diff.con commit Merge branch 'db/make-with-curl' (e425521)
   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;
  33static const char *diff_order_file_cfg;
  34int diff_auto_refresh_index = 1;
  35static int diff_mnemonic_prefix;
  36static int diff_no_prefix;
  37static int diff_stat_graph_width;
  38static int diff_dirstat_permille_default = 30;
  39static struct diff_options default_diff_options;
  40static long diff_algorithm;
  41
  42static char diff_colors[][COLOR_MAXLEN] = {
  43        GIT_COLOR_RESET,
  44        GIT_COLOR_NORMAL,       /* PLAIN */
  45        GIT_COLOR_BOLD,         /* METAINFO */
  46        GIT_COLOR_CYAN,         /* FRAGINFO */
  47        GIT_COLOR_RED,          /* OLD */
  48        GIT_COLOR_GREEN,        /* NEW */
  49        GIT_COLOR_YELLOW,       /* COMMIT */
  50        GIT_COLOR_BG_RED,       /* WHITESPACE */
  51        GIT_COLOR_NORMAL,       /* FUNCINFO */
  52};
  53
  54static int parse_diff_color_slot(const char *var, int ofs)
  55{
  56        if (!strcasecmp(var+ofs, "plain"))
  57                return DIFF_PLAIN;
  58        if (!strcasecmp(var+ofs, "meta"))
  59                return DIFF_METAINFO;
  60        if (!strcasecmp(var+ofs, "frag"))
  61                return DIFF_FRAGINFO;
  62        if (!strcasecmp(var+ofs, "old"))
  63                return DIFF_FILE_OLD;
  64        if (!strcasecmp(var+ofs, "new"))
  65                return DIFF_FILE_NEW;
  66        if (!strcasecmp(var+ofs, "commit"))
  67                return DIFF_COMMIT;
  68        if (!strcasecmp(var+ofs, "whitespace"))
  69                return DIFF_WHITESPACE;
  70        if (!strcasecmp(var+ofs, "func"))
  71                return DIFF_FUNCINFO;
  72        return -1;
  73}
  74
  75static int parse_dirstat_params(struct diff_options *options, const char *params_string,
  76                                struct strbuf *errmsg)
  77{
  78        char *params_copy = xstrdup(params_string);
  79        struct string_list params = STRING_LIST_INIT_NODUP;
  80        int ret = 0;
  81        int i;
  82
  83        if (*params_copy)
  84                string_list_split_in_place(&params, params_copy, ',', -1);
  85        for (i = 0; i < params.nr; i++) {
  86                const char *p = params.items[i].string;
  87                if (!strcmp(p, "changes")) {
  88                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  89                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  90                } else if (!strcmp(p, "lines")) {
  91                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  92                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  93                } else if (!strcmp(p, "files")) {
  94                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  95                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  96                } else if (!strcmp(p, "noncumulative")) {
  97                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
  98                } else if (!strcmp(p, "cumulative")) {
  99                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
 100                } else if (isdigit(*p)) {
 101                        char *end;
 102                        int permille = strtoul(p, &end, 10) * 10;
 103                        if (*end == '.' && isdigit(*++end)) {
 104                                /* only use first digit */
 105                                permille += *end - '0';
 106                                /* .. and ignore any further digits */
 107                                while (isdigit(*++end))
 108                                        ; /* nothing */
 109                        }
 110                        if (!*end)
 111                                options->dirstat_permille = permille;
 112                        else {
 113                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%s'\n"),
 114                                            p);
 115                                ret++;
 116                        }
 117                } else {
 118                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%s'\n"), p);
 119                        ret++;
 120                }
 121
 122        }
 123        string_list_clear(&params, 0);
 124        free(params_copy);
 125        return ret;
 126}
 127
 128static int parse_submodule_params(struct diff_options *options, const char *value)
 129{
 130        if (!strcmp(value, "log"))
 131                DIFF_OPT_SET(options, SUBMODULE_LOG);
 132        else if (!strcmp(value, "short"))
 133                DIFF_OPT_CLR(options, SUBMODULE_LOG);
 134        else
 135                return -1;
 136        return 0;
 137}
 138
 139static int git_config_rename(const char *var, const char *value)
 140{
 141        if (!value)
 142                return DIFF_DETECT_RENAME;
 143        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 144                return  DIFF_DETECT_COPY;
 145        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 146}
 147
 148long parse_algorithm_value(const char *value)
 149{
 150        if (!value)
 151                return -1;
 152        else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
 153                return 0;
 154        else if (!strcasecmp(value, "minimal"))
 155                return XDF_NEED_MINIMAL;
 156        else if (!strcasecmp(value, "patience"))
 157                return XDF_PATIENCE_DIFF;
 158        else if (!strcasecmp(value, "histogram"))
 159                return XDF_HISTOGRAM_DIFF;
 160        return -1;
 161}
 162
 163/*
 164 * These are to give UI layer defaults.
 165 * The core-level commands such as git-diff-files should
 166 * never be affected by the setting of diff.renames
 167 * the user happens to have in the configuration file.
 168 */
 169int git_diff_ui_config(const char *var, const char *value, void *cb)
 170{
 171        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 172                diff_use_color_default = git_config_colorbool(var, value);
 173                return 0;
 174        }
 175        if (!strcmp(var, "diff.context")) {
 176                diff_context_default = git_config_int(var, value);
 177                if (diff_context_default < 0)
 178                        return -1;
 179                return 0;
 180        }
 181        if (!strcmp(var, "diff.renames")) {
 182                diff_detect_rename_default = git_config_rename(var, value);
 183                return 0;
 184        }
 185        if (!strcmp(var, "diff.autorefreshindex")) {
 186                diff_auto_refresh_index = git_config_bool(var, value);
 187                return 0;
 188        }
 189        if (!strcmp(var, "diff.mnemonicprefix")) {
 190                diff_mnemonic_prefix = git_config_bool(var, value);
 191                return 0;
 192        }
 193        if (!strcmp(var, "diff.noprefix")) {
 194                diff_no_prefix = git_config_bool(var, value);
 195                return 0;
 196        }
 197        if (!strcmp(var, "diff.statgraphwidth")) {
 198                diff_stat_graph_width = git_config_int(var, value);
 199                return 0;
 200        }
 201        if (!strcmp(var, "diff.external"))
 202                return git_config_string(&external_diff_cmd_cfg, var, value);
 203        if (!strcmp(var, "diff.wordregex"))
 204                return git_config_string(&diff_word_regex_cfg, var, value);
 205        if (!strcmp(var, "diff.orderfile"))
 206                return git_config_pathname(&diff_order_file_cfg, var, value);
 207
 208        if (!strcmp(var, "diff.ignoresubmodules"))
 209                handle_ignore_submodules_arg(&default_diff_options, value);
 210
 211        if (!strcmp(var, "diff.submodule")) {
 212                if (parse_submodule_params(&default_diff_options, value))
 213                        warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
 214                                value);
 215                return 0;
 216        }
 217
 218        if (!strcmp(var, "diff.algorithm")) {
 219                diff_algorithm = parse_algorithm_value(value);
 220                if (diff_algorithm < 0)
 221                        return -1;
 222                return 0;
 223        }
 224
 225        if (git_color_config(var, value, cb) < 0)
 226                return -1;
 227
 228        return git_diff_basic_config(var, value, cb);
 229}
 230
 231int git_diff_basic_config(const char *var, const char *value, void *cb)
 232{
 233        if (!strcmp(var, "diff.renamelimit")) {
 234                diff_rename_limit_default = git_config_int(var, value);
 235                return 0;
 236        }
 237
 238        if (userdiff_config(var, value) < 0)
 239                return -1;
 240
 241        if (starts_with(var, "diff.color.") || starts_with(var, "color.diff.")) {
 242                int slot = parse_diff_color_slot(var, 11);
 243                if (slot < 0)
 244                        return 0;
 245                if (!value)
 246                        return config_error_nonbool(var);
 247                color_parse(value, var, diff_colors[slot]);
 248                return 0;
 249        }
 250
 251        /* like GNU diff's --suppress-blank-empty option  */
 252        if (!strcmp(var, "diff.suppressblankempty") ||
 253                        /* for backwards compatibility */
 254                        !strcmp(var, "diff.suppress-blank-empty")) {
 255                diff_suppress_blank_empty = git_config_bool(var, value);
 256                return 0;
 257        }
 258
 259        if (!strcmp(var, "diff.dirstat")) {
 260                struct strbuf errmsg = STRBUF_INIT;
 261                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 262                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 263                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 264                                errmsg.buf);
 265                strbuf_release(&errmsg);
 266                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 267                return 0;
 268        }
 269
 270        if (starts_with(var, "submodule."))
 271                return parse_submodule_config_option(var, value);
 272
 273        return git_default_config(var, value, cb);
 274}
 275
 276static char *quote_two(const char *one, const char *two)
 277{
 278        int need_one = quote_c_style(one, NULL, NULL, 1);
 279        int need_two = quote_c_style(two, NULL, NULL, 1);
 280        struct strbuf res = STRBUF_INIT;
 281
 282        if (need_one + need_two) {
 283                strbuf_addch(&res, '"');
 284                quote_c_style(one, &res, NULL, 1);
 285                quote_c_style(two, &res, NULL, 1);
 286                strbuf_addch(&res, '"');
 287        } else {
 288                strbuf_addstr(&res, one);
 289                strbuf_addstr(&res, two);
 290        }
 291        return strbuf_detach(&res, NULL);
 292}
 293
 294static const char *external_diff(void)
 295{
 296        static const char *external_diff_cmd = NULL;
 297        static int done_preparing = 0;
 298
 299        if (done_preparing)
 300                return external_diff_cmd;
 301        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 302        if (!external_diff_cmd)
 303                external_diff_cmd = external_diff_cmd_cfg;
 304        done_preparing = 1;
 305        return external_diff_cmd;
 306}
 307
 308static struct diff_tempfile {
 309        const char *name; /* filename external diff should read from */
 310        char hex[41];
 311        char mode[10];
 312        char tmp_path[PATH_MAX];
 313} diff_temp[2];
 314
 315typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 316
 317struct emit_callback {
 318        int color_diff;
 319        unsigned ws_rule;
 320        int blank_at_eof_in_preimage;
 321        int blank_at_eof_in_postimage;
 322        int lno_in_preimage;
 323        int lno_in_postimage;
 324        sane_truncate_fn truncate;
 325        const char **label_path;
 326        struct diff_words_data *diff_words;
 327        struct diff_options *opt;
 328        int *found_changesp;
 329        struct strbuf *header;
 330};
 331
 332static int count_lines(const char *data, int size)
 333{
 334        int count, ch, completely_empty = 1, nl_just_seen = 0;
 335        count = 0;
 336        while (0 < size--) {
 337                ch = *data++;
 338                if (ch == '\n') {
 339                        count++;
 340                        nl_just_seen = 1;
 341                        completely_empty = 0;
 342                }
 343                else {
 344                        nl_just_seen = 0;
 345                        completely_empty = 0;
 346                }
 347        }
 348        if (completely_empty)
 349                return 0;
 350        if (!nl_just_seen)
 351                count++; /* no trailing newline */
 352        return count;
 353}
 354
 355static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 356{
 357        if (!DIFF_FILE_VALID(one)) {
 358                mf->ptr = (char *)""; /* does not matter */
 359                mf->size = 0;
 360                return 0;
 361        }
 362        else if (diff_populate_filespec(one, 0))
 363                return -1;
 364
 365        mf->ptr = one->data;
 366        mf->size = one->size;
 367        return 0;
 368}
 369
 370/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 371static unsigned long diff_filespec_size(struct diff_filespec *one)
 372{
 373        if (!DIFF_FILE_VALID(one))
 374                return 0;
 375        diff_populate_filespec(one, 1);
 376        return one->size;
 377}
 378
 379static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 380{
 381        char *ptr = mf->ptr;
 382        long size = mf->size;
 383        int cnt = 0;
 384
 385        if (!size)
 386                return cnt;
 387        ptr += size - 1; /* pointing at the very end */
 388        if (*ptr != '\n')
 389                ; /* incomplete line */
 390        else
 391                ptr--; /* skip the last LF */
 392        while (mf->ptr < ptr) {
 393                char *prev_eol;
 394                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 395                        if (*prev_eol == '\n')
 396                                break;
 397                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 398                        break;
 399                cnt++;
 400                ptr = prev_eol - 1;
 401        }
 402        return cnt;
 403}
 404
 405static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 406                               struct emit_callback *ecbdata)
 407{
 408        int l1, l2, at;
 409        unsigned ws_rule = ecbdata->ws_rule;
 410        l1 = count_trailing_blank(mf1, ws_rule);
 411        l2 = count_trailing_blank(mf2, ws_rule);
 412        if (l2 <= l1) {
 413                ecbdata->blank_at_eof_in_preimage = 0;
 414                ecbdata->blank_at_eof_in_postimage = 0;
 415                return;
 416        }
 417        at = count_lines(mf1->ptr, mf1->size);
 418        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 419
 420        at = count_lines(mf2->ptr, mf2->size);
 421        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 422}
 423
 424static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 425                        int first, const char *line, int len)
 426{
 427        int has_trailing_newline, has_trailing_carriage_return;
 428        int nofirst;
 429        FILE *file = o->file;
 430
 431        fputs(diff_line_prefix(o), file);
 432
 433        if (len == 0) {
 434                has_trailing_newline = (first == '\n');
 435                has_trailing_carriage_return = (!has_trailing_newline &&
 436                                                (first == '\r'));
 437                nofirst = has_trailing_newline || has_trailing_carriage_return;
 438        } else {
 439                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 440                if (has_trailing_newline)
 441                        len--;
 442                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 443                if (has_trailing_carriage_return)
 444                        len--;
 445                nofirst = 0;
 446        }
 447
 448        if (len || !nofirst) {
 449                fputs(set, file);
 450                if (!nofirst)
 451                        fputc(first, file);
 452                fwrite(line, len, 1, file);
 453                fputs(reset, file);
 454        }
 455        if (has_trailing_carriage_return)
 456                fputc('\r', file);
 457        if (has_trailing_newline)
 458                fputc('\n', file);
 459}
 460
 461static void emit_line(struct diff_options *o, const char *set, const char *reset,
 462                      const char *line, int len)
 463{
 464        emit_line_0(o, set, reset, line[0], line+1, len-1);
 465}
 466
 467static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 468{
 469        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 470              ecbdata->blank_at_eof_in_preimage &&
 471              ecbdata->blank_at_eof_in_postimage &&
 472              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 473              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 474                return 0;
 475        return ws_blank_line(line, len, ecbdata->ws_rule);
 476}
 477
 478static void emit_add_line(const char *reset,
 479                          struct emit_callback *ecbdata,
 480                          const char *line, int len)
 481{
 482        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 483        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 484
 485        if (!*ws)
 486                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 487        else if (new_blank_line_at_eof(ecbdata, line, len))
 488                /* Blank line at EOF - paint '+' as well */
 489                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 490        else {
 491                /* Emit just the prefix, then the rest. */
 492                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 493                ws_check_emit(line, len, ecbdata->ws_rule,
 494                              ecbdata->opt->file, set, reset, ws);
 495        }
 496}
 497
 498static void emit_hunk_header(struct emit_callback *ecbdata,
 499                             const char *line, int len)
 500{
 501        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 502        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 503        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 504        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 505        static const char atat[2] = { '@', '@' };
 506        const char *cp, *ep;
 507        struct strbuf msgbuf = STRBUF_INIT;
 508        int org_len = len;
 509        int i = 1;
 510
 511        /*
 512         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 513         * it always is at least 10 bytes long.
 514         */
 515        if (len < 10 ||
 516            memcmp(line, atat, 2) ||
 517            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 518                emit_line(ecbdata->opt, plain, reset, line, len);
 519                return;
 520        }
 521        ep += 2; /* skip over @@ */
 522
 523        /* The hunk header in fraginfo color */
 524        strbuf_add(&msgbuf, frag, strlen(frag));
 525        strbuf_add(&msgbuf, line, ep - line);
 526        strbuf_add(&msgbuf, reset, strlen(reset));
 527
 528        /*
 529         * trailing "\r\n"
 530         */
 531        for ( ; i < 3; i++)
 532                if (line[len - i] == '\r' || line[len - i] == '\n')
 533                        len--;
 534
 535        /* blank before the func header */
 536        for (cp = ep; ep - line < len; ep++)
 537                if (*ep != ' ' && *ep != '\t')
 538                        break;
 539        if (ep != cp) {
 540                strbuf_add(&msgbuf, plain, strlen(plain));
 541                strbuf_add(&msgbuf, cp, ep - cp);
 542                strbuf_add(&msgbuf, reset, strlen(reset));
 543        }
 544
 545        if (ep < line + len) {
 546                strbuf_add(&msgbuf, func, strlen(func));
 547                strbuf_add(&msgbuf, ep, line + len - ep);
 548                strbuf_add(&msgbuf, reset, strlen(reset));
 549        }
 550
 551        strbuf_add(&msgbuf, line + len, org_len - len);
 552        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 553        strbuf_release(&msgbuf);
 554}
 555
 556static struct diff_tempfile *claim_diff_tempfile(void) {
 557        int i;
 558        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 559                if (!diff_temp[i].name)
 560                        return diff_temp + i;
 561        die("BUG: diff is failing to clean up its tempfiles");
 562}
 563
 564static int remove_tempfile_installed;
 565
 566static void remove_tempfile(void)
 567{
 568        int i;
 569        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 570                if (diff_temp[i].name == diff_temp[i].tmp_path)
 571                        unlink_or_warn(diff_temp[i].name);
 572                diff_temp[i].name = NULL;
 573        }
 574}
 575
 576static void remove_tempfile_on_signal(int signo)
 577{
 578        remove_tempfile();
 579        sigchain_pop(signo);
 580        raise(signo);
 581}
 582
 583static void print_line_count(FILE *file, int count)
 584{
 585        switch (count) {
 586        case 0:
 587                fprintf(file, "0,0");
 588                break;
 589        case 1:
 590                fprintf(file, "1");
 591                break;
 592        default:
 593                fprintf(file, "1,%d", count);
 594                break;
 595        }
 596}
 597
 598static void emit_rewrite_lines(struct emit_callback *ecb,
 599                               int prefix, const char *data, int size)
 600{
 601        const char *endp = NULL;
 602        static const char *nneof = " No newline at end of file\n";
 603        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 604        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 605
 606        while (0 < size) {
 607                int len;
 608
 609                endp = memchr(data, '\n', size);
 610                len = endp ? (endp - data + 1) : size;
 611                if (prefix != '+') {
 612                        ecb->lno_in_preimage++;
 613                        emit_line_0(ecb->opt, old, reset, '-',
 614                                    data, len);
 615                } else {
 616                        ecb->lno_in_postimage++;
 617                        emit_add_line(reset, ecb, data, len);
 618                }
 619                size -= len;
 620                data += len;
 621        }
 622        if (!endp) {
 623                const char *plain = diff_get_color(ecb->color_diff,
 624                                                   DIFF_PLAIN);
 625                putc('\n', ecb->opt->file);
 626                emit_line_0(ecb->opt, plain, reset, '\\',
 627                            nneof, strlen(nneof));
 628        }
 629}
 630
 631static void emit_rewrite_diff(const char *name_a,
 632                              const char *name_b,
 633                              struct diff_filespec *one,
 634                              struct diff_filespec *two,
 635                              struct userdiff_driver *textconv_one,
 636                              struct userdiff_driver *textconv_two,
 637                              struct diff_options *o)
 638{
 639        int lc_a, lc_b;
 640        const char *name_a_tab, *name_b_tab;
 641        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 642        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 643        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 644        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 645        const char *a_prefix, *b_prefix;
 646        char *data_one, *data_two;
 647        size_t size_one, size_two;
 648        struct emit_callback ecbdata;
 649        const char *line_prefix = diff_line_prefix(o);
 650
 651        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 652                a_prefix = o->b_prefix;
 653                b_prefix = o->a_prefix;
 654        } else {
 655                a_prefix = o->a_prefix;
 656                b_prefix = o->b_prefix;
 657        }
 658
 659        name_a += (*name_a == '/');
 660        name_b += (*name_b == '/');
 661        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 662        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 663
 664        strbuf_reset(&a_name);
 665        strbuf_reset(&b_name);
 666        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 667        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 668
 669        size_one = fill_textconv(textconv_one, one, &data_one);
 670        size_two = fill_textconv(textconv_two, two, &data_two);
 671
 672        memset(&ecbdata, 0, sizeof(ecbdata));
 673        ecbdata.color_diff = want_color(o->use_color);
 674        ecbdata.found_changesp = &o->found_changes;
 675        ecbdata.ws_rule = whitespace_rule(name_b);
 676        ecbdata.opt = o;
 677        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 678                mmfile_t mf1, mf2;
 679                mf1.ptr = (char *)data_one;
 680                mf2.ptr = (char *)data_two;
 681                mf1.size = size_one;
 682                mf2.size = size_two;
 683                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 684        }
 685        ecbdata.lno_in_preimage = 1;
 686        ecbdata.lno_in_postimage = 1;
 687
 688        lc_a = count_lines(data_one, size_one);
 689        lc_b = count_lines(data_two, size_two);
 690        fprintf(o->file,
 691                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 692                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 693                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 694                line_prefix, fraginfo);
 695        if (!o->irreversible_delete)
 696                print_line_count(o->file, lc_a);
 697        else
 698                fprintf(o->file, "?,?");
 699        fprintf(o->file, " +");
 700        print_line_count(o->file, lc_b);
 701        fprintf(o->file, " @@%s\n", reset);
 702        if (lc_a && !o->irreversible_delete)
 703                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 704        if (lc_b)
 705                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 706        if (textconv_one)
 707                free((char *)data_one);
 708        if (textconv_two)
 709                free((char *)data_two);
 710}
 711
 712struct diff_words_buffer {
 713        mmfile_t text;
 714        long alloc;
 715        struct diff_words_orig {
 716                const char *begin, *end;
 717        } *orig;
 718        int orig_nr, orig_alloc;
 719};
 720
 721static void diff_words_append(char *line, unsigned long len,
 722                struct diff_words_buffer *buffer)
 723{
 724        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 725        line++;
 726        len--;
 727        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 728        buffer->text.size += len;
 729        buffer->text.ptr[buffer->text.size] = '\0';
 730}
 731
 732struct diff_words_style_elem {
 733        const char *prefix;
 734        const char *suffix;
 735        const char *color; /* NULL; filled in by the setup code if
 736                            * color is enabled */
 737};
 738
 739struct diff_words_style {
 740        enum diff_words_type type;
 741        struct diff_words_style_elem new, old, ctx;
 742        const char *newline;
 743};
 744
 745static struct diff_words_style diff_words_styles[] = {
 746        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 747        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 748        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 749};
 750
 751struct diff_words_data {
 752        struct diff_words_buffer minus, plus;
 753        const char *current_plus;
 754        int last_minus;
 755        struct diff_options *opt;
 756        regex_t *word_regex;
 757        enum diff_words_type type;
 758        struct diff_words_style *style;
 759};
 760
 761static int fn_out_diff_words_write_helper(FILE *fp,
 762                                          struct diff_words_style_elem *st_el,
 763                                          const char *newline,
 764                                          size_t count, const char *buf,
 765                                          const char *line_prefix)
 766{
 767        int print = 0;
 768
 769        while (count) {
 770                char *p = memchr(buf, '\n', count);
 771                if (print)
 772                        fputs(line_prefix, fp);
 773                if (p != buf) {
 774                        if (st_el->color && fputs(st_el->color, fp) < 0)
 775                                return -1;
 776                        if (fputs(st_el->prefix, fp) < 0 ||
 777                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 778                            fputs(st_el->suffix, fp) < 0)
 779                                return -1;
 780                        if (st_el->color && *st_el->color
 781                            && fputs(GIT_COLOR_RESET, fp) < 0)
 782                                return -1;
 783                }
 784                if (!p)
 785                        return 0;
 786                if (fputs(newline, fp) < 0)
 787                        return -1;
 788                count -= p + 1 - buf;
 789                buf = p + 1;
 790                print = 1;
 791        }
 792        return 0;
 793}
 794
 795/*
 796 * '--color-words' algorithm can be described as:
 797 *
 798 *   1. collect a the minus/plus lines of a diff hunk, divided into
 799 *      minus-lines and plus-lines;
 800 *
 801 *   2. break both minus-lines and plus-lines into words and
 802 *      place them into two mmfile_t with one word for each line;
 803 *
 804 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 805 *
 806 * And for the common parts of the both file, we output the plus side text.
 807 * diff_words->current_plus is used to trace the current position of the plus file
 808 * which printed. diff_words->last_minus is used to trace the last minus word
 809 * printed.
 810 *
 811 * For '--graph' to work with '--color-words', we need to output the graph prefix
 812 * on each line of color words output. Generally, there are two conditions on
 813 * which we should output the prefix.
 814 *
 815 *   1. diff_words->last_minus == 0 &&
 816 *      diff_words->current_plus == diff_words->plus.text.ptr
 817 *
 818 *      that is: the plus text must start as a new line, and if there is no minus
 819 *      word printed, a graph prefix must be printed.
 820 *
 821 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 822 *      *(diff_words->current_plus - 1) == '\n'
 823 *
 824 *      that is: a graph prefix must be printed following a '\n'
 825 */
 826static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 827{
 828        if ((diff_words->last_minus == 0 &&
 829                diff_words->current_plus == diff_words->plus.text.ptr) ||
 830                (diff_words->current_plus > diff_words->plus.text.ptr &&
 831                *(diff_words->current_plus - 1) == '\n')) {
 832                return 1;
 833        } else {
 834                return 0;
 835        }
 836}
 837
 838static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 839{
 840        struct diff_words_data *diff_words = priv;
 841        struct diff_words_style *style = diff_words->style;
 842        int minus_first, minus_len, plus_first, plus_len;
 843        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 844        struct diff_options *opt = diff_words->opt;
 845        const char *line_prefix;
 846
 847        if (line[0] != '@' || parse_hunk_header(line, len,
 848                        &minus_first, &minus_len, &plus_first, &plus_len))
 849                return;
 850
 851        assert(opt);
 852        line_prefix = diff_line_prefix(opt);
 853
 854        /* POSIX requires that first be decremented by one if len == 0... */
 855        if (minus_len) {
 856                minus_begin = diff_words->minus.orig[minus_first].begin;
 857                minus_end =
 858                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 859        } else
 860                minus_begin = minus_end =
 861                        diff_words->minus.orig[minus_first].end;
 862
 863        if (plus_len) {
 864                plus_begin = diff_words->plus.orig[plus_first].begin;
 865                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 866        } else
 867                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 868
 869        if (color_words_output_graph_prefix(diff_words)) {
 870                fputs(line_prefix, diff_words->opt->file);
 871        }
 872        if (diff_words->current_plus != plus_begin) {
 873                fn_out_diff_words_write_helper(diff_words->opt->file,
 874                                &style->ctx, style->newline,
 875                                plus_begin - diff_words->current_plus,
 876                                diff_words->current_plus, line_prefix);
 877                if (*(plus_begin - 1) == '\n')
 878                        fputs(line_prefix, diff_words->opt->file);
 879        }
 880        if (minus_begin != minus_end) {
 881                fn_out_diff_words_write_helper(diff_words->opt->file,
 882                                &style->old, style->newline,
 883                                minus_end - minus_begin, minus_begin,
 884                                line_prefix);
 885        }
 886        if (plus_begin != plus_end) {
 887                fn_out_diff_words_write_helper(diff_words->opt->file,
 888                                &style->new, style->newline,
 889                                plus_end - plus_begin, plus_begin,
 890                                line_prefix);
 891        }
 892
 893        diff_words->current_plus = plus_end;
 894        diff_words->last_minus = minus_first;
 895}
 896
 897/* This function starts looking at *begin, and returns 0 iff a word was found. */
 898static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 899                int *begin, int *end)
 900{
 901        if (word_regex && *begin < buffer->size) {
 902                regmatch_t match[1];
 903                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 904                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 905                                        '\n', match[0].rm_eo - match[0].rm_so);
 906                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 907                        *begin += match[0].rm_so;
 908                        return *begin >= *end;
 909                }
 910                return -1;
 911        }
 912
 913        /* find the next word */
 914        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 915                (*begin)++;
 916        if (*begin >= buffer->size)
 917                return -1;
 918
 919        /* find the end of the word */
 920        *end = *begin + 1;
 921        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 922                (*end)++;
 923
 924        return 0;
 925}
 926
 927/*
 928 * This function splits the words in buffer->text, stores the list with
 929 * newline separator into out, and saves the offsets of the original words
 930 * in buffer->orig.
 931 */
 932static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 933                regex_t *word_regex)
 934{
 935        int i, j;
 936        long alloc = 0;
 937
 938        out->size = 0;
 939        out->ptr = NULL;
 940
 941        /* fake an empty "0th" word */
 942        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 943        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 944        buffer->orig_nr = 1;
 945
 946        for (i = 0; i < buffer->text.size; i++) {
 947                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 948                        return;
 949
 950                /* store original boundaries */
 951                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 952                                buffer->orig_alloc);
 953                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 954                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 955                buffer->orig_nr++;
 956
 957                /* store one word */
 958                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 959                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 960                out->ptr[out->size + j - i] = '\n';
 961                out->size += j - i + 1;
 962
 963                i = j - 1;
 964        }
 965}
 966
 967/* this executes the word diff on the accumulated buffers */
 968static void diff_words_show(struct diff_words_data *diff_words)
 969{
 970        xpparam_t xpp;
 971        xdemitconf_t xecfg;
 972        mmfile_t minus, plus;
 973        struct diff_words_style *style = diff_words->style;
 974
 975        struct diff_options *opt = diff_words->opt;
 976        const char *line_prefix;
 977
 978        assert(opt);
 979        line_prefix = diff_line_prefix(opt);
 980
 981        /* special case: only removal */
 982        if (!diff_words->plus.text.size) {
 983                fputs(line_prefix, diff_words->opt->file);
 984                fn_out_diff_words_write_helper(diff_words->opt->file,
 985                        &style->old, style->newline,
 986                        diff_words->minus.text.size,
 987                        diff_words->minus.text.ptr, line_prefix);
 988                diff_words->minus.text.size = 0;
 989                return;
 990        }
 991
 992        diff_words->current_plus = diff_words->plus.text.ptr;
 993        diff_words->last_minus = 0;
 994
 995        memset(&xpp, 0, sizeof(xpp));
 996        memset(&xecfg, 0, sizeof(xecfg));
 997        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 998        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 999        xpp.flags = 0;
1000        /* as only the hunk header will be parsed, we need a 0-context */
1001        xecfg.ctxlen = 0;
1002        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1003                      &xpp, &xecfg);
1004        free(minus.ptr);
1005        free(plus.ptr);
1006        if (diff_words->current_plus != diff_words->plus.text.ptr +
1007                        diff_words->plus.text.size) {
1008                if (color_words_output_graph_prefix(diff_words))
1009                        fputs(line_prefix, diff_words->opt->file);
1010                fn_out_diff_words_write_helper(diff_words->opt->file,
1011                        &style->ctx, style->newline,
1012                        diff_words->plus.text.ptr + diff_words->plus.text.size
1013                        - diff_words->current_plus, diff_words->current_plus,
1014                        line_prefix);
1015        }
1016        diff_words->minus.text.size = diff_words->plus.text.size = 0;
1017}
1018
1019/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1020static void diff_words_flush(struct emit_callback *ecbdata)
1021{
1022        if (ecbdata->diff_words->minus.text.size ||
1023            ecbdata->diff_words->plus.text.size)
1024                diff_words_show(ecbdata->diff_words);
1025}
1026
1027static void diff_filespec_load_driver(struct diff_filespec *one)
1028{
1029        /* Use already-loaded driver */
1030        if (one->driver)
1031                return;
1032
1033        if (S_ISREG(one->mode))
1034                one->driver = userdiff_find_by_path(one->path);
1035
1036        /* Fallback to default settings */
1037        if (!one->driver)
1038                one->driver = userdiff_find_by_name("default");
1039}
1040
1041static const char *userdiff_word_regex(struct diff_filespec *one)
1042{
1043        diff_filespec_load_driver(one);
1044        return one->driver->word_regex;
1045}
1046
1047static void init_diff_words_data(struct emit_callback *ecbdata,
1048                                 struct diff_options *orig_opts,
1049                                 struct diff_filespec *one,
1050                                 struct diff_filespec *two)
1051{
1052        int i;
1053        struct diff_options *o = xmalloc(sizeof(struct diff_options));
1054        memcpy(o, orig_opts, sizeof(struct diff_options));
1055
1056        ecbdata->diff_words =
1057                xcalloc(1, sizeof(struct diff_words_data));
1058        ecbdata->diff_words->type = o->word_diff;
1059        ecbdata->diff_words->opt = o;
1060        if (!o->word_regex)
1061                o->word_regex = userdiff_word_regex(one);
1062        if (!o->word_regex)
1063                o->word_regex = userdiff_word_regex(two);
1064        if (!o->word_regex)
1065                o->word_regex = diff_word_regex_cfg;
1066        if (o->word_regex) {
1067                ecbdata->diff_words->word_regex = (regex_t *)
1068                        xmalloc(sizeof(regex_t));
1069                if (regcomp(ecbdata->diff_words->word_regex,
1070                            o->word_regex,
1071                            REG_EXTENDED | REG_NEWLINE))
1072                        die ("Invalid regular expression: %s",
1073                             o->word_regex);
1074        }
1075        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1076                if (o->word_diff == diff_words_styles[i].type) {
1077                        ecbdata->diff_words->style =
1078                                &diff_words_styles[i];
1079                        break;
1080                }
1081        }
1082        if (want_color(o->use_color)) {
1083                struct diff_words_style *st = ecbdata->diff_words->style;
1084                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1085                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1086                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1087        }
1088}
1089
1090static void free_diff_words_data(struct emit_callback *ecbdata)
1091{
1092        if (ecbdata->diff_words) {
1093                diff_words_flush(ecbdata);
1094                free (ecbdata->diff_words->opt);
1095                free (ecbdata->diff_words->minus.text.ptr);
1096                free (ecbdata->diff_words->minus.orig);
1097                free (ecbdata->diff_words->plus.text.ptr);
1098                free (ecbdata->diff_words->plus.orig);
1099                if (ecbdata->diff_words->word_regex) {
1100                        regfree(ecbdata->diff_words->word_regex);
1101                        free(ecbdata->diff_words->word_regex);
1102                }
1103                free(ecbdata->diff_words);
1104                ecbdata->diff_words = NULL;
1105        }
1106}
1107
1108const char *diff_get_color(int diff_use_color, enum color_diff ix)
1109{
1110        if (want_color(diff_use_color))
1111                return diff_colors[ix];
1112        return "";
1113}
1114
1115const char *diff_line_prefix(struct diff_options *opt)
1116{
1117        struct strbuf *msgbuf;
1118        if (!opt->output_prefix)
1119                return "";
1120
1121        msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1122        return msgbuf->buf;
1123}
1124
1125static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1126{
1127        const char *cp;
1128        unsigned long allot;
1129        size_t l = len;
1130
1131        if (ecb->truncate)
1132                return ecb->truncate(line, len);
1133        cp = line;
1134        allot = l;
1135        while (0 < l) {
1136                (void) utf8_width(&cp, &l);
1137                if (!cp)
1138                        break; /* truncated in the middle? */
1139        }
1140        return allot - l;
1141}
1142
1143static void find_lno(const char *line, struct emit_callback *ecbdata)
1144{
1145        const char *p;
1146        ecbdata->lno_in_preimage = 0;
1147        ecbdata->lno_in_postimage = 0;
1148        p = strchr(line, '-');
1149        if (!p)
1150                return; /* cannot happen */
1151        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1152        p = strchr(p, '+');
1153        if (!p)
1154                return; /* cannot happen */
1155        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1156}
1157
1158static void fn_out_consume(void *priv, char *line, unsigned long len)
1159{
1160        struct emit_callback *ecbdata = priv;
1161        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1162        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1163        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1164        struct diff_options *o = ecbdata->opt;
1165        const char *line_prefix = diff_line_prefix(o);
1166
1167        if (ecbdata->header) {
1168                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1169                strbuf_reset(ecbdata->header);
1170                ecbdata->header = NULL;
1171        }
1172        *(ecbdata->found_changesp) = 1;
1173
1174        if (ecbdata->label_path[0]) {
1175                const char *name_a_tab, *name_b_tab;
1176
1177                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1178                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1179
1180                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1181                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1182                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1183                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1184                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1185        }
1186
1187        if (diff_suppress_blank_empty
1188            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1189                line[0] = '\n';
1190                len = 1;
1191        }
1192
1193        if (line[0] == '@') {
1194                if (ecbdata->diff_words)
1195                        diff_words_flush(ecbdata);
1196                len = sane_truncate_line(ecbdata, line, len);
1197                find_lno(line, ecbdata);
1198                emit_hunk_header(ecbdata, line, len);
1199                if (line[len-1] != '\n')
1200                        putc('\n', ecbdata->opt->file);
1201                return;
1202        }
1203
1204        if (len < 1) {
1205                emit_line(ecbdata->opt, reset, reset, line, len);
1206                if (ecbdata->diff_words
1207                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1208                        fputs("~\n", ecbdata->opt->file);
1209                return;
1210        }
1211
1212        if (ecbdata->diff_words) {
1213                if (line[0] == '-') {
1214                        diff_words_append(line, len,
1215                                          &ecbdata->diff_words->minus);
1216                        return;
1217                } else if (line[0] == '+') {
1218                        diff_words_append(line, len,
1219                                          &ecbdata->diff_words->plus);
1220                        return;
1221                } else if (starts_with(line, "\\ ")) {
1222                        /*
1223                         * Eat the "no newline at eof" marker as if we
1224                         * saw a "+" or "-" line with nothing on it,
1225                         * and return without diff_words_flush() to
1226                         * defer processing. If this is the end of
1227                         * preimage, more "+" lines may come after it.
1228                         */
1229                        return;
1230                }
1231                diff_words_flush(ecbdata);
1232                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1233                        emit_line(ecbdata->opt, plain, reset, line, len);
1234                        fputs("~\n", ecbdata->opt->file);
1235                } else {
1236                        /*
1237                         * Skip the prefix character, if any.  With
1238                         * diff_suppress_blank_empty, there may be
1239                         * none.
1240                         */
1241                        if (line[0] != '\n') {
1242                              line++;
1243                              len--;
1244                        }
1245                        emit_line(ecbdata->opt, plain, reset, line, len);
1246                }
1247                return;
1248        }
1249
1250        if (line[0] != '+') {
1251                const char *color =
1252                        diff_get_color(ecbdata->color_diff,
1253                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1254                ecbdata->lno_in_preimage++;
1255                if (line[0] == ' ')
1256                        ecbdata->lno_in_postimage++;
1257                emit_line(ecbdata->opt, color, reset, line, len);
1258        } else {
1259                ecbdata->lno_in_postimage++;
1260                emit_add_line(reset, ecbdata, line + 1, len - 1);
1261        }
1262}
1263
1264static char *pprint_rename(const char *a, const char *b)
1265{
1266        const char *old = a;
1267        const char *new = b;
1268        struct strbuf name = STRBUF_INIT;
1269        int pfx_length, sfx_length;
1270        int pfx_adjust_for_slash;
1271        int len_a = strlen(a);
1272        int len_b = strlen(b);
1273        int a_midlen, b_midlen;
1274        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1275        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1276
1277        if (qlen_a || qlen_b) {
1278                quote_c_style(a, &name, NULL, 0);
1279                strbuf_addstr(&name, " => ");
1280                quote_c_style(b, &name, NULL, 0);
1281                return strbuf_detach(&name, NULL);
1282        }
1283
1284        /* Find common prefix */
1285        pfx_length = 0;
1286        while (*old && *new && *old == *new) {
1287                if (*old == '/')
1288                        pfx_length = old - a + 1;
1289                old++;
1290                new++;
1291        }
1292
1293        /* Find common suffix */
1294        old = a + len_a;
1295        new = b + len_b;
1296        sfx_length = 0;
1297        /*
1298         * If there is a common prefix, it must end in a slash.  In
1299         * that case we let this loop run 1 into the prefix to see the
1300         * same slash.
1301         *
1302         * If there is no common prefix, we cannot do this as it would
1303         * underrun the input strings.
1304         */
1305        pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1306        while (a + pfx_length - pfx_adjust_for_slash <= old &&
1307               b + pfx_length - pfx_adjust_for_slash <= new &&
1308               *old == *new) {
1309                if (*old == '/')
1310                        sfx_length = len_a - (old - a);
1311                old--;
1312                new--;
1313        }
1314
1315        /*
1316         * pfx{mid-a => mid-b}sfx
1317         * {pfx-a => pfx-b}sfx
1318         * pfx{sfx-a => sfx-b}
1319         * name-a => name-b
1320         */
1321        a_midlen = len_a - pfx_length - sfx_length;
1322        b_midlen = len_b - pfx_length - sfx_length;
1323        if (a_midlen < 0)
1324                a_midlen = 0;
1325        if (b_midlen < 0)
1326                b_midlen = 0;
1327
1328        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1329        if (pfx_length + sfx_length) {
1330                strbuf_add(&name, a, pfx_length);
1331                strbuf_addch(&name, '{');
1332        }
1333        strbuf_add(&name, a + pfx_length, a_midlen);
1334        strbuf_addstr(&name, " => ");
1335        strbuf_add(&name, b + pfx_length, b_midlen);
1336        if (pfx_length + sfx_length) {
1337                strbuf_addch(&name, '}');
1338                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1339        }
1340        return strbuf_detach(&name, NULL);
1341}
1342
1343struct diffstat_t {
1344        int nr;
1345        int alloc;
1346        struct diffstat_file {
1347                char *from_name;
1348                char *name;
1349                char *print_name;
1350                unsigned is_unmerged:1;
1351                unsigned is_binary:1;
1352                unsigned is_renamed:1;
1353                unsigned is_interesting:1;
1354                uintmax_t added, deleted;
1355        } **files;
1356};
1357
1358static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1359                                          const char *name_a,
1360                                          const char *name_b)
1361{
1362        struct diffstat_file *x;
1363        x = xcalloc(sizeof (*x), 1);
1364        ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1365        diffstat->files[diffstat->nr++] = x;
1366        if (name_b) {
1367                x->from_name = xstrdup(name_a);
1368                x->name = xstrdup(name_b);
1369                x->is_renamed = 1;
1370        }
1371        else {
1372                x->from_name = NULL;
1373                x->name = xstrdup(name_a);
1374        }
1375        return x;
1376}
1377
1378static void diffstat_consume(void *priv, char *line, unsigned long len)
1379{
1380        struct diffstat_t *diffstat = priv;
1381        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1382
1383        if (line[0] == '+')
1384                x->added++;
1385        else if (line[0] == '-')
1386                x->deleted++;
1387}
1388
1389const char mime_boundary_leader[] = "------------";
1390
1391static int scale_linear(int it, int width, int max_change)
1392{
1393        if (!it)
1394                return 0;
1395        /*
1396         * make sure that at least one '-' or '+' is printed if
1397         * there is any change to this path. The easiest way is to
1398         * scale linearly as if the alloted width is one column shorter
1399         * than it is, and then add 1 to the result.
1400         */
1401        return 1 + (it * (width - 1) / max_change);
1402}
1403
1404static void show_name(FILE *file,
1405                      const char *prefix, const char *name, int len)
1406{
1407        fprintf(file, " %s%-*s |", prefix, len, name);
1408}
1409
1410static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1411{
1412        if (cnt <= 0)
1413                return;
1414        fprintf(file, "%s", set);
1415        while (cnt--)
1416                putc(ch, file);
1417        fprintf(file, "%s", reset);
1418}
1419
1420static void fill_print_name(struct diffstat_file *file)
1421{
1422        char *pname;
1423
1424        if (file->print_name)
1425                return;
1426
1427        if (!file->is_renamed) {
1428                struct strbuf buf = STRBUF_INIT;
1429                if (quote_c_style(file->name, &buf, NULL, 0)) {
1430                        pname = strbuf_detach(&buf, NULL);
1431                } else {
1432                        pname = file->name;
1433                        strbuf_release(&buf);
1434                }
1435        } else {
1436                pname = pprint_rename(file->from_name, file->name);
1437        }
1438        file->print_name = pname;
1439}
1440
1441int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1442{
1443        struct strbuf sb = STRBUF_INIT;
1444        int ret;
1445
1446        if (!files) {
1447                assert(insertions == 0 && deletions == 0);
1448                return fprintf(fp, "%s\n", " 0 files changed");
1449        }
1450
1451        strbuf_addf(&sb,
1452                    (files == 1) ? " %d file changed" : " %d files changed",
1453                    files);
1454
1455        /*
1456         * For binary diff, the caller may want to print "x files
1457         * changed" with insertions == 0 && deletions == 0.
1458         *
1459         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1460         * is probably less confusing (i.e skip over "2 files changed
1461         * but nothing about added/removed lines? Is this a bug in Git?").
1462         */
1463        if (insertions || deletions == 0) {
1464                strbuf_addf(&sb,
1465                            (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1466                            insertions);
1467        }
1468
1469        if (deletions || insertions == 0) {
1470                strbuf_addf(&sb,
1471                            (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1472                            deletions);
1473        }
1474        strbuf_addch(&sb, '\n');
1475        ret = fputs(sb.buf, fp);
1476        strbuf_release(&sb);
1477        return ret;
1478}
1479
1480static void show_stats(struct diffstat_t *data, struct diff_options *options)
1481{
1482        int i, len, add, del, adds = 0, dels = 0;
1483        uintmax_t max_change = 0, max_len = 0;
1484        int total_files = data->nr, count;
1485        int width, name_width, graph_width, number_width = 0, bin_width = 0;
1486        const char *reset, *add_c, *del_c;
1487        const char *line_prefix = "";
1488        int extra_shown = 0;
1489
1490        if (data->nr == 0)
1491                return;
1492
1493        line_prefix = diff_line_prefix(options);
1494        count = options->stat_count ? options->stat_count : data->nr;
1495
1496        reset = diff_get_color_opt(options, DIFF_RESET);
1497        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1498        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1499
1500        /*
1501         * Find the longest filename and max number of changes
1502         */
1503        for (i = 0; (i < count) && (i < data->nr); i++) {
1504                struct diffstat_file *file = data->files[i];
1505                uintmax_t change = file->added + file->deleted;
1506
1507                if (!file->is_interesting && (change == 0)) {
1508                        count++; /* not shown == room for one more */
1509                        continue;
1510                }
1511                fill_print_name(file);
1512                len = strlen(file->print_name);
1513                if (max_len < len)
1514                        max_len = len;
1515
1516                if (file->is_unmerged) {
1517                        /* "Unmerged" is 8 characters */
1518                        bin_width = bin_width < 8 ? 8 : bin_width;
1519                        continue;
1520                }
1521                if (file->is_binary) {
1522                        /* "Bin XXX -> YYY bytes" */
1523                        int w = 14 + decimal_width(file->added)
1524                                + decimal_width(file->deleted);
1525                        bin_width = bin_width < w ? w : bin_width;
1526                        /* Display change counts aligned with "Bin" */
1527                        number_width = 3;
1528                        continue;
1529                }
1530
1531                if (max_change < change)
1532                        max_change = change;
1533        }
1534        count = i; /* where we can stop scanning in data->files[] */
1535
1536        /*
1537         * We have width = stat_width or term_columns() columns total.
1538         * We want a maximum of min(max_len, stat_name_width) for the name part.
1539         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1540         * We also need 1 for " " and 4 + decimal_width(max_change)
1541         * for " | NNNN " and one the empty column at the end, altogether
1542         * 6 + decimal_width(max_change).
1543         *
1544         * If there's not enough space, we will use the smaller of
1545         * stat_name_width (if set) and 5/8*width for the filename,
1546         * and the rest for constant elements + graph part, but no more
1547         * than stat_graph_width for the graph part.
1548         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1549         * for the standard terminal size).
1550         *
1551         * In other words: stat_width limits the maximum width, and
1552         * stat_name_width fixes the maximum width of the filename,
1553         * and is also used to divide available columns if there
1554         * aren't enough.
1555         *
1556         * Binary files are displayed with "Bin XXX -> YYY bytes"
1557         * instead of the change count and graph. This part is treated
1558         * similarly to the graph part, except that it is not
1559         * "scaled". If total width is too small to accommodate the
1560         * guaranteed minimum width of the filename part and the
1561         * separators and this message, this message will "overflow"
1562         * making the line longer than the maximum width.
1563         */
1564
1565        if (options->stat_width == -1)
1566                width = term_columns() - options->output_prefix_length;
1567        else
1568                width = options->stat_width ? options->stat_width : 80;
1569        number_width = decimal_width(max_change) > number_width ?
1570                decimal_width(max_change) : number_width;
1571
1572        if (options->stat_graph_width == -1)
1573                options->stat_graph_width = diff_stat_graph_width;
1574
1575        /*
1576         * Guarantee 3/8*16==6 for the graph part
1577         * and 5/8*16==10 for the filename part
1578         */
1579        if (width < 16 + 6 + number_width)
1580                width = 16 + 6 + number_width;
1581
1582        /*
1583         * First assign sizes that are wanted, ignoring available width.
1584         * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1585         * starting from "XXX" should fit in graph_width.
1586         */
1587        graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1588        if (options->stat_graph_width &&
1589            options->stat_graph_width < graph_width)
1590                graph_width = options->stat_graph_width;
1591
1592        name_width = (options->stat_name_width > 0 &&
1593                      options->stat_name_width < max_len) ?
1594                options->stat_name_width : max_len;
1595
1596        /*
1597         * Adjust adjustable widths not to exceed maximum width
1598         */
1599        if (name_width + number_width + 6 + graph_width > width) {
1600                if (graph_width > width * 3/8 - number_width - 6) {
1601                        graph_width = width * 3/8 - number_width - 6;
1602                        if (graph_width < 6)
1603                                graph_width = 6;
1604                }
1605
1606                if (options->stat_graph_width &&
1607                    graph_width > options->stat_graph_width)
1608                        graph_width = options->stat_graph_width;
1609                if (name_width > width - number_width - 6 - graph_width)
1610                        name_width = width - number_width - 6 - graph_width;
1611                else
1612                        graph_width = width - number_width - 6 - name_width;
1613        }
1614
1615        /*
1616         * From here name_width is the width of the name area,
1617         * and graph_width is the width of the graph area.
1618         * max_change is used to scale graph properly.
1619         */
1620        for (i = 0; i < count; i++) {
1621                const char *prefix = "";
1622                struct diffstat_file *file = data->files[i];
1623                char *name = file->print_name;
1624                uintmax_t added = file->added;
1625                uintmax_t deleted = file->deleted;
1626                int name_len;
1627
1628                if (!file->is_interesting && (added + deleted == 0))
1629                        continue;
1630
1631                /*
1632                 * "scale" the filename
1633                 */
1634                len = name_width;
1635                name_len = strlen(name);
1636                if (name_width < name_len) {
1637                        char *slash;
1638                        prefix = "...";
1639                        len -= 3;
1640                        name += name_len - len;
1641                        slash = strchr(name, '/');
1642                        if (slash)
1643                                name = slash;
1644                }
1645
1646                if (file->is_binary) {
1647                        fprintf(options->file, "%s", line_prefix);
1648                        show_name(options->file, prefix, name, len);
1649                        fprintf(options->file, " %*s", number_width, "Bin");
1650                        if (!added && !deleted) {
1651                                putc('\n', options->file);
1652                                continue;
1653                        }
1654                        fprintf(options->file, " %s%"PRIuMAX"%s",
1655                                del_c, deleted, reset);
1656                        fprintf(options->file, " -> ");
1657                        fprintf(options->file, "%s%"PRIuMAX"%s",
1658                                add_c, added, reset);
1659                        fprintf(options->file, " bytes");
1660                        fprintf(options->file, "\n");
1661                        continue;
1662                }
1663                else if (file->is_unmerged) {
1664                        fprintf(options->file, "%s", line_prefix);
1665                        show_name(options->file, prefix, name, len);
1666                        fprintf(options->file, " Unmerged\n");
1667                        continue;
1668                }
1669
1670                /*
1671                 * scale the add/delete
1672                 */
1673                add = added;
1674                del = deleted;
1675
1676                if (graph_width <= max_change) {
1677                        int total = scale_linear(add + del, graph_width, max_change);
1678                        if (total < 2 && add && del)
1679                                /* width >= 2 due to the sanity check */
1680                                total = 2;
1681                        if (add < del) {
1682                                add = scale_linear(add, graph_width, max_change);
1683                                del = total - add;
1684                        } else {
1685                                del = scale_linear(del, graph_width, max_change);
1686                                add = total - del;
1687                        }
1688                }
1689                fprintf(options->file, "%s", line_prefix);
1690                show_name(options->file, prefix, name, len);
1691                fprintf(options->file, " %*"PRIuMAX"%s",
1692                        number_width, added + deleted,
1693                        added + deleted ? " " : "");
1694                show_graph(options->file, '+', add, add_c, reset);
1695                show_graph(options->file, '-', del, del_c, reset);
1696                fprintf(options->file, "\n");
1697        }
1698
1699        for (i = 0; i < data->nr; i++) {
1700                struct diffstat_file *file = data->files[i];
1701                uintmax_t added = file->added;
1702                uintmax_t deleted = file->deleted;
1703
1704                if (file->is_unmerged ||
1705                    (!file->is_interesting && (added + deleted == 0))) {
1706                        total_files--;
1707                        continue;
1708                }
1709
1710                if (!file->is_binary) {
1711                        adds += added;
1712                        dels += deleted;
1713                }
1714                if (i < count)
1715                        continue;
1716                if (!extra_shown)
1717                        fprintf(options->file, "%s ...\n", line_prefix);
1718                extra_shown = 1;
1719        }
1720        fprintf(options->file, "%s", line_prefix);
1721        print_stat_summary(options->file, total_files, adds, dels);
1722}
1723
1724static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1725{
1726        int i, adds = 0, dels = 0, total_files = data->nr;
1727
1728        if (data->nr == 0)
1729                return;
1730
1731        for (i = 0; i < data->nr; i++) {
1732                int added = data->files[i]->added;
1733                int deleted= data->files[i]->deleted;
1734
1735                if (data->files[i]->is_unmerged ||
1736                    (!data->files[i]->is_interesting && (added + deleted == 0))) {
1737                        total_files--;
1738                } else if (!data->files[i]->is_binary) { /* don't count bytes */
1739                        adds += added;
1740                        dels += deleted;
1741                }
1742        }
1743        fprintf(options->file, "%s", diff_line_prefix(options));
1744        print_stat_summary(options->file, total_files, adds, dels);
1745}
1746
1747static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1748{
1749        int i;
1750
1751        if (data->nr == 0)
1752                return;
1753
1754        for (i = 0; i < data->nr; i++) {
1755                struct diffstat_file *file = data->files[i];
1756
1757                fprintf(options->file, "%s", diff_line_prefix(options));
1758
1759                if (file->is_binary)
1760                        fprintf(options->file, "-\t-\t");
1761                else
1762                        fprintf(options->file,
1763                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1764                                file->added, file->deleted);
1765                if (options->line_termination) {
1766                        fill_print_name(file);
1767                        if (!file->is_renamed)
1768                                write_name_quoted(file->name, options->file,
1769                                                  options->line_termination);
1770                        else {
1771                                fputs(file->print_name, options->file);
1772                                putc(options->line_termination, options->file);
1773                        }
1774                } else {
1775                        if (file->is_renamed) {
1776                                putc('\0', options->file);
1777                                write_name_quoted(file->from_name, options->file, '\0');
1778                        }
1779                        write_name_quoted(file->name, options->file, '\0');
1780                }
1781        }
1782}
1783
1784struct dirstat_file {
1785        const char *name;
1786        unsigned long changed;
1787};
1788
1789struct dirstat_dir {
1790        struct dirstat_file *files;
1791        int alloc, nr, permille, cumulative;
1792};
1793
1794static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1795                unsigned long changed, const char *base, int baselen)
1796{
1797        unsigned long this_dir = 0;
1798        unsigned int sources = 0;
1799        const char *line_prefix = diff_line_prefix(opt);
1800
1801        while (dir->nr) {
1802                struct dirstat_file *f = dir->files;
1803                int namelen = strlen(f->name);
1804                unsigned long this;
1805                char *slash;
1806
1807                if (namelen < baselen)
1808                        break;
1809                if (memcmp(f->name, base, baselen))
1810                        break;
1811                slash = strchr(f->name + baselen, '/');
1812                if (slash) {
1813                        int newbaselen = slash + 1 - f->name;
1814                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1815                        sources++;
1816                } else {
1817                        this = f->changed;
1818                        dir->files++;
1819                        dir->nr--;
1820                        sources += 2;
1821                }
1822                this_dir += this;
1823        }
1824
1825        /*
1826         * We don't report dirstat's for
1827         *  - the top level
1828         *  - or cases where everything came from a single directory
1829         *    under this directory (sources == 1).
1830         */
1831        if (baselen && sources != 1) {
1832                if (this_dir) {
1833                        int permille = this_dir * 1000 / changed;
1834                        if (permille >= dir->permille) {
1835                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1836                                        permille / 10, permille % 10, baselen, base);
1837                                if (!dir->cumulative)
1838                                        return 0;
1839                        }
1840                }
1841        }
1842        return this_dir;
1843}
1844
1845static int dirstat_compare(const void *_a, const void *_b)
1846{
1847        const struct dirstat_file *a = _a;
1848        const struct dirstat_file *b = _b;
1849        return strcmp(a->name, b->name);
1850}
1851
1852static void show_dirstat(struct diff_options *options)
1853{
1854        int i;
1855        unsigned long changed;
1856        struct dirstat_dir dir;
1857        struct diff_queue_struct *q = &diff_queued_diff;
1858
1859        dir.files = NULL;
1860        dir.alloc = 0;
1861        dir.nr = 0;
1862        dir.permille = options->dirstat_permille;
1863        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1864
1865        changed = 0;
1866        for (i = 0; i < q->nr; i++) {
1867                struct diff_filepair *p = q->queue[i];
1868                const char *name;
1869                unsigned long copied, added, damage;
1870                int content_changed;
1871
1872                name = p->two->path ? p->two->path : p->one->path;
1873
1874                if (p->one->sha1_valid && p->two->sha1_valid)
1875                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1876                else
1877                        content_changed = 1;
1878
1879                if (!content_changed) {
1880                        /*
1881                         * The SHA1 has not changed, so pre-/post-content is
1882                         * identical. We can therefore skip looking at the
1883                         * file contents altogether.
1884                         */
1885                        damage = 0;
1886                        goto found_damage;
1887                }
1888
1889                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1890                        /*
1891                         * In --dirstat-by-file mode, we don't really need to
1892                         * look at the actual file contents at all.
1893                         * The fact that the SHA1 changed is enough for us to
1894                         * add this file to the list of results
1895                         * (with each file contributing equal damage).
1896                         */
1897                        damage = 1;
1898                        goto found_damage;
1899                }
1900
1901                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1902                        diff_populate_filespec(p->one, 0);
1903                        diff_populate_filespec(p->two, 0);
1904                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1905                                               &copied, &added);
1906                        diff_free_filespec_data(p->one);
1907                        diff_free_filespec_data(p->two);
1908                } else if (DIFF_FILE_VALID(p->one)) {
1909                        diff_populate_filespec(p->one, 1);
1910                        copied = added = 0;
1911                        diff_free_filespec_data(p->one);
1912                } else if (DIFF_FILE_VALID(p->two)) {
1913                        diff_populate_filespec(p->two, 1);
1914                        copied = 0;
1915                        added = p->two->size;
1916                        diff_free_filespec_data(p->two);
1917                } else
1918                        continue;
1919
1920                /*
1921                 * Original minus copied is the removed material,
1922                 * added is the new material.  They are both damages
1923                 * made to the preimage.
1924                 * If the resulting damage is zero, we know that
1925                 * diffcore_count_changes() considers the two entries to
1926                 * be identical, but since content_changed is true, we
1927                 * know that there must have been _some_ kind of change,
1928                 * so we force all entries to have damage > 0.
1929                 */
1930                damage = (p->one->size - copied) + added;
1931                if (!damage)
1932                        damage = 1;
1933
1934found_damage:
1935                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1936                dir.files[dir.nr].name = name;
1937                dir.files[dir.nr].changed = damage;
1938                changed += damage;
1939                dir.nr++;
1940        }
1941
1942        /* This can happen even with many files, if everything was renames */
1943        if (!changed)
1944                return;
1945
1946        /* Show all directories with more than x% of the changes */
1947        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1948        gather_dirstat(options, &dir, changed, "", 0);
1949}
1950
1951static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1952{
1953        int i;
1954        unsigned long changed;
1955        struct dirstat_dir dir;
1956
1957        if (data->nr == 0)
1958                return;
1959
1960        dir.files = NULL;
1961        dir.alloc = 0;
1962        dir.nr = 0;
1963        dir.permille = options->dirstat_permille;
1964        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1965
1966        changed = 0;
1967        for (i = 0; i < data->nr; i++) {
1968                struct diffstat_file *file = data->files[i];
1969                unsigned long damage = file->added + file->deleted;
1970                if (file->is_binary)
1971                        /*
1972                         * binary files counts bytes, not lines. Must find some
1973                         * way to normalize binary bytes vs. textual lines.
1974                         * The following heuristic assumes that there are 64
1975                         * bytes per "line".
1976                         * This is stupid and ugly, but very cheap...
1977                         */
1978                        damage = (damage + 63) / 64;
1979                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1980                dir.files[dir.nr].name = file->name;
1981                dir.files[dir.nr].changed = damage;
1982                changed += damage;
1983                dir.nr++;
1984        }
1985
1986        /* This can happen even with many files, if everything was renames */
1987        if (!changed)
1988                return;
1989
1990        /* Show all directories with more than x% of the changes */
1991        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1992        gather_dirstat(options, &dir, changed, "", 0);
1993}
1994
1995static void free_diffstat_info(struct diffstat_t *diffstat)
1996{
1997        int i;
1998        for (i = 0; i < diffstat->nr; i++) {
1999                struct diffstat_file *f = diffstat->files[i];
2000                if (f->name != f->print_name)
2001                        free(f->print_name);
2002                free(f->name);
2003                free(f->from_name);
2004                free(f);
2005        }
2006        free(diffstat->files);
2007}
2008
2009struct checkdiff_t {
2010        const char *filename;
2011        int lineno;
2012        int conflict_marker_size;
2013        struct diff_options *o;
2014        unsigned ws_rule;
2015        unsigned status;
2016};
2017
2018static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2019{
2020        char firstchar;
2021        int cnt;
2022
2023        if (len < marker_size + 1)
2024                return 0;
2025        firstchar = line[0];
2026        switch (firstchar) {
2027        case '=': case '>': case '<': case '|':
2028                break;
2029        default:
2030                return 0;
2031        }
2032        for (cnt = 1; cnt < marker_size; cnt++)
2033                if (line[cnt] != firstchar)
2034                        return 0;
2035        /* line[1] thru line[marker_size-1] are same as firstchar */
2036        if (len < marker_size + 1 || !isspace(line[marker_size]))
2037                return 0;
2038        return 1;
2039}
2040
2041static void checkdiff_consume(void *priv, char *line, unsigned long len)
2042{
2043        struct checkdiff_t *data = priv;
2044        int marker_size = data->conflict_marker_size;
2045        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2046        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2047        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2048        char *err;
2049        const char *line_prefix;
2050
2051        assert(data->o);
2052        line_prefix = diff_line_prefix(data->o);
2053
2054        if (line[0] == '+') {
2055                unsigned bad;
2056                data->lineno++;
2057                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2058                        data->status |= 1;
2059                        fprintf(data->o->file,
2060                                "%s%s:%d: leftover conflict marker\n",
2061                                line_prefix, data->filename, data->lineno);
2062                }
2063                bad = ws_check(line + 1, len - 1, data->ws_rule);
2064                if (!bad)
2065                        return;
2066                data->status |= bad;
2067                err = whitespace_error_string(bad);
2068                fprintf(data->o->file, "%s%s:%d: %s.\n",
2069                        line_prefix, data->filename, data->lineno, err);
2070                free(err);
2071                emit_line(data->o, set, reset, line, 1);
2072                ws_check_emit(line + 1, len - 1, data->ws_rule,
2073                              data->o->file, set, reset, ws);
2074        } else if (line[0] == ' ') {
2075                data->lineno++;
2076        } else if (line[0] == '@') {
2077                char *plus = strchr(line, '+');
2078                if (plus)
2079                        data->lineno = strtol(plus, NULL, 10) - 1;
2080                else
2081                        die("invalid diff");
2082        }
2083}
2084
2085static unsigned char *deflate_it(char *data,
2086                                 unsigned long size,
2087                                 unsigned long *result_size)
2088{
2089        int bound;
2090        unsigned char *deflated;
2091        git_zstream stream;
2092
2093        memset(&stream, 0, sizeof(stream));
2094        git_deflate_init(&stream, zlib_compression_level);
2095        bound = git_deflate_bound(&stream, size);
2096        deflated = xmalloc(bound);
2097        stream.next_out = deflated;
2098        stream.avail_out = bound;
2099
2100        stream.next_in = (unsigned char *)data;
2101        stream.avail_in = size;
2102        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2103                ; /* nothing */
2104        git_deflate_end(&stream);
2105        *result_size = stream.total_out;
2106        return deflated;
2107}
2108
2109static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2110                                  const char *prefix)
2111{
2112        void *cp;
2113        void *delta;
2114        void *deflated;
2115        void *data;
2116        unsigned long orig_size;
2117        unsigned long delta_size;
2118        unsigned long deflate_size;
2119        unsigned long data_size;
2120
2121        /* We could do deflated delta, or we could do just deflated two,
2122         * whichever is smaller.
2123         */
2124        delta = NULL;
2125        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2126        if (one->size && two->size) {
2127                delta = diff_delta(one->ptr, one->size,
2128                                   two->ptr, two->size,
2129                                   &delta_size, deflate_size);
2130                if (delta) {
2131                        void *to_free = delta;
2132                        orig_size = delta_size;
2133                        delta = deflate_it(delta, delta_size, &delta_size);
2134                        free(to_free);
2135                }
2136        }
2137
2138        if (delta && delta_size < deflate_size) {
2139                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2140                free(deflated);
2141                data = delta;
2142                data_size = delta_size;
2143        }
2144        else {
2145                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2146                free(delta);
2147                data = deflated;
2148                data_size = deflate_size;
2149        }
2150
2151        /* emit data encoded in base85 */
2152        cp = data;
2153        while (data_size) {
2154                int bytes = (52 < data_size) ? 52 : data_size;
2155                char line[70];
2156                data_size -= bytes;
2157                if (bytes <= 26)
2158                        line[0] = bytes + 'A' - 1;
2159                else
2160                        line[0] = bytes - 26 + 'a' - 1;
2161                encode_85(line + 1, cp, bytes);
2162                cp = (char *) cp + bytes;
2163                fprintf(file, "%s", prefix);
2164                fputs(line, file);
2165                fputc('\n', file);
2166        }
2167        fprintf(file, "%s\n", prefix);
2168        free(data);
2169}
2170
2171static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2172                             const char *prefix)
2173{
2174        fprintf(file, "%sGIT binary patch\n", prefix);
2175        emit_binary_diff_body(file, one, two, prefix);
2176        emit_binary_diff_body(file, two, one, prefix);
2177}
2178
2179int diff_filespec_is_binary(struct diff_filespec *one)
2180{
2181        if (one->is_binary == -1) {
2182                diff_filespec_load_driver(one);
2183                if (one->driver->binary != -1)
2184                        one->is_binary = one->driver->binary;
2185                else {
2186                        if (!one->data && DIFF_FILE_VALID(one))
2187                                diff_populate_filespec(one, 0);
2188                        if (one->data)
2189                                one->is_binary = buffer_is_binary(one->data,
2190                                                one->size);
2191                        if (one->is_binary == -1)
2192                                one->is_binary = 0;
2193                }
2194        }
2195        return one->is_binary;
2196}
2197
2198static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2199{
2200        diff_filespec_load_driver(one);
2201        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2202}
2203
2204void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2205{
2206        if (!options->a_prefix)
2207                options->a_prefix = a;
2208        if (!options->b_prefix)
2209                options->b_prefix = b;
2210}
2211
2212struct userdiff_driver *get_textconv(struct diff_filespec *one)
2213{
2214        if (!DIFF_FILE_VALID(one))
2215                return NULL;
2216
2217        diff_filespec_load_driver(one);
2218        return userdiff_get_textconv(one->driver);
2219}
2220
2221static void builtin_diff(const char *name_a,
2222                         const char *name_b,
2223                         struct diff_filespec *one,
2224                         struct diff_filespec *two,
2225                         const char *xfrm_msg,
2226                         int must_show_header,
2227                         struct diff_options *o,
2228                         int complete_rewrite)
2229{
2230        mmfile_t mf1, mf2;
2231        const char *lbl[2];
2232        char *a_one, *b_two;
2233        const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2234        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2235        const char *a_prefix, *b_prefix;
2236        struct userdiff_driver *textconv_one = NULL;
2237        struct userdiff_driver *textconv_two = NULL;
2238        struct strbuf header = STRBUF_INIT;
2239        const char *line_prefix = diff_line_prefix(o);
2240
2241        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2242                        (!one->mode || S_ISGITLINK(one->mode)) &&
2243                        (!two->mode || S_ISGITLINK(two->mode))) {
2244                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2245                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2246                show_submodule_summary(o->file, one->path ? one->path : two->path,
2247                                line_prefix,
2248                                one->sha1, two->sha1, two->dirty_submodule,
2249                                meta, del, add, reset);
2250                return;
2251        }
2252
2253        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2254                textconv_one = get_textconv(one);
2255                textconv_two = get_textconv(two);
2256        }
2257
2258        diff_set_mnemonic_prefix(o, "a/", "b/");
2259        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2260                a_prefix = o->b_prefix;
2261                b_prefix = o->a_prefix;
2262        } else {
2263                a_prefix = o->a_prefix;
2264                b_prefix = o->b_prefix;
2265        }
2266
2267        /* Never use a non-valid filename anywhere if at all possible */
2268        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2269        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2270
2271        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2272        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2273        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2274        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2275        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2276        if (lbl[0][0] == '/') {
2277                /* /dev/null */
2278                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2279                if (xfrm_msg)
2280                        strbuf_addstr(&header, xfrm_msg);
2281                must_show_header = 1;
2282        }
2283        else if (lbl[1][0] == '/') {
2284                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2285                if (xfrm_msg)
2286                        strbuf_addstr(&header, xfrm_msg);
2287                must_show_header = 1;
2288        }
2289        else {
2290                if (one->mode != two->mode) {
2291                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2292                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2293                        must_show_header = 1;
2294                }
2295                if (xfrm_msg)
2296                        strbuf_addstr(&header, xfrm_msg);
2297
2298                /*
2299                 * we do not run diff between different kind
2300                 * of objects.
2301                 */
2302                if ((one->mode ^ two->mode) & S_IFMT)
2303                        goto free_ab_and_return;
2304                if (complete_rewrite &&
2305                    (textconv_one || !diff_filespec_is_binary(one)) &&
2306                    (textconv_two || !diff_filespec_is_binary(two))) {
2307                        fprintf(o->file, "%s", header.buf);
2308                        strbuf_reset(&header);
2309                        emit_rewrite_diff(name_a, name_b, one, two,
2310                                                textconv_one, textconv_two, o);
2311                        o->found_changes = 1;
2312                        goto free_ab_and_return;
2313                }
2314        }
2315
2316        if (o->irreversible_delete && lbl[1][0] == '/') {
2317                fprintf(o->file, "%s", header.buf);
2318                strbuf_reset(&header);
2319                goto free_ab_and_return;
2320        } else if (!DIFF_OPT_TST(o, TEXT) &&
2321            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2322              (!textconv_two && diff_filespec_is_binary(two)) )) {
2323                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2324                        die("unable to read files to diff");
2325                /* Quite common confusing case */
2326                if (mf1.size == mf2.size &&
2327                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2328                        if (must_show_header)
2329                                fprintf(o->file, "%s", header.buf);
2330                        goto free_ab_and_return;
2331                }
2332                fprintf(o->file, "%s", header.buf);
2333                strbuf_reset(&header);
2334                if (DIFF_OPT_TST(o, BINARY))
2335                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2336                else
2337                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2338                                line_prefix, lbl[0], lbl[1]);
2339                o->found_changes = 1;
2340        } else {
2341                /* Crazy xdl interfaces.. */
2342                const char *diffopts = getenv("GIT_DIFF_OPTS");
2343                xpparam_t xpp;
2344                xdemitconf_t xecfg;
2345                struct emit_callback ecbdata;
2346                const struct userdiff_funcname *pe;
2347
2348                if (must_show_header) {
2349                        fprintf(o->file, "%s", header.buf);
2350                        strbuf_reset(&header);
2351                }
2352
2353                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2354                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2355
2356                pe = diff_funcname_pattern(one);
2357                if (!pe)
2358                        pe = diff_funcname_pattern(two);
2359
2360                memset(&xpp, 0, sizeof(xpp));
2361                memset(&xecfg, 0, sizeof(xecfg));
2362                memset(&ecbdata, 0, sizeof(ecbdata));
2363                ecbdata.label_path = lbl;
2364                ecbdata.color_diff = want_color(o->use_color);
2365                ecbdata.found_changesp = &o->found_changes;
2366                ecbdata.ws_rule = whitespace_rule(name_b);
2367                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2368                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2369                ecbdata.opt = o;
2370                ecbdata.header = header.len ? &header : NULL;
2371                xpp.flags = o->xdl_opts;
2372                xecfg.ctxlen = o->context;
2373                xecfg.interhunkctxlen = o->interhunkcontext;
2374                xecfg.flags = XDL_EMIT_FUNCNAMES;
2375                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2376                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2377                if (pe)
2378                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2379                if (!diffopts)
2380                        ;
2381                else if (starts_with(diffopts, "--unified="))
2382                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2383                else if (starts_with(diffopts, "-u"))
2384                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2385                if (o->word_diff)
2386                        init_diff_words_data(&ecbdata, o, one, two);
2387                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2388                              &xpp, &xecfg);
2389                if (o->word_diff)
2390                        free_diff_words_data(&ecbdata);
2391                if (textconv_one)
2392                        free(mf1.ptr);
2393                if (textconv_two)
2394                        free(mf2.ptr);
2395                xdiff_clear_find_func(&xecfg);
2396        }
2397
2398 free_ab_and_return:
2399        strbuf_release(&header);
2400        diff_free_filespec_data(one);
2401        diff_free_filespec_data(two);
2402        free(a_one);
2403        free(b_two);
2404        return;
2405}
2406
2407static void builtin_diffstat(const char *name_a, const char *name_b,
2408                             struct diff_filespec *one,
2409                             struct diff_filespec *two,
2410                             struct diffstat_t *diffstat,
2411                             struct diff_options *o,
2412                             struct diff_filepair *p)
2413{
2414        mmfile_t mf1, mf2;
2415        struct diffstat_file *data;
2416        int same_contents;
2417        int complete_rewrite = 0;
2418
2419        if (!DIFF_PAIR_UNMERGED(p)) {
2420                if (p->status == DIFF_STATUS_MODIFIED && p->score)
2421                        complete_rewrite = 1;
2422        }
2423
2424        data = diffstat_add(diffstat, name_a, name_b);
2425        data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2426
2427        if (!one || !two) {
2428                data->is_unmerged = 1;
2429                return;
2430        }
2431
2432        same_contents = !hashcmp(one->sha1, two->sha1);
2433
2434        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2435                data->is_binary = 1;
2436                if (same_contents) {
2437                        data->added = 0;
2438                        data->deleted = 0;
2439                } else {
2440                        data->added = diff_filespec_size(two);
2441                        data->deleted = diff_filespec_size(one);
2442                }
2443        }
2444
2445        else if (complete_rewrite) {
2446                diff_populate_filespec(one, 0);
2447                diff_populate_filespec(two, 0);
2448                data->deleted = count_lines(one->data, one->size);
2449                data->added = count_lines(two->data, two->size);
2450        }
2451
2452        else if (!same_contents) {
2453                /* Crazy xdl interfaces.. */
2454                xpparam_t xpp;
2455                xdemitconf_t xecfg;
2456
2457                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2458                        die("unable to read files to diff");
2459
2460                memset(&xpp, 0, sizeof(xpp));
2461                memset(&xecfg, 0, sizeof(xecfg));
2462                xpp.flags = o->xdl_opts;
2463                xecfg.ctxlen = o->context;
2464                xecfg.interhunkctxlen = o->interhunkcontext;
2465                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2466                              &xpp, &xecfg);
2467        }
2468
2469        diff_free_filespec_data(one);
2470        diff_free_filespec_data(two);
2471}
2472
2473static void builtin_checkdiff(const char *name_a, const char *name_b,
2474                              const char *attr_path,
2475                              struct diff_filespec *one,
2476                              struct diff_filespec *two,
2477                              struct diff_options *o)
2478{
2479        mmfile_t mf1, mf2;
2480        struct checkdiff_t data;
2481
2482        if (!two)
2483                return;
2484
2485        memset(&data, 0, sizeof(data));
2486        data.filename = name_b ? name_b : name_a;
2487        data.lineno = 0;
2488        data.o = o;
2489        data.ws_rule = whitespace_rule(attr_path);
2490        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2491
2492        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2493                die("unable to read files to diff");
2494
2495        /*
2496         * All the other codepaths check both sides, but not checking
2497         * the "old" side here is deliberate.  We are checking the newly
2498         * introduced changes, and as long as the "new" side is text, we
2499         * can and should check what it introduces.
2500         */
2501        if (diff_filespec_is_binary(two))
2502                goto free_and_return;
2503        else {
2504                /* Crazy xdl interfaces.. */
2505                xpparam_t xpp;
2506                xdemitconf_t xecfg;
2507
2508                memset(&xpp, 0, sizeof(xpp));
2509                memset(&xecfg, 0, sizeof(xecfg));
2510                xecfg.ctxlen = 1; /* at least one context line */
2511                xpp.flags = 0;
2512                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2513                              &xpp, &xecfg);
2514
2515                if (data.ws_rule & WS_BLANK_AT_EOF) {
2516                        struct emit_callback ecbdata;
2517                        int blank_at_eof;
2518
2519                        ecbdata.ws_rule = data.ws_rule;
2520                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2521                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2522
2523                        if (blank_at_eof) {
2524                                static char *err;
2525                                if (!err)
2526                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2527                                fprintf(o->file, "%s:%d: %s.\n",
2528                                        data.filename, blank_at_eof, err);
2529                                data.status = 1; /* report errors */
2530                        }
2531                }
2532        }
2533 free_and_return:
2534        diff_free_filespec_data(one);
2535        diff_free_filespec_data(two);
2536        if (data.status)
2537                DIFF_OPT_SET(o, CHECK_FAILED);
2538}
2539
2540struct diff_filespec *alloc_filespec(const char *path)
2541{
2542        int namelen = strlen(path);
2543        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2544
2545        memset(spec, 0, sizeof(*spec));
2546        spec->path = (char *)(spec + 1);
2547        memcpy(spec->path, path, namelen+1);
2548        spec->count = 1;
2549        spec->is_binary = -1;
2550        return spec;
2551}
2552
2553void free_filespec(struct diff_filespec *spec)
2554{
2555        if (!--spec->count) {
2556                diff_free_filespec_data(spec);
2557                free(spec);
2558        }
2559}
2560
2561void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2562                   int sha1_valid, unsigned short mode)
2563{
2564        if (mode) {
2565                spec->mode = canon_mode(mode);
2566                hashcpy(spec->sha1, sha1);
2567                spec->sha1_valid = sha1_valid;
2568        }
2569}
2570
2571/*
2572 * Given a name and sha1 pair, if the index tells us the file in
2573 * the work tree has that object contents, return true, so that
2574 * prepare_temp_file() does not have to inflate and extract.
2575 */
2576static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2577{
2578        const struct cache_entry *ce;
2579        struct stat st;
2580        int pos, len;
2581
2582        /*
2583         * We do not read the cache ourselves here, because the
2584         * benchmark with my previous version that always reads cache
2585         * shows that it makes things worse for diff-tree comparing
2586         * two linux-2.6 kernel trees in an already checked out work
2587         * tree.  This is because most diff-tree comparisons deal with
2588         * only a small number of files, while reading the cache is
2589         * expensive for a large project, and its cost outweighs the
2590         * savings we get by not inflating the object to a temporary
2591         * file.  Practically, this code only helps when we are used
2592         * by diff-cache --cached, which does read the cache before
2593         * calling us.
2594         */
2595        if (!active_cache)
2596                return 0;
2597
2598        /* We want to avoid the working directory if our caller
2599         * doesn't need the data in a normal file, this system
2600         * is rather slow with its stat/open/mmap/close syscalls,
2601         * and the object is contained in a pack file.  The pack
2602         * is probably already open and will be faster to obtain
2603         * the data through than the working directory.  Loose
2604         * objects however would tend to be slower as they need
2605         * to be individually opened and inflated.
2606         */
2607        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2608                return 0;
2609
2610        len = strlen(name);
2611        pos = cache_name_pos(name, len);
2612        if (pos < 0)
2613                return 0;
2614        ce = active_cache[pos];
2615
2616        /*
2617         * This is not the sha1 we are looking for, or
2618         * unreusable because it is not a regular file.
2619         */
2620        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2621                return 0;
2622
2623        /*
2624         * If ce is marked as "assume unchanged", there is no
2625         * guarantee that work tree matches what we are looking for.
2626         */
2627        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2628                return 0;
2629
2630        /*
2631         * If ce matches the file in the work tree, we can reuse it.
2632         */
2633        if (ce_uptodate(ce) ||
2634            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2635                return 1;
2636
2637        return 0;
2638}
2639
2640static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2641{
2642        int len;
2643        char *data = xmalloc(100), *dirty = "";
2644
2645        /* Are we looking at the work tree? */
2646        if (s->dirty_submodule)
2647                dirty = "-dirty";
2648
2649        len = snprintf(data, 100,
2650                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2651        s->data = data;
2652        s->size = len;
2653        s->should_free = 1;
2654        if (size_only) {
2655                s->data = NULL;
2656                free(data);
2657        }
2658        return 0;
2659}
2660
2661/*
2662 * While doing rename detection and pickaxe operation, we may need to
2663 * grab the data for the blob (or file) for our own in-core comparison.
2664 * diff_filespec has data and size fields for this purpose.
2665 */
2666int diff_populate_filespec(struct diff_filespec *s, int size_only)
2667{
2668        int err = 0;
2669        /*
2670         * demote FAIL to WARN to allow inspecting the situation
2671         * instead of refusing.
2672         */
2673        enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2674                                    ? SAFE_CRLF_WARN
2675                                    : safe_crlf);
2676
2677        if (!DIFF_FILE_VALID(s))
2678                die("internal error: asking to populate invalid file.");
2679        if (S_ISDIR(s->mode))
2680                return -1;
2681
2682        if (s->data)
2683                return 0;
2684
2685        if (size_only && 0 < s->size)
2686                return 0;
2687
2688        if (S_ISGITLINK(s->mode))
2689                return diff_populate_gitlink(s, size_only);
2690
2691        if (!s->sha1_valid ||
2692            reuse_worktree_file(s->path, s->sha1, 0)) {
2693                struct strbuf buf = STRBUF_INIT;
2694                struct stat st;
2695                int fd;
2696
2697                if (lstat(s->path, &st) < 0) {
2698                        if (errno == ENOENT) {
2699                        err_empty:
2700                                err = -1;
2701                        empty:
2702                                s->data = (char *)"";
2703                                s->size = 0;
2704                                return err;
2705                        }
2706                }
2707                s->size = xsize_t(st.st_size);
2708                if (!s->size)
2709                        goto empty;
2710                if (S_ISLNK(st.st_mode)) {
2711                        struct strbuf sb = STRBUF_INIT;
2712
2713                        if (strbuf_readlink(&sb, s->path, s->size))
2714                                goto err_empty;
2715                        s->size = sb.len;
2716                        s->data = strbuf_detach(&sb, NULL);
2717                        s->should_free = 1;
2718                        return 0;
2719                }
2720                if (size_only)
2721                        return 0;
2722                fd = open(s->path, O_RDONLY);
2723                if (fd < 0)
2724                        goto err_empty;
2725                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2726                close(fd);
2727                s->should_munmap = 1;
2728
2729                /*
2730                 * Convert from working tree format to canonical git format
2731                 */
2732                if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2733                        size_t size = 0;
2734                        munmap(s->data, s->size);
2735                        s->should_munmap = 0;
2736                        s->data = strbuf_detach(&buf, &size);
2737                        s->size = size;
2738                        s->should_free = 1;
2739                }
2740        }
2741        else {
2742                enum object_type type;
2743                if (size_only) {
2744                        type = sha1_object_info(s->sha1, &s->size);
2745                        if (type < 0)
2746                                die("unable to read %s", sha1_to_hex(s->sha1));
2747                } else {
2748                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2749                        if (!s->data)
2750                                die("unable to read %s", sha1_to_hex(s->sha1));
2751                        s->should_free = 1;
2752                }
2753        }
2754        return 0;
2755}
2756
2757void diff_free_filespec_blob(struct diff_filespec *s)
2758{
2759        if (s->should_free)
2760                free(s->data);
2761        else if (s->should_munmap)
2762                munmap(s->data, s->size);
2763
2764        if (s->should_free || s->should_munmap) {
2765                s->should_free = s->should_munmap = 0;
2766                s->data = NULL;
2767        }
2768}
2769
2770void diff_free_filespec_data(struct diff_filespec *s)
2771{
2772        diff_free_filespec_blob(s);
2773        free(s->cnt_data);
2774        s->cnt_data = NULL;
2775}
2776
2777static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2778                           void *blob,
2779                           unsigned long size,
2780                           const unsigned char *sha1,
2781                           int mode)
2782{
2783        int fd;
2784        struct strbuf buf = STRBUF_INIT;
2785        struct strbuf template = STRBUF_INIT;
2786        char *path_dup = xstrdup(path);
2787        const char *base = basename(path_dup);
2788
2789        /* Generate "XXXXXX_basename.ext" */
2790        strbuf_addstr(&template, "XXXXXX_");
2791        strbuf_addstr(&template, base);
2792
2793        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2794                        strlen(base) + 1);
2795        if (fd < 0)
2796                die_errno("unable to create temp-file");
2797        if (convert_to_working_tree(path,
2798                        (const char *)blob, (size_t)size, &buf)) {
2799                blob = buf.buf;
2800                size = buf.len;
2801        }
2802        if (write_in_full(fd, blob, size) != size)
2803                die_errno("unable to write temp-file");
2804        close(fd);
2805        temp->name = temp->tmp_path;
2806        strcpy(temp->hex, sha1_to_hex(sha1));
2807        temp->hex[40] = 0;
2808        sprintf(temp->mode, "%06o", mode);
2809        strbuf_release(&buf);
2810        strbuf_release(&template);
2811        free(path_dup);
2812}
2813
2814static struct diff_tempfile *prepare_temp_file(const char *name,
2815                struct diff_filespec *one)
2816{
2817        struct diff_tempfile *temp = claim_diff_tempfile();
2818
2819        if (!DIFF_FILE_VALID(one)) {
2820        not_a_valid_file:
2821                /* A '-' entry produces this for file-2, and
2822                 * a '+' entry produces this for file-1.
2823                 */
2824                temp->name = "/dev/null";
2825                strcpy(temp->hex, ".");
2826                strcpy(temp->mode, ".");
2827                return temp;
2828        }
2829
2830        if (!remove_tempfile_installed) {
2831                atexit(remove_tempfile);
2832                sigchain_push_common(remove_tempfile_on_signal);
2833                remove_tempfile_installed = 1;
2834        }
2835
2836        if (!S_ISGITLINK(one->mode) &&
2837            (!one->sha1_valid ||
2838             reuse_worktree_file(name, one->sha1, 1))) {
2839                struct stat st;
2840                if (lstat(name, &st) < 0) {
2841                        if (errno == ENOENT)
2842                                goto not_a_valid_file;
2843                        die_errno("stat(%s)", name);
2844                }
2845                if (S_ISLNK(st.st_mode)) {
2846                        struct strbuf sb = STRBUF_INIT;
2847                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2848                                die_errno("readlink(%s)", name);
2849                        prep_temp_blob(name, temp, sb.buf, sb.len,
2850                                       (one->sha1_valid ?
2851                                        one->sha1 : null_sha1),
2852                                       (one->sha1_valid ?
2853                                        one->mode : S_IFLNK));
2854                        strbuf_release(&sb);
2855                }
2856                else {
2857                        /* we can borrow from the file in the work tree */
2858                        temp->name = name;
2859                        if (!one->sha1_valid)
2860                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2861                        else
2862                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2863                        /* Even though we may sometimes borrow the
2864                         * contents from the work tree, we always want
2865                         * one->mode.  mode is trustworthy even when
2866                         * !(one->sha1_valid), as long as
2867                         * DIFF_FILE_VALID(one).
2868                         */
2869                        sprintf(temp->mode, "%06o", one->mode);
2870                }
2871                return temp;
2872        }
2873        else {
2874                if (diff_populate_filespec(one, 0))
2875                        die("cannot read data blob for %s", one->path);
2876                prep_temp_blob(name, temp, one->data, one->size,
2877                               one->sha1, one->mode);
2878        }
2879        return temp;
2880}
2881
2882/* An external diff command takes:
2883 *
2884 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2885 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2886 *
2887 */
2888static void run_external_diff(const char *pgm,
2889                              const char *name,
2890                              const char *other,
2891                              struct diff_filespec *one,
2892                              struct diff_filespec *two,
2893                              const char *xfrm_msg,
2894                              int complete_rewrite,
2895                              struct diff_options *o)
2896{
2897        const char *spawn_arg[10];
2898        int retval;
2899        const char **arg = &spawn_arg[0];
2900        struct diff_queue_struct *q = &diff_queued_diff;
2901        const char *env[3] = { NULL };
2902        char env_counter[50];
2903        char env_total[50];
2904
2905        if (one && two) {
2906                struct diff_tempfile *temp_one, *temp_two;
2907                const char *othername = (other ? other : name);
2908                temp_one = prepare_temp_file(name, one);
2909                temp_two = prepare_temp_file(othername, two);
2910                *arg++ = pgm;
2911                *arg++ = name;
2912                *arg++ = temp_one->name;
2913                *arg++ = temp_one->hex;
2914                *arg++ = temp_one->mode;
2915                *arg++ = temp_two->name;
2916                *arg++ = temp_two->hex;
2917                *arg++ = temp_two->mode;
2918                if (other) {
2919                        *arg++ = other;
2920                        *arg++ = xfrm_msg;
2921                }
2922        } else {
2923                *arg++ = pgm;
2924                *arg++ = name;
2925        }
2926        *arg = NULL;
2927        fflush(NULL);
2928
2929        env[0] = env_counter;
2930        snprintf(env_counter, sizeof(env_counter), "GIT_DIFF_PATH_COUNTER=%d",
2931                 ++o->diff_path_counter);
2932        env[1] = env_total;
2933        snprintf(env_total, sizeof(env_total), "GIT_DIFF_PATH_TOTAL=%d", q->nr);
2934
2935        retval = run_command_v_opt_cd_env(spawn_arg, RUN_USING_SHELL, NULL, env);
2936        remove_tempfile();
2937        if (retval) {
2938                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2939                exit(1);
2940        }
2941}
2942
2943static int similarity_index(struct diff_filepair *p)
2944{
2945        return p->score * 100 / MAX_SCORE;
2946}
2947
2948static void fill_metainfo(struct strbuf *msg,
2949                          const char *name,
2950                          const char *other,
2951                          struct diff_filespec *one,
2952                          struct diff_filespec *two,
2953                          struct diff_options *o,
2954                          struct diff_filepair *p,
2955                          int *must_show_header,
2956                          int use_color)
2957{
2958        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2959        const char *reset = diff_get_color(use_color, DIFF_RESET);
2960        const char *line_prefix = diff_line_prefix(o);
2961
2962        *must_show_header = 1;
2963        strbuf_init(msg, PATH_MAX * 2 + 300);
2964        switch (p->status) {
2965        case DIFF_STATUS_COPIED:
2966                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2967                            line_prefix, set, similarity_index(p));
2968                strbuf_addf(msg, "%s\n%s%scopy from ",
2969                            reset,  line_prefix, set);
2970                quote_c_style(name, msg, NULL, 0);
2971                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2972                quote_c_style(other, msg, NULL, 0);
2973                strbuf_addf(msg, "%s\n", reset);
2974                break;
2975        case DIFF_STATUS_RENAMED:
2976                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2977                            line_prefix, set, similarity_index(p));
2978                strbuf_addf(msg, "%s\n%s%srename from ",
2979                            reset, line_prefix, set);
2980                quote_c_style(name, msg, NULL, 0);
2981                strbuf_addf(msg, "%s\n%s%srename to ",
2982                            reset, line_prefix, set);
2983                quote_c_style(other, msg, NULL, 0);
2984                strbuf_addf(msg, "%s\n", reset);
2985                break;
2986        case DIFF_STATUS_MODIFIED:
2987                if (p->score) {
2988                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2989                                    line_prefix,
2990                                    set, similarity_index(p), reset);
2991                        break;
2992                }
2993                /* fallthru */
2994        default:
2995                *must_show_header = 0;
2996        }
2997        if (one && two && hashcmp(one->sha1, two->sha1)) {
2998                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2999
3000                if (DIFF_OPT_TST(o, BINARY)) {
3001                        mmfile_t mf;
3002                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3003                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3004                                abbrev = 40;
3005                }
3006                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3007                            find_unique_abbrev(one->sha1, abbrev));
3008                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3009                if (one->mode == two->mode)
3010                        strbuf_addf(msg, " %06o", one->mode);
3011                strbuf_addf(msg, "%s\n", reset);
3012        }
3013}
3014
3015static void run_diff_cmd(const char *pgm,
3016                         const char *name,
3017                         const char *other,
3018                         const char *attr_path,
3019                         struct diff_filespec *one,
3020                         struct diff_filespec *two,
3021                         struct strbuf *msg,
3022                         struct diff_options *o,
3023                         struct diff_filepair *p)
3024{
3025        const char *xfrm_msg = NULL;
3026        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3027        int must_show_header = 0;
3028
3029
3030        if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3031                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3032                if (drv && drv->external)
3033                        pgm = drv->external;
3034        }
3035
3036        if (msg) {
3037                /*
3038                 * don't use colors when the header is intended for an
3039                 * external diff driver
3040                 */
3041                fill_metainfo(msg, name, other, one, two, o, p,
3042                              &must_show_header,
3043                              want_color(o->use_color) && !pgm);
3044                xfrm_msg = msg->len ? msg->buf : NULL;
3045        }
3046
3047        if (pgm) {
3048                run_external_diff(pgm, name, other, one, two, xfrm_msg,
3049                                  complete_rewrite, o);
3050                return;
3051        }
3052        if (one && two)
3053                builtin_diff(name, other ? other : name,
3054                             one, two, xfrm_msg, must_show_header,
3055                             o, complete_rewrite);
3056        else
3057                fprintf(o->file, "* Unmerged path %s\n", name);
3058}
3059
3060static void diff_fill_sha1_info(struct diff_filespec *one)
3061{
3062        if (DIFF_FILE_VALID(one)) {
3063                if (!one->sha1_valid) {
3064                        struct stat st;
3065                        if (one->is_stdin) {
3066                                hashcpy(one->sha1, null_sha1);
3067                                return;
3068                        }
3069                        if (lstat(one->path, &st) < 0)
3070                                die_errno("stat '%s'", one->path);
3071                        if (index_path(one->sha1, one->path, &st, 0))
3072                                die("cannot hash %s", one->path);
3073                }
3074        }
3075        else
3076                hashclr(one->sha1);
3077}
3078
3079static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3080{
3081        /* Strip the prefix but do not molest /dev/null and absolute paths */
3082        if (*namep && **namep != '/') {
3083                *namep += prefix_length;
3084                if (**namep == '/')
3085                        ++*namep;
3086        }
3087        if (*otherp && **otherp != '/') {
3088                *otherp += prefix_length;
3089                if (**otherp == '/')
3090                        ++*otherp;
3091        }
3092}
3093
3094static void run_diff(struct diff_filepair *p, struct diff_options *o)
3095{
3096        const char *pgm = external_diff();
3097        struct strbuf msg;
3098        struct diff_filespec *one = p->one;
3099        struct diff_filespec *two = p->two;
3100        const char *name;
3101        const char *other;
3102        const char *attr_path;
3103
3104        name  = p->one->path;
3105        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3106        attr_path = name;
3107        if (o->prefix_length)
3108                strip_prefix(o->prefix_length, &name, &other);
3109
3110        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3111                pgm = NULL;
3112
3113        if (DIFF_PAIR_UNMERGED(p)) {
3114                run_diff_cmd(pgm, name, NULL, attr_path,
3115                             NULL, NULL, NULL, o, p);
3116                return;
3117        }
3118
3119        diff_fill_sha1_info(one);
3120        diff_fill_sha1_info(two);
3121
3122        if (!pgm &&
3123            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3124            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3125                /*
3126                 * a filepair that changes between file and symlink
3127                 * needs to be split into deletion and creation.
3128                 */
3129                struct diff_filespec *null = alloc_filespec(two->path);
3130                run_diff_cmd(NULL, name, other, attr_path,
3131                             one, null, &msg, o, p);
3132                free(null);
3133                strbuf_release(&msg);
3134
3135                null = alloc_filespec(one->path);
3136                run_diff_cmd(NULL, name, other, attr_path,
3137                             null, two, &msg, o, p);
3138                free(null);
3139        }
3140        else
3141                run_diff_cmd(pgm, name, other, attr_path,
3142                             one, two, &msg, o, p);
3143
3144        strbuf_release(&msg);
3145}
3146
3147static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3148                         struct diffstat_t *diffstat)
3149{
3150        const char *name;
3151        const char *other;
3152
3153        if (DIFF_PAIR_UNMERGED(p)) {
3154                /* unmerged */
3155                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3156                return;
3157        }
3158
3159        name = p->one->path;
3160        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3161
3162        if (o->prefix_length)
3163                strip_prefix(o->prefix_length, &name, &other);
3164
3165        diff_fill_sha1_info(p->one);
3166        diff_fill_sha1_info(p->two);
3167
3168        builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3169}
3170
3171static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3172{
3173        const char *name;
3174        const char *other;
3175        const char *attr_path;
3176
3177        if (DIFF_PAIR_UNMERGED(p)) {
3178                /* unmerged */
3179                return;
3180        }
3181
3182        name = p->one->path;
3183        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3184        attr_path = other ? other : name;
3185
3186        if (o->prefix_length)
3187                strip_prefix(o->prefix_length, &name, &other);
3188
3189        diff_fill_sha1_info(p->one);
3190        diff_fill_sha1_info(p->two);
3191
3192        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3193}
3194
3195void diff_setup(struct diff_options *options)
3196{
3197        memcpy(options, &default_diff_options, sizeof(*options));
3198
3199        options->file = stdout;
3200
3201        options->line_termination = '\n';
3202        options->break_opt = -1;
3203        options->rename_limit = -1;
3204        options->dirstat_permille = diff_dirstat_permille_default;
3205        options->context = diff_context_default;
3206        DIFF_OPT_SET(options, RENAME_EMPTY);
3207
3208        options->change = diff_change;
3209        options->add_remove = diff_addremove;
3210        options->use_color = diff_use_color_default;
3211        options->detect_rename = diff_detect_rename_default;
3212        options->xdl_opts |= diff_algorithm;
3213
3214        options->orderfile = diff_order_file_cfg;
3215
3216        if (diff_no_prefix) {
3217                options->a_prefix = options->b_prefix = "";
3218        } else if (!diff_mnemonic_prefix) {
3219                options->a_prefix = "a/";
3220                options->b_prefix = "b/";
3221        }
3222}
3223
3224void diff_setup_done(struct diff_options *options)
3225{
3226        int count = 0;
3227
3228        if (options->set_default)
3229                options->set_default(options);
3230
3231        if (options->output_format & DIFF_FORMAT_NAME)
3232                count++;
3233        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3234                count++;
3235        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3236                count++;
3237        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3238                count++;
3239        if (count > 1)
3240                die("--name-only, --name-status, --check and -s are mutually exclusive");
3241
3242        /*
3243         * Most of the time we can say "there are changes"
3244         * only by checking if there are changed paths, but
3245         * --ignore-whitespace* options force us to look
3246         * inside contents.
3247         */
3248
3249        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3250            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3251            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3252                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3253        else
3254                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3255
3256        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3257                options->detect_rename = DIFF_DETECT_COPY;
3258
3259        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3260                options->prefix = NULL;
3261        if (options->prefix)
3262                options->prefix_length = strlen(options->prefix);
3263        else
3264                options->prefix_length = 0;
3265
3266        if (options->output_format & (DIFF_FORMAT_NAME |
3267                                      DIFF_FORMAT_NAME_STATUS |
3268                                      DIFF_FORMAT_CHECKDIFF |
3269                                      DIFF_FORMAT_NO_OUTPUT))
3270                options->output_format &= ~(DIFF_FORMAT_RAW |
3271                                            DIFF_FORMAT_NUMSTAT |
3272                                            DIFF_FORMAT_DIFFSTAT |
3273                                            DIFF_FORMAT_SHORTSTAT |
3274                                            DIFF_FORMAT_DIRSTAT |
3275                                            DIFF_FORMAT_SUMMARY |
3276                                            DIFF_FORMAT_PATCH);
3277
3278        /*
3279         * These cases always need recursive; we do not drop caller-supplied
3280         * recursive bits for other formats here.
3281         */
3282        if (options->output_format & (DIFF_FORMAT_PATCH |
3283                                      DIFF_FORMAT_NUMSTAT |
3284                                      DIFF_FORMAT_DIFFSTAT |
3285                                      DIFF_FORMAT_SHORTSTAT |
3286                                      DIFF_FORMAT_DIRSTAT |
3287                                      DIFF_FORMAT_SUMMARY |
3288                                      DIFF_FORMAT_CHECKDIFF))
3289                DIFF_OPT_SET(options, RECURSIVE);
3290        /*
3291         * Also pickaxe would not work very well if you do not say recursive
3292         */
3293        if (options->pickaxe)
3294                DIFF_OPT_SET(options, RECURSIVE);
3295        /*
3296         * When patches are generated, submodules diffed against the work tree
3297         * must be checked for dirtiness too so it can be shown in the output
3298         */
3299        if (options->output_format & DIFF_FORMAT_PATCH)
3300                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3301
3302        if (options->detect_rename && options->rename_limit < 0)
3303                options->rename_limit = diff_rename_limit_default;
3304        if (options->setup & DIFF_SETUP_USE_CACHE) {
3305                if (!active_cache)
3306                        /* read-cache does not die even when it fails
3307                         * so it is safe for us to do this here.  Also
3308                         * it does not smudge active_cache or active_nr
3309                         * when it fails, so we do not have to worry about
3310                         * cleaning it up ourselves either.
3311                         */
3312                        read_cache();
3313        }
3314        if (options->abbrev <= 0 || 40 < options->abbrev)
3315                options->abbrev = 40; /* full */
3316
3317        /*
3318         * It does not make sense to show the first hit we happened
3319         * to have found.  It does not make sense not to return with
3320         * exit code in such a case either.
3321         */
3322        if (DIFF_OPT_TST(options, QUICK)) {
3323                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3324                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3325        }
3326
3327        options->diff_path_counter = 0;
3328}
3329
3330static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3331{
3332        char c, *eq;
3333        int len;
3334
3335        if (*arg != '-')
3336                return 0;
3337        c = *++arg;
3338        if (!c)
3339                return 0;
3340        if (c == arg_short) {
3341                c = *++arg;
3342                if (!c)
3343                        return 1;
3344                if (val && isdigit(c)) {
3345                        char *end;
3346                        int n = strtoul(arg, &end, 10);
3347                        if (*end)
3348                                return 0;
3349                        *val = n;
3350                        return 1;
3351                }
3352                return 0;
3353        }
3354        if (c != '-')
3355                return 0;
3356        arg++;
3357        eq = strchrnul(arg, '=');
3358        len = eq - arg;
3359        if (!len || strncmp(arg, arg_long, len))
3360                return 0;
3361        if (*eq) {
3362                int n;
3363                char *end;
3364                if (!isdigit(*++eq))
3365                        return 0;
3366                n = strtoul(eq, &end, 10);
3367                if (*end)
3368                        return 0;
3369                *val = n;
3370        }
3371        return 1;
3372}
3373
3374static int diff_scoreopt_parse(const char *opt);
3375
3376static inline int short_opt(char opt, const char **argv,
3377                            const char **optarg)
3378{
3379        const char *arg = argv[0];
3380        if (arg[0] != '-' || arg[1] != opt)
3381                return 0;
3382        if (arg[2] != '\0') {
3383                *optarg = arg + 2;
3384                return 1;
3385        }
3386        if (!argv[1])
3387                die("Option '%c' requires a value", opt);
3388        *optarg = argv[1];
3389        return 2;
3390}
3391
3392int parse_long_opt(const char *opt, const char **argv,
3393                   const char **optarg)
3394{
3395        const char *arg = argv[0];
3396        if (arg[0] != '-' || arg[1] != '-')
3397                return 0;
3398        arg += strlen("--");
3399        if (!starts_with(arg, opt))
3400                return 0;
3401        arg += strlen(opt);
3402        if (*arg == '=') { /* stuck form: --option=value */
3403                *optarg = arg + 1;
3404                return 1;
3405        }
3406        if (*arg != '\0')
3407                return 0;
3408        /* separate form: --option value */
3409        if (!argv[1])
3410                die("Option '--%s' requires a value", opt);
3411        *optarg = argv[1];
3412        return 2;
3413}
3414
3415static int stat_opt(struct diff_options *options, const char **av)
3416{
3417        const char *arg = av[0];
3418        char *end;
3419        int width = options->stat_width;
3420        int name_width = options->stat_name_width;
3421        int graph_width = options->stat_graph_width;
3422        int count = options->stat_count;
3423        int argcount = 1;
3424
3425        arg += strlen("--stat");
3426        end = (char *)arg;
3427
3428        switch (*arg) {
3429        case '-':
3430                if (starts_with(arg, "-width")) {
3431                        arg += strlen("-width");
3432                        if (*arg == '=')
3433                                width = strtoul(arg + 1, &end, 10);
3434                        else if (!*arg && !av[1])
3435                                die("Option '--stat-width' requires a value");
3436                        else if (!*arg) {
3437                                width = strtoul(av[1], &end, 10);
3438                                argcount = 2;
3439                        }
3440                } else if (starts_with(arg, "-name-width")) {
3441                        arg += strlen("-name-width");
3442                        if (*arg == '=')
3443                                name_width = strtoul(arg + 1, &end, 10);
3444                        else if (!*arg && !av[1])
3445                                die("Option '--stat-name-width' requires a value");
3446                        else if (!*arg) {
3447                                name_width = strtoul(av[1], &end, 10);
3448                                argcount = 2;
3449                        }
3450                } else if (starts_with(arg, "-graph-width")) {
3451                        arg += strlen("-graph-width");
3452                        if (*arg == '=')
3453                                graph_width = strtoul(arg + 1, &end, 10);
3454                        else if (!*arg && !av[1])
3455                                die("Option '--stat-graph-width' requires a value");
3456                        else if (!*arg) {
3457                                graph_width = strtoul(av[1], &end, 10);
3458                                argcount = 2;
3459                        }
3460                } else if (starts_with(arg, "-count")) {
3461                        arg += strlen("-count");
3462                        if (*arg == '=')
3463                                count = strtoul(arg + 1, &end, 10);
3464                        else if (!*arg && !av[1])
3465                                die("Option '--stat-count' requires a value");
3466                        else if (!*arg) {
3467                                count = strtoul(av[1], &end, 10);
3468                                argcount = 2;
3469                        }
3470                }
3471                break;
3472        case '=':
3473                width = strtoul(arg+1, &end, 10);
3474                if (*end == ',')
3475                        name_width = strtoul(end+1, &end, 10);
3476                if (*end == ',')
3477                        count = strtoul(end+1, &end, 10);
3478        }
3479
3480        /* Important! This checks all the error cases! */
3481        if (*end)
3482                return 0;
3483        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3484        options->stat_name_width = name_width;
3485        options->stat_graph_width = graph_width;
3486        options->stat_width = width;
3487        options->stat_count = count;
3488        return argcount;
3489}
3490
3491static int parse_dirstat_opt(struct diff_options *options, const char *params)
3492{
3493        struct strbuf errmsg = STRBUF_INIT;
3494        if (parse_dirstat_params(options, params, &errmsg))
3495                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3496                    errmsg.buf);
3497        strbuf_release(&errmsg);
3498        /*
3499         * The caller knows a dirstat-related option is given from the command
3500         * line; allow it to say "return this_function();"
3501         */
3502        options->output_format |= DIFF_FORMAT_DIRSTAT;
3503        return 1;
3504}
3505
3506static int parse_submodule_opt(struct diff_options *options, const char *value)
3507{
3508        if (parse_submodule_params(options, value))
3509                die(_("Failed to parse --submodule option parameter: '%s'"),
3510                        value);
3511        return 1;
3512}
3513
3514static const char diff_status_letters[] = {
3515        DIFF_STATUS_ADDED,
3516        DIFF_STATUS_COPIED,
3517        DIFF_STATUS_DELETED,
3518        DIFF_STATUS_MODIFIED,
3519        DIFF_STATUS_RENAMED,
3520        DIFF_STATUS_TYPE_CHANGED,
3521        DIFF_STATUS_UNKNOWN,
3522        DIFF_STATUS_UNMERGED,
3523        DIFF_STATUS_FILTER_AON,
3524        DIFF_STATUS_FILTER_BROKEN,
3525        '\0',
3526};
3527
3528static unsigned int filter_bit['Z' + 1];
3529
3530static void prepare_filter_bits(void)
3531{
3532        int i;
3533
3534        if (!filter_bit[DIFF_STATUS_ADDED]) {
3535                for (i = 0; diff_status_letters[i]; i++)
3536                        filter_bit[(int) diff_status_letters[i]] = (1 << i);
3537        }
3538}
3539
3540static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3541{
3542        return opt->filter & filter_bit[(int) status];
3543}
3544
3545static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3546{
3547        int i, optch;
3548
3549        prepare_filter_bits();
3550
3551        /*
3552         * If there is a negation e.g. 'd' in the input, and we haven't
3553         * initialized the filter field with another --diff-filter, start
3554         * from full set of bits, except for AON.
3555         */
3556        if (!opt->filter) {
3557                for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3558                        if (optch < 'a' || 'z' < optch)
3559                                continue;
3560                        opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3561                        opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3562                        break;
3563                }
3564        }
3565
3566        for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3567                unsigned int bit;
3568                int negate;
3569
3570                if ('a' <= optch && optch <= 'z') {
3571                        negate = 1;
3572                        optch = toupper(optch);
3573                } else {
3574                        negate = 0;
3575                }
3576
3577                bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3578                if (!bit)
3579                        return optarg[i];
3580                if (negate)
3581                        opt->filter &= ~bit;
3582                else
3583                        opt->filter |= bit;
3584        }
3585        return 0;
3586}
3587
3588static void enable_patch_output(int *fmt) {
3589        *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3590        *fmt |= DIFF_FORMAT_PATCH;
3591}
3592
3593int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3594{
3595        const char *arg = av[0];
3596        const char *optarg;
3597        int argcount;
3598
3599        /* Output format options */
3600        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3601            || opt_arg(arg, 'U', "unified", &options->context))
3602                enable_patch_output(&options->output_format);
3603        else if (!strcmp(arg, "--raw"))
3604                options->output_format |= DIFF_FORMAT_RAW;
3605        else if (!strcmp(arg, "--patch-with-raw")) {
3606                enable_patch_output(&options->output_format);
3607                options->output_format |= DIFF_FORMAT_RAW;
3608        } else if (!strcmp(arg, "--numstat"))
3609                options->output_format |= DIFF_FORMAT_NUMSTAT;
3610        else if (!strcmp(arg, "--shortstat"))
3611                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3612        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3613                return parse_dirstat_opt(options, "");
3614        else if (starts_with(arg, "-X"))
3615                return parse_dirstat_opt(options, arg + 2);
3616        else if (starts_with(arg, "--dirstat="))
3617                return parse_dirstat_opt(options, arg + 10);
3618        else if (!strcmp(arg, "--cumulative"))
3619                return parse_dirstat_opt(options, "cumulative");
3620        else if (!strcmp(arg, "--dirstat-by-file"))
3621                return parse_dirstat_opt(options, "files");
3622        else if (starts_with(arg, "--dirstat-by-file=")) {
3623                parse_dirstat_opt(options, "files");
3624                return parse_dirstat_opt(options, arg + 18);
3625        }
3626        else if (!strcmp(arg, "--check"))
3627                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3628        else if (!strcmp(arg, "--summary"))
3629                options->output_format |= DIFF_FORMAT_SUMMARY;
3630        else if (!strcmp(arg, "--patch-with-stat")) {
3631                enable_patch_output(&options->output_format);
3632                options->output_format |= DIFF_FORMAT_DIFFSTAT;
3633        } else if (!strcmp(arg, "--name-only"))
3634                options->output_format |= DIFF_FORMAT_NAME;
3635        else if (!strcmp(arg, "--name-status"))
3636                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3637        else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3638                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3639        else if (starts_with(arg, "--stat"))
3640                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3641                return stat_opt(options, av);
3642
3643        /* renames options */
3644        else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3645                 !strcmp(arg, "--break-rewrites")) {
3646                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3647                        return error("invalid argument to -B: %s", arg+2);
3648        }
3649        else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3650                 !strcmp(arg, "--find-renames")) {
3651                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3652                        return error("invalid argument to -M: %s", arg+2);
3653                options->detect_rename = DIFF_DETECT_RENAME;
3654        }
3655        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3656                options->irreversible_delete = 1;
3657        }
3658        else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3659                 !strcmp(arg, "--find-copies")) {
3660                if (options->detect_rename == DIFF_DETECT_COPY)
3661                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3662                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3663                        return error("invalid argument to -C: %s", arg+2);
3664                options->detect_rename = DIFF_DETECT_COPY;
3665        }
3666        else if (!strcmp(arg, "--no-renames"))
3667                options->detect_rename = 0;
3668        else if (!strcmp(arg, "--rename-empty"))
3669                DIFF_OPT_SET(options, RENAME_EMPTY);
3670        else if (!strcmp(arg, "--no-rename-empty"))
3671                DIFF_OPT_CLR(options, RENAME_EMPTY);
3672        else if (!strcmp(arg, "--relative"))
3673                DIFF_OPT_SET(options, RELATIVE_NAME);
3674        else if (starts_with(arg, "--relative=")) {
3675                DIFF_OPT_SET(options, RELATIVE_NAME);
3676                options->prefix = arg + 11;
3677        }
3678
3679        /* xdiff options */
3680        else if (!strcmp(arg, "--minimal"))
3681                DIFF_XDL_SET(options, NEED_MINIMAL);
3682        else if (!strcmp(arg, "--no-minimal"))
3683                DIFF_XDL_CLR(options, NEED_MINIMAL);
3684        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3685                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3686        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3687                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3688        else if (!strcmp(arg, "--ignore-space-at-eol"))
3689                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3690        else if (!strcmp(arg, "--ignore-blank-lines"))
3691                DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3692        else if (!strcmp(arg, "--patience"))
3693                options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3694        else if (!strcmp(arg, "--histogram"))
3695                options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3696        else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3697                long value = parse_algorithm_value(optarg);
3698                if (value < 0)
3699                        return error("option diff-algorithm accepts \"myers\", "
3700                                     "\"minimal\", \"patience\" and \"histogram\"");
3701                /* clear out previous settings */
3702                DIFF_XDL_CLR(options, NEED_MINIMAL);
3703                options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3704                options->xdl_opts |= value;
3705                return argcount;
3706        }
3707
3708        /* flags options */
3709        else if (!strcmp(arg, "--binary")) {
3710                enable_patch_output(&options->output_format);
3711                DIFF_OPT_SET(options, BINARY);
3712        }
3713        else if (!strcmp(arg, "--full-index"))
3714                DIFF_OPT_SET(options, FULL_INDEX);
3715        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3716                DIFF_OPT_SET(options, TEXT);
3717        else if (!strcmp(arg, "-R"))
3718                DIFF_OPT_SET(options, REVERSE_DIFF);
3719        else if (!strcmp(arg, "--find-copies-harder"))
3720                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3721        else if (!strcmp(arg, "--follow"))
3722                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3723        else if (!strcmp(arg, "--no-follow"))
3724                DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3725        else if (!strcmp(arg, "--color"))
3726                options->use_color = 1;
3727        else if (starts_with(arg, "--color=")) {
3728                int value = git_config_colorbool(NULL, arg+8);
3729                if (value < 0)
3730                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3731                options->use_color = value;
3732        }
3733        else if (!strcmp(arg, "--no-color"))
3734                options->use_color = 0;
3735        else if (!strcmp(arg, "--color-words")) {
3736                options->use_color = 1;
3737                options->word_diff = DIFF_WORDS_COLOR;
3738        }
3739        else if (starts_with(arg, "--color-words=")) {
3740                options->use_color = 1;
3741                options->word_diff = DIFF_WORDS_COLOR;
3742                options->word_regex = arg + 14;
3743        }
3744        else if (!strcmp(arg, "--word-diff")) {
3745                if (options->word_diff == DIFF_WORDS_NONE)
3746                        options->word_diff = DIFF_WORDS_PLAIN;
3747        }
3748        else if (starts_with(arg, "--word-diff=")) {
3749                const char *type = arg + 12;
3750                if (!strcmp(type, "plain"))
3751                        options->word_diff = DIFF_WORDS_PLAIN;
3752                else if (!strcmp(type, "color")) {
3753                        options->use_color = 1;
3754                        options->word_diff = DIFF_WORDS_COLOR;
3755                }
3756                else if (!strcmp(type, "porcelain"))
3757                        options->word_diff = DIFF_WORDS_PORCELAIN;
3758                else if (!strcmp(type, "none"))
3759                        options->word_diff = DIFF_WORDS_NONE;
3760                else
3761                        die("bad --word-diff argument: %s", type);
3762        }
3763        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3764                if (options->word_diff == DIFF_WORDS_NONE)
3765                        options->word_diff = DIFF_WORDS_PLAIN;
3766                options->word_regex = optarg;
3767                return argcount;
3768        }
3769        else if (!strcmp(arg, "--exit-code"))
3770                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3771        else if (!strcmp(arg, "--quiet"))
3772                DIFF_OPT_SET(options, QUICK);
3773        else if (!strcmp(arg, "--ext-diff"))
3774                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3775        else if (!strcmp(arg, "--no-ext-diff"))
3776                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3777        else if (!strcmp(arg, "--textconv"))
3778                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3779        else if (!strcmp(arg, "--no-textconv"))
3780                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3781        else if (!strcmp(arg, "--ignore-submodules")) {
3782                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3783                handle_ignore_submodules_arg(options, "all");
3784        } else if (starts_with(arg, "--ignore-submodules=")) {
3785                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3786                handle_ignore_submodules_arg(options, arg + 20);
3787        } else if (!strcmp(arg, "--submodule"))
3788                DIFF_OPT_SET(options, SUBMODULE_LOG);
3789        else if (starts_with(arg, "--submodule="))
3790                return parse_submodule_opt(options, arg + 12);
3791
3792        /* misc options */
3793        else if (!strcmp(arg, "-z"))
3794                options->line_termination = 0;
3795        else if ((argcount = short_opt('l', av, &optarg))) {
3796                options->rename_limit = strtoul(optarg, NULL, 10);
3797                return argcount;
3798        }
3799        else if ((argcount = short_opt('S', av, &optarg))) {
3800                options->pickaxe = optarg;
3801                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3802                return argcount;
3803        } else if ((argcount = short_opt('G', av, &optarg))) {
3804                options->pickaxe = optarg;
3805                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3806                return argcount;
3807        }
3808        else if (!strcmp(arg, "--pickaxe-all"))
3809                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3810        else if (!strcmp(arg, "--pickaxe-regex"))
3811                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3812        else if ((argcount = short_opt('O', av, &optarg))) {
3813                options->orderfile = optarg;
3814                return argcount;
3815        }
3816        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3817                int offending = parse_diff_filter_opt(optarg, options);
3818                if (offending)
3819                        die("unknown change class '%c' in --diff-filter=%s",
3820                            offending, optarg);
3821                return argcount;
3822        }
3823        else if (!strcmp(arg, "--abbrev"))
3824                options->abbrev = DEFAULT_ABBREV;
3825        else if (starts_with(arg, "--abbrev=")) {
3826                options->abbrev = strtoul(arg + 9, NULL, 10);
3827                if (options->abbrev < MINIMUM_ABBREV)
3828                        options->abbrev = MINIMUM_ABBREV;
3829                else if (40 < options->abbrev)
3830                        options->abbrev = 40;
3831        }
3832        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3833                options->a_prefix = optarg;
3834                return argcount;
3835        }
3836        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3837                options->b_prefix = optarg;
3838                return argcount;
3839        }
3840        else if (!strcmp(arg, "--no-prefix"))
3841                options->a_prefix = options->b_prefix = "";
3842        else if (opt_arg(arg, '\0', "inter-hunk-context",
3843                         &options->interhunkcontext))
3844                ;
3845        else if (!strcmp(arg, "-W"))
3846                DIFF_OPT_SET(options, FUNCCONTEXT);
3847        else if (!strcmp(arg, "--function-context"))
3848                DIFF_OPT_SET(options, FUNCCONTEXT);
3849        else if (!strcmp(arg, "--no-function-context"))
3850                DIFF_OPT_CLR(options, FUNCCONTEXT);
3851        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3852                options->file = fopen(optarg, "w");
3853                if (!options->file)
3854                        die_errno("Could not open '%s'", optarg);
3855                options->close_file = 1;
3856                return argcount;
3857        } else
3858                return 0;
3859        return 1;
3860}
3861
3862int parse_rename_score(const char **cp_p)
3863{
3864        unsigned long num, scale;
3865        int ch, dot;
3866        const char *cp = *cp_p;
3867
3868        num = 0;
3869        scale = 1;
3870        dot = 0;
3871        for (;;) {
3872                ch = *cp;
3873                if ( !dot && ch == '.' ) {
3874                        scale = 1;
3875                        dot = 1;
3876                } else if ( ch == '%' ) {
3877                        scale = dot ? scale*100 : 100;
3878                        cp++;   /* % is always at the end */
3879                        break;
3880                } else if ( ch >= '0' && ch <= '9' ) {
3881                        if ( scale < 100000 ) {
3882                                scale *= 10;
3883                                num = (num*10) + (ch-'0');
3884                        }
3885                } else {
3886                        break;
3887                }
3888                cp++;
3889        }
3890        *cp_p = cp;
3891
3892        /* user says num divided by scale and we say internally that
3893         * is MAX_SCORE * num / scale.
3894         */
3895        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3896}
3897
3898static int diff_scoreopt_parse(const char *opt)
3899{
3900        int opt1, opt2, cmd;
3901
3902        if (*opt++ != '-')
3903                return -1;
3904        cmd = *opt++;
3905        if (cmd == '-') {
3906                /* convert the long-form arguments into short-form versions */
3907                if (starts_with(opt, "break-rewrites")) {
3908                        opt += strlen("break-rewrites");
3909                        if (*opt == 0 || *opt++ == '=')
3910                                cmd = 'B';
3911                } else if (starts_with(opt, "find-copies")) {
3912                        opt += strlen("find-copies");
3913                        if (*opt == 0 || *opt++ == '=')
3914                                cmd = 'C';
3915                } else if (starts_with(opt, "find-renames")) {
3916                        opt += strlen("find-renames");
3917                        if (*opt == 0 || *opt++ == '=')
3918                                cmd = 'M';
3919                }
3920        }
3921        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3922                return -1; /* that is not a -M, -C, or -B option */
3923
3924        opt1 = parse_rename_score(&opt);
3925        if (cmd != 'B')
3926                opt2 = 0;
3927        else {
3928                if (*opt == 0)
3929                        opt2 = 0;
3930                else if (*opt != '/')
3931                        return -1; /* we expect -B80/99 or -B80 */
3932                else {
3933                        opt++;
3934                        opt2 = parse_rename_score(&opt);
3935                }
3936        }
3937        if (*opt != 0)
3938                return -1;
3939        return opt1 | (opt2 << 16);
3940}
3941
3942struct diff_queue_struct diff_queued_diff;
3943
3944void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3945{
3946        ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
3947        queue->queue[queue->nr++] = dp;
3948}
3949
3950struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3951                                 struct diff_filespec *one,
3952                                 struct diff_filespec *two)
3953{
3954        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3955        dp->one = one;
3956        dp->two = two;
3957        if (queue)
3958                diff_q(queue, dp);
3959        return dp;
3960}
3961
3962void diff_free_filepair(struct diff_filepair *p)
3963{
3964        free_filespec(p->one);
3965        free_filespec(p->two);
3966        free(p);
3967}
3968
3969/* This is different from find_unique_abbrev() in that
3970 * it stuffs the result with dots for alignment.
3971 */
3972const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3973{
3974        int abblen;
3975        const char *abbrev;
3976        if (len == 40)
3977                return sha1_to_hex(sha1);
3978
3979        abbrev = find_unique_abbrev(sha1, len);
3980        abblen = strlen(abbrev);
3981        if (abblen < 37) {
3982                static char hex[41];
3983                if (len < abblen && abblen <= len + 2)
3984                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3985                else
3986                        sprintf(hex, "%s...", abbrev);
3987                return hex;
3988        }
3989        return sha1_to_hex(sha1);
3990}
3991
3992static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3993{
3994        int line_termination = opt->line_termination;
3995        int inter_name_termination = line_termination ? '\t' : '\0';
3996
3997        fprintf(opt->file, "%s", diff_line_prefix(opt));
3998        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3999                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4000                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
4001                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
4002        }
4003        if (p->score) {
4004                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4005                        inter_name_termination);
4006        } else {
4007                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4008        }
4009
4010        if (p->status == DIFF_STATUS_COPIED ||
4011            p->status == DIFF_STATUS_RENAMED) {
4012                const char *name_a, *name_b;
4013                name_a = p->one->path;
4014                name_b = p->two->path;
4015                strip_prefix(opt->prefix_length, &name_a, &name_b);
4016                write_name_quoted(name_a, opt->file, inter_name_termination);
4017                write_name_quoted(name_b, opt->file, line_termination);
4018        } else {
4019                const char *name_a, *name_b;
4020                name_a = p->one->mode ? p->one->path : p->two->path;
4021                name_b = NULL;
4022                strip_prefix(opt->prefix_length, &name_a, &name_b);
4023                write_name_quoted(name_a, opt->file, line_termination);
4024        }
4025}
4026
4027int diff_unmodified_pair(struct diff_filepair *p)
4028{
4029        /* This function is written stricter than necessary to support
4030         * the currently implemented transformers, but the idea is to
4031         * let transformers to produce diff_filepairs any way they want,
4032         * and filter and clean them up here before producing the output.
4033         */
4034        struct diff_filespec *one = p->one, *two = p->two;
4035
4036        if (DIFF_PAIR_UNMERGED(p))
4037                return 0; /* unmerged is interesting */
4038
4039        /* deletion, addition, mode or type change
4040         * and rename are all interesting.
4041         */
4042        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4043            DIFF_PAIR_MODE_CHANGED(p) ||
4044            strcmp(one->path, two->path))
4045                return 0;
4046
4047        /* both are valid and point at the same path.  that is, we are
4048         * dealing with a change.
4049         */
4050        if (one->sha1_valid && two->sha1_valid &&
4051            !hashcmp(one->sha1, two->sha1) &&
4052            !one->dirty_submodule && !two->dirty_submodule)
4053                return 1; /* no change */
4054        if (!one->sha1_valid && !two->sha1_valid)
4055                return 1; /* both look at the same file on the filesystem. */
4056        return 0;
4057}
4058
4059static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4060{
4061        if (diff_unmodified_pair(p))
4062                return;
4063
4064        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4065            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4066                return; /* no tree diffs in patch format */
4067
4068        run_diff(p, o);
4069}
4070
4071static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4072                            struct diffstat_t *diffstat)
4073{
4074        if (diff_unmodified_pair(p))
4075                return;
4076
4077        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4078            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4079                return; /* no useful stat for tree diffs */
4080
4081        run_diffstat(p, o, diffstat);
4082}
4083
4084static void diff_flush_checkdiff(struct diff_filepair *p,
4085                struct diff_options *o)
4086{
4087        if (diff_unmodified_pair(p))
4088                return;
4089
4090        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4091            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4092                return; /* nothing to check in tree diffs */
4093
4094        run_checkdiff(p, o);
4095}
4096
4097int diff_queue_is_empty(void)
4098{
4099        struct diff_queue_struct *q = &diff_queued_diff;
4100        int i;
4101        for (i = 0; i < q->nr; i++)
4102                if (!diff_unmodified_pair(q->queue[i]))
4103                        return 0;
4104        return 1;
4105}
4106
4107#if DIFF_DEBUG
4108void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4109{
4110        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4111                x, one ? one : "",
4112                s->path,
4113                DIFF_FILE_VALID(s) ? "valid" : "invalid",
4114                s->mode,
4115                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4116        fprintf(stderr, "queue[%d] %s size %lu\n",
4117                x, one ? one : "",
4118                s->size);
4119}
4120
4121void diff_debug_filepair(const struct diff_filepair *p, int i)
4122{
4123        diff_debug_filespec(p->one, i, "one");
4124        diff_debug_filespec(p->two, i, "two");
4125        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4126                p->score, p->status ? p->status : '?',
4127                p->one->rename_used, p->broken_pair);
4128}
4129
4130void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4131{
4132        int i;
4133        if (msg)
4134                fprintf(stderr, "%s\n", msg);
4135        fprintf(stderr, "q->nr = %d\n", q->nr);
4136        for (i = 0; i < q->nr; i++) {
4137                struct diff_filepair *p = q->queue[i];
4138                diff_debug_filepair(p, i);
4139        }
4140}
4141#endif
4142
4143static void diff_resolve_rename_copy(void)
4144{
4145        int i;
4146        struct diff_filepair *p;
4147        struct diff_queue_struct *q = &diff_queued_diff;
4148
4149        diff_debug_queue("resolve-rename-copy", q);
4150
4151        for (i = 0; i < q->nr; i++) {
4152                p = q->queue[i];
4153                p->status = 0; /* undecided */
4154                if (DIFF_PAIR_UNMERGED(p))
4155                        p->status = DIFF_STATUS_UNMERGED;
4156                else if (!DIFF_FILE_VALID(p->one))
4157                        p->status = DIFF_STATUS_ADDED;
4158                else if (!DIFF_FILE_VALID(p->two))
4159                        p->status = DIFF_STATUS_DELETED;
4160                else if (DIFF_PAIR_TYPE_CHANGED(p))
4161                        p->status = DIFF_STATUS_TYPE_CHANGED;
4162
4163                /* from this point on, we are dealing with a pair
4164                 * whose both sides are valid and of the same type, i.e.
4165                 * either in-place edit or rename/copy edit.
4166                 */
4167                else if (DIFF_PAIR_RENAME(p)) {
4168                        /*
4169                         * A rename might have re-connected a broken
4170                         * pair up, causing the pathnames to be the
4171                         * same again. If so, that's not a rename at
4172                         * all, just a modification..
4173                         *
4174                         * Otherwise, see if this source was used for
4175                         * multiple renames, in which case we decrement
4176                         * the count, and call it a copy.
4177                         */
4178                        if (!strcmp(p->one->path, p->two->path))
4179                                p->status = DIFF_STATUS_MODIFIED;
4180                        else if (--p->one->rename_used > 0)
4181                                p->status = DIFF_STATUS_COPIED;
4182                        else
4183                                p->status = DIFF_STATUS_RENAMED;
4184                }
4185                else if (hashcmp(p->one->sha1, p->two->sha1) ||
4186                         p->one->mode != p->two->mode ||
4187                         p->one->dirty_submodule ||
4188                         p->two->dirty_submodule ||
4189                         is_null_sha1(p->one->sha1))
4190                        p->status = DIFF_STATUS_MODIFIED;
4191                else {
4192                        /* This is a "no-change" entry and should not
4193                         * happen anymore, but prepare for broken callers.
4194                         */
4195                        error("feeding unmodified %s to diffcore",
4196                              p->one->path);
4197                        p->status = DIFF_STATUS_UNKNOWN;
4198                }
4199        }
4200        diff_debug_queue("resolve-rename-copy done", q);
4201}
4202
4203static int check_pair_status(struct diff_filepair *p)
4204{
4205        switch (p->status) {
4206        case DIFF_STATUS_UNKNOWN:
4207                return 0;
4208        case 0:
4209                die("internal error in diff-resolve-rename-copy");
4210        default:
4211                return 1;
4212        }
4213}
4214
4215static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4216{
4217        int fmt = opt->output_format;
4218
4219        if (fmt & DIFF_FORMAT_CHECKDIFF)
4220                diff_flush_checkdiff(p, opt);
4221        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4222                diff_flush_raw(p, opt);
4223        else if (fmt & DIFF_FORMAT_NAME) {
4224                const char *name_a, *name_b;
4225                name_a = p->two->path;
4226                name_b = NULL;
4227                strip_prefix(opt->prefix_length, &name_a, &name_b);
4228                write_name_quoted(name_a, opt->file, opt->line_termination);
4229        }
4230}
4231
4232static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4233{
4234        if (fs->mode)
4235                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4236        else
4237                fprintf(file, " %s ", newdelete);
4238        write_name_quoted(fs->path, file, '\n');
4239}
4240
4241
4242static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4243                const char *line_prefix)
4244{
4245        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4246                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4247                        p->two->mode, show_name ? ' ' : '\n');
4248                if (show_name) {
4249                        write_name_quoted(p->two->path, file, '\n');
4250                }
4251        }
4252}
4253
4254static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4255                        const char *line_prefix)
4256{
4257        char *names = pprint_rename(p->one->path, p->two->path);
4258
4259        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4260        free(names);
4261        show_mode_change(file, p, 0, line_prefix);
4262}
4263
4264static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4265{
4266        FILE *file = opt->file;
4267        const char *line_prefix = diff_line_prefix(opt);
4268
4269        switch(p->status) {
4270        case DIFF_STATUS_DELETED:
4271                fputs(line_prefix, file);
4272                show_file_mode_name(file, "delete", p->one);
4273                break;
4274        case DIFF_STATUS_ADDED:
4275                fputs(line_prefix, file);
4276                show_file_mode_name(file, "create", p->two);
4277                break;
4278        case DIFF_STATUS_COPIED:
4279                fputs(line_prefix, file);
4280                show_rename_copy(file, "copy", p, line_prefix);
4281                break;
4282        case DIFF_STATUS_RENAMED:
4283                fputs(line_prefix, file);
4284                show_rename_copy(file, "rename", p, line_prefix);
4285                break;
4286        default:
4287                if (p->score) {
4288                        fprintf(file, "%s rewrite ", line_prefix);
4289                        write_name_quoted(p->two->path, file, ' ');
4290                        fprintf(file, "(%d%%)\n", similarity_index(p));
4291                }
4292                show_mode_change(file, p, !p->score, line_prefix);
4293                break;
4294        }
4295}
4296
4297struct patch_id_t {
4298        git_SHA_CTX *ctx;
4299        int patchlen;
4300};
4301
4302static int remove_space(char *line, int len)
4303{
4304        int i;
4305        char *dst = line;
4306        unsigned char c;
4307
4308        for (i = 0; i < len; i++)
4309                if (!isspace((c = line[i])))
4310                        *dst++ = c;
4311
4312        return dst - line;
4313}
4314
4315static void patch_id_consume(void *priv, char *line, unsigned long len)
4316{
4317        struct patch_id_t *data = priv;
4318        int new_len;
4319
4320        /* Ignore line numbers when computing the SHA1 of the patch */
4321        if (starts_with(line, "@@ -"))
4322                return;
4323
4324        new_len = remove_space(line, len);
4325
4326        git_SHA1_Update(data->ctx, line, new_len);
4327        data->patchlen += new_len;
4328}
4329
4330/* returns 0 upon success, and writes result into sha1 */
4331static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4332{
4333        struct diff_queue_struct *q = &diff_queued_diff;
4334        int i;
4335        git_SHA_CTX ctx;
4336        struct patch_id_t data;
4337        char buffer[PATH_MAX * 4 + 20];
4338
4339        git_SHA1_Init(&ctx);
4340        memset(&data, 0, sizeof(struct patch_id_t));
4341        data.ctx = &ctx;
4342
4343        for (i = 0; i < q->nr; i++) {
4344                xpparam_t xpp;
4345                xdemitconf_t xecfg;
4346                mmfile_t mf1, mf2;
4347                struct diff_filepair *p = q->queue[i];
4348                int len1, len2;
4349
4350                memset(&xpp, 0, sizeof(xpp));
4351                memset(&xecfg, 0, sizeof(xecfg));
4352                if (p->status == 0)
4353                        return error("internal diff status error");
4354                if (p->status == DIFF_STATUS_UNKNOWN)
4355                        continue;
4356                if (diff_unmodified_pair(p))
4357                        continue;
4358                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4359                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4360                        continue;
4361                if (DIFF_PAIR_UNMERGED(p))
4362                        continue;
4363
4364                diff_fill_sha1_info(p->one);
4365                diff_fill_sha1_info(p->two);
4366                if (fill_mmfile(&mf1, p->one) < 0 ||
4367                                fill_mmfile(&mf2, p->two) < 0)
4368                        return error("unable to read files to diff");
4369
4370                len1 = remove_space(p->one->path, strlen(p->one->path));
4371                len2 = remove_space(p->two->path, strlen(p->two->path));
4372                if (p->one->mode == 0)
4373                        len1 = snprintf(buffer, sizeof(buffer),
4374                                        "diff--gita/%.*sb/%.*s"
4375                                        "newfilemode%06o"
4376                                        "---/dev/null"
4377                                        "+++b/%.*s",
4378                                        len1, p->one->path,
4379                                        len2, p->two->path,
4380                                        p->two->mode,
4381                                        len2, p->two->path);
4382                else if (p->two->mode == 0)
4383                        len1 = snprintf(buffer, sizeof(buffer),
4384                                        "diff--gita/%.*sb/%.*s"
4385                                        "deletedfilemode%06o"
4386                                        "---a/%.*s"
4387                                        "+++/dev/null",
4388                                        len1, p->one->path,
4389                                        len2, p->two->path,
4390                                        p->one->mode,
4391                                        len1, p->one->path);
4392                else
4393                        len1 = snprintf(buffer, sizeof(buffer),
4394                                        "diff--gita/%.*sb/%.*s"
4395                                        "---a/%.*s"
4396                                        "+++b/%.*s",
4397                                        len1, p->one->path,
4398                                        len2, p->two->path,
4399                                        len1, p->one->path,
4400                                        len2, p->two->path);
4401                git_SHA1_Update(&ctx, buffer, len1);
4402
4403                if (diff_filespec_is_binary(p->one) ||
4404                    diff_filespec_is_binary(p->two)) {
4405                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4406                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4407                        continue;
4408                }
4409
4410                xpp.flags = 0;
4411                xecfg.ctxlen = 3;
4412                xecfg.flags = 0;
4413                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4414                              &xpp, &xecfg);
4415        }
4416
4417        git_SHA1_Final(sha1, &ctx);
4418        return 0;
4419}
4420
4421int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4422{
4423        struct diff_queue_struct *q = &diff_queued_diff;
4424        int i;
4425        int result = diff_get_patch_id(options, sha1);
4426
4427        for (i = 0; i < q->nr; i++)
4428                diff_free_filepair(q->queue[i]);
4429
4430        free(q->queue);
4431        DIFF_QUEUE_CLEAR(q);
4432
4433        return result;
4434}
4435
4436static int is_summary_empty(const struct diff_queue_struct *q)
4437{
4438        int i;
4439
4440        for (i = 0; i < q->nr; i++) {
4441                const struct diff_filepair *p = q->queue[i];
4442
4443                switch (p->status) {
4444                case DIFF_STATUS_DELETED:
4445                case DIFF_STATUS_ADDED:
4446                case DIFF_STATUS_COPIED:
4447                case DIFF_STATUS_RENAMED:
4448                        return 0;
4449                default:
4450                        if (p->score)
4451                                return 0;
4452                        if (p->one->mode && p->two->mode &&
4453                            p->one->mode != p->two->mode)
4454                                return 0;
4455                        break;
4456                }
4457        }
4458        return 1;
4459}
4460
4461static const char rename_limit_warning[] =
4462"inexact rename detection was skipped due to too many files.";
4463
4464static const char degrade_cc_to_c_warning[] =
4465"only found copies from modified paths due to too many files.";
4466
4467static const char rename_limit_advice[] =
4468"you may want to set your %s variable to at least "
4469"%d and retry the command.";
4470
4471void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4472{
4473        if (degraded_cc)
4474                warning(degrade_cc_to_c_warning);
4475        else if (needed)
4476                warning(rename_limit_warning);
4477        else
4478                return;
4479        if (0 < needed && needed < 32767)
4480                warning(rename_limit_advice, varname, needed);
4481}
4482
4483void diff_flush(struct diff_options *options)
4484{
4485        struct diff_queue_struct *q = &diff_queued_diff;
4486        int i, output_format = options->output_format;
4487        int separator = 0;
4488        int dirstat_by_line = 0;
4489
4490        /*
4491         * Order: raw, stat, summary, patch
4492         * or:    name/name-status/checkdiff (other bits clear)
4493         */
4494        if (!q->nr)
4495                goto free_queue;
4496
4497        if (output_format & (DIFF_FORMAT_RAW |
4498                             DIFF_FORMAT_NAME |
4499                             DIFF_FORMAT_NAME_STATUS |
4500                             DIFF_FORMAT_CHECKDIFF)) {
4501                for (i = 0; i < q->nr; i++) {
4502                        struct diff_filepair *p = q->queue[i];
4503                        if (check_pair_status(p))
4504                                flush_one_pair(p, options);
4505                }
4506                separator++;
4507        }
4508
4509        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4510                dirstat_by_line = 1;
4511
4512        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4513            dirstat_by_line) {
4514                struct diffstat_t diffstat;
4515
4516                memset(&diffstat, 0, sizeof(struct diffstat_t));
4517                for (i = 0; i < q->nr; i++) {
4518                        struct diff_filepair *p = q->queue[i];
4519                        if (check_pair_status(p))
4520                                diff_flush_stat(p, options, &diffstat);
4521                }
4522                if (output_format & DIFF_FORMAT_NUMSTAT)
4523                        show_numstat(&diffstat, options);
4524                if (output_format & DIFF_FORMAT_DIFFSTAT)
4525                        show_stats(&diffstat, options);
4526                if (output_format & DIFF_FORMAT_SHORTSTAT)
4527                        show_shortstats(&diffstat, options);
4528                if (output_format & DIFF_FORMAT_DIRSTAT)
4529                        show_dirstat_by_line(&diffstat, options);
4530                free_diffstat_info(&diffstat);
4531                separator++;
4532        }
4533        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4534                show_dirstat(options);
4535
4536        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4537                for (i = 0; i < q->nr; i++) {
4538                        diff_summary(options, q->queue[i]);
4539                }
4540                separator++;
4541        }
4542
4543        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4544            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4545            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4546                /*
4547                 * run diff_flush_patch for the exit status. setting
4548                 * options->file to /dev/null should be safe, because we
4549                 * aren't supposed to produce any output anyway.
4550                 */
4551                if (options->close_file)
4552                        fclose(options->file);
4553                options->file = fopen("/dev/null", "w");
4554                if (!options->file)
4555                        die_errno("Could not open /dev/null");
4556                options->close_file = 1;
4557                for (i = 0; i < q->nr; i++) {
4558                        struct diff_filepair *p = q->queue[i];
4559                        if (check_pair_status(p))
4560                                diff_flush_patch(p, options);
4561                        if (options->found_changes)
4562                                break;
4563                }
4564        }
4565
4566        if (output_format & DIFF_FORMAT_PATCH) {
4567                if (separator) {
4568                        fprintf(options->file, "%s%c",
4569                                diff_line_prefix(options),
4570                                options->line_termination);
4571                        if (options->stat_sep) {
4572                                /* attach patch instead of inline */
4573                                fputs(options->stat_sep, options->file);
4574                        }
4575                }
4576
4577                for (i = 0; i < q->nr; i++) {
4578                        struct diff_filepair *p = q->queue[i];
4579                        if (check_pair_status(p))
4580                                diff_flush_patch(p, options);
4581                }
4582        }
4583
4584        if (output_format & DIFF_FORMAT_CALLBACK)
4585                options->format_callback(q, options, options->format_callback_data);
4586
4587        for (i = 0; i < q->nr; i++)
4588                diff_free_filepair(q->queue[i]);
4589free_queue:
4590        free(q->queue);
4591        DIFF_QUEUE_CLEAR(q);
4592        if (options->close_file)
4593                fclose(options->file);
4594
4595        /*
4596         * Report the content-level differences with HAS_CHANGES;
4597         * diff_addremove/diff_change does not set the bit when
4598         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4599         */
4600        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4601                if (options->found_changes)
4602                        DIFF_OPT_SET(options, HAS_CHANGES);
4603                else
4604                        DIFF_OPT_CLR(options, HAS_CHANGES);
4605        }
4606}
4607
4608static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4609{
4610        return (((p->status == DIFF_STATUS_MODIFIED) &&
4611                 ((p->score &&
4612                   filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4613                  (!p->score &&
4614                   filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4615                ((p->status != DIFF_STATUS_MODIFIED) &&
4616                 filter_bit_tst(p->status, options)));
4617}
4618
4619static void diffcore_apply_filter(struct diff_options *options)
4620{
4621        int i;
4622        struct diff_queue_struct *q = &diff_queued_diff;
4623        struct diff_queue_struct outq;
4624
4625        DIFF_QUEUE_CLEAR(&outq);
4626
4627        if (!options->filter)
4628                return;
4629
4630        if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4631                int found;
4632                for (i = found = 0; !found && i < q->nr; i++) {
4633                        if (match_filter(options, q->queue[i]))
4634                                found++;
4635                }
4636                if (found)
4637                        return;
4638
4639                /* otherwise we will clear the whole queue
4640                 * by copying the empty outq at the end of this
4641                 * function, but first clear the current entries
4642                 * in the queue.
4643                 */
4644                for (i = 0; i < q->nr; i++)
4645                        diff_free_filepair(q->queue[i]);
4646        }
4647        else {
4648                /* Only the matching ones */
4649                for (i = 0; i < q->nr; i++) {
4650                        struct diff_filepair *p = q->queue[i];
4651                        if (match_filter(options, p))
4652                                diff_q(&outq, p);
4653                        else
4654                                diff_free_filepair(p);
4655                }
4656        }
4657        free(q->queue);
4658        *q = outq;
4659}
4660
4661/* Check whether two filespecs with the same mode and size are identical */
4662static int diff_filespec_is_identical(struct diff_filespec *one,
4663                                      struct diff_filespec *two)
4664{
4665        if (S_ISGITLINK(one->mode))
4666                return 0;
4667        if (diff_populate_filespec(one, 0))
4668                return 0;
4669        if (diff_populate_filespec(two, 0))
4670                return 0;
4671        return !memcmp(one->data, two->data, one->size);
4672}
4673
4674static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4675{
4676        if (p->done_skip_stat_unmatch)
4677                return p->skip_stat_unmatch_result;
4678
4679        p->done_skip_stat_unmatch = 1;
4680        p->skip_stat_unmatch_result = 0;
4681        /*
4682         * 1. Entries that come from stat info dirtiness
4683         *    always have both sides (iow, not create/delete),
4684         *    one side of the object name is unknown, with
4685         *    the same mode and size.  Keep the ones that
4686         *    do not match these criteria.  They have real
4687         *    differences.
4688         *
4689         * 2. At this point, the file is known to be modified,
4690         *    with the same mode and size, and the object
4691         *    name of one side is unknown.  Need to inspect
4692         *    the identical contents.
4693         */
4694        if (!DIFF_FILE_VALID(p->one) || /* (1) */
4695            !DIFF_FILE_VALID(p->two) ||
4696            (p->one->sha1_valid && p->two->sha1_valid) ||
4697            (p->one->mode != p->two->mode) ||
4698            diff_populate_filespec(p->one, 1) ||
4699            diff_populate_filespec(p->two, 1) ||
4700            (p->one->size != p->two->size) ||
4701            !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4702                p->skip_stat_unmatch_result = 1;
4703        return p->skip_stat_unmatch_result;
4704}
4705
4706static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4707{
4708        int i;
4709        struct diff_queue_struct *q = &diff_queued_diff;
4710        struct diff_queue_struct outq;
4711        DIFF_QUEUE_CLEAR(&outq);
4712
4713        for (i = 0; i < q->nr; i++) {
4714                struct diff_filepair *p = q->queue[i];
4715
4716                if (diff_filespec_check_stat_unmatch(p))
4717                        diff_q(&outq, p);
4718                else {
4719                        /*
4720                         * The caller can subtract 1 from skip_stat_unmatch
4721                         * to determine how many paths were dirty only
4722                         * due to stat info mismatch.
4723                         */
4724                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4725                                diffopt->skip_stat_unmatch++;
4726                        diff_free_filepair(p);
4727                }
4728        }
4729        free(q->queue);
4730        *q = outq;
4731}
4732
4733static int diffnamecmp(const void *a_, const void *b_)
4734{
4735        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4736        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4737        const char *name_a, *name_b;
4738
4739        name_a = a->one ? a->one->path : a->two->path;
4740        name_b = b->one ? b->one->path : b->two->path;
4741        return strcmp(name_a, name_b);
4742}
4743
4744void diffcore_fix_diff_index(struct diff_options *options)
4745{
4746        struct diff_queue_struct *q = &diff_queued_diff;
4747        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4748}
4749
4750void diffcore_std(struct diff_options *options)
4751{
4752        if (options->skip_stat_unmatch)
4753                diffcore_skip_stat_unmatch(options);
4754        if (!options->found_follow) {
4755                /* See try_to_follow_renames() in tree-diff.c */
4756                if (options->break_opt != -1)
4757                        diffcore_break(options->break_opt);
4758                if (options->detect_rename)
4759                        diffcore_rename(options);
4760                if (options->break_opt != -1)
4761                        diffcore_merge_broken();
4762        }
4763        if (options->pickaxe)
4764                diffcore_pickaxe(options);
4765        if (options->orderfile)
4766                diffcore_order(options->orderfile);
4767        if (!options->found_follow)
4768                /* See try_to_follow_renames() in tree-diff.c */
4769                diff_resolve_rename_copy();
4770        diffcore_apply_filter(options);
4771
4772        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4773                DIFF_OPT_SET(options, HAS_CHANGES);
4774        else
4775                DIFF_OPT_CLR(options, HAS_CHANGES);
4776
4777        options->found_follow = 0;
4778}
4779
4780int diff_result_code(struct diff_options *opt, int status)
4781{
4782        int result = 0;
4783
4784        diff_warn_rename_limit("diff.renameLimit",
4785                               opt->needed_rename_limit,
4786                               opt->degraded_cc_to_c);
4787        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4788            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4789                return status;
4790        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4791            DIFF_OPT_TST(opt, HAS_CHANGES))
4792                result |= 01;
4793        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4794            DIFF_OPT_TST(opt, CHECK_FAILED))
4795                result |= 02;
4796        return result;
4797}
4798
4799int diff_can_quit_early(struct diff_options *opt)
4800{
4801        return (DIFF_OPT_TST(opt, QUICK) &&
4802                !opt->filter &&
4803                DIFF_OPT_TST(opt, HAS_CHANGES));
4804}
4805
4806/*
4807 * Shall changes to this submodule be ignored?
4808 *
4809 * Submodule changes can be configured to be ignored separately for each path,
4810 * but that configuration can be overridden from the command line.
4811 */
4812static int is_submodule_ignored(const char *path, struct diff_options *options)
4813{
4814        int ignored = 0;
4815        unsigned orig_flags = options->flags;
4816        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4817                set_diffopt_flags_from_submodule_config(options, path);
4818        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4819                ignored = 1;
4820        options->flags = orig_flags;
4821        return ignored;
4822}
4823
4824void diff_addremove(struct diff_options *options,
4825                    int addremove, unsigned mode,
4826                    const unsigned char *sha1,
4827                    int sha1_valid,
4828                    const char *concatpath, unsigned dirty_submodule)
4829{
4830        struct diff_filespec *one, *two;
4831
4832        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4833                return;
4834
4835        /* This may look odd, but it is a preparation for
4836         * feeding "there are unchanged files which should
4837         * not produce diffs, but when you are doing copy
4838         * detection you would need them, so here they are"
4839         * entries to the diff-core.  They will be prefixed
4840         * with something like '=' or '*' (I haven't decided
4841         * which but should not make any difference).
4842         * Feeding the same new and old to diff_change()
4843         * also has the same effect.
4844         * Before the final output happens, they are pruned after
4845         * merged into rename/copy pairs as appropriate.
4846         */
4847        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4848                addremove = (addremove == '+' ? '-' :
4849                             addremove == '-' ? '+' : addremove);
4850
4851        if (options->prefix &&
4852            strncmp(concatpath, options->prefix, options->prefix_length))
4853                return;
4854
4855        one = alloc_filespec(concatpath);
4856        two = alloc_filespec(concatpath);
4857
4858        if (addremove != '+')
4859                fill_filespec(one, sha1, sha1_valid, mode);
4860        if (addremove != '-') {
4861                fill_filespec(two, sha1, sha1_valid, mode);
4862                two->dirty_submodule = dirty_submodule;
4863        }
4864
4865        diff_queue(&diff_queued_diff, one, two);
4866        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4867                DIFF_OPT_SET(options, HAS_CHANGES);
4868}
4869
4870void diff_change(struct diff_options *options,
4871                 unsigned old_mode, unsigned new_mode,
4872                 const unsigned char *old_sha1,
4873                 const unsigned char *new_sha1,
4874                 int old_sha1_valid, int new_sha1_valid,
4875                 const char *concatpath,
4876                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4877{
4878        struct diff_filespec *one, *two;
4879        struct diff_filepair *p;
4880
4881        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4882            is_submodule_ignored(concatpath, options))
4883                return;
4884
4885        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4886                unsigned tmp;
4887                const unsigned char *tmp_c;
4888                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4889                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4890                tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4891                        new_sha1_valid = tmp;
4892                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4893                        new_dirty_submodule = tmp;
4894        }
4895
4896        if (options->prefix &&
4897            strncmp(concatpath, options->prefix, options->prefix_length))
4898                return;
4899
4900        one = alloc_filespec(concatpath);
4901        two = alloc_filespec(concatpath);
4902        fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4903        fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4904        one->dirty_submodule = old_dirty_submodule;
4905        two->dirty_submodule = new_dirty_submodule;
4906        p = diff_queue(&diff_queued_diff, one, two);
4907
4908        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4909                return;
4910
4911        if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
4912            !diff_filespec_check_stat_unmatch(p))
4913                return;
4914
4915        DIFF_OPT_SET(options, HAS_CHANGES);
4916}
4917
4918struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4919{
4920        struct diff_filepair *pair;
4921        struct diff_filespec *one, *two;
4922
4923        if (options->prefix &&
4924            strncmp(path, options->prefix, options->prefix_length))
4925                return NULL;
4926
4927        one = alloc_filespec(path);
4928        two = alloc_filespec(path);
4929        pair = diff_queue(&diff_queued_diff, one, two);
4930        pair->is_unmerged = 1;
4931        return pair;
4932}
4933
4934static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4935                size_t *outsize)
4936{
4937        struct diff_tempfile *temp;
4938        const char *argv[3];
4939        const char **arg = argv;
4940        struct child_process child;
4941        struct strbuf buf = STRBUF_INIT;
4942        int err = 0;
4943
4944        temp = prepare_temp_file(spec->path, spec);
4945        *arg++ = pgm;
4946        *arg++ = temp->name;
4947        *arg = NULL;
4948
4949        memset(&child, 0, sizeof(child));
4950        child.use_shell = 1;
4951        child.argv = argv;
4952        child.out = -1;
4953        if (start_command(&child)) {
4954                remove_tempfile();
4955                return NULL;
4956        }
4957
4958        if (strbuf_read(&buf, child.out, 0) < 0)
4959                err = error("error reading from textconv command '%s'", pgm);
4960        close(child.out);
4961
4962        if (finish_command(&child) || err) {
4963                strbuf_release(&buf);
4964                remove_tempfile();
4965                return NULL;
4966        }
4967        remove_tempfile();
4968
4969        return strbuf_detach(&buf, outsize);
4970}
4971
4972size_t fill_textconv(struct userdiff_driver *driver,
4973                     struct diff_filespec *df,
4974                     char **outbuf)
4975{
4976        size_t size;
4977
4978        if (!driver || !driver->textconv) {
4979                if (!DIFF_FILE_VALID(df)) {
4980                        *outbuf = "";
4981                        return 0;
4982                }
4983                if (diff_populate_filespec(df, 0))
4984                        die("unable to read files to diff");
4985                *outbuf = df->data;
4986                return df->size;
4987        }
4988
4989        if (driver->textconv_cache && df->sha1_valid) {
4990                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4991                                          &size);
4992                if (*outbuf)
4993                        return size;
4994        }
4995
4996        *outbuf = run_textconv(driver->textconv, df, &size);
4997        if (!*outbuf)
4998                die("unable to read files to diff");
4999
5000        if (driver->textconv_cache && df->sha1_valid) {
5001                /* ignore errors, as we might be in a readonly repository */
5002                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
5003                                size);
5004                /*
5005                 * we could save up changes and flush them all at the end,
5006                 * but we would need an extra call after all diffing is done.
5007                 * Since generating a cache entry is the slow path anyway,
5008                 * this extra overhead probably isn't a big deal.
5009                 */
5010                notes_cache_write(driver->textconv_cache);
5011        }
5012
5013        return size;
5014}
5015
5016void setup_diff_pager(struct diff_options *opt)
5017{
5018        /*
5019         * If the user asked for our exit code, then either they want --quiet
5020         * or --exit-code. We should definitely not bother with a pager in the
5021         * former case, as we will generate no output. Since we still properly
5022         * report our exit code even when a pager is run, we _could_ run a
5023         * pager with --exit-code. But since we have not done so historically,
5024         * and because it is easy to find people oneline advising "git diff
5025         * --exit-code" in hooks and other scripts, we do not do so.
5026         */
5027        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5028            check_pager_config("diff") != 0)
5029                setup_pager();
5030}