
1665                        } else {
1666                                del = scale_linear(del, graph_width, max_change);
1667                                add = total - del;
1668                        }
1669                }
1670                fprintf(options->file, "%s", line_prefix);
1671                show_name(options->file, prefix, name, len);
1672                fprintf(options->file, " %*"PRIuMAX"%s",
1673                        number_width, added + deleted,
1674                        added + deleted ? " " : "");
1675                show_graph(options->file, '+', add, add_c, reset);
1676                show_graph(options->file, '-', del, del_c, reset);
1677                fprintf(options->file, "\n");
1678        }
1679        for (i = count; i < data->nr; i++) {
1680                uintmax_t added = data->files[i]->added;
1681                uintmax_t deleted = data->files[i]->deleted;
1682                if (!data->files[i]->is_interesting &&
1683                         (added + deleted == 0)) {
1684                        total_files--;
1685                        continue;
1686                }
1687                adds += added;
1688                dels += deleted;
1689                if (!extra_shown)
1690                        fprintf(options->file, "%s ...\n", line_prefix);
1691                extra_shown = 1;
1692        }
1693        fprintf(options->file, "%s", line_prefix);
1694        print_stat_summary(options->file, total_files, adds, dels);
1695}
1696
1697static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1698{
1699        int i, adds = 0, dels = 0, total_files = data->nr;
1700
1701        if (data->nr == 0)
1702                return;
1703
1704        for (i = 0; i < data->nr; i++) {
1705                int added = data->files[i]->added;
1706                int deleted= data->files[i]->deleted;
1707
1708                if (data->files[i]->is_unmerged)
1709                        continue;
1710                if (!data->files[i]->is_interesting && (added + deleted == 0)) {
1711                        total_files--;
1712                } else if (!data->files[i]->is_binary) { /* don't count bytes */
1713                        adds += added;
1714                        dels += deleted;
1715                }
1716        }
1717        if (options->output_prefix) {
1718                struct strbuf *msg = NULL;
1719                msg = options->output_prefix(options,
1720                                options->output_prefix_data);
1721                fprintf(options->file, "%s", msg->buf);
1722        }
1723        print_stat_summary(options->file, total_files, adds, dels);
1724}
1725
1726static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1727{
1728        int i;
1729
1730        if (data->nr == 0)
1731                return;
1732
1733        for (i = 0; i < data->nr; i++) {
1734                struct diffstat_file *file = data->files[i];
1735
1736                if (options->output_prefix) {
1737                        struct strbuf *msg = NULL;
1738                        msg = options->output_prefix(options,
1739                                        options->output_prefix_data);
1740                        fprintf(options->file, "%s", msg->buf);
1741                }
1742
1743                if (file->is_binary)
1744                        fprintf(options->file, "-\t-\t");
1745                else
1746                        fprintf(options->file,
1747                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1748                                file->added, file->deleted);
1749                if (options->line_termination) {
1750                        fill_print_name(file);
1751                        if (!file->is_renamed)
1752                                write_name_quoted(file->name, options->file,
1753                                                  options->line_termination);
1754                        else {
1755                                fputs(file->print_name, options->file);
1756                                putc(options->line_termination, options->file);
1757                        }
1758                } else {
1759                        if (file->is_renamed) {
1760                                putc('\0', options->file);
1761                                write_name_quoted(file->from_name, options->file, '\0');
1762                        }
1763                        write_name_quoted(file->name, options->file, '\0');
1764                }
1765        }
1766}
1767
1768struct dirstat_file {
1769        const char *name;
1770        unsigned long changed;
1771};
1772
1773struct dirstat_dir {
1774        struct dirstat_file *files;
1775        int alloc, nr, permille, cumulative;
1776};
1777
1778static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1779                unsigned long changed, const char *base, int baselen)
1780{
1781        unsigned long this_dir = 0;
1782        unsigned int sources = 0;
1783        const char *line_prefix = "";
1784        struct strbuf *msg = NULL;
1785
1786        if (opt->output_prefix) {
1787                msg = opt->output_prefix(opt, opt->output_prefix_data);
1788                line_prefix = msg->buf;
1789        }
1790
1791        while (dir->nr) {
1792                struct dirstat_file *f = dir->files;
1793                int namelen = strlen(f->name);
1794                unsigned long this;
1795                char *slash;
1796
1797                if (namelen < baselen)
1798                        break;
1799                if (memcmp(f->name, base, baselen))
1800                        break;
1801                slash = strchr(f->name + baselen, '/');
1802                if (slash) {
1803                        int newbaselen = slash + 1 - f->name;
1804                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1805                        sources++;
1806                } else {
1807                        this = f->changed;
1808                        dir->files++;
1809                        dir->nr--;
1810                        sources += 2;
1811                }
1812                this_dir += this;
1813        }
1814
1815        /*
1816         * We don't report dirstat's for
1817         *  - the top level
1818         *  - or cases where everything came from a single directory
1819         *    under this directory (sources == 1).
1820         */
1821        if (baselen && sources != 1) {
1822                if (this_dir) {
1823                        int permille = this_dir * 1000 / changed;
1824                        if (permille >= dir->permille) {
1825                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1826                                        permille / 10, permille % 10, baselen, base);
1827                                if (!dir->cumulative)
1828                                        return 0;
1829                        }
1830                }
1831        }
1832        return this_dir;
1833}
1834
1835static int dirstat_compare(const void *_a, const void *_b)
1836{
1837        const struct dirstat_file *a = _a;
1838        const struct dirstat_file *b = _b;
1839        return strcmp(a->name, b->name);
1840}
1841
1842static void show_dirstat(struct diff_options *options)
1843{
1844        int i;
1845        unsigned long changed;
1846        struct dirstat_dir dir;
1847        struct diff_queue_struct *q = &diff_queued_diff;
1848
1849        dir.files = NULL;
1850        dir.alloc = 0;
1851        dir.nr = 0;
1852        dir.permille = options->dirstat_permille;
1853        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1854
1855        changed = 0;
1856        for (i = 0; i < q->nr; i++) {
1857                struct diff_filepair *p = q->queue[i];
1858                const char *name;
1859                unsigned long copied, added, damage;
1860                int content_changed;
1861
1862                name = p->two->path ? p->two->path : p->one->path;
1863
1864                if (p->one->sha1_valid && p->two->sha1_valid)
1865                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1866                else
1867                        content_changed = 1;
1868
1869                if (!content_changed) {
1870                        /*
1871                         * The SHA1 has not changed, so pre-/post-content is
1872                         * identical. We can therefore skip looking at the
1873                         * file contents altogether.
1874                         */
1875                        damage = 0;
1876                        goto found_damage;
1877                }
1878
1879                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1880                        /*
1881                         * In --dirstat-by-file mode, we don't really need to
1882                         * look at the actual file contents at all.
1883                         * The fact that the SHA1 changed is enough for us to
1884                         * add this file to the list of results
1885                         * (with each file contributing equal damage).
1886                         */
1887                        damage = 1;
1888                        goto found_damage;
1889                }
1890
1891                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1892                        diff_populate_filespec(p->one, 0);
1893                        diff_populate_filespec(p->two, 0);
1894                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1895                                               &copied, &added);
1896                        diff_free_filespec_data(p->one);
1897                        diff_free_filespec_data(p->two);
1898                } else if (DIFF_FILE_VALID(p->one)) {
1899                        diff_populate_filespec(p->one, 1);
1900                        copied = added = 0;
1901                        diff_free_filespec_data(p->one);
1902                } else if (DIFF_FILE_VALID(p->two)) {
1903                        diff_populate_filespec(p->two, 1);
1904                        copied = 0;
1905                        added = p->two->size;
1906                        diff_free_filespec_data(p->two);
1907                } else
1908                        continue;
1909
1910                /*
1911                 * Original minus copied is the removed material,
1912                 * added is the new material.  They are both damages
1913                 * made to the preimage.
1914                 * If the resulting damage is zero, we know that
1915                 * diffcore_count_changes() considers the two entries to
1916                 * be identical, but since content_changed is true, we
1917                 * know that there must have been _some_ kind of change,
1918                 * so we force all entries to have damage > 0.
1919                 */
1920                damage = (p->one->size - copied) + added;
1921                if (!damage)
1922                        damage = 1;
1923
1924found_damage:
1925                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1926                dir.files[dir.nr].name = name;
1927                dir.files[dir.nr].changed = damage;
1928                changed += damage;
1929                dir.nr++;
1930        }
1931
1932        /* This can happen even with many files, if everything was renames */
1933        if (!changed)
1934                return;
1935
1936        /* Show all directories with more than x% of the changes */
1937        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1938        gather_dirstat(options, &dir, changed, "", 0);
1939}
1940
1941static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1942{
1943        int i;
1944        unsigned long changed;
1945        struct dirstat_dir dir;
1946
1947        if (data->nr == 0)
1948                return;
1949
1950        dir.files = NULL;
1951        dir.alloc = 0;
1952        dir.nr = 0;
1953        dir.permille = options->dirstat_permille;
1954        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1955
1956        changed = 0;
1957        for (i = 0; i < data->nr; i++) {
1958                struct diffstat_file *file = data->files[i];
1959                unsigned long damage = file->added + file->deleted;
1960                if (file->is_binary)
1961                        /*
1962                         * binary files counts bytes, not lines. Must find some
1963                         * way to normalize binary bytes vs. textual lines.
1964                         * The following heuristic assumes that there are 64
1965                         * bytes per "line".
1966                         * This is stupid and ugly, but very cheap...
1967                         */
1968                        damage = (damage + 63) / 64;
1969                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1970                dir.files[dir.nr].name = file->name;
1971                dir.files[dir.nr].changed = damage;
1972                changed += damage;
1973                dir.nr++;
1974        }
1975
1976        /* This can happen even with many files, if everything was renames */
1977        if (!changed)
1978                return;
1979
1980        /* Show all directories with more than x% of the changes */
1981        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1982        gather_dirstat(options, &dir, changed, "", 0);
1983}
1984
1985static void free_diffstat_info(struct diffstat_t *diffstat)
1986{
1987        int i;
1988        for (i = 0; i < diffstat->nr; i++) {
1989                struct diffstat_file *f = diffstat->files[i];
1990                if (f->name != f->print_name)
1991                        free(f->print_name);
1992                free(f->name);
1993                free(f->from_name);
1994                free(f);
1995        }
1996        free(diffstat->files);
1997}
1998
1999struct checkdiff_t {
2000        const char *filename;
2001        int lineno;
2002        int conflict_marker_size;
2003        struct diff_options *o;
2004        unsigned ws_rule;
2005        unsigned status;
2006};
2007
2008static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2009{
2010        char firstchar;
2011        int cnt;
2012
2013        if (len < marker_size + 1)
2014                return 0;
2015        firstchar = line[0];
2016        switch (firstchar) {
2017        case '=': case '>': case '<': case '|':
2018                break;
2019        default:
2020                return 0;
2021        }
2022        for (cnt = 1; cnt < marker_size; cnt++)
2023                if (line[cnt] != firstchar)
2024                        return 0;
2025        /* line[1] thru line[marker_size-1] are same as firstchar */
2026        if (len < marker_size + 1 || !isspace(line[marker_size]))
2027                return 0;
2028        return 1;
2029}
2030
2031static void checkdiff_consume(void *priv, char *line, unsigned long len)
2032{
2033        struct checkdiff_t *data = priv;
2034        int marker_size = data->conflict_marker_size;
2035        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2036        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2037        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2038        char *err;
2039        char *line_prefix = "";
2040        struct strbuf *msgbuf;
2041
2042        assert(data->o);
2043        if (data->o->output_prefix) {
2044                msgbuf = data->o->output_prefix(data->o,
2045                        data->o->output_prefix_data);
2046                line_prefix = msgbuf->buf;
2047        }
2048
2049        if (line[0] == '+') {
2050                unsigned bad;
2051                data->lineno++;
2052                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2053                        data->status |= 1;
2054                        fprintf(data->o->file,
2055                                "%s%s:%d: leftover conflict marker\n",
2056                                line_prefix, data->filename, data->lineno);
2057                }
2058                bad = ws_check(line + 1, len - 1, data->ws_rule);
2059                if (!bad)
2060                        return;
2061                data->status |= bad;
2062                err = whitespace_error_string(bad);
2063                fprintf(data->o->file, "%s%s:%d: %s.\n",
2064                        line_prefix, data->filename, data->lineno, err);
2065                free(err);
2066                emit_line(data->o, set, reset, line, 1);
2067                ws_check_emit(line + 1, len - 1, data->ws_rule,
2068                              data->o->file, set, reset, ws);
2069        } else if (line[0] == ' ') {
2070                data->lineno++;
2071        } else if (line[0] == '@') {
2072                char *plus = strchr(line, '+');
2073                if (plus)
2074                        data->lineno = strtol(plus, NULL, 10) - 1;
2075                else
2076                        die("invalid diff");
2077        }
2078}
2079
2080static unsigned char *deflate_it(char *data,
2081                                 unsigned long size,
2082                                 unsigned long *result_size)
2083{
2084        int bound;
2085        unsigned char *deflated;
2086        git_zstream stream;
2087
2088        memset(&stream, 0, sizeof(stream));
2089        git_deflate_init(&stream, zlib_compression_level);
2090        bound = git_deflate_bound(&stream, size);
2091        deflated = xmalloc(bound);
2092        stream.next_out = deflated;
2093        stream.avail_out = bound;
2094
2095        stream.next_in = (unsigned char *)data;
2096        stream.avail_in = size;
2097        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2098                ; /* nothing */
2099        git_deflate_end(&stream);
2100        *result_size = stream.total_out;
2101        return deflated;
2102}
2103
2104static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2105{
2106        void *cp;
2107        void *delta;
2108        void *deflated;
2109        void *data;
2110        unsigned long orig_size;
2111        unsigned long delta_size;
2112        unsigned long deflate_size;
2113        unsigned long data_size;
2114
2115        /* We could do deflated delta, or we could do just deflated two,
2116         * whichever is smaller.
2117         */
2118        delta = NULL;
2119        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2120        if (one->size && two->size) {
2121                delta = diff_delta(one->ptr, one->size,
2122                                   two->ptr, two->size,
2123                                   &delta_size, deflate_size);
2124                if (delta) {
2125                        void *to_free = delta;
2126                        orig_size = delta_size;
2127                        delta = deflate_it(delta, delta_size, &delta_size);
2128                        free(to_free);
2129                }
2130        }
2131
2132        if (delta && delta_size < deflate_size) {
2133                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2134                free(deflated);
2135                data = delta;
2136                data_size = delta_size;
2137        }
2138        else {
2139                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2140                free(delta);
2141                data = deflated;
2142                data_size = deflate_size;
2143        }
2144
2145        /* emit data encoded in base85 */
2146        cp = data;
2147        while (data_size) {
2148                int bytes = (52 < data_size) ? 52 : data_size;
2149                char line[70];
2150                data_size -= bytes;
2151                if (bytes <= 26)
2152                        line[0] = bytes + 'A' - 1;
2153                else
2154                        line[0] = bytes - 26 + 'a' - 1;
2155                encode_85(line + 1, cp, bytes);
2156                cp = (char *) cp + bytes;
2157                fprintf(file, "%s", prefix);
2158                fputs(line, file);
2159                fputc('\n', file);
2160        }
2161        fprintf(file, "%s\n", prefix);
2162        free(data);
2163}
2164
2165static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2166{
2167        fprintf(file, "%sGIT binary patch\n", prefix);
2168        emit_binary_diff_body(file, one, two, prefix);
2169        emit_binary_diff_body(file, two, one, prefix);
2170}
2171
2172int diff_filespec_is_binary(struct diff_filespec *one)
2173{
2174        if (one->is_binary == -1) {
2175                diff_filespec_load_driver(one);
2176                if (one->driver->binary != -1)
2177                        one->is_binary = one->driver->binary;
2178                else {
2179                        if (!one->data && DIFF_FILE_VALID(one))
2180                                diff_populate_filespec(one, 0);
2181                        if (one->data)
2182                                one->is_binary = buffer_is_binary(one->data,
2183                                                one->size);
2184                        if (one->is_binary == -1)
2185                                one->is_binary = 0;
2186                }
2187        }
2188        return one->is_binary;
2189}
2190
2191static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2192{
2193        diff_filespec_load_driver(one);
2194        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2195}
2196
2197void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2198{
2199        if (!options->a_prefix)
2200                options->a_prefix = a;
2201        if (!options->b_prefix)
2202                options->b_prefix = b;
2203}
2204
2205struct userdiff_driver *get_textconv(struct diff_filespec *one)
2206{
2207        if (!DIFF_FILE_VALID(one))
2208                return NULL;
2209
2210        diff_filespec_load_driver(one);
2211        return userdiff_get_textconv(one->driver);
2212}
2213
2214static void builtin_diff(const char *name_a,
2215                         const char *name_b,
2216                         struct diff_filespec *one,
2217                         struct diff_filespec *two,
2218                         const char *xfrm_msg,
2219                         int must_show_header,
2220                         struct diff_options *o,
2221                         int complete_rewrite)
2222{
2223        mmfile_t mf1, mf2;
2224        const char *lbl[2];
2225        char *a_one, *b_two;
2226        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2227        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2228        const char *a_prefix, *b_prefix;
2229        struct userdiff_driver *textconv_one = NULL;
2230        struct userdiff_driver *textconv_two = NULL;
2231        struct strbuf header = STRBUF_INIT;
2232        struct strbuf *msgbuf;
2233        char *line_prefix = "";
2234
2235        if (o->output_prefix) {
2236                msgbuf = o->output_prefix(o, o->output_prefix_data);
2237                line_prefix = msgbuf->buf;
2238        }
2239
2240        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2241                        (!one->mode || S_ISGITLINK(one->mode)) &&
2242                        (!two->mode || S_ISGITLINK(two->mode))) {
2243                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2244                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2245                show_submodule_summary(o->file, one ? one->path : two->path,
2246                                one->sha1, two->sha1, two->dirty_submodule,
2247                                del, add, reset);
2248                return;
2249        }
2250
2251        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2252                textconv_one = get_textconv(one);
2253                textconv_two = get_textconv(two);
2254        }
2255
2256        diff_set_mnemonic_prefix(o, "a/", "b/");
2257        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2258                a_prefix = o->b_prefix;
2259                b_prefix = o->a_prefix;
2260        } else {
2261                a_prefix = o->a_prefix;
2262                b_prefix = o->b_prefix;
2263        }
2264
2265        /* Never use a non-valid filename anywhere if at all possible */
2266        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2267        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2268
2269        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2270        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2271        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2272        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2273        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2274        if (lbl[0][0] == '/') {
2275                /* /dev/null */
2276                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2277                if (xfrm_msg)
2278                        strbuf_addstr(&header, xfrm_msg);
2279                must_show_header = 1;
2280        }
2281        else if (lbl[1][0] == '/') {
2282                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2283                if (xfrm_msg)
2284                        strbuf_addstr(&header, xfrm_msg);
2285                must_show_header = 1;
2286        }
2287        else {
2288                if (one->mode != two->mode) {
2289                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2290                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2291                        must_show_header = 1;
2292                }
2293                if (xfrm_msg)
2294                        strbuf_addstr(&header, xfrm_msg);
2295
2296                /*
2297                 * we do not run diff between different kind
2298                 * of objects.
2299                 */
2300                if ((one->mode ^ two->mode) & S_IFMT)
2301                        goto free_ab_and_return;
2302                if (complete_rewrite &&
2303                    (textconv_one || !diff_filespec_is_binary(one)) &&
2304                    (textconv_two || !diff_filespec_is_binary(two))) {
2305                        fprintf(o->file, "%s", header.buf);
2306                        strbuf_reset(&header);
2307                        emit_rewrite_diff(name_a, name_b, one, two,
2308                                                textconv_one, textconv_two, o);
2309                        o->found_changes = 1;
2310                        goto free_ab_and_return;
2311                }
2312        }
2313
2314        if (o->irreversible_delete && lbl[1][0] == '/') {
2315                fprintf(o->file, "%s", header.buf);
2316                strbuf_reset(&header);
2317                goto free_ab_and_return;
2318        } else if (!DIFF_OPT_TST(o, TEXT) &&
2319            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2320              (!textconv_two && diff_filespec_is_binary(two)) )) {
2321                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2322                        die("unable to read files to diff");
2323                /* Quite common confusing case */
2324                if (mf1.size == mf2.size &&
2325                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2326                        if (must_show_header)
2327                                fprintf(o->file, "%s", header.buf);
2328                        goto free_ab_and_return;
2329                }
2330                fprintf(o->file, "%s", header.buf);
2331                strbuf_reset(&header);
2332                if (DIFF_OPT_TST(o, BINARY))
2333                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2334                else
2335                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2336                                line_prefix, lbl[0], lbl[1]);
2337                o->found_changes = 1;
2338        } else {
2339                /* Crazy xdl interfaces.. */
2340                const char *diffopts = getenv("GIT_DIFF_OPTS");
2341                xpparam_t xpp;
2342                xdemitconf_t xecfg;
2343                struct emit_callback ecbdata;
2344                const struct userdiff_funcname *pe;
2345
2346                if (must_show_header) {
2347                        fprintf(o->file, "%s", header.buf);
2348                        strbuf_reset(&header);
2349                }
2350
2351                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2352                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2353
2354                pe = diff_funcname_pattern(one);
2355                if (!pe)
2356                        pe = diff_funcname_pattern(two);
2357
2358                memset(&xpp, 0, sizeof(xpp));
2359                memset(&xecfg, 0, sizeof(xecfg));
2360                memset(&ecbdata, 0, sizeof(ecbdata));
2361                ecbdata.label_path = lbl;
2362                ecbdata.color_diff = want_color(o->use_color);
2363                ecbdata.found_changesp = &o->found_changes;
2364                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2365                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2366                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2367                ecbdata.opt = o;
2368                ecbdata.header = header.len ? &header : NULL;
2369                xpp.flags = o->xdl_opts;
2370                xecfg.ctxlen = o->context;
2371                xecfg.interhunkctxlen = o->interhunkcontext;
2372                xecfg.flags = XDL_EMIT_FUNCNAMES;
2373                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2374                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2375                if (pe)
2376                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2377                if (!diffopts)
2378                        ;
2379                else if (!prefixcmp(diffopts, "--unified="))
2380                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2381                else if (!prefixcmp(diffopts, "-u"))
2382                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2383                if (o->word_diff)
2384                        init_diff_words_data(&ecbdata, o, one, two);
2385                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2386                              &xpp, &xecfg);
2387                if (o->word_diff)
2388                        free_diff_words_data(&ecbdata);
2389                if (textconv_one)
2390                        free(mf1.ptr);
2391                if (textconv_two)
2392                        free(mf2.ptr);
2393                xdiff_clear_find_func(&xecfg);
2394        }
2395
2396 free_ab_and_return:
2397        strbuf_release(&header);
2398        diff_free_filespec_data(one);
2399        diff_free_filespec_data(two);
2400        free(a_one);
2401        free(b_two);
2402        return;
2403}
2404
2405static void builtin_diffstat(const char *name_a, const char *name_b,
2406                             struct diff_filespec *one,
2407                             struct diff_filespec *two,
2408                             struct diffstat_t *diffstat,
2409                             struct diff_options *o,
2410                             struct diff_filepair *p)
2411{
2412        mmfile_t mf1, mf2;
2413        struct diffstat_file *data;
2414        int same_contents;
2415        int complete_rewrite = 0;
2416
2417        if (!DIFF_PAIR_UNMERGED(p)) {
2418                if (p->status == DIFF_STATUS_MODIFIED && p->score)
2419                        complete_rewrite = 1;
2420        }
2421
2422        data = diffstat_add(diffstat, name_a, name_b);
2423        data->is_interesting = p->status != 0;
2424
2425        if (!one || !two) {
2426                data->is_unmerged = 1;
2427                return;
2428        }
2429
2430        same_contents = !hashcmp(one->sha1, two->sha1);
2431
2432        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2433                data->is_binary = 1;
2434                if (same_contents) {
2435                        data->added = 0;
2436                        data->deleted = 0;
2437                } else {
2438                        data->added = diff_filespec_size(two);
2439                        data->deleted = diff_filespec_size(one);
2440                }
2441        }
2442
2443        else if (complete_rewrite) {
2444                diff_populate_filespec(one, 0);
2445                diff_populate_filespec(two, 0);
2446                data->deleted = count_lines(one->data, one->size);
2447                data->added = count_lines(two->data, two->size);
2448        }
2449
2450        else if (!same_contents) {
2451                /* Crazy xdl interfaces.. */
2452                xpparam_t xpp;
2453                xdemitconf_t xecfg;
2454
2455                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2456                        die("unable to read files to diff");
2457
2458                memset(&xpp, 0, sizeof(xpp));
2459                memset(&xecfg, 0, sizeof(xecfg));
2460                xpp.flags = o->xdl_opts;
2461                xecfg.ctxlen = o->context;
2462                xecfg.interhunkctxlen = o->interhunkcontext;
2463                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2464                              &xpp, &xecfg);
2465        }
2466
2467        diff_free_filespec_data(one);
2468        diff_free_filespec_data(two);
2469}
2470
2471static void builtin_checkdiff(const char *name_a, const char *name_b,
2472                              const char *attr_path,
2473                              struct diff_filespec *one,
2474                              struct diff_filespec *two,
2475                              struct diff_options *o)
2476{
2477        mmfile_t mf1, mf2;
2478        struct checkdiff_t data;
2479
2480        if (!two)
2481                return;
2482
2483        memset(&data, 0, sizeof(data));
2484        data.filename = name_b ? name_b : name_a;
2485        data.lineno = 0;
2486        data.o = o;
2487        data.ws_rule = whitespace_rule(attr_path);
2488        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2489
2490        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2491                die("unable to read files to diff");
2492
2493        /*
2494         * All the other codepaths check both sides, but not checking
2495         * the "old" side here is deliberate.  We are checking the newly
2496         * introduced changes, and as long as the "new" side is text, we
2497         * can and should check what it introduces.
2498         */
2499        if (diff_filespec_is_binary(two))
2500                goto free_and_return;
2501        else {
2502                /* Crazy xdl interfaces.. */
2503                xpparam_t xpp;
2504                xdemitconf_t xecfg;
2505
2506                memset(&xpp, 0, sizeof(xpp));
2507                memset(&xecfg, 0, sizeof(xecfg));
2508                xecfg.ctxlen = 1; /* at least one context line */
2509                xpp.flags = 0;
2510                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2511                              &xpp, &xecfg);
2512
2513                if (data.ws_rule & WS_BLANK_AT_EOF) {
2514                        struct emit_callback ecbdata;
2515                        int blank_at_eof;
2516
2517                        ecbdata.ws_rule = data.ws_rule;
2518                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2519                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2520
2521                        if (blank_at_eof) {
2522                                static char *err;
2523                                if (!err)
2524                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2525                                fprintf(o->file, "%s:%d: %s.\n",
2526                                        data.filename, blank_at_eof, err);
2527                                data.status = 1; /* report errors */
2528                        }
2529                }
2530        }
2531 free_and_return:
2532        diff_free_filespec_data(one);
2533        diff_free_filespec_data(two);
2534        if (data.status)
2535                DIFF_OPT_SET(o, CHECK_FAILED);
2536}
2537
2538struct diff_filespec *alloc_filespec(const char *path)
2539{
2540        int namelen = strlen(path);
2541        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2542
2543        memset(spec, 0, sizeof(*spec));
2544        spec->path = (char *)(spec + 1);
2545        memcpy(spec->path, path, namelen+1);
2546        spec->count = 1;
2547        spec->is_binary = -1;
2548        return spec;
2549}
2550
2551void free_filespec(struct diff_filespec *spec)
2552{
2553        if (!--spec->count) {
2554                diff_free_filespec_data(spec);
2555                free(spec);
2556        }
2557}
2558
2559void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2560                   int sha1_valid, unsigned short mode)
2561{
2562        if (mode) {
2563                spec->mode = canon_mode(mode);
2564                hashcpy(spec->sha1, sha1);
2565                spec->sha1_valid = sha1_valid;
2566        }
2567}
2568
2569/*
2570 * Given a name and sha1 pair, if the index tells us the file in
2571 * the work tree has that object contents, return true, so that
2572 * prepare_temp_file() does not have to inflate and extract.
2573 */
2574static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2575{
2576        struct cache_entry *ce;
2577        struct stat st;
2578        int pos, len;
2579
2580        /*
2581         * We do not read the cache ourselves here, because the
2582         * benchmark with my previous version that always reads cache
2583         * shows that it makes things worse for diff-tree comparing
2584         * two linux-2.6 kernel trees in an already checked out work
2585         * tree.  This is because most diff-tree comparisons deal with
2586         * only a small number of files, while reading the cache is
2587         * expensive for a large project, and its cost outweighs the
2588         * savings we get by not inflating the object to a temporary
2589         * file.  Practically, this code only helps when we are used
2590         * by diff-cache --cached, which does read the cache before
2591         * calling us.
2592         */
2593        if (!active_cache)
2594                return 0;
2595
2596        /* We want to avoid the working directory if our caller
2597         * doesn't need the data in a normal file, this system
2598         * is rather slow with its stat/open/mmap/close syscalls,
2599         * and the object is contained in a pack file.  The pack
2600         * is probably already open and will be faster to obtain
2601         * the data through than the working directory.  Loose
2602         * objects however would tend to be slower as they need
2603         * to be individually opened and inflated.
2604         */
2605        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2606                return 0;
2607
2608        len = strlen(name);
2609        pos = cache_name_pos(name, len);
2610        if (pos < 0)
2611                return 0;
2612        ce = active_cache[pos];
2613
2614        /*
2615         * This is not the sha1 we are looking for, or
2616         * unreusable because it is not a regular file.
2617         */
2618        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2619                return 0;
2620
2621        /*
2622         * If ce is marked as "assume unchanged", there is no
2623         * guarantee that work tree matches what we are looking for.
2624         */
2625        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2626                return 0;
2627
2628        /*
2629         * If ce matches the file in the work tree, we can reuse it.
2630         */
2631        if (ce_uptodate(ce) ||
2632            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2633                return 1;
2634
2635        return 0;
2636}
2637
2638static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2639{
2640        int len;
2641        char *data = xmalloc(100), *dirty = "";
2642
2643        /* Are we looking at the work tree? */
2644        if (s->dirty_submodule)
2645                dirty = "-dirty";
2646
2647        len = snprintf(data, 100,
2648                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2649        s->data = data;
2650        s->size = len;
2651        s->should_free = 1;
2652        if (size_only) {
2653                s->data = NULL;
2654                free(data);
2655        }
2656        return 0;
2657}
2658
2659/*
2660 * While doing rename detection and pickaxe operation, we may need to
2661 * grab the data for the blob (or file) for our own in-core comparison.
2662 * diff_filespec has data and size fields for this purpose.
2663 */
2664int diff_populate_filespec(struct diff_filespec *s, int size_only)
2665{
2666        int err = 0;
2667        if (!DIFF_FILE_VALID(s))
2668                die("internal error: asking to populate invalid file.");
2669        if (S_ISDIR(s->mode))
2670                return -1;
2671
2672        if (s->data)
2673                return 0;
2674
2675        if (size_only && 0 < s->size)
2676                return 0;
2677
2678        if (S_ISGITLINK(s->mode))
2679                return diff_populate_gitlink(s, size_only);
2680
2681        if (!s->sha1_valid ||
2682            reuse_worktree_file(s->path, s->sha1, 0)) {
2683                struct strbuf buf = STRBUF_INIT;
2684                struct stat st;
2685                int fd;
2686
2687                if (lstat(s->path, &st) < 0) {
2688                        if (errno == ENOENT) {
2689                        err_empty:
2690                                err = -1;
2691                        empty:
2692                                s->data = (char *)"";
2693                                s->size = 0;
2694                                return err;
2695                        }
2696                }
2697                s->size = xsize_t(st.st_size);
2698                if (!s->size)
2699                        goto empty;
2700                if (S_ISLNK(st.st_mode)) {
2701                        struct strbuf sb = STRBUF_INIT;
2702
2703                        if (strbuf_readlink(&sb, s->path, s->size))
2704                                goto err_empty;
2705                        s->size = sb.len;
2706                        s->data = strbuf_detach(&sb, NULL);
2707                        s->should_free = 1;
2708                        return 0;
2709                }
2710                if (size_only)
2711                        return 0;
2712                fd = open(s->path, O_RDONLY);
2713                if (fd < 0)
2714                        goto err_empty;
2715                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2716                close(fd);
2717                s->should_munmap = 1;
2718
2719                /*
2720                 * Convert from working tree format to canonical git format
2721                 */
2722                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2723                        size_t size = 0;
2724                        munmap(s->data, s->size);
2725                        s->should_munmap = 0;
2726                        s->data = strbuf_detach(&buf, &size);
2727                        s->size = size;
2728                        s->should_free = 1;
2729                }
2730        }
2731        else {
2732                enum object_type type;
2733                if (size_only) {
2734                        type = sha1_object_info(s->sha1, &s->size);
2735                        if (type < 0)
2736                                die("unable to read %s", sha1_to_hex(s->sha1));
2737                } else {
2738                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2739                        if (!s->data)
2740                                die("unable to read %s", sha1_to_hex(s->sha1));
2741                        s->should_free = 1;
2742                }
2743        }
2744        return 0;
2745}
2746
2747void diff_free_filespec_blob(struct diff_filespec *s)
2748{
2749        if (s->should_free)
2750                free(s->data);
2751        else if (s->should_munmap)
2752                munmap(s->data, s->size);
2753
2754        if (s->should_free || s->should_munmap) {
2755                s->should_free = s->should_munmap = 0;
2756                s->data = NULL;
2757        }
2758}
2759
2760void diff_free_filespec_data(struct diff_filespec *s)
2761{
2762        diff_free_filespec_blob(s);
2763        free(s->cnt_data);
2764        s->cnt_data = NULL;
2765}
2766
2767static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2768                           void *blob,
2769                           unsigned long size,
2770                           const unsigned char *sha1,
2771                           int mode)
2772{
2773        int fd;
2774        struct strbuf buf = STRBUF_INIT;
2775        struct strbuf template = STRBUF_INIT;
2776        char *path_dup = xstrdup(path);
2777        const char *base = basename(path_dup);
2778
2779        /* Generate "XXXXXX_basename.ext" */
2780        strbuf_addstr(&template, "XXXXXX_");
2781        strbuf_addstr(&template, base);
2782
2783        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2784                        strlen(base) + 1);
2785        if (fd < 0)
2786                die_errno("unable to create temp-file");
2787        if (convert_to_working_tree(path,
2788                        (const char *)blob, (size_t)size, &buf)) {
2789                blob = buf.buf;
2790                size = buf.len;
2791        }
2792        if (write_in_full(fd, blob, size) != size)
2793                die_errno("unable to write temp-file");
2794        close(fd);
2795        temp->name = temp->tmp_path;
2796        strcpy(temp->hex, sha1_to_hex(sha1));
2797        temp->hex[40] = 0;
2798        sprintf(temp->mode, "%06o", mode);
2799        strbuf_release(&buf);
2800        strbuf_release(&template);
2801        free(path_dup);
2802}
2803
2804static struct diff_tempfile *prepare_temp_file(const char *name,
2805                struct diff_filespec *one)
2806{
2807        struct diff_tempfile *temp = claim_diff_tempfile();
2808
2809        if (!DIFF_FILE_VALID(one)) {
2810        not_a_valid_file:
2811                /* A '-' entry produces this for file-2, and
2812                 * a '+' entry produces this for file-1.
2813                 */
2814                temp->name = "/dev/null";
2815                strcpy(temp->hex, ".");
2816                strcpy(temp->mode, ".");
2817                return temp;
2818        }
2819
2820        if (!remove_tempfile_installed) {
2821                atexit(remove_tempfile);
2822                sigchain_push_common(remove_tempfile_on_signal);
2823                remove_tempfile_installed = 1;
2824        }
2825
2826        if (!one->sha1_valid ||
2827            reuse_worktree_file(name, one->sha1, 1)) {
2828                struct stat st;
2829                if (lstat(name, &st) < 0) {
2830                        if (errno == ENOENT)
2831                                goto not_a_valid_file;
2832                        die_errno("stat(%s)", name);
2833                }
2834                if (S_ISLNK(st.st_mode)) {
2835                        struct strbuf sb = STRBUF_INIT;
2836                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2837                                die_errno("readlink(%s)", name);
2838                        prep_temp_blob(name, temp, sb.buf, sb.len,
2839                                       (one->sha1_valid ?
2840                                        one->sha1 : null_sha1),
2841                                       (one->sha1_valid ?
2842                                        one->mode : S_IFLNK));
2843                        strbuf_release(&sb);
2844                }
2845                else {
2846                        /* we can borrow from the file in the work tree */
2847                        temp->name = name;
2848                        if (!one->sha1_valid)
2849                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2850                        else
2851                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2852                        /* Even though we may sometimes borrow the
2853                         * contents from the work tree, we always want
2854                         * one->mode.  mode is trustworthy even when
2855                         * !(one->sha1_valid), as long as
2856                         * DIFF_FILE_VALID(one).
2857                         */
2858                        sprintf(temp->mode, "%06o", one->mode);
2859                }
2860                return temp;
2861        }
2862        else {
2863                if (diff_populate_filespec(one, 0))
2864                        die("cannot read data blob for %s", one->path);
2865                prep_temp_blob(name, temp, one->data, one->size,
2866                               one->sha1, one->mode);
2867        }
2868        return temp;
2869}
2870
2871/* An external diff command takes:
2872 *
2873 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2874 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2875 *
2876 */
2877static void run_external_diff(const char *pgm,
2878                              const char *name,
2879                              const char *other,
2880                              struct diff_filespec *one,
2881                              struct diff_filespec *two,
2882                              const char *xfrm_msg,
2883                              int complete_rewrite)
2884{
2885        const char *spawn_arg[10];
2886        int retval;
2887        const char **arg = &spawn_arg[0];
2888
2889        if (one && two) {
2890                struct diff_tempfile *temp_one, *temp_two;
2891                const char *othername = (other ? other : name);
2892                temp_one = prepare_temp_file(name, one);
2893                temp_two = prepare_temp_file(othername, two);
2894                *arg++ = pgm;
2895                *arg++ = name;
2896                *arg++ = temp_one->name;
2897                *arg++ = temp_one->hex;
2898                *arg++ = temp_one->mode;
2899                *arg++ = temp_two->name;
2900                *arg++ = temp_two->hex;
2901                *arg++ = temp_two->mode;
2902                if (other) {
2903                        *arg++ = other;
2904                        *arg++ = xfrm_msg;
2905                }
2906        } else {
2907                *arg++ = pgm;
2908                *arg++ = name;
2909        }
2910        *arg = NULL;
2911        fflush(NULL);
2912        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2913        remove_tempfile();
2914        if (retval) {
2915                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2916                exit(1);
2917        }
2918}
2919
2920static int similarity_index(struct diff_filepair *p)
2921{
2922        return p->score * 100 / MAX_SCORE;
2923}
2924
2925static void fill_metainfo(struct strbuf *msg,
2926                          const char *name,
2927                          const char *other,
2928                          struct diff_filespec *one,
2929                          struct diff_filespec *two,
2930                          struct diff_options *o,
2931                          struct diff_filepair *p,
2932                          int *must_show_header,
2933                          int use_color)
2934{
2935        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2936        const char *reset = diff_get_color(use_color, DIFF_RESET);
2937        struct strbuf *msgbuf;
2938        char *line_prefix = "";
2939
2940        *must_show_header = 1;
2941        if (o->output_prefix) {
2942                msgbuf = o->output_prefix(o, o->output_prefix_data);
2943                line_prefix = msgbuf->buf;
2944        }
2945        strbuf_init(msg, PATH_MAX * 2 + 300);
2946        switch (p->status) {
2947        case DIFF_STATUS_COPIED:
2948                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2949                            line_prefix, set, similarity_index(p));
2950                strbuf_addf(msg, "%s\n%s%scopy from ",
2951                            reset,  line_prefix, set);
2952                quote_c_style(name, msg, NULL, 0);
2953                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2954                quote_c_style(other, msg, NULL, 0);
2955                strbuf_addf(msg, "%s\n", reset);
2956                break;
2957        case DIFF_STATUS_RENAMED:
2958                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2959                            line_prefix, set, similarity_index(p));
2960                strbuf_addf(msg, "%s\n%s%srename from ",
2961                            reset, line_prefix, set);
2962                quote_c_style(name, msg, NULL, 0);
2963                strbuf_addf(msg, "%s\n%s%srename to ",
2964                            reset, line_prefix, set);
2965                quote_c_style(other, msg, NULL, 0);
2966                strbuf_addf(msg, "%s\n", reset);
2967                break;
2968        case DIFF_STATUS_MODIFIED:
2969                if (p->score) {
2970                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2971                                    line_prefix,
2972                                    set, similarity_index(p), reset);
2973                        break;
2974                }
2975                /* fallthru */
2976        default:
2977                *must_show_header = 0;
2978        }
2979        if (one && two && hashcmp(one->sha1, two->sha1)) {
2980                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2981
2982                if (DIFF_OPT_TST(o, BINARY)) {
2983                        mmfile_t mf;
2984                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2985                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2986                                abbrev = 40;
2987                }
2988                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2989                            find_unique_abbrev(one->sha1, abbrev));
2990                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2991                if (one->mode == two->mode)
2992                        strbuf_addf(msg, " %06o", one->mode);
2993                strbuf_addf(msg, "%s\n", reset);
2994        }
2995}
2996
2997static void run_diff_cmd(const char *pgm,
2998                         const char *name,
2999                         const char *other,
3000                         const char *attr_path,
3001                         struct diff_filespec *one,
3002                         struct diff_filespec *two,
3003                         struct strbuf *msg,
3004                         struct diff_options *o,
3005                         struct diff_filepair *p)
3006{
3007        const char *xfrm_msg = NULL;
3008        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3009        int must_show_header = 0;
3010
3011
3012        if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3013                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3014                if (drv && drv->external)
3015                        pgm = drv->external;
3016        }
3017
3018        if (msg) {
3019                /*
3020                 * don't use colors when the header is intended for an
3021                 * external diff driver
3022                 */
3023                fill_metainfo(msg, name, other, one, two, o, p,
3024                              &must_show_header,
3025                              want_color(o->use_color) && !pgm);
3026                xfrm_msg = msg->len ? msg->buf : NULL;
3027        }
3028
3029        if (pgm) {
3030                run_external_diff(pgm, name, other, one, two, xfrm_msg,
3031                                  complete_rewrite);
3032                return;
3033        }
3034        if (one && two)
3035                builtin_diff(name, other ? other : name,
3036                             one, two, xfrm_msg, must_show_header,
3037                             o, complete_rewrite);
3038        else
3039                fprintf(o->file, "* Unmerged path %s\n", name);
3040}
3041
3042static void diff_fill_sha1_info(struct diff_filespec *one)
3043{
3044        if (DIFF_FILE_VALID(one)) {
3045                if (!one->sha1_valid) {
3046                        struct stat st;
3047                        if (one->is_stdin) {
3048                                hashcpy(one->sha1, null_sha1);
3049                                return;
3050                        }
3051                        if (lstat(one->path, &st) < 0)
3052                                die_errno("stat '%s'", one->path);
3053                        if (index_path(one->sha1, one->path, &st, 0))
3054                                die("cannot hash %s", one->path);
3055                }
3056        }
3057        else
3058                hashclr(one->sha1);
3059}
3060
3061static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3062{
3063        /* Strip the prefix but do not molest /dev/null and absolute paths */
3064        if (*namep && **namep != '/') {
3065                *namep += prefix_length;
3066                if (**namep == '/')
3067                        ++*namep;
3068        }
3069        if (*otherp && **otherp != '/') {
3070                *otherp += prefix_length;
3071                if (**otherp == '/')
3072                        ++*otherp;
3073        }
3074}
3075
3076static void run_diff(struct diff_filepair *p, struct diff_options *o)
3077{
3078        const char *pgm = external_diff();
3079        struct strbuf msg;
3080        struct diff_filespec *one = p->one;
3081        struct diff_filespec *two = p->two;
3082        const char *name;
3083        const char *other;
3084        const char *attr_path;
3085
3086        name  = p->one->path;
3087        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3088        attr_path = name;
3089        if (o->prefix_length)
3090                strip_prefix(o->prefix_length, &name, &other);
3091
3092        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3093                pgm = NULL;
3094
3095        if (DIFF_PAIR_UNMERGED(p)) {
3096                run_diff_cmd(pgm, name, NULL, attr_path,
3097                             NULL, NULL, NULL, o, p);
3098                return;
3099        }
3100
3101        diff_fill_sha1_info(one);
3102        diff_fill_sha1_info(two);
3103
3104        if (!pgm &&
3105            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3106            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3107                /*
3108                 * a filepair that changes between file and symlink
3109                 * needs to be split into deletion and creation.
3110                 */
3111                struct diff_filespec *null = alloc_filespec(two->path);
3112                run_diff_cmd(NULL, name, other, attr_path,
3113                             one, null, &msg, o, p);
3114                free(null);
3115                strbuf_release(&msg);
3116
3117                null = alloc_filespec(one->path);
3118                run_diff_cmd(NULL, name, other, attr_path,
3119                             null, two, &msg, o, p);
3120                free(null);
3121        }
3122        else
3123                run_diff_cmd(pgm, name, other, attr_path,
3124                             one, two, &msg, o, p);
3125
3126        strbuf_release(&msg);
3127}
3128
3129static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3130                         struct diffstat_t *diffstat)
3131{
3132        const char *name;
3133        const char *other;
3134
3135        if (DIFF_PAIR_UNMERGED(p)) {
3136                /* unmerged */
3137                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3138                return;
3139        }
3140
3141        name = p->one->path;
3142        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3143
3144        if (o->prefix_length)
3145                strip_prefix(o->prefix_length, &name, &other);
3146
3147        diff_fill_sha1_info(p->one);
3148        diff_fill_sha1_info(p->two);
3149
3150        builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3151}
3152
3153static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3154{
3155        const char *name;
3156        const char *other;
3157        const char *attr_path;
3158
3159        if (DIFF_PAIR_UNMERGED(p)) {
3160                /* unmerged */
3161                return;
3162        }
3163
3164        name = p->one->path;
3165        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3166        attr_path = other ? other : name;
3167
3168        if (o->prefix_length)
3169                strip_prefix(o->prefix_length, &name, &other);
3170
3171        diff_fill_sha1_info(p->one);
3172        diff_fill_sha1_info(p->two);
3173
3174        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3175}
3176
3177void diff_setup(struct diff_options *options)
3178{
3179        memcpy(options, &default_diff_options, sizeof(*options));
3180
3181        options->file = stdout;
3182
3183        options->line_termination = '\n';
3184        options->break_opt = -1;
3185        options->rename_limit = -1;
3186        options->dirstat_permille = diff_dirstat_permille_default;
3187        options->context = diff_context_default;
3188        DIFF_OPT_SET(options, RENAME_EMPTY);
3189
3190        options->change = diff_change;
3191        options->add_remove = diff_addremove;
3192        options->use_color = diff_use_color_default;
3193        options->detect_rename = diff_detect_rename_default;
3194
3195        if (diff_no_prefix) {
3196                options->a_prefix = options->b_prefix = "";
3197        } else if (!diff_mnemonic_prefix) {
3198                options->a_prefix = "a/";
3199                options->b_prefix = "b/";
3200        }
3201}
3202
3203void diff_setup_done(struct diff_options *options)
3204{
3205        int count = 0;
3206
3207        if (options->output_format & DIFF_FORMAT_NAME)
3208                count++;
3209        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3210                count++;
3211        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3212                count++;
3213        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3214                count++;
3215        if (count > 1)
3216                die("--name-only, --name-status, --check and -s are mutually exclusive");
3217
3218        /*
3219         * Most of the time we can say "there are changes"
3220         * only by checking if there are changed paths, but
3221         * --ignore-whitespace* options force us to look
3222         * inside contents.
3223         */
3224
3225        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3226            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3227            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3228                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3229        else
3230                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3231
3232        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3233                options->detect_rename = DIFF_DETECT_COPY;
3234
3235        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3236                options->prefix = NULL;
3237        if (options->prefix)
3238                options->prefix_length = strlen(options->prefix);
3239        else
3240                options->prefix_length = 0;
3241
3242        if (options->output_format & (DIFF_FORMAT_NAME |
3243                                      DIFF_FORMAT_NAME_STATUS |
3244                                      DIFF_FORMAT_CHECKDIFF |
3245                                      DIFF_FORMAT_NO_OUTPUT))
3246                options->output_format &= ~(DIFF_FORMAT_RAW |
3247                                            DIFF_FORMAT_NUMSTAT |
3248                                            DIFF_FORMAT_DIFFSTAT |
3249                                            DIFF_FORMAT_SHORTSTAT |
3250                                            DIFF_FORMAT_DIRSTAT |
3251                                            DIFF_FORMAT_SUMMARY |
3252                                            DIFF_FORMAT_PATCH);
3253
3254        /*
3255         * These cases always need recursive; we do not drop caller-supplied
3256         * recursive bits for other formats here.
3257         */
3258        if (options->output_format & (DIFF_FORMAT_PATCH |
3259                                      DIFF_FORMAT_NUMSTAT |
3260                                      DIFF_FORMAT_DIFFSTAT |
3261                                      DIFF_FORMAT_SHORTSTAT |
3262                                      DIFF_FORMAT_DIRSTAT |
3263                                      DIFF_FORMAT_SUMMARY |
3264                                      DIFF_FORMAT_CHECKDIFF))
3265                DIFF_OPT_SET(options, RECURSIVE);
3266        /*
3267         * Also pickaxe would not work very well if you do not say recursive
3268         */
3269        if (options->pickaxe)
3270                DIFF_OPT_SET(options, RECURSIVE);
3271        /*
3272         * When patches are generated, submodules diffed against the work tree
3273         * must be checked for dirtiness too so it can be shown in the output
3274         */
3275        if (options->output_format & DIFF_FORMAT_PATCH)
3276                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3277
3278        if (options->detect_rename && options->rename_limit < 0)
3279                options->rename_limit = diff_rename_limit_default;
3280        if (options->setup & DIFF_SETUP_USE_CACHE) {
3281                if (!active_cache)
3282                        /* read-cache does not die even when it fails
3283                         * so it is safe for us to do this here.  Also
3284                         * it does not smudge active_cache or active_nr
3285                         * when it fails, so we do not have to worry about
3286                         * cleaning it up ourselves either.
3287                         */
3288                        read_cache();
3289        }
3290        if (options->abbrev <= 0 || 40 < options->abbrev)
3291                options->abbrev = 40; /* full */
3292
3293        /*
3294         * It does not make sense to show the first hit we happened
3295         * to have found.  It does not make sense not to return with
3296         * exit code in such a case either.
3297         */
3298        if (DIFF_OPT_TST(options, QUICK)) {
3299                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3300                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3301        }
3302}
3303
3304static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3305{
3306        char c, *eq;
3307        int len;
3308
3309        if (*arg != '-')
3310                return 0;
3311        c = *++arg;
3312        if (!c)
3313                return 0;
3314        if (c == arg_short) {
3315                c = *++arg;
3316                if (!c)
3317                        return 1;
3318                if (val && isdigit(c)) {
3319                        char *end;
3320                        int n = strtoul(arg, &end, 10);
3321                        if (*end)
3322                                return 0;
3323                        *val = n;
3324                        return 1;
3325                }
3326                return 0;
3327        }
3328        if (c != '-')
3329                return 0;
3330        arg++;
3331        eq = strchr(arg, '=');
3332        if (eq)
3333                len = eq - arg;
3334        else
3335                len = strlen(arg);
3336        if (!len || strncmp(arg, arg_long, len))
3337                return 0;
3338        if (eq) {
3339                int n;
3340                char *end;
3341                if (!isdigit(*++eq))
3342                        return 0;
3343                n = strtoul(eq, &end, 10);
3344                if (*end)
3345                        return 0;
3346                *val = n;
3347        }
3348        return 1;
3349}
3350
3351static int diff_scoreopt_parse(const char *opt);
3352
3353static inline int short_opt(char opt, const char **argv,
3354                            const char **optarg)
3355{
3356        const char *arg = argv[0];
3357        if (arg[0] != '-' || arg[1] != opt)
3358                return 0;
3359        if (arg[2] != '\0') {
3360                *optarg = arg + 2;
3361                return 1;
3362        }
3363        if (!argv[1])
3364                die("Option '%c' requires a value", opt);
3365        *optarg = argv[1];
3366        return 2;
3367}
3368
3369int parse_long_opt(const char *opt, const char **argv,
3370                   const char **optarg)
3371{
3372        const char *arg = argv[0];
3373        if (arg[0] != '-' || arg[1] != '-')
3374                return 0;
3375        arg += strlen("--");
3376        if (prefixcmp(arg, opt))
3377                return 0;
3378        arg += strlen(opt);
3379        if (*arg == '=') { /* sticked form: --option=value */
3380                *optarg = arg + 1;
3381                return 1;
3382        }
3383        if (*arg != '\0')
3384                return 0;
3385        /* separate form: --option value */
3386        if (!argv[1])
3387                die("Option '--%s' requires a value", opt);
3388        *optarg = argv[1];
3389        return 2;
3390}
3391
3392static int stat_opt(struct diff_options *options, const char **av)
3393{
3394        const char *arg = av[0];
3395        char *end;
3396        int width = options->stat_width;
3397        int name_width = options->stat_name_width;
3398        int graph_width = options->stat_graph_width;
3399        int count = options->stat_count;
3400        int argcount = 1;
3401
3402        arg += strlen("--stat");
3403        end = (char *)arg;
3404
3405        switch (*arg) {
3406        case '-':
3407                if (!prefixcmp(arg, "-width")) {
3408                        arg += strlen("-width");
3409                        if (*arg == '=')
3410                                width = strtoul(arg + 1, &end, 10);
3411                        else if (!*arg && !av[1])
3412                                die("Option '--stat-width' requires a value");
3413                        else if (!*arg) {
3414                                width = strtoul(av[1], &end, 10);
3415                                argcount = 2;
3416                        }
3417                } else if (!prefixcmp(arg, "-name-width")) {
3418                        arg += strlen("-name-width");
3419                        if (*arg == '=')
3420                                name_width = strtoul(arg + 1, &end, 10);
3421                        else if (!*arg && !av[1])
3422                                die("Option '--stat-name-width' requires a value");
3423                        else if (!*arg) {
3424                                name_width = strtoul(av[1], &end, 10);
3425                                argcount = 2;
3426                        }
3427                } else if (!prefixcmp(arg, "-graph-width")) {
3428                        arg += strlen("-graph-width");
3429                        if (*arg == '=')
3430                                graph_width = strtoul(arg + 1, &end, 10);
3431                        else if (!*arg && !av[1])
3432                                die("Option '--stat-graph-width' requires a value");
3433                        else if (!*arg) {
3434                                graph_width = strtoul(av[1], &end, 10);
3435                                argcount = 2;
3436                        }
3437                } else if (!prefixcmp(arg, "-count")) {
3438                        arg += strlen("-count");
3439                        if (*arg == '=')
3440                                count = strtoul(arg + 1, &end, 10);
3441                        else if (!*arg && !av[1])
3442                                die("Option '--stat-count' requires a value");
3443                        else if (!*arg) {
3444                                count = strtoul(av[1], &end, 10);
3445                                argcount = 2;
3446                        }
3447                }
3448                break;
3449        case '=':
3450                width = strtoul(arg+1, &end, 10);
3451                if (*end == ',')
3452                        name_width = strtoul(end+1, &end, 10);
3453                if (*end == ',')
3454                        count = strtoul(end+1, &end, 10);
3455        }
3456
3457        /* Important! This checks all the error cases! */
3458        if (*end)
3459                return 0;
3460        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3461        options->stat_name_width = name_width;
3462        options->stat_graph_width = graph_width;
3463        options->stat_width = width;
3464        options->stat_count = count;
3465        return argcount;
3466}
3467
3468static int parse_dirstat_opt(struct diff_options *options, const char *params)
3469{
3470        struct strbuf errmsg = STRBUF_INIT;
3471        if (parse_dirstat_params(options, params, &errmsg))
3472                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3473                    errmsg.buf);
3474        strbuf_release(&errmsg);
3475        /*
3476         * The caller knows a dirstat-related option is given from the command
3477         * line; allow it to say "return this_function();"
3478         */
3479        options->output_format |= DIFF_FORMAT_DIRSTAT;
3480        return 1;
3481}
3482
3483int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3484{
3485        const char *arg = av[0];
3486        const char *optarg;
3487        int argcount;
3488
3489        /* Output format options */
3490        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3491                options->output_format |= DIFF_FORMAT_PATCH;
3492        else if (opt_arg(arg, 'U', "unified", &options->context))
3493                options->output_format |= DIFF_FORMAT_PATCH;
3494        else if (!strcmp(arg, "--raw"))
3495                options->output_format |= DIFF_FORMAT_RAW;
3496        else if (!strcmp(arg, "--patch-with-raw"))
3497                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3498        else if (!strcmp(arg, "--numstat"))
3499                options->output_format |= DIFF_FORMAT_NUMSTAT;
3500        else if (!strcmp(arg, "--shortstat"))
3501                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3502        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3503                return parse_dirstat_opt(options, "");
3504        else if (!prefixcmp(arg, "-X"))
3505                return parse_dirstat_opt(options, arg + 2);
3506        else if (!prefixcmp(arg, "--dirstat="))
3507                return parse_dirstat_opt(options, arg + 10);
3508        else if (!strcmp(arg, "--cumulative"))
3509                return parse_dirstat_opt(options, "cumulative");
3510        else if (!strcmp(arg, "--dirstat-by-file"))
3511                return parse_dirstat_opt(options, "files");
3512        else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3513                parse_dirstat_opt(options, "files");
3514                return parse_dirstat_opt(options, arg + 18);
3515        }
3516        else if (!strcmp(arg, "--check"))
3517                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3518        else if (!strcmp(arg, "--summary"))
3519                options->output_format |= DIFF_FORMAT_SUMMARY;
3520        else if (!strcmp(arg, "--patch-with-stat"))
3521                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3522        else if (!strcmp(arg, "--name-only"))
3523                options->output_format |= DIFF_FORMAT_NAME;
3524        else if (!strcmp(arg, "--name-status"))
3525                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3526        else if (!strcmp(arg, "-s"))
3527                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3528        else if (!prefixcmp(arg, "--stat"))
3529                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3530                return stat_opt(options, av);
3531
3532        /* renames options */
3533        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3534                 !strcmp(arg, "--break-rewrites")) {
3535                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3536                        return error("invalid argument to -B: %s", arg+2);
3537        }
3538        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3539                 !strcmp(arg, "--find-renames")) {
3540                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3541                        return error("invalid argument to -M: %s", arg+2);
3542                options->detect_rename = DIFF_DETECT_RENAME;
3543        }
3544        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3545                options->irreversible_delete = 1;
3546        }
3547        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3548                 !strcmp(arg, "--find-copies")) {
3549                if (options->detect_rename == DIFF_DETECT_COPY)
3550                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3551                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3552                        return error("invalid argument to -C: %s", arg+2);
3553                options->detect_rename = DIFF_DETECT_COPY;
3554        }
3555        else if (!strcmp(arg, "--no-renames"))
3556                options->detect_rename = 0;
3557        else if (!strcmp(arg, "--rename-empty"))
3558                DIFF_OPT_SET(options, RENAME_EMPTY);
3559        else if (!strcmp(arg, "--no-rename-empty"))
3560                DIFF_OPT_CLR(options, RENAME_EMPTY);
3561        else if (!strcmp(arg, "--relative"))
3562                DIFF_OPT_SET(options, RELATIVE_NAME);
3563        else if (!prefixcmp(arg, "--relative=")) {
3564                DIFF_OPT_SET(options, RELATIVE_NAME);
3565                options->prefix = arg + 11;
3566        }
3567
3568        /* xdiff options */
3569        else if (!strcmp(arg, "--minimal"))
3570                DIFF_XDL_SET(options, NEED_MINIMAL);
3571        else if (!strcmp(arg, "--no-minimal"))
3572                DIFF_XDL_CLR(options, NEED_MINIMAL);
3573        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3574                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3575        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3576                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3577        else if (!strcmp(arg, "--ignore-space-at-eol"))
3578                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3579        else if (!strcmp(arg, "--patience"))
3580                options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3581        else if (!strcmp(arg, "--histogram"))
3582                options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3583
3584        /* flags options */
3585        else if (!strcmp(arg, "--binary")) {
3586                options->output_format |= DIFF_FORMAT_PATCH;
3587                DIFF_OPT_SET(options, BINARY);
3588        }
3589        else if (!strcmp(arg, "--full-index"))
3590                DIFF_OPT_SET(options, FULL_INDEX);
3591        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3592                DIFF_OPT_SET(options, TEXT);
3593        else if (!strcmp(arg, "-R"))
3594                DIFF_OPT_SET(options, REVERSE_DIFF);
3595        else if (!strcmp(arg, "--find-copies-harder"))
3596                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3597        else if (!strcmp(arg, "--follow"))
3598                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3599        else if (!strcmp(arg, "--color"))
3600                options->use_color = 1;
3601        else if (!prefixcmp(arg, "--color=")) {
3602                int value = git_config_colorbool(NULL, arg+8);
3603                if (value < 0)
3604                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3605                options->use_color = value;
3606        }
3607        else if (!strcmp(arg, "--no-color"))
3608                options->use_color = 0;
3609        else if (!strcmp(arg, "--color-words")) {
3610                options->use_color = 1;
3611                options->word_diff = DIFF_WORDS_COLOR;
3612        }
3613        else if (!prefixcmp(arg, "--color-words=")) {
3614                options->use_color = 1;
3615                options->word_diff = DIFF_WORDS_COLOR;
3616                options->word_regex = arg + 14;
3617        }
3618        else if (!strcmp(arg, "--word-diff")) {
3619                if (options->word_diff == DIFF_WORDS_NONE)
3620                        options->word_diff = DIFF_WORDS_PLAIN;
3621        }
3622        else if (!prefixcmp(arg, "--word-diff=")) {
3623                const char *type = arg + 12;
3624                if (!strcmp(type, "plain"))
3625                        options->word_diff = DIFF_WORDS_PLAIN;
3626                else if (!strcmp(type, "color")) {
3627                        options->use_color = 1;
3628                        options->word_diff = DIFF_WORDS_COLOR;
3629                }
3630                else if (!strcmp(type, "porcelain"))
3631                        options->word_diff = DIFF_WORDS_PORCELAIN;
3632                else if (!strcmp(type, "none"))
3633                        options->word_diff = DIFF_WORDS_NONE;
3634                else
3635                        die("bad --word-diff argument: %s", type);
3636        }
3637        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3638                if (options->word_diff == DIFF_WORDS_NONE)
3639                        options->word_diff = DIFF_WORDS_PLAIN;
3640                options->word_regex = optarg;
3641                return argcount;
3642        }
3643        else if (!strcmp(arg, "--exit-code"))
3644                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3645        else if (!strcmp(arg, "--quiet"))
3646                DIFF_OPT_SET(options, QUICK);
3647        else if (!strcmp(arg, "--ext-diff"))
3648                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3649        else if (!strcmp(arg, "--no-ext-diff"))
3650                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3651        else if (!strcmp(arg, "--textconv"))
3652                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3653        else if (!strcmp(arg, "--no-textconv"))
3654                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3655        else if (!strcmp(arg, "--ignore-submodules")) {
3656                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3657                handle_ignore_submodules_arg(options, "all");
3658        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3659                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3660                handle_ignore_submodules_arg(options, arg + 20);
3661        } else if (!strcmp(arg, "--submodule"))
3662                DIFF_OPT_SET(options, SUBMODULE_LOG);
3663        else if (!prefixcmp(arg, "--submodule=")) {
3664                if (!strcmp(arg + 12, "log"))
3665                        DIFF_OPT_SET(options, SUBMODULE_LOG);
3666        }
3667
3668        /* misc options */
3669        else if (!strcmp(arg, "-z"))
3670                options->line_termination = 0;
3671        else if ((argcount = short_opt('l', av, &optarg))) {
3672                options->rename_limit = strtoul(optarg, NULL, 10);
3673                return argcount;
3674        }
3675        else if ((argcount = short_opt('S', av, &optarg))) {
3676                options->pickaxe = optarg;
3677                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3678                return argcount;
3679        } else if ((argcount = short_opt('G', av, &optarg))) {
3680                options->pickaxe = optarg;
3681                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3682                return argcount;
3683        }
3684        else if (!strcmp(arg, "--pickaxe-all"))
3685                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3686        else if (!strcmp(arg, "--pickaxe-regex"))
3687                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3688        else if ((argcount = short_opt('O', av, &optarg))) {
3689                options->orderfile = optarg;
3690                return argcount;
3691        }
3692        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3693                options->filter = optarg;
3694                return argcount;
3695        }
3696        else if (!strcmp(arg, "--abbrev"))
3697                options->abbrev = DEFAULT_ABBREV;
3698        else if (!prefixcmp(arg, "--abbrev=")) {
3699                options->abbrev = strtoul(arg + 9, NULL, 10);
3700                if (options->abbrev < MINIMUM_ABBREV)
3701                        options->abbrev = MINIMUM_ABBREV;
3702                else if (40 < options->abbrev)
3703                        options->abbrev = 40;
3704        }
3705        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3706                options->a_prefix = optarg;
3707                return argcount;
3708        }
3709        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3710                options->b_prefix = optarg;
3711                return argcount;
3712        }
3713        else if (!strcmp(arg, "--no-prefix"))
3714                options->a_prefix = options->b_prefix = "";
3715        else if (opt_arg(arg, '\0', "inter-hunk-context",
3716                         &options->interhunkcontext))
3717                ;
3718        else if (!strcmp(arg, "-W"))
3719                DIFF_OPT_SET(options, FUNCCONTEXT);
3720        else if (!strcmp(arg, "--function-context"))
3721                DIFF_OPT_SET(options, FUNCCONTEXT);
3722        else if (!strcmp(arg, "--no-function-context"))
3723                DIFF_OPT_CLR(options, FUNCCONTEXT);
3724        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3725                options->file = fopen(optarg, "w");
3726                if (!options->file)
3727                        die_errno("Could not open '%s'", optarg);
3728                options->close_file = 1;
3729                return argcount;
3730        } else
3731                return 0;
3732        return 1;
3733}
3734
3735int parse_rename_score(const char **cp_p)
3736{
3737        unsigned long num, scale;
3738        int ch, dot;
3739        const char *cp = *cp_p;
3740
3741        num = 0;
3742        scale = 1;
3743        dot = 0;
3744        for (;;) {
3745                ch = *cp;
3746                if ( !dot && ch == '.' ) {
3747                        scale = 1;
3748                        dot = 1;
3749                } else if ( ch == '%' ) {
3750                        scale = dot ? scale*100 : 100;
3751                        cp++;   /* % is always at the end */
3752                        break;
3753                } else if ( ch >= '0' && ch <= '9' ) {
3754                        if ( scale < 100000 ) {
3755                                scale *= 10;
3756                                num = (num*10) + (ch-'0');
3757                        }
3758                } else {
3759                        break;
3760                }
3761                cp++;
3762        }
3763        *cp_p = cp;
3764
3765        /* user says num divided by scale and we say internally that
3766         * is MAX_SCORE * num / scale.
3767         */
3768        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3769}
3770
3771static int diff_scoreopt_parse(const char *opt)
3772{
3773        int opt1, opt2, cmd;
3774
3775        if (*opt++ != '-')
3776                return -1;
3777        cmd = *opt++;
3778        if (cmd == '-') {
3779                /* convert the long-form arguments into short-form versions */
3780                if (!prefixcmp(opt, "break-rewrites")) {
3781                        opt += strlen("break-rewrites");
3782                        if (*opt == 0 || *opt++ == '=')
3783                                cmd = 'B';
3784                } else if (!prefixcmp(opt, "find-copies")) {
3785                        opt += strlen("find-copies");
3786                        if (*opt == 0 || *opt++ == '=')
3787                                cmd = 'C';
3788                } else if (!prefixcmp(opt, "find-renames")) {
3789                        opt += strlen("find-renames");
3790                        if (*opt == 0 || *opt++ == '=')
3791                                cmd = 'M';
3792                }
3793        }
3794        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3795                return -1; /* that is not a -M, -C nor -B option */
3796
3797        opt1 = parse_rename_score(&opt);
3798        if (cmd != 'B')
3799                opt2 = 0;
3800        else {
3801                if (*opt == 0)
3802                        opt2 = 0;
3803                else if (*opt != '/')
3804                        return -1; /* we expect -B80/99 or -B80 */
3805                else {
3806                        opt++;
3807                        opt2 = parse_rename_score(&opt);
3808                }
3809        }
3810        if (*opt != 0)
3811                return -1;
3812        return opt1 | (opt2 << 16);
3813}
3814
3815struct diff_queue_struct diff_queued_diff;
3816
3817void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3818{
3819        if (queue->alloc <= queue->nr) {
3820                queue->alloc = alloc_nr(queue->alloc);
3821                queue->queue = xrealloc(queue->queue,
3822                                        sizeof(dp) * queue->alloc);
3823        }
3824        queue->queue[queue->nr++] = dp;
3825}
3826
3827struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3828                                 struct diff_filespec *one,
3829                                 struct diff_filespec *two)
3830{
3831        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3832        dp->one = one;
3833        dp->two = two;
3834        if (queue)
3835                diff_q(queue, dp);
3836        return dp;
3837}
3838
3839void diff_free_filepair(struct diff_filepair *p)
3840{
3841        free_filespec(p->one);
3842        free_filespec(p->two);
3843        free(p);
3844}
3845
3846/* This is different from find_unique_abbrev() in that
3847 * it stuffs the result with dots for alignment.
3848 */
3849const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3850{
3851        int abblen;
3852        const char *abbrev;
3853        if (len == 40)
3854                return sha1_to_hex(sha1);
3855
3856        abbrev = find_unique_abbrev(sha1, len);
3857        abblen = strlen(abbrev);
3858        if (abblen < 37) {
3859                static char hex[41];
3860                if (len < abblen && abblen <= len + 2)
3861                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3862                else
3863                        sprintf(hex, "%s...", abbrev);
3864                return hex;
3865        }
3866        return sha1_to_hex(sha1);
3867}
3868
3869static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3870{
3871        int line_termination = opt->line_termination;
3872        int inter_name_termination = line_termination ? '\t' : '\0';
3873        if (opt->output_prefix) {
3874                struct strbuf *msg = NULL;
3875                msg = opt->output_prefix(opt, opt->output_prefix_data);
3876                fprintf(opt->file, "%s", msg->buf);
3877        }
3878
3879        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3880                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3881                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3882                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3883        }
3884        if (p->score) {
3885                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3886                        inter_name_termination);
3887        } else {
3888                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3889        }
3890
3891        if (p->status == DIFF_STATUS_COPIED ||
3892            p->status == DIFF_STATUS_RENAMED) {
3893                const char *name_a, *name_b;
3894                name_a = p->one->path;
3895                name_b = p->two->path;
3896                strip_prefix(opt->prefix_length, &name_a, &name_b);
3897                write_name_quoted(name_a, opt->file, inter_name_termination);
3898                write_name_quoted(name_b, opt->file, line_termination);
3899        } else {
3900                const char *name_a, *name_b;
3901                name_a = p->one->mode ? p->one->path : p->two->path;
3902                name_b = NULL;
3903                strip_prefix(opt->prefix_length, &name_a, &name_b);
3904                write_name_quoted(name_a, opt->file, line_termination);
3905        }
3906}
3907
3908int diff_unmodified_pair(struct diff_filepair *p)
3909{
3910        /* This function is written stricter than necessary to support
3911         * the currently implemented transformers, but the idea is to
3912         * let transformers to produce diff_filepairs any way they want,
3913         * and filter and clean them up here before producing the output.
3914         */
3915        struct diff_filespec *one = p->one, *two = p->two;
3916
3917        if (DIFF_PAIR_UNMERGED(p))
3918                return 0; /* unmerged is interesting */
3919
3920        /* deletion, addition, mode or type change
3921         * and rename are all interesting.
3922         */
3923        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3924            DIFF_PAIR_MODE_CHANGED(p) ||
3925            strcmp(one->path, two->path))
3926                return 0;
3927
3928        /* both are valid and point at the same path.  that is, we are
3929         * dealing with a change.
3930         */
3931        if (one->sha1_valid && two->sha1_valid &&
3932            !hashcmp(one->sha1, two->sha1) &&
3933            !one->dirty_submodule && !two->dirty_submodule)
3934                return 1; /* no change */
3935        if (!one->sha1_valid && !two->sha1_valid)
3936                return 1; /* both look at the same file on the filesystem. */
3937        return 0;
3938}
3939
3940static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3941{
3942        if (diff_unmodified_pair(p))
3943                return;
3944
3945        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3946            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3947                return; /* no tree diffs in patch format */
3948
3949        run_diff(p, o);
3950}
3951
3952static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3953                            struct diffstat_t *diffstat)
3954{
3955        if (diff_unmodified_pair(p))
3956                return;
3957
3958        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3959            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3960                return; /* no useful stat for tree diffs */
3961
3962        run_diffstat(p, o, diffstat);
3963}
3964
3965static void diff_flush_checkdiff(struct diff_filepair *p,
3966                struct diff_options *o)
3967{
3968        if (diff_unmodified_pair(p))
3969                return;
3970
3971        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3972            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3973                return; /* nothing to check in tree diffs */
3974
3975        run_checkdiff(p, o);
3976}
3977
3978int diff_queue_is_empty(void)
3979{
3980        struct diff_queue_struct *q = &diff_queued_diff;
3981        int i;
3982        for (i = 0; i < q->nr; i++)
3983                if (!diff_unmodified_pair(q->queue[i]))
3984                        return 0;
3985        return 1;
3986}
3987
3988#if DIFF_DEBUG
3989void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3990{
3991        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3992                x, one ? one : "",
3993                s->path,
3994                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3995                s->mode,
3996                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3997        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3998                x, one ? one : "",
3999                s->size, s->xfrm_flags);
4000}
4001
4002void diff_debug_filepair(const struct diff_filepair *p, int i)
4003{
4004        diff_debug_filespec(p->one, i, "one");
4005        diff_debug_filespec(p->two, i, "two");
4006        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4007                p->score, p->status ? p->status : '?',
4008                p->one->rename_used, p->broken_pair);
4009}
4010
4011void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4012{
4013        int i;
4014        if (msg)
4015                fprintf(stderr, "%s\n", msg);
4016        fprintf(stderr, "q->nr = %d\n", q->nr);
4017        for (i = 0; i < q->nr; i++) {
4018                struct diff_filepair *p = q->queue[i];
4019                diff_debug_filepair(p, i);
4020        }
4021}
4022#endif
4023
4024static void diff_resolve_rename_copy(void)
4025{
4026        int i;
4027        struct diff_filepair *p;
4028        struct diff_queue_struct *q = &diff_queued_diff;
4029
4030        diff_debug_queue("resolve-rename-copy", q);
4031
4032        for (i = 0; i < q->nr; i++) {
4033                p = q->queue[i];
4034                p->status = 0; /* undecided */
4035                if (DIFF_PAIR_UNMERGED(p))
4036                        p->status = DIFF_STATUS_UNMERGED;
4037                else if (!DIFF_FILE_VALID(p->one))
4038                        p->status = DIFF_STATUS_ADDED;
4039                else if (!DIFF_FILE_VALID(p->two))
4040                        p->status = DIFF_STATUS_DELETED;
4041                else if (DIFF_PAIR_TYPE_CHANGED(p))
4042                        p->status = DIFF_STATUS_TYPE_CHANGED;
4043
4044                /* from this point on, we are dealing with a pair
4045                 * whose both sides are valid and of the same type, i.e.
4046                 * either in-place edit or rename/copy edit.
4047                 */
4048                else if (DIFF_PAIR_RENAME(p)) {
4049                        /*
4050                         * A rename might have re-connected a broken
4051                         * pair up, causing the pathnames to be the
4052                         * same again. If so, that's not a rename at
4053                         * all, just a modification..
4054                         *
4055                         * Otherwise, see if this source was used for
4056                         * multiple renames, in which case we decrement
4057                         * the count, and call it a copy.
4058                         */
4059                        if (!strcmp(p->one->path, p->two->path))
4060                                p->status = DIFF_STATUS_MODIFIED;
4061                        else if (--p->one->rename_used > 0)
4062                                p->status = DIFF_STATUS_COPIED;
4063                        else
4064                                p->status = DIFF_STATUS_RENAMED;
4065                }
4066                else if (hashcmp(p->one->sha1, p->two->sha1) ||
4067                         p->one->mode != p->two->mode ||
4068                         p->one->dirty_submodule ||
4069                         p->two->dirty_submodule ||
4070                         is_null_sha1(p->one->sha1))
4071                        p->status = DIFF_STATUS_MODIFIED;
4072                else {
4073                        /* This is a "no-change" entry and should not
4074                         * happen anymore, but prepare for broken callers.
4075                         */
4076                        error("feeding unmodified %s to diffcore",
4077                              p->one->path);
4078                        p->status = DIFF_STATUS_UNKNOWN;
4079                }
4080        }
4081        diff_debug_queue("resolve-rename-copy done", q);
4082}
4083
4084static int check_pair_status(struct diff_filepair *p)
4085{
4086        switch (p->status) {
4087        case DIFF_STATUS_UNKNOWN:
4088                return 0;
4089        case 0:
4090                die("internal error in diff-resolve-rename-copy");
4091        default:
4092                return 1;
4093        }
4094}
4095
4096static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4097{
4098        int fmt = opt->output_format;
4099
4100        if (fmt & DIFF_FORMAT_CHECKDIFF)
4101                diff_flush_checkdiff(p, opt);
4102        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4103                diff_flush_raw(p, opt);
4104        else if (fmt & DIFF_FORMAT_NAME) {
4105                const char *name_a, *name_b;
4106                name_a = p->two->path;
4107                name_b = NULL;
4108                strip_prefix(opt->prefix_length, &name_a, &name_b);
4109                write_name_quoted(name_a, opt->file, opt->line_termination);
4110        }
4111}
4112
4113static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4114{
4115        if (fs->mode)
4116                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4117        else
4118                fprintf(file, " %s ", newdelete);
4119        write_name_quoted(fs->path, file, '\n');
4120}
4121
4122
4123static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4124                const char *line_prefix)
4125{
4126        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4127                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4128                        p->two->mode, show_name ? ' ' : '\n');
4129                if (show_name) {
4130                        write_name_quoted(p->two->path, file, '\n');
4131                }
4132        }
4133}
4134
4135static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4136                        const char *line_prefix)
4137{
4138        char *names = pprint_rename(p->one->path, p->two->path);
4139
4140        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4141        free(names);
4142        show_mode_change(file, p, 0, line_prefix);
4143}
4144
4145static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4146{
4147        FILE *file = opt->file;
4148        char *line_prefix = "";
4149
4150        if (opt->output_prefix) {
4151                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4152                line_prefix = buf->buf;
4153        }
4154
4155        switch(p->status) {
4156        case DIFF_STATUS_DELETED:
4157                fputs(line_prefix, file);
4158                show_file_mode_name(file, "delete", p->one);
4159                break;
4160        case DIFF_STATUS_ADDED:
4161                fputs(line_prefix, file);
4162                show_file_mode_name(file, "create", p->two);
4163                break;
4164        case DIFF_STATUS_COPIED:
4165                fputs(line_prefix, file);
4166                show_rename_copy(file, "copy", p, line_prefix);
4167                break;
4168        case DIFF_STATUS_RENAMED:
4169                fputs(line_prefix, file);
4170                show_rename_copy(file, "rename", p, line_prefix);
4171                break;
4172        default:
4173                if (p->score) {
4174                        fprintf(file, "%s rewrite ", line_prefix);
4175                        write_name_quoted(p->two->path, file, ' ');
4176                        fprintf(file, "(%d%%)\n", similarity_index(p));
4177                }
4178                show_mode_change(file, p, !p->score, line_prefix);
4179                break;
4180        }
4181}
4182
4183struct patch_id_t {
4184        git_SHA_CTX *ctx;
4185        int patchlen;
4186};
4187
4188static int remove_space(char *line, int len)
4189{
4190        int i;
4191        char *dst = line;
4192        unsigned char c;
4193
4194        for (i = 0; i < len; i++)
4195                if (!isspace((c = line[i])))
4196                        *dst++ = c;
4197
4198        return dst - line;
4199}
4200
4201static void patch_id_consume(void *priv, char *line, unsigned long len)
4202{
4203        struct patch_id_t *data = priv;
4204        int new_len;
4205
4206        /* Ignore line numbers when computing the SHA1 of the patch */
4207        if (!prefixcmp(line, "@@ -"))
4208                return;
4209
4210        new_len = remove_space(line, len);
4211
4212        git_SHA1_Update(data->ctx, line, new_len);
4213        data->patchlen += new_len;
4214}
4215
4216/* returns 0 upon success, and writes result into sha1 */
4217static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4218{
4219        struct diff_queue_struct *q = &diff_queued_diff;
4220        int i;
4221        git_SHA_CTX ctx;
4222        struct patch_id_t data;
4223        char buffer[PATH_MAX * 4 + 20];
4224
4225        git_SHA1_Init(&ctx);
4226        memset(&data, 0, sizeof(struct patch_id_t));
4227        data.ctx = &ctx;
4228
4229        for (i = 0; i < q->nr; i++) {
4230                xpparam_t xpp;
4231                xdemitconf_t xecfg;
4232                mmfile_t mf1, mf2;
4233                struct diff_filepair *p = q->queue[i];
4234                int len1, len2;
4235
4236                memset(&xpp, 0, sizeof(xpp));
4237                memset(&xecfg, 0, sizeof(xecfg));
4238                if (p->status == 0)
4239                        return error("internal diff status error");
4240                if (p->status == DIFF_STATUS_UNKNOWN)
4241                        continue;
4242                if (diff_unmodified_pair(p))
4243                        continue;
4244                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4245                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4246                        continue;
4247                if (DIFF_PAIR_UNMERGED(p))
4248                        continue;
4249
4250                diff_fill_sha1_info(p->one);
4251                diff_fill_sha1_info(p->two);
4252                if (fill_mmfile(&mf1, p->one) < 0 ||
4253                                fill_mmfile(&mf2, p->two) < 0)
4254                        return error("unable to read files to diff");
4255
4256                len1 = remove_space(p->one->path, strlen(p->one->path));
4257                len2 = remove_space(p->two->path, strlen(p->two->path));
4258                if (p->one->mode == 0)
4259                        len1 = snprintf(buffer, sizeof(buffer),
4260                                        "diff--gita/%.*sb/%.*s"
4261                                        "newfilemode%06o"
4262                                        "---/dev/null"
4263                                        "+++b/%.*s",
4264                                        len1, p->one->path,
4265                                        len2, p->two->path,
4266                                        p->two->mode,
4267                                        len2, p->two->path);
4268                else if (p->two->mode == 0)
4269                        len1 = snprintf(buffer, sizeof(buffer),
4270                                        "diff--gita/%.*sb/%.*s"
4271                                        "deletedfilemode%06o"
4272                                        "---a/%.*s"
4273                                        "+++/dev/null",
4274                                        len1, p->one->path,
4275                                        len2, p->two->path,
4276                                        p->one->mode,
4277                                        len1, p->one->path);
4278                else
4279                        len1 = snprintf(buffer, sizeof(buffer),
4280                                        "diff--gita/%.*sb/%.*s"
4281                                        "---a/%.*s"
4282                                        "+++b/%.*s",
4283                                        len1, p->one->path,
4284                                        len2, p->two->path,
4285                                        len1, p->one->path,
4286                                        len2, p->two->path);
4287                git_SHA1_Update(&ctx, buffer, len1);
4288
4289                if (diff_filespec_is_binary(p->one) ||
4290                    diff_filespec_is_binary(p->two)) {
4291                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4292                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4293                        continue;
4294                }
4295
4296                xpp.flags = 0;
4297                xecfg.ctxlen = 3;
4298                xecfg.flags = 0;
4299                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4300                              &xpp, &xecfg);
4301        }
4302
4303        git_SHA1_Final(sha1, &ctx);
4304        return 0;
4305}
4306
4307int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4308{
4309        struct diff_queue_struct *q = &diff_queued_diff;
4310        int i;
4311        int result = diff_get_patch_id(options, sha1);
4312
4313        for (i = 0; i < q->nr; i++)
4314                diff_free_filepair(q->queue[i]);
4315
4316        free(q->queue);
4317        DIFF_QUEUE_CLEAR(q);
4318
4319        return result;
4320}
4321
4322static int is_summary_empty(const struct diff_queue_struct *q)
4323{
4324        int i;
4325
4326        for (i = 0; i < q->nr; i++) {
4327                const struct diff_filepair *p = q->queue[i];
4328
4329                switch (p->status) {
4330                case DIFF_STATUS_DELETED:
4331                case DIFF_STATUS_ADDED:
4332                case DIFF_STATUS_COPIED:
4333                case DIFF_STATUS_RENAMED:
4334                        return 0;
4335                default:
4336                        if (p->score)
4337                                return 0;
4338                        if (p->one->mode && p->two->mode &&
4339                            p->one->mode != p->two->mode)
4340                                return 0;
4341                        break;
4342                }
4343        }
4344        return 1;
4345}
4346
4347static const char rename_limit_warning[] =
4348"inexact rename detection was skipped due to too many files.";
4349
4350static const char degrade_cc_to_c_warning[] =
4351"only found copies from modified paths due to too many files.";
4352
4353static const char rename_limit_advice[] =
4354"you may want to set your %s variable to at least "
4355"%d and retry the command.";
4356
4357void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4358{
4359        if (degraded_cc)
4360                warning(degrade_cc_to_c_warning);
4361        else if (needed)
4362                warning(rename_limit_warning);
4363        else
4364                return;
4365        if (0 < needed && needed < 32767)
4366                warning(rename_limit_advice, varname, needed);
4367}
4368
4369void diff_flush(struct diff_options *options)
4370{
4371        struct diff_queue_struct *q = &diff_queued_diff;
4372        int i, output_format = options->output_format;
4373        int separator = 0;
4374        int dirstat_by_line = 0;
4375
4376        /*
4377         * Order: raw, stat, summary, patch
4378         * or:    name/name-status/checkdiff (other bits clear)
4379         */
4380        if (!q->nr)
4381                goto free_queue;
4382
4383        if (output_format & (DIFF_FORMAT_RAW |
4384                             DIFF_FORMAT_NAME |
4385                             DIFF_FORMAT_NAME_STATUS |
4386                             DIFF_FORMAT_CHECKDIFF)) {
4387                for (i = 0; i < q->nr; i++) {
4388                        struct diff_filepair *p = q->queue[i];
4389                        if (check_pair_status(p))
4390                                flush_one_pair(p, options);
4391                }
4392                separator++;
4393        }
4394
4395        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4396                dirstat_by_line = 1;
4397
4398        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4399            dirstat_by_line) {
4400                struct diffstat_t diffstat;
4401
4402                memset(&diffstat, 0, sizeof(struct diffstat_t));
4403                for (i = 0; i < q->nr; i++) {
4404                        struct diff_filepair *p = q->queue[i];
4405                        if (check_pair_status(p))
4406                                diff_flush_stat(p, options, &diffstat);
4407                }
4408                if (output_format & DIFF_FORMAT_NUMSTAT)
4409                        show_numstat(&diffstat, options);
4410                if (output_format & DIFF_FORMAT_DIFFSTAT)
4411                        show_stats(&diffstat, options);
4412                if (output_format & DIFF_FORMAT_SHORTSTAT)
4413                        show_shortstats(&diffstat, options);
4414                if (output_format & DIFF_FORMAT_DIRSTAT)
4415                        show_dirstat_by_line(&diffstat, options);
4416                free_diffstat_info(&diffstat);
4417                separator++;
4418        }
4419        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4420                show_dirstat(options);
4421
4422        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4423                for (i = 0; i < q->nr; i++) {
4424                        diff_summary(options, q->queue[i]);
4425                }
4426                separator++;
4427        }
4428
4429        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4430            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4431            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4432                /*
4433                 * run diff_flush_patch for the exit status. setting
4434                 * options->file to /dev/null should be safe, becaue we
4435                 * aren't supposed to produce any output anyway.
4436                 */
4437                if (options->close_file)
4438                        fclose(options->file);
4439                options->file = fopen("/dev/null", "w");
4440                if (!options->file)
4441                        die_errno("Could not open /dev/null");
4442                options->close_file = 1;
4443                for (i = 0; i < q->nr; i++) {
4444                        struct diff_filepair *p = q->queue[i];
4445                        if (check_pair_status(p))
4446                                diff_flush_patch(p, options);
4447                        if (options->found_changes)
4448                                break;
4449                }
4450        }
4451
4452        if (output_format & DIFF_FORMAT_PATCH) {
4453                if (separator) {
4454                        if (options->output_prefix) {
4455                                struct strbuf *msg = NULL;
4456                                msg = options->output_prefix(options,
4457                                        options->output_prefix_data);
4458                                fwrite(msg->buf, msg->len, 1, stdout);
4459                        }
4460                        putc(options->line_termination, options->file);
4461                        if (options->stat_sep) {
4462                                /* attach patch instead of inline */
4463                                fputs(options->stat_sep, options->file);
4464                        }
4465                }
4466
4467                for (i = 0; i < q->nr; i++) {
4468                        struct diff_filepair *p = q->queue[i];
4469                        if (check_pair_status(p))
4470                                diff_flush_patch(p, options);
4471                }
4472        }
4473
4474        if (output_format & DIFF_FORMAT_CALLBACK)
4475                options->format_callback(q, options, options->format_callback_data);
4476
4477        for (i = 0; i < q->nr; i++)
4478                diff_free_filepair(q->queue[i]);
4479free_queue:
4480        free(q->queue);
4481        DIFF_QUEUE_CLEAR(q);
4482        if (options->close_file)
4483                fclose(options->file);
4484
4485        /*
4486         * Report the content-level differences with HAS_CHANGES;
4487         * diff_addremove/diff_change does not set the bit when
4488         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4489         */
4490        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4491                if (options->found_changes)
4492                        DIFF_OPT_SET(options, HAS_CHANGES);
4493                else
4494                        DIFF_OPT_CLR(options, HAS_CHANGES);
4495        }
4496}
4497
4498static void diffcore_apply_filter(const char *filter)
4499{
4500        int i;
4501        struct diff_queue_struct *q = &diff_queued_diff;
4502        struct diff_queue_struct outq;
4503        DIFF_QUEUE_CLEAR(&outq);
4504
4505        if (!filter)
4506                return;
4507
4508        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4509                int found;
4510                for (i = found = 0; !found && i < q->nr; i++) {
4511                        struct diff_filepair *p = q->queue[i];
4512                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4513                             ((p->score &&
4514                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4515                              (!p->score &&
4516                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4517                            ((p->status != DIFF_STATUS_MODIFIED) &&
4518                             strchr(filter, p->status)))
4519                                found++;
4520                }
4521                if (found)
4522                        return;
4523
4524                /* otherwise we will clear the whole queue
4525                 * by copying the empty outq at the end of this
4526                 * function, but first clear the current entries
4527                 * in the queue.
4528                 */
4529                for (i = 0; i < q->nr; i++)
4530                        diff_free_filepair(q->queue[i]);
4531        }
4532        else {
4533                /* Only the matching ones */
4534                for (i = 0; i < q->nr; i++) {
4535                        struct diff_filepair *p = q->queue[i];
4536
4537                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4538                             ((p->score &&
4539                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4540                              (!p->score &&
4541                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4542                            ((p->status != DIFF_STATUS_MODIFIED) &&
4543                             strchr(filter, p->status)))
4544                                diff_q(&outq, p);
4545                        else
4546                                diff_free_filepair(p);
4547                }
4548        }
4549        free(q->queue);
4550        *q = outq;
4551}
4552
4553/* Check whether two filespecs with the same mode and size are identical */
4554static int diff_filespec_is_identical(struct diff_filespec *one,
4555                                      struct diff_filespec *two)
4556{
4557        if (S_ISGITLINK(one->mode))
4558                return 0;
4559        if (diff_populate_filespec(one, 0))
4560                return 0;
4561        if (diff_populate_filespec(two, 0))
4562                return 0;
4563        return !memcmp(one->data, two->data, one->size);
4564}
4565
4566static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4567{
4568        int i;
4569        struct diff_queue_struct *q = &diff_queued_diff;
4570        struct diff_queue_struct outq;
4571        DIFF_QUEUE_CLEAR(&outq);
4572
4573        for (i = 0; i < q->nr; i++) {
4574                struct diff_filepair *p = q->queue[i];
4575
4576                /*
4577                 * 1. Entries that come from stat info dirtiness
4578                 *    always have both sides (iow, not create/delete),
4579                 *    one side of the object name is unknown, with
4580                 *    the same mode and size.  Keep the ones that
4581                 *    do not match these criteria.  They have real
4582                 *    differences.
4583                 *
4584                 * 2. At this point, the file is known to be modified,
4585                 *    with the same mode and size, and the object
4586                 *    name of one side is unknown.  Need to inspect
4587                 *    the identical contents.
4588                 */
4589                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4590                    !DIFF_FILE_VALID(p->two) ||
4591                    (p->one->sha1_valid && p->two->sha1_valid) ||
4592                    (p->one->mode != p->two->mode) ||
4593                    diff_populate_filespec(p->one, 1) ||
4594                    diff_populate_filespec(p->two, 1) ||
4595                    (p->one->size != p->two->size) ||
4596                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4597                        diff_q(&outq, p);
4598                else {
4599                        /*
4600                         * The caller can subtract 1 from skip_stat_unmatch
4601                         * to determine how many paths were dirty only
4602                         * due to stat info mismatch.
4603                         */
4604                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4605                                diffopt->skip_stat_unmatch++;
4606                        diff_free_filepair(p);
4607                }
4608        }
4609        free(q->queue);
4610        *q = outq;
4611}
4612
4613static int diffnamecmp(const void *a_, const void *b_)
4614{
4615        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4616        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4617        const char *name_a, *name_b;
4618
4619        name_a = a->one ? a->one->path : a->two->path;
4620        name_b = b->one ? b->one->path : b->two->path;
4621        return strcmp(name_a, name_b);
4622}
4623
4624void diffcore_fix_diff_index(struct diff_options *options)
4625{
4626        struct diff_queue_struct *q = &diff_queued_diff;
4627        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4628}
4629
4630void diffcore_std(struct diff_options *options)
4631{
4632        if (options->skip_stat_unmatch)
4633                diffcore_skip_stat_unmatch(options);
4634        if (!options->found_follow) {
4635                /* See try_to_follow_renames() in tree-diff.c */
4636                if (options->break_opt != -1)
4637                        diffcore_break(options->break_opt);
4638                if (options->detect_rename)
4639                        diffcore_rename(options);
4640                if (options->break_opt != -1)
4641                        diffcore_merge_broken();
4642        }
4643        if (options->pickaxe)
4644                diffcore_pickaxe(options);
4645        if (options->orderfile)
4646                diffcore_order(options->orderfile);
4647        if (!options->found_follow)
4648                /* See try_to_follow_renames() in tree-diff.c */
4649                diff_resolve_rename_copy();
4650        diffcore_apply_filter(options->filter);
4651
4652        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4653                DIFF_OPT_SET(options, HAS_CHANGES);
4654        else
4655                DIFF_OPT_CLR(options, HAS_CHANGES);
4656
4657        options->found_follow = 0;
4658}
4659
4660int diff_result_code(struct diff_options *opt, int status)
4661{
4662        int result = 0;
4663
4664        diff_warn_rename_limit("diff.renamelimit",
4665                               opt->needed_rename_limit,
4666                               opt->degraded_cc_to_c);
4667        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4668            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4669                return status;
4670        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4671            DIFF_OPT_TST(opt, HAS_CHANGES))
4672                result |= 01;
4673        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4674            DIFF_OPT_TST(opt, CHECK_FAILED))
4675                result |= 02;
4676        return result;
4677}
4678
4679int diff_can_quit_early(struct diff_options *opt)
4680{
4681        return (DIFF_OPT_TST(opt, QUICK) &&
4682                !opt->filter &&
4683                DIFF_OPT_TST(opt, HAS_CHANGES));
4684}
4685
4686/*
4687 * Shall changes to this submodule be ignored?
4688 *
4689 * Submodule changes can be configured to be ignored separately for each path,
4690 * but that configuration can be overridden from the command line.
4691 */
4692static int is_submodule_ignored(const char *path, struct diff_options *options)
4693{
4694        int ignored = 0;
4695        unsigned orig_flags = options->flags;
4696        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4697                set_diffopt_flags_from_submodule_config(options, path);
4698        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4699                ignored = 1;
4700        options->flags = orig_flags;
4701        return ignored;
4702}
4703
4704void diff_addremove(struct diff_options *options,
4705                    int addremove, unsigned mode,
4706                    const unsigned char *sha1,
4707                    int sha1_valid,
4708                    const char *concatpath, unsigned dirty_submodule)
4709{
4710        struct diff_filespec *one, *two;
4711
4712        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4713                return;
4714
4715        /* This may look odd, but it is a preparation for
4716         * feeding "there are unchanged files which should
4717         * not produce diffs, but when you are doing copy
4718         * detection you would need them, so here they are"
4719         * entries to the diff-core.  They will be prefixed
4720         * with something like '=' or '*' (I haven't decided
4721         * which but should not make any difference).
4722         * Feeding the same new and old to diff_change()
4723         * also has the same effect.
4724         * Before the final output happens, they are pruned after
4725         * merged into rename/copy pairs as appropriate.
4726         */
4727        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4728                addremove = (addremove == '+' ? '-' :
4729                             addremove == '-' ? '+' : addremove);
4730
4731        if (options->prefix &&
4732            strncmp(concatpath, options->prefix, options->prefix_length))
4733                return;
4734
4735        one = alloc_filespec(concatpath);
4736        two = alloc_filespec(concatpath);
4737
4738        if (addremove != '+')
4739                fill_filespec(one, sha1, sha1_valid, mode);
4740        if (addremove != '-') {
4741                fill_filespec(two, sha1, sha1_valid, mode);
4742                two->dirty_submodule = dirty_submodule;
4743        }
4744
4745        diff_queue(&diff_queued_diff, one, two);
4746        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4747                DIFF_OPT_SET(options, HAS_CHANGES);
4748}
4749
4750void diff_change(struct diff_options *options,
4751                 unsigned old_mode, unsigned new_mode,
4752                 const unsigned char *old_sha1,
4753                 const unsigned char *new_sha1,
4754                 int old_sha1_valid, int new_sha1_valid,
4755                 const char *concatpath,
4756                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4757{
4758        struct diff_filespec *one, *two;
4759
4760        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4761            is_submodule_ignored(concatpath, options))
4762                return;
4763
4764        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4765                unsigned tmp;
4766                const unsigned char *tmp_c;
4767                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4768                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4769                tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4770                        new_sha1_valid = tmp;
4771                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4772                        new_dirty_submodule = tmp;
4773        }
4774
4775        if (options->prefix &&
4776            strncmp(concatpath, options->prefix, options->prefix_length))
4777                return;
4778
4779        one = alloc_filespec(concatpath);
4780        two = alloc_filespec(concatpath);
4781        fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4782        fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4783        one->dirty_submodule = old_dirty_submodule;
4784        two->dirty_submodule = new_dirty_submodule;
4785
4786        diff_queue(&diff_queued_diff, one, two);
4787        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4788                DIFF_OPT_SET(options, HAS_CHANGES);
4789}
4790
4791struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4792{
4793        struct diff_filepair *pair;
4794        struct diff_filespec *one, *two;
4795
4796        if (options->prefix &&
4797            strncmp(path, options->prefix, options->prefix_length))
4798                return NULL;
4799
4800        one = alloc_filespec(path);
4801        two = alloc_filespec(path);
4802        pair = diff_queue(&diff_queued_diff, one, two);
4803        pair->is_unmerged = 1;
4804        return pair;
4805}
4806
4807static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4808                size_t *outsize)
4809{
4810        struct diff_tempfile *temp;
4811        const char *argv[3];
4812        const char **arg = argv;
4813        struct child_process child;
4814        struct strbuf buf = STRBUF_INIT;
4815        int err = 0;
4816
4817        temp = prepare_temp_file(spec->path, spec);
4818        *arg++ = pgm;
4819        *arg++ = temp->name;
4820        *arg = NULL;
4821
4822        memset(&child, 0, sizeof(child));
4823        child.use_shell = 1;
4824        child.argv = argv;
4825        child.out = -1;
4826        if (start_command(&child)) {
4827                remove_tempfile();
4828                return NULL;
4829        }
4830
4831        if (strbuf_read(&buf, child.out, 0) < 0)
4832                err = error("error reading from textconv command '%s'", pgm);
4833        close(child.out);
4834
4835        if (finish_command(&child) || err) {
4836                strbuf_release(&buf);
4837                remove_tempfile();
4838                return NULL;
4839        }
4840        remove_tempfile();
4841
4842        return strbuf_detach(&buf, outsize);
4843}
4844
4845size_t fill_textconv(struct userdiff_driver *driver,
4846                     struct diff_filespec *df,
4847                     char **outbuf)
4848{
4849        size_t size;
4850
4851        if (!driver || !driver->textconv) {
4852                if (!DIFF_FILE_VALID(df)) {
4853                        *outbuf = "";
4854                        return 0;
4855                }
4856                if (diff_populate_filespec(df, 0))
4857                        die("unable to read files to diff");
4858                *outbuf = df->data;
4859                return df->size;
4860        }
4861
4862        if (driver->textconv_cache && df->sha1_valid) {
4863                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4864                                          &size);
4865                if (*outbuf)
4866                        return size;
4867        }
4868
4869        *outbuf = run_textconv(driver->textconv, df, &size);
4870        if (!*outbuf)
4871                die("unable to read files to diff");
4872
4873        if (driver->textconv_cache && df->sha1_valid) {
4874                /* ignore errors, as we might be in a readonly repository */
4875                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4876                                size);
4877                /*
4878                 * we could save up changes and flush them all at the end,
4879                 * but we would need an extra call after all diffing is done.
4880                 * Since generating a cache entry is the slow path anyway,
4881                 * this extra overhead probably isn't a big deal.
4882                 */
4883                notes_cache_write(driver->textconv_cache);
4884        }
4885
4886        return size;
4887}
4888
4889void setup_diff_pager(struct diff_options *opt)
4890{
4891        /*
4892         * If the user asked for our exit code, then either they want --quiet
4893         * or --exit-code. We should definitely not bother with a pager in the
4894         * former case, as we will generate no output. Since we still properly
4895         * report our exit code even when a pager is run, we _could_ run a
4896         * pager with --exit-code. But since we have not done so historically,
4897         * and because it is easy to find people oneline advising "git diff
4898         * --exit-code" in hooks and other scripts, we do not do so.
4899         */
4900        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4901            check_pager_config("diff") != 0)
4902                setup_pager();
4903}