diff.con commit diff.c: emit_diff_symbol learns DIFF_SYMBOL_HEADER (a29b0a1)
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "config.h"
   6#include "tempfile.h"
   7#include "quote.h"
   8#include "diff.h"
   9#include "diffcore.h"
  10#include "delta.h"
  11#include "xdiff-interface.h"
  12#include "color.h"
  13#include "attr.h"
  14#include "run-command.h"
  15#include "utf8.h"
  16#include "userdiff.h"
  17#include "submodule-config.h"
  18#include "submodule.h"
  19#include "ll-merge.h"
  20#include "string-list.h"
  21#include "argv-array.h"
  22#include "graph.h"
  23
  24#ifdef NO_FAST_WORKING_DIRECTORY
  25#define FAST_WORKING_DIRECTORY 0
  26#else
  27#define FAST_WORKING_DIRECTORY 1
  28#endif
  29
  30static int diff_detect_rename_default;
  31static int diff_indent_heuristic = 1;
  32static int diff_rename_limit_default = 400;
  33static int diff_suppress_blank_empty;
  34static int diff_use_color_default = -1;
  35static int diff_context_default = 3;
  36static int diff_interhunk_context_default;
  37static const char *diff_word_regex_cfg;
  38static const char *external_diff_cmd_cfg;
  39static const char *diff_order_file_cfg;
  40int diff_auto_refresh_index = 1;
  41static int diff_mnemonic_prefix;
  42static int diff_no_prefix;
  43static int diff_stat_graph_width;
  44static int diff_dirstat_permille_default = 30;
  45static struct diff_options default_diff_options;
  46static long diff_algorithm;
  47static unsigned ws_error_highlight_default = WSEH_NEW;
  48
  49static char diff_colors[][COLOR_MAXLEN] = {
  50        GIT_COLOR_RESET,
  51        GIT_COLOR_NORMAL,       /* CONTEXT */
  52        GIT_COLOR_BOLD,         /* METAINFO */
  53        GIT_COLOR_CYAN,         /* FRAGINFO */
  54        GIT_COLOR_RED,          /* OLD */
  55        GIT_COLOR_GREEN,        /* NEW */
  56        GIT_COLOR_YELLOW,       /* COMMIT */
  57        GIT_COLOR_BG_RED,       /* WHITESPACE */
  58        GIT_COLOR_NORMAL,       /* FUNCINFO */
  59};
  60
  61static NORETURN void die_want_option(const char *option_name)
  62{
  63        die(_("option '%s' requires a value"), option_name);
  64}
  65
  66static int parse_diff_color_slot(const char *var)
  67{
  68        if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
  69                return DIFF_CONTEXT;
  70        if (!strcasecmp(var, "meta"))
  71                return DIFF_METAINFO;
  72        if (!strcasecmp(var, "frag"))
  73                return DIFF_FRAGINFO;
  74        if (!strcasecmp(var, "old"))
  75                return DIFF_FILE_OLD;
  76        if (!strcasecmp(var, "new"))
  77                return DIFF_FILE_NEW;
  78        if (!strcasecmp(var, "commit"))
  79                return DIFF_COMMIT;
  80        if (!strcasecmp(var, "whitespace"))
  81                return DIFF_WHITESPACE;
  82        if (!strcasecmp(var, "func"))
  83                return DIFF_FUNCINFO;
  84        return -1;
  85}
  86
  87static int parse_dirstat_params(struct diff_options *options, const char *params_string,
  88                                struct strbuf *errmsg)
  89{
  90        char *params_copy = xstrdup(params_string);
  91        struct string_list params = STRING_LIST_INIT_NODUP;
  92        int ret = 0;
  93        int i;
  94
  95        if (*params_copy)
  96                string_list_split_in_place(&params, params_copy, ',', -1);
  97        for (i = 0; i < params.nr; i++) {
  98                const char *p = params.items[i].string;
  99                if (!strcmp(p, "changes")) {
 100                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
 101                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
 102                } else if (!strcmp(p, "lines")) {
 103                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
 104                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
 105                } else if (!strcmp(p, "files")) {
 106                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
 107                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
 108                } else if (!strcmp(p, "noncumulative")) {
 109                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
 110                } else if (!strcmp(p, "cumulative")) {
 111                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
 112                } else if (isdigit(*p)) {
 113                        char *end;
 114                        int permille = strtoul(p, &end, 10) * 10;
 115                        if (*end == '.' && isdigit(*++end)) {
 116                                /* only use first digit */
 117                                permille += *end - '0';
 118                                /* .. and ignore any further digits */
 119                                while (isdigit(*++end))
 120                                        ; /* nothing */
 121                        }
 122                        if (!*end)
 123                                options->dirstat_permille = permille;
 124                        else {
 125                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%s'\n"),
 126                                            p);
 127                                ret++;
 128                        }
 129                } else {
 130                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%s'\n"), p);
 131                        ret++;
 132                }
 133
 134        }
 135        string_list_clear(&params, 0);
 136        free(params_copy);
 137        return ret;
 138}
 139
 140static int parse_submodule_params(struct diff_options *options, const char *value)
 141{
 142        if (!strcmp(value, "log"))
 143                options->submodule_format = DIFF_SUBMODULE_LOG;
 144        else if (!strcmp(value, "short"))
 145                options->submodule_format = DIFF_SUBMODULE_SHORT;
 146        else if (!strcmp(value, "diff"))
 147                options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
 148        else
 149                return -1;
 150        return 0;
 151}
 152
 153static int git_config_rename(const char *var, const char *value)
 154{
 155        if (!value)
 156                return DIFF_DETECT_RENAME;
 157        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 158                return  DIFF_DETECT_COPY;
 159        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 160}
 161
 162long parse_algorithm_value(const char *value)
 163{
 164        if (!value)
 165                return -1;
 166        else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
 167                return 0;
 168        else if (!strcasecmp(value, "minimal"))
 169                return XDF_NEED_MINIMAL;
 170        else if (!strcasecmp(value, "patience"))
 171                return XDF_PATIENCE_DIFF;
 172        else if (!strcasecmp(value, "histogram"))
 173                return XDF_HISTOGRAM_DIFF;
 174        return -1;
 175}
 176
 177static int parse_one_token(const char **arg, const char *token)
 178{
 179        const char *rest;
 180        if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
 181                *arg = rest;
 182                return 1;
 183        }
 184        return 0;
 185}
 186
 187static int parse_ws_error_highlight(const char *arg)
 188{
 189        const char *orig_arg = arg;
 190        unsigned val = 0;
 191
 192        while (*arg) {
 193                if (parse_one_token(&arg, "none"))
 194                        val = 0;
 195                else if (parse_one_token(&arg, "default"))
 196                        val = WSEH_NEW;
 197                else if (parse_one_token(&arg, "all"))
 198                        val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
 199                else if (parse_one_token(&arg, "new"))
 200                        val |= WSEH_NEW;
 201                else if (parse_one_token(&arg, "old"))
 202                        val |= WSEH_OLD;
 203                else if (parse_one_token(&arg, "context"))
 204                        val |= WSEH_CONTEXT;
 205                else {
 206                        return -1 - (int)(arg - orig_arg);
 207                }
 208                if (*arg)
 209                        arg++;
 210        }
 211        return val;
 212}
 213
 214/*
 215 * These are to give UI layer defaults.
 216 * The core-level commands such as git-diff-files should
 217 * never be affected by the setting of diff.renames
 218 * the user happens to have in the configuration file.
 219 */
 220void init_diff_ui_defaults(void)
 221{
 222        diff_detect_rename_default = 1;
 223}
 224
 225int git_diff_heuristic_config(const char *var, const char *value, void *cb)
 226{
 227        if (!strcmp(var, "diff.indentheuristic"))
 228                diff_indent_heuristic = git_config_bool(var, value);
 229        return 0;
 230}
 231
 232int git_diff_ui_config(const char *var, const char *value, void *cb)
 233{
 234        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 235                diff_use_color_default = git_config_colorbool(var, value);
 236                return 0;
 237        }
 238        if (!strcmp(var, "diff.context")) {
 239                diff_context_default = git_config_int(var, value);
 240                if (diff_context_default < 0)
 241                        return -1;
 242                return 0;
 243        }
 244        if (!strcmp(var, "diff.interhunkcontext")) {
 245                diff_interhunk_context_default = git_config_int(var, value);
 246                if (diff_interhunk_context_default < 0)
 247                        return -1;
 248                return 0;
 249        }
 250        if (!strcmp(var, "diff.renames")) {
 251                diff_detect_rename_default = git_config_rename(var, value);
 252                return 0;
 253        }
 254        if (!strcmp(var, "diff.autorefreshindex")) {
 255                diff_auto_refresh_index = git_config_bool(var, value);
 256                return 0;
 257        }
 258        if (!strcmp(var, "diff.mnemonicprefix")) {
 259                diff_mnemonic_prefix = git_config_bool(var, value);
 260                return 0;
 261        }
 262        if (!strcmp(var, "diff.noprefix")) {
 263                diff_no_prefix = git_config_bool(var, value);
 264                return 0;
 265        }
 266        if (!strcmp(var, "diff.statgraphwidth")) {
 267                diff_stat_graph_width = git_config_int(var, value);
 268                return 0;
 269        }
 270        if (!strcmp(var, "diff.external"))
 271                return git_config_string(&external_diff_cmd_cfg, var, value);
 272        if (!strcmp(var, "diff.wordregex"))
 273                return git_config_string(&diff_word_regex_cfg, var, value);
 274        if (!strcmp(var, "diff.orderfile"))
 275                return git_config_pathname(&diff_order_file_cfg, var, value);
 276
 277        if (!strcmp(var, "diff.ignoresubmodules"))
 278                handle_ignore_submodules_arg(&default_diff_options, value);
 279
 280        if (!strcmp(var, "diff.submodule")) {
 281                if (parse_submodule_params(&default_diff_options, value))
 282                        warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
 283                                value);
 284                return 0;
 285        }
 286
 287        if (!strcmp(var, "diff.algorithm")) {
 288                diff_algorithm = parse_algorithm_value(value);
 289                if (diff_algorithm < 0)
 290                        return -1;
 291                return 0;
 292        }
 293
 294        if (!strcmp(var, "diff.wserrorhighlight")) {
 295                int val = parse_ws_error_highlight(value);
 296                if (val < 0)
 297                        return -1;
 298                ws_error_highlight_default = val;
 299                return 0;
 300        }
 301
 302        if (git_color_config(var, value, cb) < 0)
 303                return -1;
 304
 305        return git_diff_basic_config(var, value, cb);
 306}
 307
 308int git_diff_basic_config(const char *var, const char *value, void *cb)
 309{
 310        const char *name;
 311
 312        if (!strcmp(var, "diff.renamelimit")) {
 313                diff_rename_limit_default = git_config_int(var, value);
 314                return 0;
 315        }
 316
 317        if (userdiff_config(var, value) < 0)
 318                return -1;
 319
 320        if (skip_prefix(var, "diff.color.", &name) ||
 321            skip_prefix(var, "color.diff.", &name)) {
 322                int slot = parse_diff_color_slot(name);
 323                if (slot < 0)
 324                        return 0;
 325                if (!value)
 326                        return config_error_nonbool(var);
 327                return color_parse(value, diff_colors[slot]);
 328        }
 329
 330        /* like GNU diff's --suppress-blank-empty option  */
 331        if (!strcmp(var, "diff.suppressblankempty") ||
 332                        /* for backwards compatibility */
 333                        !strcmp(var, "diff.suppress-blank-empty")) {
 334                diff_suppress_blank_empty = git_config_bool(var, value);
 335                return 0;
 336        }
 337
 338        if (!strcmp(var, "diff.dirstat")) {
 339                struct strbuf errmsg = STRBUF_INIT;
 340                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 341                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 342                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 343                                errmsg.buf);
 344                strbuf_release(&errmsg);
 345                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 346                return 0;
 347        }
 348
 349        if (starts_with(var, "submodule."))
 350                return parse_submodule_config_option(var, value);
 351
 352        if (git_diff_heuristic_config(var, value, cb) < 0)
 353                return -1;
 354
 355        return git_default_config(var, value, cb);
 356}
 357
 358static char *quote_two(const char *one, const char *two)
 359{
 360        int need_one = quote_c_style(one, NULL, NULL, 1);
 361        int need_two = quote_c_style(two, NULL, NULL, 1);
 362        struct strbuf res = STRBUF_INIT;
 363
 364        if (need_one + need_two) {
 365                strbuf_addch(&res, '"');
 366                quote_c_style(one, &res, NULL, 1);
 367                quote_c_style(two, &res, NULL, 1);
 368                strbuf_addch(&res, '"');
 369        } else {
 370                strbuf_addstr(&res, one);
 371                strbuf_addstr(&res, two);
 372        }
 373        return strbuf_detach(&res, NULL);
 374}
 375
 376static const char *external_diff(void)
 377{
 378        static const char *external_diff_cmd = NULL;
 379        static int done_preparing = 0;
 380
 381        if (done_preparing)
 382                return external_diff_cmd;
 383        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 384        if (!external_diff_cmd)
 385                external_diff_cmd = external_diff_cmd_cfg;
 386        done_preparing = 1;
 387        return external_diff_cmd;
 388}
 389
 390/*
 391 * Keep track of files used for diffing. Sometimes such an entry
 392 * refers to a temporary file, sometimes to an existing file, and
 393 * sometimes to "/dev/null".
 394 */
 395static struct diff_tempfile {
 396        /*
 397         * filename external diff should read from, or NULL if this
 398         * entry is currently not in use:
 399         */
 400        const char *name;
 401
 402        char hex[GIT_MAX_HEXSZ + 1];
 403        char mode[10];
 404
 405        /*
 406         * If this diff_tempfile instance refers to a temporary file,
 407         * this tempfile object is used to manage its lifetime.
 408         */
 409        struct tempfile tempfile;
 410} diff_temp[2];
 411
 412typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 413
 414struct emit_callback {
 415        int color_diff;
 416        unsigned ws_rule;
 417        int blank_at_eof_in_preimage;
 418        int blank_at_eof_in_postimage;
 419        int lno_in_preimage;
 420        int lno_in_postimage;
 421        sane_truncate_fn truncate;
 422        const char **label_path;
 423        struct diff_words_data *diff_words;
 424        struct diff_options *opt;
 425        struct strbuf *header;
 426};
 427
 428static int count_lines(const char *data, int size)
 429{
 430        int count, ch, completely_empty = 1, nl_just_seen = 0;
 431        count = 0;
 432        while (0 < size--) {
 433                ch = *data++;
 434                if (ch == '\n') {
 435                        count++;
 436                        nl_just_seen = 1;
 437                        completely_empty = 0;
 438                }
 439                else {
 440                        nl_just_seen = 0;
 441                        completely_empty = 0;
 442                }
 443        }
 444        if (completely_empty)
 445                return 0;
 446        if (!nl_just_seen)
 447                count++; /* no trailing newline */
 448        return count;
 449}
 450
 451static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 452{
 453        if (!DIFF_FILE_VALID(one)) {
 454                mf->ptr = (char *)""; /* does not matter */
 455                mf->size = 0;
 456                return 0;
 457        }
 458        else if (diff_populate_filespec(one, 0))
 459                return -1;
 460
 461        mf->ptr = one->data;
 462        mf->size = one->size;
 463        return 0;
 464}
 465
 466/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 467static unsigned long diff_filespec_size(struct diff_filespec *one)
 468{
 469        if (!DIFF_FILE_VALID(one))
 470                return 0;
 471        diff_populate_filespec(one, CHECK_SIZE_ONLY);
 472        return one->size;
 473}
 474
 475static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 476{
 477        char *ptr = mf->ptr;
 478        long size = mf->size;
 479        int cnt = 0;
 480
 481        if (!size)
 482                return cnt;
 483        ptr += size - 1; /* pointing at the very end */
 484        if (*ptr != '\n')
 485                ; /* incomplete line */
 486        else
 487                ptr--; /* skip the last LF */
 488        while (mf->ptr < ptr) {
 489                char *prev_eol;
 490                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 491                        if (*prev_eol == '\n')
 492                                break;
 493                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 494                        break;
 495                cnt++;
 496                ptr = prev_eol - 1;
 497        }
 498        return cnt;
 499}
 500
 501static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 502                               struct emit_callback *ecbdata)
 503{
 504        int l1, l2, at;
 505        unsigned ws_rule = ecbdata->ws_rule;
 506        l1 = count_trailing_blank(mf1, ws_rule);
 507        l2 = count_trailing_blank(mf2, ws_rule);
 508        if (l2 <= l1) {
 509                ecbdata->blank_at_eof_in_preimage = 0;
 510                ecbdata->blank_at_eof_in_postimage = 0;
 511                return;
 512        }
 513        at = count_lines(mf1->ptr, mf1->size);
 514        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 515
 516        at = count_lines(mf2->ptr, mf2->size);
 517        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 518}
 519
 520static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 521                        int first, const char *line, int len)
 522{
 523        int has_trailing_newline, has_trailing_carriage_return;
 524        int nofirst;
 525        FILE *file = o->file;
 526
 527        fputs(diff_line_prefix(o), file);
 528
 529        if (len == 0) {
 530                has_trailing_newline = (first == '\n');
 531                has_trailing_carriage_return = (!has_trailing_newline &&
 532                                                (first == '\r'));
 533                nofirst = has_trailing_newline || has_trailing_carriage_return;
 534        } else {
 535                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 536                if (has_trailing_newline)
 537                        len--;
 538                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 539                if (has_trailing_carriage_return)
 540                        len--;
 541                nofirst = 0;
 542        }
 543
 544        if (len || !nofirst) {
 545                fputs(set, file);
 546                if (!nofirst)
 547                        fputc(first, file);
 548                fwrite(line, len, 1, file);
 549                fputs(reset, file);
 550        }
 551        if (has_trailing_carriage_return)
 552                fputc('\r', file);
 553        if (has_trailing_newline)
 554                fputc('\n', file);
 555}
 556
 557static void emit_line(struct diff_options *o, const char *set, const char *reset,
 558                      const char *line, int len)
 559{
 560        emit_line_0(o, set, reset, line[0], line+1, len-1);
 561}
 562
 563enum diff_symbol {
 564        DIFF_SYMBOL_HEADER,
 565        DIFF_SYMBOL_FILEPAIR_PLUS,
 566        DIFF_SYMBOL_FILEPAIR_MINUS,
 567        DIFF_SYMBOL_WORDS_PORCELAIN,
 568        DIFF_SYMBOL_WORDS,
 569        DIFF_SYMBOL_CONTEXT,
 570        DIFF_SYMBOL_CONTEXT_INCOMPLETE,
 571        DIFF_SYMBOL_PLUS,
 572        DIFF_SYMBOL_MINUS,
 573        DIFF_SYMBOL_NO_LF_EOF,
 574        DIFF_SYMBOL_CONTEXT_FRAGINFO,
 575        DIFF_SYMBOL_CONTEXT_MARKER,
 576        DIFF_SYMBOL_SEPARATOR
 577};
 578/*
 579 * Flags for content lines:
 580 * 0..12 are whitespace rules
 581 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
 582 * 16 is marking if the line is blank at EOF
 583 */
 584#define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
 585#define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
 586
 587static void emit_line_ws_markup(struct diff_options *o,
 588                                const char *set, const char *reset,
 589                                const char *line, int len, char sign,
 590                                unsigned ws_rule, int blank_at_eof)
 591{
 592        const char *ws = NULL;
 593
 594        if (o->ws_error_highlight & ws_rule) {
 595                ws = diff_get_color_opt(o, DIFF_WHITESPACE);
 596                if (!*ws)
 597                        ws = NULL;
 598        }
 599
 600        if (!ws)
 601                emit_line_0(o, set, reset, sign, line, len);
 602        else if (blank_at_eof)
 603                /* Blank line at EOF - paint '+' as well */
 604                emit_line_0(o, ws, reset, sign, line, len);
 605        else {
 606                /* Emit just the prefix, then the rest. */
 607                emit_line_0(o, set, reset, sign, "", 0);
 608                ws_check_emit(line, len, ws_rule,
 609                              o->file, set, reset, ws);
 610        }
 611}
 612
 613static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
 614                             const char *line, int len, unsigned flags)
 615{
 616        static const char *nneof = " No newline at end of file\n";
 617        const char *context, *reset, *set, *meta;
 618        switch (s) {
 619        case DIFF_SYMBOL_NO_LF_EOF:
 620                context = diff_get_color_opt(o, DIFF_CONTEXT);
 621                reset = diff_get_color_opt(o, DIFF_RESET);
 622                putc('\n', o->file);
 623                emit_line_0(o, context, reset, '\\',
 624                            nneof, strlen(nneof));
 625                break;
 626        case DIFF_SYMBOL_CONTEXT_FRAGINFO:
 627                emit_line(o, "", "", line, len);
 628                break;
 629        case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
 630        case DIFF_SYMBOL_CONTEXT_MARKER:
 631                context = diff_get_color_opt(o, DIFF_CONTEXT);
 632                reset = diff_get_color_opt(o, DIFF_RESET);
 633                emit_line(o, context, reset, line, len);
 634                break;
 635        case DIFF_SYMBOL_SEPARATOR:
 636                fprintf(o->file, "%s%c",
 637                        diff_line_prefix(o),
 638                        o->line_termination);
 639                break;
 640        case DIFF_SYMBOL_CONTEXT:
 641                set = diff_get_color_opt(o, DIFF_CONTEXT);
 642                reset = diff_get_color_opt(o, DIFF_RESET);
 643                emit_line_ws_markup(o, set, reset, line, len, ' ',
 644                                    flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
 645                break;
 646        case DIFF_SYMBOL_PLUS:
 647                set = diff_get_color_opt(o, DIFF_FILE_NEW);
 648                reset = diff_get_color_opt(o, DIFF_RESET);
 649                emit_line_ws_markup(o, set, reset, line, len, '+',
 650                                    flags & DIFF_SYMBOL_CONTENT_WS_MASK,
 651                                    flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
 652                break;
 653        case DIFF_SYMBOL_MINUS:
 654                set = diff_get_color_opt(o, DIFF_FILE_OLD);
 655                reset = diff_get_color_opt(o, DIFF_RESET);
 656                emit_line_ws_markup(o, set, reset, line, len, '-',
 657                                    flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
 658                break;
 659        case DIFF_SYMBOL_WORDS_PORCELAIN:
 660                context = diff_get_color_opt(o, DIFF_CONTEXT);
 661                reset = diff_get_color_opt(o, DIFF_RESET);
 662                emit_line(o, context, reset, line, len);
 663                fputs("~\n", o->file);
 664                break;
 665        case DIFF_SYMBOL_WORDS:
 666                context = diff_get_color_opt(o, DIFF_CONTEXT);
 667                reset = diff_get_color_opt(o, DIFF_RESET);
 668                /*
 669                 * Skip the prefix character, if any.  With
 670                 * diff_suppress_blank_empty, there may be
 671                 * none.
 672                 */
 673                if (line[0] != '\n') {
 674                        line++;
 675                        len--;
 676                }
 677                emit_line(o, context, reset, line, len);
 678                break;
 679        case DIFF_SYMBOL_FILEPAIR_PLUS:
 680                meta = diff_get_color_opt(o, DIFF_METAINFO);
 681                reset = diff_get_color_opt(o, DIFF_RESET);
 682                fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
 683                        line, reset,
 684                        strchr(line, ' ') ? "\t" : "");
 685                break;
 686        case DIFF_SYMBOL_FILEPAIR_MINUS:
 687                meta = diff_get_color_opt(o, DIFF_METAINFO);
 688                reset = diff_get_color_opt(o, DIFF_RESET);
 689                fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
 690                        line, reset,
 691                        strchr(line, ' ') ? "\t" : "");
 692                break;
 693        case DIFF_SYMBOL_HEADER:
 694                fprintf(o->file, "%s", line);
 695                break;
 696        default:
 697                die("BUG: unknown diff symbol");
 698        }
 699}
 700
 701static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 702{
 703        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 704              ecbdata->blank_at_eof_in_preimage &&
 705              ecbdata->blank_at_eof_in_postimage &&
 706              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 707              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 708                return 0;
 709        return ws_blank_line(line, len, ecbdata->ws_rule);
 710}
 711
 712static void emit_add_line(const char *reset,
 713                          struct emit_callback *ecbdata,
 714                          const char *line, int len)
 715{
 716        unsigned flags = WSEH_NEW | ecbdata->ws_rule;
 717        if (new_blank_line_at_eof(ecbdata, line, len))
 718                flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
 719
 720        emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
 721}
 722
 723static void emit_del_line(const char *reset,
 724                          struct emit_callback *ecbdata,
 725                          const char *line, int len)
 726{
 727        unsigned flags = WSEH_OLD | ecbdata->ws_rule;
 728        emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
 729}
 730
 731static void emit_context_line(const char *reset,
 732                              struct emit_callback *ecbdata,
 733                              const char *line, int len)
 734{
 735        unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
 736        emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
 737}
 738
 739static void emit_hunk_header(struct emit_callback *ecbdata,
 740                             const char *line, int len)
 741{
 742        const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
 743        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 744        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 745        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 746        static const char atat[2] = { '@', '@' };
 747        const char *cp, *ep;
 748        struct strbuf msgbuf = STRBUF_INIT;
 749        int org_len = len;
 750        int i = 1;
 751
 752        /*
 753         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 754         * it always is at least 10 bytes long.
 755         */
 756        if (len < 10 ||
 757            memcmp(line, atat, 2) ||
 758            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 759                emit_diff_symbol(ecbdata->opt,
 760                                 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
 761                return;
 762        }
 763        ep += 2; /* skip over @@ */
 764
 765        /* The hunk header in fraginfo color */
 766        strbuf_addstr(&msgbuf, frag);
 767        strbuf_add(&msgbuf, line, ep - line);
 768        strbuf_addstr(&msgbuf, reset);
 769
 770        /*
 771         * trailing "\r\n"
 772         */
 773        for ( ; i < 3; i++)
 774                if (line[len - i] == '\r' || line[len - i] == '\n')
 775                        len--;
 776
 777        /* blank before the func header */
 778        for (cp = ep; ep - line < len; ep++)
 779                if (*ep != ' ' && *ep != '\t')
 780                        break;
 781        if (ep != cp) {
 782                strbuf_addstr(&msgbuf, context);
 783                strbuf_add(&msgbuf, cp, ep - cp);
 784                strbuf_addstr(&msgbuf, reset);
 785        }
 786
 787        if (ep < line + len) {
 788                strbuf_addstr(&msgbuf, func);
 789                strbuf_add(&msgbuf, ep, line + len - ep);
 790                strbuf_addstr(&msgbuf, reset);
 791        }
 792
 793        strbuf_add(&msgbuf, line + len, org_len - len);
 794        strbuf_complete_line(&msgbuf);
 795        emit_diff_symbol(ecbdata->opt,
 796                         DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
 797        strbuf_release(&msgbuf);
 798}
 799
 800static struct diff_tempfile *claim_diff_tempfile(void) {
 801        int i;
 802        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 803                if (!diff_temp[i].name)
 804                        return diff_temp + i;
 805        die("BUG: diff is failing to clean up its tempfiles");
 806}
 807
 808static void remove_tempfile(void)
 809{
 810        int i;
 811        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 812                if (is_tempfile_active(&diff_temp[i].tempfile))
 813                        delete_tempfile(&diff_temp[i].tempfile);
 814                diff_temp[i].name = NULL;
 815        }
 816}
 817
 818static void print_line_count(FILE *file, int count)
 819{
 820        switch (count) {
 821        case 0:
 822                fprintf(file, "0,0");
 823                break;
 824        case 1:
 825                fprintf(file, "1");
 826                break;
 827        default:
 828                fprintf(file, "1,%d", count);
 829                break;
 830        }
 831}
 832
 833static void emit_rewrite_lines(struct emit_callback *ecb,
 834                               int prefix, const char *data, int size)
 835{
 836        const char *endp = NULL;
 837        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 838
 839        while (0 < size) {
 840                int len;
 841
 842                endp = memchr(data, '\n', size);
 843                len = endp ? (endp - data + 1) : size;
 844                if (prefix != '+') {
 845                        ecb->lno_in_preimage++;
 846                        emit_del_line(reset, ecb, data, len);
 847                } else {
 848                        ecb->lno_in_postimage++;
 849                        emit_add_line(reset, ecb, data, len);
 850                }
 851                size -= len;
 852                data += len;
 853        }
 854        if (!endp)
 855                emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
 856}
 857
 858static void emit_rewrite_diff(const char *name_a,
 859                              const char *name_b,
 860                              struct diff_filespec *one,
 861                              struct diff_filespec *two,
 862                              struct userdiff_driver *textconv_one,
 863                              struct userdiff_driver *textconv_two,
 864                              struct diff_options *o)
 865{
 866        int lc_a, lc_b;
 867        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 868        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 869        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 870        const char *a_prefix, *b_prefix;
 871        char *data_one, *data_two;
 872        size_t size_one, size_two;
 873        struct emit_callback ecbdata;
 874        const char *line_prefix = diff_line_prefix(o);
 875
 876        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 877                a_prefix = o->b_prefix;
 878                b_prefix = o->a_prefix;
 879        } else {
 880                a_prefix = o->a_prefix;
 881                b_prefix = o->b_prefix;
 882        }
 883
 884        name_a += (*name_a == '/');
 885        name_b += (*name_b == '/');
 886
 887        strbuf_reset(&a_name);
 888        strbuf_reset(&b_name);
 889        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 890        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 891
 892        size_one = fill_textconv(textconv_one, one, &data_one);
 893        size_two = fill_textconv(textconv_two, two, &data_two);
 894
 895        memset(&ecbdata, 0, sizeof(ecbdata));
 896        ecbdata.color_diff = want_color(o->use_color);
 897        ecbdata.ws_rule = whitespace_rule(name_b);
 898        ecbdata.opt = o;
 899        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 900                mmfile_t mf1, mf2;
 901                mf1.ptr = (char *)data_one;
 902                mf2.ptr = (char *)data_two;
 903                mf1.size = size_one;
 904                mf2.size = size_two;
 905                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 906        }
 907        ecbdata.lno_in_preimage = 1;
 908        ecbdata.lno_in_postimage = 1;
 909
 910        lc_a = count_lines(data_one, size_one);
 911        lc_b = count_lines(data_two, size_two);
 912
 913        emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
 914                         a_name.buf, a_name.len, 0);
 915        emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
 916                         b_name.buf, b_name.len, 0);
 917
 918        fprintf(o->file, "%s%s@@ -", line_prefix, fraginfo);
 919        if (!o->irreversible_delete)
 920                print_line_count(o->file, lc_a);
 921        else
 922                fprintf(o->file, "?,?");
 923        fprintf(o->file, " +");
 924        print_line_count(o->file, lc_b);
 925        fprintf(o->file, " @@%s\n", reset);
 926        if (lc_a && !o->irreversible_delete)
 927                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 928        if (lc_b)
 929                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 930        if (textconv_one)
 931                free((char *)data_one);
 932        if (textconv_two)
 933                free((char *)data_two);
 934}
 935
 936struct diff_words_buffer {
 937        mmfile_t text;
 938        long alloc;
 939        struct diff_words_orig {
 940                const char *begin, *end;
 941        } *orig;
 942        int orig_nr, orig_alloc;
 943};
 944
 945static void diff_words_append(char *line, unsigned long len,
 946                struct diff_words_buffer *buffer)
 947{
 948        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 949        line++;
 950        len--;
 951        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 952        buffer->text.size += len;
 953        buffer->text.ptr[buffer->text.size] = '\0';
 954}
 955
 956struct diff_words_style_elem {
 957        const char *prefix;
 958        const char *suffix;
 959        const char *color; /* NULL; filled in by the setup code if
 960                            * color is enabled */
 961};
 962
 963struct diff_words_style {
 964        enum diff_words_type type;
 965        struct diff_words_style_elem new, old, ctx;
 966        const char *newline;
 967};
 968
 969static struct diff_words_style diff_words_styles[] = {
 970        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 971        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 972        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 973};
 974
 975struct diff_words_data {
 976        struct diff_words_buffer minus, plus;
 977        const char *current_plus;
 978        int last_minus;
 979        struct diff_options *opt;
 980        regex_t *word_regex;
 981        enum diff_words_type type;
 982        struct diff_words_style *style;
 983};
 984
 985static int fn_out_diff_words_write_helper(FILE *fp,
 986                                          struct diff_words_style_elem *st_el,
 987                                          const char *newline,
 988                                          size_t count, const char *buf,
 989                                          const char *line_prefix)
 990{
 991        int print = 0;
 992
 993        while (count) {
 994                char *p = memchr(buf, '\n', count);
 995                if (print)
 996                        fputs(line_prefix, fp);
 997                if (p != buf) {
 998                        if (st_el->color && fputs(st_el->color, fp) < 0)
 999                                return -1;
1000                        if (fputs(st_el->prefix, fp) < 0 ||
1001                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
1002                            fputs(st_el->suffix, fp) < 0)
1003                                return -1;
1004                        if (st_el->color && *st_el->color
1005                            && fputs(GIT_COLOR_RESET, fp) < 0)
1006                                return -1;
1007                }
1008                if (!p)
1009                        return 0;
1010                if (fputs(newline, fp) < 0)
1011                        return -1;
1012                count -= p + 1 - buf;
1013                buf = p + 1;
1014                print = 1;
1015        }
1016        return 0;
1017}
1018
1019/*
1020 * '--color-words' algorithm can be described as:
1021 *
1022 *   1. collect the minus/plus lines of a diff hunk, divided into
1023 *      minus-lines and plus-lines;
1024 *
1025 *   2. break both minus-lines and plus-lines into words and
1026 *      place them into two mmfile_t with one word for each line;
1027 *
1028 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1029 *
1030 * And for the common parts of the both file, we output the plus side text.
1031 * diff_words->current_plus is used to trace the current position of the plus file
1032 * which printed. diff_words->last_minus is used to trace the last minus word
1033 * printed.
1034 *
1035 * For '--graph' to work with '--color-words', we need to output the graph prefix
1036 * on each line of color words output. Generally, there are two conditions on
1037 * which we should output the prefix.
1038 *
1039 *   1. diff_words->last_minus == 0 &&
1040 *      diff_words->current_plus == diff_words->plus.text.ptr
1041 *
1042 *      that is: the plus text must start as a new line, and if there is no minus
1043 *      word printed, a graph prefix must be printed.
1044 *
1045 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
1046 *      *(diff_words->current_plus - 1) == '\n'
1047 *
1048 *      that is: a graph prefix must be printed following a '\n'
1049 */
1050static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1051{
1052        if ((diff_words->last_minus == 0 &&
1053                diff_words->current_plus == diff_words->plus.text.ptr) ||
1054                (diff_words->current_plus > diff_words->plus.text.ptr &&
1055                *(diff_words->current_plus - 1) == '\n')) {
1056                return 1;
1057        } else {
1058                return 0;
1059        }
1060}
1061
1062static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1063{
1064        struct diff_words_data *diff_words = priv;
1065        struct diff_words_style *style = diff_words->style;
1066        int minus_first, minus_len, plus_first, plus_len;
1067        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1068        struct diff_options *opt = diff_words->opt;
1069        const char *line_prefix;
1070
1071        if (line[0] != '@' || parse_hunk_header(line, len,
1072                        &minus_first, &minus_len, &plus_first, &plus_len))
1073                return;
1074
1075        assert(opt);
1076        line_prefix = diff_line_prefix(opt);
1077
1078        /* POSIX requires that first be decremented by one if len == 0... */
1079        if (minus_len) {
1080                minus_begin = diff_words->minus.orig[minus_first].begin;
1081                minus_end =
1082                        diff_words->minus.orig[minus_first + minus_len - 1].end;
1083        } else
1084                minus_begin = minus_end =
1085                        diff_words->minus.orig[minus_first].end;
1086
1087        if (plus_len) {
1088                plus_begin = diff_words->plus.orig[plus_first].begin;
1089                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1090        } else
1091                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1092
1093        if (color_words_output_graph_prefix(diff_words)) {
1094                fputs(line_prefix, diff_words->opt->file);
1095        }
1096        if (diff_words->current_plus != plus_begin) {
1097                fn_out_diff_words_write_helper(diff_words->opt->file,
1098                                &style->ctx, style->newline,
1099                                plus_begin - diff_words->current_plus,
1100                                diff_words->current_plus, line_prefix);
1101                if (*(plus_begin - 1) == '\n')
1102                        fputs(line_prefix, diff_words->opt->file);
1103        }
1104        if (minus_begin != minus_end) {
1105                fn_out_diff_words_write_helper(diff_words->opt->file,
1106                                &style->old, style->newline,
1107                                minus_end - minus_begin, minus_begin,
1108                                line_prefix);
1109        }
1110        if (plus_begin != plus_end) {
1111                fn_out_diff_words_write_helper(diff_words->opt->file,
1112                                &style->new, style->newline,
1113                                plus_end - plus_begin, plus_begin,
1114                                line_prefix);
1115        }
1116
1117        diff_words->current_plus = plus_end;
1118        diff_words->last_minus = minus_first;
1119}
1120
1121/* This function starts looking at *begin, and returns 0 iff a word was found. */
1122static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1123                int *begin, int *end)
1124{
1125        if (word_regex && *begin < buffer->size) {
1126                regmatch_t match[1];
1127                if (!regexec_buf(word_regex, buffer->ptr + *begin,
1128                                 buffer->size - *begin, 1, match, 0)) {
1129                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1130                                        '\n', match[0].rm_eo - match[0].rm_so);
1131                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1132                        *begin += match[0].rm_so;
1133                        return *begin >= *end;
1134                }
1135                return -1;
1136        }
1137
1138        /* find the next word */
1139        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1140                (*begin)++;
1141        if (*begin >= buffer->size)
1142                return -1;
1143
1144        /* find the end of the word */
1145        *end = *begin + 1;
1146        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1147                (*end)++;
1148
1149        return 0;
1150}
1151
1152/*
1153 * This function splits the words in buffer->text, stores the list with
1154 * newline separator into out, and saves the offsets of the original words
1155 * in buffer->orig.
1156 */
1157static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1158                regex_t *word_regex)
1159{
1160        int i, j;
1161        long alloc = 0;
1162
1163        out->size = 0;
1164        out->ptr = NULL;
1165
1166        /* fake an empty "0th" word */
1167        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1168        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1169        buffer->orig_nr = 1;
1170
1171        for (i = 0; i < buffer->text.size; i++) {
1172                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1173                        return;
1174
1175                /* store original boundaries */
1176                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1177                                buffer->orig_alloc);
1178                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1179                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1180                buffer->orig_nr++;
1181
1182                /* store one word */
1183                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1184                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1185                out->ptr[out->size + j - i] = '\n';
1186                out->size += j - i + 1;
1187
1188                i = j - 1;
1189        }
1190}
1191
1192/* this executes the word diff on the accumulated buffers */
1193static void diff_words_show(struct diff_words_data *diff_words)
1194{
1195        xpparam_t xpp;
1196        xdemitconf_t xecfg;
1197        mmfile_t minus, plus;
1198        struct diff_words_style *style = diff_words->style;
1199
1200        struct diff_options *opt = diff_words->opt;
1201        const char *line_prefix;
1202
1203        assert(opt);
1204        line_prefix = diff_line_prefix(opt);
1205
1206        /* special case: only removal */
1207        if (!diff_words->plus.text.size) {
1208                fputs(line_prefix, diff_words->opt->file);
1209                fn_out_diff_words_write_helper(diff_words->opt->file,
1210                        &style->old, style->newline,
1211                        diff_words->minus.text.size,
1212                        diff_words->minus.text.ptr, line_prefix);
1213                diff_words->minus.text.size = 0;
1214                return;
1215        }
1216
1217        diff_words->current_plus = diff_words->plus.text.ptr;
1218        diff_words->last_minus = 0;
1219
1220        memset(&xpp, 0, sizeof(xpp));
1221        memset(&xecfg, 0, sizeof(xecfg));
1222        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1223        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1224        xpp.flags = 0;
1225        /* as only the hunk header will be parsed, we need a 0-context */
1226        xecfg.ctxlen = 0;
1227        if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1228                          &xpp, &xecfg))
1229                die("unable to generate word diff");
1230        free(minus.ptr);
1231        free(plus.ptr);
1232        if (diff_words->current_plus != diff_words->plus.text.ptr +
1233                        diff_words->plus.text.size) {
1234                if (color_words_output_graph_prefix(diff_words))
1235                        fputs(line_prefix, diff_words->opt->file);
1236                fn_out_diff_words_write_helper(diff_words->opt->file,
1237                        &style->ctx, style->newline,
1238                        diff_words->plus.text.ptr + diff_words->plus.text.size
1239                        - diff_words->current_plus, diff_words->current_plus,
1240                        line_prefix);
1241        }
1242        diff_words->minus.text.size = diff_words->plus.text.size = 0;
1243}
1244
1245/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1246static void diff_words_flush(struct emit_callback *ecbdata)
1247{
1248        if (ecbdata->diff_words->minus.text.size ||
1249            ecbdata->diff_words->plus.text.size)
1250                diff_words_show(ecbdata->diff_words);
1251}
1252
1253static void diff_filespec_load_driver(struct diff_filespec *one)
1254{
1255        /* Use already-loaded driver */
1256        if (one->driver)
1257                return;
1258
1259        if (S_ISREG(one->mode))
1260                one->driver = userdiff_find_by_path(one->path);
1261
1262        /* Fallback to default settings */
1263        if (!one->driver)
1264                one->driver = userdiff_find_by_name("default");
1265}
1266
1267static const char *userdiff_word_regex(struct diff_filespec *one)
1268{
1269        diff_filespec_load_driver(one);
1270        return one->driver->word_regex;
1271}
1272
1273static void init_diff_words_data(struct emit_callback *ecbdata,
1274                                 struct diff_options *orig_opts,
1275                                 struct diff_filespec *one,
1276                                 struct diff_filespec *two)
1277{
1278        int i;
1279        struct diff_options *o = xmalloc(sizeof(struct diff_options));
1280        memcpy(o, orig_opts, sizeof(struct diff_options));
1281
1282        ecbdata->diff_words =
1283                xcalloc(1, sizeof(struct diff_words_data));
1284        ecbdata->diff_words->type = o->word_diff;
1285        ecbdata->diff_words->opt = o;
1286        if (!o->word_regex)
1287                o->word_regex = userdiff_word_regex(one);
1288        if (!o->word_regex)
1289                o->word_regex = userdiff_word_regex(two);
1290        if (!o->word_regex)
1291                o->word_regex = diff_word_regex_cfg;
1292        if (o->word_regex) {
1293                ecbdata->diff_words->word_regex = (regex_t *)
1294                        xmalloc(sizeof(regex_t));
1295                if (regcomp(ecbdata->diff_words->word_regex,
1296                            o->word_regex,
1297                            REG_EXTENDED | REG_NEWLINE))
1298                        die ("Invalid regular expression: %s",
1299                             o->word_regex);
1300        }
1301        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1302                if (o->word_diff == diff_words_styles[i].type) {
1303                        ecbdata->diff_words->style =
1304                                &diff_words_styles[i];
1305                        break;
1306                }
1307        }
1308        if (want_color(o->use_color)) {
1309                struct diff_words_style *st = ecbdata->diff_words->style;
1310                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1311                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1312                st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1313        }
1314}
1315
1316static void free_diff_words_data(struct emit_callback *ecbdata)
1317{
1318        if (ecbdata->diff_words) {
1319                diff_words_flush(ecbdata);
1320                free (ecbdata->diff_words->opt);
1321                free (ecbdata->diff_words->minus.text.ptr);
1322                free (ecbdata->diff_words->minus.orig);
1323                free (ecbdata->diff_words->plus.text.ptr);
1324                free (ecbdata->diff_words->plus.orig);
1325                if (ecbdata->diff_words->word_regex) {
1326                        regfree(ecbdata->diff_words->word_regex);
1327                        free(ecbdata->diff_words->word_regex);
1328                }
1329                FREE_AND_NULL(ecbdata->diff_words);
1330        }
1331}
1332
1333const char *diff_get_color(int diff_use_color, enum color_diff ix)
1334{
1335        if (want_color(diff_use_color))
1336                return diff_colors[ix];
1337        return "";
1338}
1339
1340const char *diff_line_prefix(struct diff_options *opt)
1341{
1342        struct strbuf *msgbuf;
1343        if (!opt->output_prefix)
1344                return "";
1345
1346        msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1347        return msgbuf->buf;
1348}
1349
1350static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1351{
1352        const char *cp;
1353        unsigned long allot;
1354        size_t l = len;
1355
1356        if (ecb->truncate)
1357                return ecb->truncate(line, len);
1358        cp = line;
1359        allot = l;
1360        while (0 < l) {
1361                (void) utf8_width(&cp, &l);
1362                if (!cp)
1363                        break; /* truncated in the middle? */
1364        }
1365        return allot - l;
1366}
1367
1368static void find_lno(const char *line, struct emit_callback *ecbdata)
1369{
1370        const char *p;
1371        ecbdata->lno_in_preimage = 0;
1372        ecbdata->lno_in_postimage = 0;
1373        p = strchr(line, '-');
1374        if (!p)
1375                return; /* cannot happen */
1376        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1377        p = strchr(p, '+');
1378        if (!p)
1379                return; /* cannot happen */
1380        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1381}
1382
1383static void fn_out_consume(void *priv, char *line, unsigned long len)
1384{
1385        struct emit_callback *ecbdata = priv;
1386        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1387        struct diff_options *o = ecbdata->opt;
1388
1389        o->found_changes = 1;
1390
1391        if (ecbdata->header) {
1392                emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
1393                                 ecbdata->header->buf, ecbdata->header->len, 0);
1394                strbuf_reset(ecbdata->header);
1395                ecbdata->header = NULL;
1396        }
1397
1398        if (ecbdata->label_path[0]) {
1399                emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1400                                 ecbdata->label_path[0],
1401                                 strlen(ecbdata->label_path[0]), 0);
1402                emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1403                                 ecbdata->label_path[1],
1404                                 strlen(ecbdata->label_path[1]), 0);
1405                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1406        }
1407
1408        if (diff_suppress_blank_empty
1409            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1410                line[0] = '\n';
1411                len = 1;
1412        }
1413
1414        if (line[0] == '@') {
1415                if (ecbdata->diff_words)
1416                        diff_words_flush(ecbdata);
1417                len = sane_truncate_line(ecbdata, line, len);
1418                find_lno(line, ecbdata);
1419                emit_hunk_header(ecbdata, line, len);
1420                return;
1421        }
1422
1423        if (ecbdata->diff_words) {
1424                enum diff_symbol s =
1425                        ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
1426                        DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
1427                if (line[0] == '-') {
1428                        diff_words_append(line, len,
1429                                          &ecbdata->diff_words->minus);
1430                        return;
1431                } else if (line[0] == '+') {
1432                        diff_words_append(line, len,
1433                                          &ecbdata->diff_words->plus);
1434                        return;
1435                } else if (starts_with(line, "\\ ")) {
1436                        /*
1437                         * Eat the "no newline at eof" marker as if we
1438                         * saw a "+" or "-" line with nothing on it,
1439                         * and return without diff_words_flush() to
1440                         * defer processing. If this is the end of
1441                         * preimage, more "+" lines may come after it.
1442                         */
1443                        return;
1444                }
1445                diff_words_flush(ecbdata);
1446                emit_diff_symbol(o, s, line, len, 0);
1447                return;
1448        }
1449
1450        switch (line[0]) {
1451        case '+':
1452                ecbdata->lno_in_postimage++;
1453                emit_add_line(reset, ecbdata, line + 1, len - 1);
1454                break;
1455        case '-':
1456                ecbdata->lno_in_preimage++;
1457                emit_del_line(reset, ecbdata, line + 1, len - 1);
1458                break;
1459        case ' ':
1460                ecbdata->lno_in_postimage++;
1461                ecbdata->lno_in_preimage++;
1462                emit_context_line(reset, ecbdata, line + 1, len - 1);
1463                break;
1464        default:
1465                /* incomplete line at the end */
1466                ecbdata->lno_in_preimage++;
1467                emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
1468                                 line, len, 0);
1469                break;
1470        }
1471}
1472
1473static char *pprint_rename(const char *a, const char *b)
1474{
1475        const char *old = a;
1476        const char *new = b;
1477        struct strbuf name = STRBUF_INIT;
1478        int pfx_length, sfx_length;
1479        int pfx_adjust_for_slash;
1480        int len_a = strlen(a);
1481        int len_b = strlen(b);
1482        int a_midlen, b_midlen;
1483        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1484        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1485
1486        if (qlen_a || qlen_b) {
1487                quote_c_style(a, &name, NULL, 0);
1488                strbuf_addstr(&name, " => ");
1489                quote_c_style(b, &name, NULL, 0);
1490                return strbuf_detach(&name, NULL);
1491        }
1492
1493        /* Find common prefix */
1494        pfx_length = 0;
1495        while (*old && *new && *old == *new) {
1496                if (*old == '/')
1497                        pfx_length = old - a + 1;
1498                old++;
1499                new++;
1500        }
1501
1502        /* Find common suffix */
1503        old = a + len_a;
1504        new = b + len_b;
1505        sfx_length = 0;
1506        /*
1507         * If there is a common prefix, it must end in a slash.  In
1508         * that case we let this loop run 1 into the prefix to see the
1509         * same slash.
1510         *
1511         * If there is no common prefix, we cannot do this as it would
1512         * underrun the input strings.
1513         */
1514        pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1515        while (a + pfx_length - pfx_adjust_for_slash <= old &&
1516               b + pfx_length - pfx_adjust_for_slash <= new &&
1517               *old == *new) {
1518                if (*old == '/')
1519                        sfx_length = len_a - (old - a);
1520                old--;
1521                new--;
1522        }
1523
1524        /*
1525         * pfx{mid-a => mid-b}sfx
1526         * {pfx-a => pfx-b}sfx
1527         * pfx{sfx-a => sfx-b}
1528         * name-a => name-b
1529         */
1530        a_midlen = len_a - pfx_length - sfx_length;
1531        b_midlen = len_b - pfx_length - sfx_length;
1532        if (a_midlen < 0)
1533                a_midlen = 0;
1534        if (b_midlen < 0)
1535                b_midlen = 0;
1536
1537        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1538        if (pfx_length + sfx_length) {
1539                strbuf_add(&name, a, pfx_length);
1540                strbuf_addch(&name, '{');
1541        }
1542        strbuf_add(&name, a + pfx_length, a_midlen);
1543        strbuf_addstr(&name, " => ");
1544        strbuf_add(&name, b + pfx_length, b_midlen);
1545        if (pfx_length + sfx_length) {
1546                strbuf_addch(&name, '}');
1547                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1548        }
1549        return strbuf_detach(&name, NULL);
1550}
1551
1552struct diffstat_t {
1553        int nr;
1554        int alloc;
1555        struct diffstat_file {
1556                char *from_name;
1557                char *name;
1558                char *print_name;
1559                unsigned is_unmerged:1;
1560                unsigned is_binary:1;
1561                unsigned is_renamed:1;
1562                unsigned is_interesting:1;
1563                uintmax_t added, deleted;
1564        } **files;
1565};
1566
1567static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1568                                          const char *name_a,
1569                                          const char *name_b)
1570{
1571        struct diffstat_file *x;
1572        x = xcalloc(1, sizeof(*x));
1573        ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1574        diffstat->files[diffstat->nr++] = x;
1575        if (name_b) {
1576                x->from_name = xstrdup(name_a);
1577                x->name = xstrdup(name_b);
1578                x->is_renamed = 1;
1579        }
1580        else {
1581                x->from_name = NULL;
1582                x->name = xstrdup(name_a);
1583        }
1584        return x;
1585}
1586
1587static void diffstat_consume(void *priv, char *line, unsigned long len)
1588{
1589        struct diffstat_t *diffstat = priv;
1590        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1591
1592        if (line[0] == '+')
1593                x->added++;
1594        else if (line[0] == '-')
1595                x->deleted++;
1596}
1597
1598const char mime_boundary_leader[] = "------------";
1599
1600static int scale_linear(int it, int width, int max_change)
1601{
1602        if (!it)
1603                return 0;
1604        /*
1605         * make sure that at least one '-' or '+' is printed if
1606         * there is any change to this path. The easiest way is to
1607         * scale linearly as if the alloted width is one column shorter
1608         * than it is, and then add 1 to the result.
1609         */
1610        return 1 + (it * (width - 1) / max_change);
1611}
1612
1613static void show_name(FILE *file,
1614                      const char *prefix, const char *name, int len)
1615{
1616        fprintf(file, " %s%-*s |", prefix, len, name);
1617}
1618
1619static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1620{
1621        if (cnt <= 0)
1622                return;
1623        fprintf(file, "%s", set);
1624        while (cnt--)
1625                putc(ch, file);
1626        fprintf(file, "%s", reset);
1627}
1628
1629static void fill_print_name(struct diffstat_file *file)
1630{
1631        char *pname;
1632
1633        if (file->print_name)
1634                return;
1635
1636        if (!file->is_renamed) {
1637                struct strbuf buf = STRBUF_INIT;
1638                if (quote_c_style(file->name, &buf, NULL, 0)) {
1639                        pname = strbuf_detach(&buf, NULL);
1640                } else {
1641                        pname = file->name;
1642                        strbuf_release(&buf);
1643                }
1644        } else {
1645                pname = pprint_rename(file->from_name, file->name);
1646        }
1647        file->print_name = pname;
1648}
1649
1650int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1651{
1652        struct strbuf sb = STRBUF_INIT;
1653        int ret;
1654
1655        if (!files) {
1656                assert(insertions == 0 && deletions == 0);
1657                return fprintf(fp, "%s\n", " 0 files changed");
1658        }
1659
1660        strbuf_addf(&sb,
1661                    (files == 1) ? " %d file changed" : " %d files changed",
1662                    files);
1663
1664        /*
1665         * For binary diff, the caller may want to print "x files
1666         * changed" with insertions == 0 && deletions == 0.
1667         *
1668         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1669         * is probably less confusing (i.e skip over "2 files changed
1670         * but nothing about added/removed lines? Is this a bug in Git?").
1671         */
1672        if (insertions || deletions == 0) {
1673                strbuf_addf(&sb,
1674                            (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1675                            insertions);
1676        }
1677
1678        if (deletions || insertions == 0) {
1679                strbuf_addf(&sb,
1680                            (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1681                            deletions);
1682        }
1683        strbuf_addch(&sb, '\n');
1684        ret = fputs(sb.buf, fp);
1685        strbuf_release(&sb);
1686        return ret;
1687}
1688
1689static void show_stats(struct diffstat_t *data, struct diff_options *options)
1690{
1691        int i, len, add, del, adds = 0, dels = 0;
1692        uintmax_t max_change = 0, max_len = 0;
1693        int total_files = data->nr, count;
1694        int width, name_width, graph_width, number_width = 0, bin_width = 0;
1695        const char *reset, *add_c, *del_c;
1696        const char *line_prefix = "";
1697        int extra_shown = 0;
1698
1699        if (data->nr == 0)
1700                return;
1701
1702        line_prefix = diff_line_prefix(options);
1703        count = options->stat_count ? options->stat_count : data->nr;
1704
1705        reset = diff_get_color_opt(options, DIFF_RESET);
1706        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1707        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1708
1709        /*
1710         * Find the longest filename and max number of changes
1711         */
1712        for (i = 0; (i < count) && (i < data->nr); i++) {
1713                struct diffstat_file *file = data->files[i];
1714                uintmax_t change = file->added + file->deleted;
1715
1716                if (!file->is_interesting && (change == 0)) {
1717                        count++; /* not shown == room for one more */
1718                        continue;
1719                }
1720                fill_print_name(file);
1721                len = strlen(file->print_name);
1722                if (max_len < len)
1723                        max_len = len;
1724
1725                if (file->is_unmerged) {
1726                        /* "Unmerged" is 8 characters */
1727                        bin_width = bin_width < 8 ? 8 : bin_width;
1728                        continue;
1729                }
1730                if (file->is_binary) {
1731                        /* "Bin XXX -> YYY bytes" */
1732                        int w = 14 + decimal_width(file->added)
1733                                + decimal_width(file->deleted);
1734                        bin_width = bin_width < w ? w : bin_width;
1735                        /* Display change counts aligned with "Bin" */
1736                        number_width = 3;
1737                        continue;
1738                }
1739
1740                if (max_change < change)
1741                        max_change = change;
1742        }
1743        count = i; /* where we can stop scanning in data->files[] */
1744
1745        /*
1746         * We have width = stat_width or term_columns() columns total.
1747         * We want a maximum of min(max_len, stat_name_width) for the name part.
1748         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1749         * We also need 1 for " " and 4 + decimal_width(max_change)
1750         * for " | NNNN " and one the empty column at the end, altogether
1751         * 6 + decimal_width(max_change).
1752         *
1753         * If there's not enough space, we will use the smaller of
1754         * stat_name_width (if set) and 5/8*width for the filename,
1755         * and the rest for constant elements + graph part, but no more
1756         * than stat_graph_width for the graph part.
1757         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1758         * for the standard terminal size).
1759         *
1760         * In other words: stat_width limits the maximum width, and
1761         * stat_name_width fixes the maximum width of the filename,
1762         * and is also used to divide available columns if there
1763         * aren't enough.
1764         *
1765         * Binary files are displayed with "Bin XXX -> YYY bytes"
1766         * instead of the change count and graph. This part is treated
1767         * similarly to the graph part, except that it is not
1768         * "scaled". If total width is too small to accommodate the
1769         * guaranteed minimum width of the filename part and the
1770         * separators and this message, this message will "overflow"
1771         * making the line longer than the maximum width.
1772         */
1773
1774        if (options->stat_width == -1)
1775                width = term_columns() - strlen(line_prefix);
1776        else
1777                width = options->stat_width ? options->stat_width : 80;
1778        number_width = decimal_width(max_change) > number_width ?
1779                decimal_width(max_change) : number_width;
1780
1781        if (options->stat_graph_width == -1)
1782                options->stat_graph_width = diff_stat_graph_width;
1783
1784        /*
1785         * Guarantee 3/8*16==6 for the graph part
1786         * and 5/8*16==10 for the filename part
1787         */
1788        if (width < 16 + 6 + number_width)
1789                width = 16 + 6 + number_width;
1790
1791        /*
1792         * First assign sizes that are wanted, ignoring available width.
1793         * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1794         * starting from "XXX" should fit in graph_width.
1795         */
1796        graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1797        if (options->stat_graph_width &&
1798            options->stat_graph_width < graph_width)
1799                graph_width = options->stat_graph_width;
1800
1801        name_width = (options->stat_name_width > 0 &&
1802                      options->stat_name_width < max_len) ?
1803                options->stat_name_width : max_len;
1804
1805        /*
1806         * Adjust adjustable widths not to exceed maximum width
1807         */
1808        if (name_width + number_width + 6 + graph_width > width) {
1809                if (graph_width > width * 3/8 - number_width - 6) {
1810                        graph_width = width * 3/8 - number_width - 6;
1811                        if (graph_width < 6)
1812                                graph_width = 6;
1813                }
1814
1815                if (options->stat_graph_width &&
1816                    graph_width > options->stat_graph_width)
1817                        graph_width = options->stat_graph_width;
1818                if (name_width > width - number_width - 6 - graph_width)
1819                        name_width = width - number_width - 6 - graph_width;
1820                else
1821                        graph_width = width - number_width - 6 - name_width;
1822        }
1823
1824        /*
1825         * From here name_width is the width of the name area,
1826         * and graph_width is the width of the graph area.
1827         * max_change is used to scale graph properly.
1828         */
1829        for (i = 0; i < count; i++) {
1830                const char *prefix = "";
1831                struct diffstat_file *file = data->files[i];
1832                char *name = file->print_name;
1833                uintmax_t added = file->added;
1834                uintmax_t deleted = file->deleted;
1835                int name_len;
1836
1837                if (!file->is_interesting && (added + deleted == 0))
1838                        continue;
1839
1840                /*
1841                 * "scale" the filename
1842                 */
1843                len = name_width;
1844                name_len = strlen(name);
1845                if (name_width < name_len) {
1846                        char *slash;
1847                        prefix = "...";
1848                        len -= 3;
1849                        name += name_len - len;
1850                        slash = strchr(name, '/');
1851                        if (slash)
1852                                name = slash;
1853                }
1854
1855                if (file->is_binary) {
1856                        fprintf(options->file, "%s", line_prefix);
1857                        show_name(options->file, prefix, name, len);
1858                        fprintf(options->file, " %*s", number_width, "Bin");
1859                        if (!added && !deleted) {
1860                                putc('\n', options->file);
1861                                continue;
1862                        }
1863                        fprintf(options->file, " %s%"PRIuMAX"%s",
1864                                del_c, deleted, reset);
1865                        fprintf(options->file, " -> ");
1866                        fprintf(options->file, "%s%"PRIuMAX"%s",
1867                                add_c, added, reset);
1868                        fprintf(options->file, " bytes");
1869                        fprintf(options->file, "\n");
1870                        continue;
1871                }
1872                else if (file->is_unmerged) {
1873                        fprintf(options->file, "%s", line_prefix);
1874                        show_name(options->file, prefix, name, len);
1875                        fprintf(options->file, " Unmerged\n");
1876                        continue;
1877                }
1878
1879                /*
1880                 * scale the add/delete
1881                 */
1882                add = added;
1883                del = deleted;
1884
1885                if (graph_width <= max_change) {
1886                        int total = scale_linear(add + del, graph_width, max_change);
1887                        if (total < 2 && add && del)
1888                                /* width >= 2 due to the sanity check */
1889                                total = 2;
1890                        if (add < del) {
1891                                add = scale_linear(add, graph_width, max_change);
1892                                del = total - add;
1893                        } else {
1894                                del = scale_linear(del, graph_width, max_change);
1895                                add = total - del;
1896                        }
1897                }
1898                fprintf(options->file, "%s", line_prefix);
1899                show_name(options->file, prefix, name, len);
1900                fprintf(options->file, " %*"PRIuMAX"%s",
1901                        number_width, added + deleted,
1902                        added + deleted ? " " : "");
1903                show_graph(options->file, '+', add, add_c, reset);
1904                show_graph(options->file, '-', del, del_c, reset);
1905                fprintf(options->file, "\n");
1906        }
1907
1908        for (i = 0; i < data->nr; i++) {
1909                struct diffstat_file *file = data->files[i];
1910                uintmax_t added = file->added;
1911                uintmax_t deleted = file->deleted;
1912
1913                if (file->is_unmerged ||
1914                    (!file->is_interesting && (added + deleted == 0))) {
1915                        total_files--;
1916                        continue;
1917                }
1918
1919                if (!file->is_binary) {
1920                        adds += added;
1921                        dels += deleted;
1922                }
1923                if (i < count)
1924                        continue;
1925                if (!extra_shown)
1926                        fprintf(options->file, "%s ...\n", line_prefix);
1927                extra_shown = 1;
1928        }
1929        fprintf(options->file, "%s", line_prefix);
1930        print_stat_summary(options->file, total_files, adds, dels);
1931}
1932
1933static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1934{
1935        int i, adds = 0, dels = 0, total_files = data->nr;
1936
1937        if (data->nr == 0)
1938                return;
1939
1940        for (i = 0; i < data->nr; i++) {
1941                int added = data->files[i]->added;
1942                int deleted= data->files[i]->deleted;
1943
1944                if (data->files[i]->is_unmerged ||
1945                    (!data->files[i]->is_interesting && (added + deleted == 0))) {
1946                        total_files--;
1947                } else if (!data->files[i]->is_binary) { /* don't count bytes */
1948                        adds += added;
1949                        dels += deleted;
1950                }
1951        }
1952        fprintf(options->file, "%s", diff_line_prefix(options));
1953        print_stat_summary(options->file, total_files, adds, dels);
1954}
1955
1956static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1957{
1958        int i;
1959
1960        if (data->nr == 0)
1961                return;
1962
1963        for (i = 0; i < data->nr; i++) {
1964                struct diffstat_file *file = data->files[i];
1965
1966                fprintf(options->file, "%s", diff_line_prefix(options));
1967
1968                if (file->is_binary)
1969                        fprintf(options->file, "-\t-\t");
1970                else
1971                        fprintf(options->file,
1972                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1973                                file->added, file->deleted);
1974                if (options->line_termination) {
1975                        fill_print_name(file);
1976                        if (!file->is_renamed)
1977                                write_name_quoted(file->name, options->file,
1978                                                  options->line_termination);
1979                        else {
1980                                fputs(file->print_name, options->file);
1981                                putc(options->line_termination, options->file);
1982                        }
1983                } else {
1984                        if (file->is_renamed) {
1985                                putc('\0', options->file);
1986                                write_name_quoted(file->from_name, options->file, '\0');
1987                        }
1988                        write_name_quoted(file->name, options->file, '\0');
1989                }
1990        }
1991}
1992
1993struct dirstat_file {
1994        const char *name;
1995        unsigned long changed;
1996};
1997
1998struct dirstat_dir {
1999        struct dirstat_file *files;
2000        int alloc, nr, permille, cumulative;
2001};
2002
2003static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2004                unsigned long changed, const char *base, int baselen)
2005{
2006        unsigned long this_dir = 0;
2007        unsigned int sources = 0;
2008        const char *line_prefix = diff_line_prefix(opt);
2009
2010        while (dir->nr) {
2011                struct dirstat_file *f = dir->files;
2012                int namelen = strlen(f->name);
2013                unsigned long this;
2014                char *slash;
2015
2016                if (namelen < baselen)
2017                        break;
2018                if (memcmp(f->name, base, baselen))
2019                        break;
2020                slash = strchr(f->name + baselen, '/');
2021                if (slash) {
2022                        int newbaselen = slash + 1 - f->name;
2023                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2024                        sources++;
2025                } else {
2026                        this = f->changed;
2027                        dir->files++;
2028                        dir->nr--;
2029                        sources += 2;
2030                }
2031                this_dir += this;
2032        }
2033
2034        /*
2035         * We don't report dirstat's for
2036         *  - the top level
2037         *  - or cases where everything came from a single directory
2038         *    under this directory (sources == 1).
2039         */
2040        if (baselen && sources != 1) {
2041                if (this_dir) {
2042                        int permille = this_dir * 1000 / changed;
2043                        if (permille >= dir->permille) {
2044                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2045                                        permille / 10, permille % 10, baselen, base);
2046                                if (!dir->cumulative)
2047                                        return 0;
2048                        }
2049                }
2050        }
2051        return this_dir;
2052}
2053
2054static int dirstat_compare(const void *_a, const void *_b)
2055{
2056        const struct dirstat_file *a = _a;
2057        const struct dirstat_file *b = _b;
2058        return strcmp(a->name, b->name);
2059}
2060
2061static void show_dirstat(struct diff_options *options)
2062{
2063        int i;
2064        unsigned long changed;
2065        struct dirstat_dir dir;
2066        struct diff_queue_struct *q = &diff_queued_diff;
2067
2068        dir.files = NULL;
2069        dir.alloc = 0;
2070        dir.nr = 0;
2071        dir.permille = options->dirstat_permille;
2072        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2073
2074        changed = 0;
2075        for (i = 0; i < q->nr; i++) {
2076                struct diff_filepair *p = q->queue[i];
2077                const char *name;
2078                unsigned long copied, added, damage;
2079                int content_changed;
2080
2081                name = p->two->path ? p->two->path : p->one->path;
2082
2083                if (p->one->oid_valid && p->two->oid_valid)
2084                        content_changed = oidcmp(&p->one->oid, &p->two->oid);
2085                else
2086                        content_changed = 1;
2087
2088                if (!content_changed) {
2089                        /*
2090                         * The SHA1 has not changed, so pre-/post-content is
2091                         * identical. We can therefore skip looking at the
2092                         * file contents altogether.
2093                         */
2094                        damage = 0;
2095                        goto found_damage;
2096                }
2097
2098                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2099                        /*
2100                         * In --dirstat-by-file mode, we don't really need to
2101                         * look at the actual file contents at all.
2102                         * The fact that the SHA1 changed is enough for us to
2103                         * add this file to the list of results
2104                         * (with each file contributing equal damage).
2105                         */
2106                        damage = 1;
2107                        goto found_damage;
2108                }
2109
2110                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2111                        diff_populate_filespec(p->one, 0);
2112                        diff_populate_filespec(p->two, 0);
2113                        diffcore_count_changes(p->one, p->two, NULL, NULL,
2114                                               &copied, &added);
2115                        diff_free_filespec_data(p->one);
2116                        diff_free_filespec_data(p->two);
2117                } else if (DIFF_FILE_VALID(p->one)) {
2118                        diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2119                        copied = added = 0;
2120                        diff_free_filespec_data(p->one);
2121                } else if (DIFF_FILE_VALID(p->two)) {
2122                        diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2123                        copied = 0;
2124                        added = p->two->size;
2125                        diff_free_filespec_data(p->two);
2126                } else
2127                        continue;
2128
2129                /*
2130                 * Original minus copied is the removed material,
2131                 * added is the new material.  They are both damages
2132                 * made to the preimage.
2133                 * If the resulting damage is zero, we know that
2134                 * diffcore_count_changes() considers the two entries to
2135                 * be identical, but since content_changed is true, we
2136                 * know that there must have been _some_ kind of change,
2137                 * so we force all entries to have damage > 0.
2138                 */
2139                damage = (p->one->size - copied) + added;
2140                if (!damage)
2141                        damage = 1;
2142
2143found_damage:
2144                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2145                dir.files[dir.nr].name = name;
2146                dir.files[dir.nr].changed = damage;
2147                changed += damage;
2148                dir.nr++;
2149        }
2150
2151        /* This can happen even with many files, if everything was renames */
2152        if (!changed)
2153                return;
2154
2155        /* Show all directories with more than x% of the changes */
2156        QSORT(dir.files, dir.nr, dirstat_compare);
2157        gather_dirstat(options, &dir, changed, "", 0);
2158}
2159
2160static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2161{
2162        int i;
2163        unsigned long changed;
2164        struct dirstat_dir dir;
2165
2166        if (data->nr == 0)
2167                return;
2168
2169        dir.files = NULL;
2170        dir.alloc = 0;
2171        dir.nr = 0;
2172        dir.permille = options->dirstat_permille;
2173        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2174
2175        changed = 0;
2176        for (i = 0; i < data->nr; i++) {
2177                struct diffstat_file *file = data->files[i];
2178                unsigned long damage = file->added + file->deleted;
2179                if (file->is_binary)
2180                        /*
2181                         * binary files counts bytes, not lines. Must find some
2182                         * way to normalize binary bytes vs. textual lines.
2183                         * The following heuristic assumes that there are 64
2184                         * bytes per "line".
2185                         * This is stupid and ugly, but very cheap...
2186                         */
2187                        damage = (damage + 63) / 64;
2188                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2189                dir.files[dir.nr].name = file->name;
2190                dir.files[dir.nr].changed = damage;
2191                changed += damage;
2192                dir.nr++;
2193        }
2194
2195        /* This can happen even with many files, if everything was renames */
2196        if (!changed)
2197                return;
2198
2199        /* Show all directories with more than x% of the changes */
2200        QSORT(dir.files, dir.nr, dirstat_compare);
2201        gather_dirstat(options, &dir, changed, "", 0);
2202}
2203
2204static void free_diffstat_info(struct diffstat_t *diffstat)
2205{
2206        int i;
2207        for (i = 0; i < diffstat->nr; i++) {
2208                struct diffstat_file *f = diffstat->files[i];
2209                if (f->name != f->print_name)
2210                        free(f->print_name);
2211                free(f->name);
2212                free(f->from_name);
2213                free(f);
2214        }
2215        free(diffstat->files);
2216}
2217
2218struct checkdiff_t {
2219        const char *filename;
2220        int lineno;
2221        int conflict_marker_size;
2222        struct diff_options *o;
2223        unsigned ws_rule;
2224        unsigned status;
2225};
2226
2227static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2228{
2229        char firstchar;
2230        int cnt;
2231
2232        if (len < marker_size + 1)
2233                return 0;
2234        firstchar = line[0];
2235        switch (firstchar) {
2236        case '=': case '>': case '<': case '|':
2237                break;
2238        default:
2239                return 0;
2240        }
2241        for (cnt = 1; cnt < marker_size; cnt++)
2242                if (line[cnt] != firstchar)
2243                        return 0;
2244        /* line[1] thru line[marker_size-1] are same as firstchar */
2245        if (len < marker_size + 1 || !isspace(line[marker_size]))
2246                return 0;
2247        return 1;
2248}
2249
2250static void checkdiff_consume(void *priv, char *line, unsigned long len)
2251{
2252        struct checkdiff_t *data = priv;
2253        int marker_size = data->conflict_marker_size;
2254        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2255        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2256        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2257        char *err;
2258        const char *line_prefix;
2259
2260        assert(data->o);
2261        line_prefix = diff_line_prefix(data->o);
2262
2263        if (line[0] == '+') {
2264                unsigned bad;
2265                data->lineno++;
2266                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2267                        data->status |= 1;
2268                        fprintf(data->o->file,
2269                                "%s%s:%d: leftover conflict marker\n",
2270                                line_prefix, data->filename, data->lineno);
2271                }
2272                bad = ws_check(line + 1, len - 1, data->ws_rule);
2273                if (!bad)
2274                        return;
2275                data->status |= bad;
2276                err = whitespace_error_string(bad);
2277                fprintf(data->o->file, "%s%s:%d: %s.\n",
2278                        line_prefix, data->filename, data->lineno, err);
2279                free(err);
2280                emit_line(data->o, set, reset, line, 1);
2281                ws_check_emit(line + 1, len - 1, data->ws_rule,
2282                              data->o->file, set, reset, ws);
2283        } else if (line[0] == ' ') {
2284                data->lineno++;
2285        } else if (line[0] == '@') {
2286                char *plus = strchr(line, '+');
2287                if (plus)
2288                        data->lineno = strtol(plus, NULL, 10) - 1;
2289                else
2290                        die("invalid diff");
2291        }
2292}
2293
2294static unsigned char *deflate_it(char *data,
2295                                 unsigned long size,
2296                                 unsigned long *result_size)
2297{
2298        int bound;
2299        unsigned char *deflated;
2300        git_zstream stream;
2301
2302        git_deflate_init(&stream, zlib_compression_level);
2303        bound = git_deflate_bound(&stream, size);
2304        deflated = xmalloc(bound);
2305        stream.next_out = deflated;
2306        stream.avail_out = bound;
2307
2308        stream.next_in = (unsigned char *)data;
2309        stream.avail_in = size;
2310        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2311                ; /* nothing */
2312        git_deflate_end(&stream);
2313        *result_size = stream.total_out;
2314        return deflated;
2315}
2316
2317static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2318                                  const char *prefix)
2319{
2320        void *cp;
2321        void *delta;
2322        void *deflated;
2323        void *data;
2324        unsigned long orig_size;
2325        unsigned long delta_size;
2326        unsigned long deflate_size;
2327        unsigned long data_size;
2328
2329        /* We could do deflated delta, or we could do just deflated two,
2330         * whichever is smaller.
2331         */
2332        delta = NULL;
2333        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2334        if (one->size && two->size) {
2335                delta = diff_delta(one->ptr, one->size,
2336                                   two->ptr, two->size,
2337                                   &delta_size, deflate_size);
2338                if (delta) {
2339                        void *to_free = delta;
2340                        orig_size = delta_size;
2341                        delta = deflate_it(delta, delta_size, &delta_size);
2342                        free(to_free);
2343                }
2344        }
2345
2346        if (delta && delta_size < deflate_size) {
2347                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2348                free(deflated);
2349                data = delta;
2350                data_size = delta_size;
2351        }
2352        else {
2353                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2354                free(delta);
2355                data = deflated;
2356                data_size = deflate_size;
2357        }
2358
2359        /* emit data encoded in base85 */
2360        cp = data;
2361        while (data_size) {
2362                int bytes = (52 < data_size) ? 52 : data_size;
2363                char line[70];
2364                data_size -= bytes;
2365                if (bytes <= 26)
2366                        line[0] = bytes + 'A' - 1;
2367                else
2368                        line[0] = bytes - 26 + 'a' - 1;
2369                encode_85(line + 1, cp, bytes);
2370                cp = (char *) cp + bytes;
2371                fprintf(file, "%s", prefix);
2372                fputs(line, file);
2373                fputc('\n', file);
2374        }
2375        fprintf(file, "%s\n", prefix);
2376        free(data);
2377}
2378
2379static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2380                             const char *prefix)
2381{
2382        fprintf(file, "%sGIT binary patch\n", prefix);
2383        emit_binary_diff_body(file, one, two, prefix);
2384        emit_binary_diff_body(file, two, one, prefix);
2385}
2386
2387int diff_filespec_is_binary(struct diff_filespec *one)
2388{
2389        if (one->is_binary == -1) {
2390                diff_filespec_load_driver(one);
2391                if (one->driver->binary != -1)
2392                        one->is_binary = one->driver->binary;
2393                else {
2394                        if (!one->data && DIFF_FILE_VALID(one))
2395                                diff_populate_filespec(one, CHECK_BINARY);
2396                        if (one->is_binary == -1 && one->data)
2397                                one->is_binary = buffer_is_binary(one->data,
2398                                                one->size);
2399                        if (one->is_binary == -1)
2400                                one->is_binary = 0;
2401                }
2402        }
2403        return one->is_binary;
2404}
2405
2406static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2407{
2408        diff_filespec_load_driver(one);
2409        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2410}
2411
2412void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2413{
2414        if (!options->a_prefix)
2415                options->a_prefix = a;
2416        if (!options->b_prefix)
2417                options->b_prefix = b;
2418}
2419
2420struct userdiff_driver *get_textconv(struct diff_filespec *one)
2421{
2422        if (!DIFF_FILE_VALID(one))
2423                return NULL;
2424
2425        diff_filespec_load_driver(one);
2426        return userdiff_get_textconv(one->driver);
2427}
2428
2429static void builtin_diff(const char *name_a,
2430                         const char *name_b,
2431                         struct diff_filespec *one,
2432                         struct diff_filespec *two,
2433                         const char *xfrm_msg,
2434                         int must_show_header,
2435                         struct diff_options *o,
2436                         int complete_rewrite)
2437{
2438        mmfile_t mf1, mf2;
2439        const char *lbl[2];
2440        char *a_one, *b_two;
2441        const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2442        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2443        const char *a_prefix, *b_prefix;
2444        struct userdiff_driver *textconv_one = NULL;
2445        struct userdiff_driver *textconv_two = NULL;
2446        struct strbuf header = STRBUF_INIT;
2447        const char *line_prefix = diff_line_prefix(o);
2448
2449        diff_set_mnemonic_prefix(o, "a/", "b/");
2450        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2451                a_prefix = o->b_prefix;
2452                b_prefix = o->a_prefix;
2453        } else {
2454                a_prefix = o->a_prefix;
2455                b_prefix = o->b_prefix;
2456        }
2457
2458        if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2459            (!one->mode || S_ISGITLINK(one->mode)) &&
2460            (!two->mode || S_ISGITLINK(two->mode))) {
2461                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2462                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2463                show_submodule_summary(o->file, one->path ? one->path : two->path,
2464                                line_prefix,
2465                                &one->oid, &two->oid,
2466                                two->dirty_submodule,
2467                                meta, del, add, reset);
2468                return;
2469        } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2470                   (!one->mode || S_ISGITLINK(one->mode)) &&
2471                   (!two->mode || S_ISGITLINK(two->mode))) {
2472                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2473                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2474                show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2475                                line_prefix,
2476                                &one->oid, &two->oid,
2477                                two->dirty_submodule,
2478                                meta, del, add, reset, o);
2479                return;
2480        }
2481
2482        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2483                textconv_one = get_textconv(one);
2484                textconv_two = get_textconv(two);
2485        }
2486
2487        /* Never use a non-valid filename anywhere if at all possible */
2488        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2489        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2490
2491        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2492        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2493        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2494        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2495        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2496        if (lbl[0][0] == '/') {
2497                /* /dev/null */
2498                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2499                if (xfrm_msg)
2500                        strbuf_addstr(&header, xfrm_msg);
2501                must_show_header = 1;
2502        }
2503        else if (lbl[1][0] == '/') {
2504                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2505                if (xfrm_msg)
2506                        strbuf_addstr(&header, xfrm_msg);
2507                must_show_header = 1;
2508        }
2509        else {
2510                if (one->mode != two->mode) {
2511                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2512                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2513                        must_show_header = 1;
2514                }
2515                if (xfrm_msg)
2516                        strbuf_addstr(&header, xfrm_msg);
2517
2518                /*
2519                 * we do not run diff between different kind
2520                 * of objects.
2521                 */
2522                if ((one->mode ^ two->mode) & S_IFMT)
2523                        goto free_ab_and_return;
2524                if (complete_rewrite &&
2525                    (textconv_one || !diff_filespec_is_binary(one)) &&
2526                    (textconv_two || !diff_filespec_is_binary(two))) {
2527                        emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2528                                         header.buf, header.len, 0);
2529                        strbuf_reset(&header);
2530                        emit_rewrite_diff(name_a, name_b, one, two,
2531                                                textconv_one, textconv_two, o);
2532                        o->found_changes = 1;
2533                        goto free_ab_and_return;
2534                }
2535        }
2536
2537        if (o->irreversible_delete && lbl[1][0] == '/') {
2538                emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
2539                                 header.len, 0);
2540                strbuf_reset(&header);
2541                goto free_ab_and_return;
2542        } else if (!DIFF_OPT_TST(o, TEXT) &&
2543            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2544              (!textconv_two && diff_filespec_is_binary(two)) )) {
2545                if (!one->data && !two->data &&
2546                    S_ISREG(one->mode) && S_ISREG(two->mode) &&
2547                    !DIFF_OPT_TST(o, BINARY)) {
2548                        if (!oidcmp(&one->oid, &two->oid)) {
2549                                if (must_show_header)
2550                                        emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2551                                                         header.buf, header.len,
2552                                                         0);
2553                                goto free_ab_and_return;
2554                        }
2555                        emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2556                                         header.buf, header.len, 0);
2557                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2558                                line_prefix, lbl[0], lbl[1]);
2559                        goto free_ab_and_return;
2560                }
2561                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2562                        die("unable to read files to diff");
2563                /* Quite common confusing case */
2564                if (mf1.size == mf2.size &&
2565                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2566                        if (must_show_header)
2567                                emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2568                                                 header.buf, header.len, 0);
2569                        goto free_ab_and_return;
2570                }
2571                emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
2572                strbuf_reset(&header);
2573                if (DIFF_OPT_TST(o, BINARY))
2574                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2575                else
2576                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2577                                line_prefix, lbl[0], lbl[1]);
2578                o->found_changes = 1;
2579        } else {
2580                /* Crazy xdl interfaces.. */
2581                const char *diffopts = getenv("GIT_DIFF_OPTS");
2582                const char *v;
2583                xpparam_t xpp;
2584                xdemitconf_t xecfg;
2585                struct emit_callback ecbdata;
2586                const struct userdiff_funcname *pe;
2587
2588                if (must_show_header) {
2589                        emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2590                                         header.buf, header.len, 0);
2591                        strbuf_reset(&header);
2592                }
2593
2594                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2595                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2596
2597                pe = diff_funcname_pattern(one);
2598                if (!pe)
2599                        pe = diff_funcname_pattern(two);
2600
2601                memset(&xpp, 0, sizeof(xpp));
2602                memset(&xecfg, 0, sizeof(xecfg));
2603                memset(&ecbdata, 0, sizeof(ecbdata));
2604                ecbdata.label_path = lbl;
2605                ecbdata.color_diff = want_color(o->use_color);
2606                ecbdata.ws_rule = whitespace_rule(name_b);
2607                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2608                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2609                ecbdata.opt = o;
2610                ecbdata.header = header.len ? &header : NULL;
2611                xpp.flags = o->xdl_opts;
2612                xecfg.ctxlen = o->context;
2613                xecfg.interhunkctxlen = o->interhunkcontext;
2614                xecfg.flags = XDL_EMIT_FUNCNAMES;
2615                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2616                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2617                if (pe)
2618                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2619                if (!diffopts)
2620                        ;
2621                else if (skip_prefix(diffopts, "--unified=", &v))
2622                        xecfg.ctxlen = strtoul(v, NULL, 10);
2623                else if (skip_prefix(diffopts, "-u", &v))
2624                        xecfg.ctxlen = strtoul(v, NULL, 10);
2625                if (o->word_diff)
2626                        init_diff_words_data(&ecbdata, o, one, two);
2627                if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2628                                  &xpp, &xecfg))
2629                        die("unable to generate diff for %s", one->path);
2630                if (o->word_diff)
2631                        free_diff_words_data(&ecbdata);
2632                if (textconv_one)
2633                        free(mf1.ptr);
2634                if (textconv_two)
2635                        free(mf2.ptr);
2636                xdiff_clear_find_func(&xecfg);
2637        }
2638
2639 free_ab_and_return:
2640        strbuf_release(&header);
2641        diff_free_filespec_data(one);
2642        diff_free_filespec_data(two);
2643        free(a_one);
2644        free(b_two);
2645        return;
2646}
2647
2648static void builtin_diffstat(const char *name_a, const char *name_b,
2649                             struct diff_filespec *one,
2650                             struct diff_filespec *two,
2651                             struct diffstat_t *diffstat,
2652                             struct diff_options *o,
2653                             struct diff_filepair *p)
2654{
2655        mmfile_t mf1, mf2;
2656        struct diffstat_file *data;
2657        int same_contents;
2658        int complete_rewrite = 0;
2659
2660        if (!DIFF_PAIR_UNMERGED(p)) {
2661                if (p->status == DIFF_STATUS_MODIFIED && p->score)
2662                        complete_rewrite = 1;
2663        }
2664
2665        data = diffstat_add(diffstat, name_a, name_b);
2666        data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2667
2668        if (!one || !two) {
2669                data->is_unmerged = 1;
2670                return;
2671        }
2672
2673        same_contents = !oidcmp(&one->oid, &two->oid);
2674
2675        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2676                data->is_binary = 1;
2677                if (same_contents) {
2678                        data->added = 0;
2679                        data->deleted = 0;
2680                } else {
2681                        data->added = diff_filespec_size(two);
2682                        data->deleted = diff_filespec_size(one);
2683                }
2684        }
2685
2686        else if (complete_rewrite) {
2687                diff_populate_filespec(one, 0);
2688                diff_populate_filespec(two, 0);
2689                data->deleted = count_lines(one->data, one->size);
2690                data->added = count_lines(two->data, two->size);
2691        }
2692
2693        else if (!same_contents) {
2694                /* Crazy xdl interfaces.. */
2695                xpparam_t xpp;
2696                xdemitconf_t xecfg;
2697
2698                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2699                        die("unable to read files to diff");
2700
2701                memset(&xpp, 0, sizeof(xpp));
2702                memset(&xecfg, 0, sizeof(xecfg));
2703                xpp.flags = o->xdl_opts;
2704                xecfg.ctxlen = o->context;
2705                xecfg.interhunkctxlen = o->interhunkcontext;
2706                if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2707                                  &xpp, &xecfg))
2708                        die("unable to generate diffstat for %s", one->path);
2709        }
2710
2711        diff_free_filespec_data(one);
2712        diff_free_filespec_data(two);
2713}
2714
2715static void builtin_checkdiff(const char *name_a, const char *name_b,
2716                              const char *attr_path,
2717                              struct diff_filespec *one,
2718                              struct diff_filespec *two,
2719                              struct diff_options *o)
2720{
2721        mmfile_t mf1, mf2;
2722        struct checkdiff_t data;
2723
2724        if (!two)
2725                return;
2726
2727        memset(&data, 0, sizeof(data));
2728        data.filename = name_b ? name_b : name_a;
2729        data.lineno = 0;
2730        data.o = o;
2731        data.ws_rule = whitespace_rule(attr_path);
2732        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2733
2734        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2735                die("unable to read files to diff");
2736
2737        /*
2738         * All the other codepaths check both sides, but not checking
2739         * the "old" side here is deliberate.  We are checking the newly
2740         * introduced changes, and as long as the "new" side is text, we
2741         * can and should check what it introduces.
2742         */
2743        if (diff_filespec_is_binary(two))
2744                goto free_and_return;
2745        else {
2746                /* Crazy xdl interfaces.. */
2747                xpparam_t xpp;
2748                xdemitconf_t xecfg;
2749
2750                memset(&xpp, 0, sizeof(xpp));
2751                memset(&xecfg, 0, sizeof(xecfg));
2752                xecfg.ctxlen = 1; /* at least one context line */
2753                xpp.flags = 0;
2754                if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2755                                  &xpp, &xecfg))
2756                        die("unable to generate checkdiff for %s", one->path);
2757
2758                if (data.ws_rule & WS_BLANK_AT_EOF) {
2759                        struct emit_callback ecbdata;
2760                        int blank_at_eof;
2761
2762                        ecbdata.ws_rule = data.ws_rule;
2763                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2764                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2765
2766                        if (blank_at_eof) {
2767                                static char *err;
2768                                if (!err)
2769                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2770                                fprintf(o->file, "%s:%d: %s.\n",
2771                                        data.filename, blank_at_eof, err);
2772                                data.status = 1; /* report errors */
2773                        }
2774                }
2775        }
2776 free_and_return:
2777        diff_free_filespec_data(one);
2778        diff_free_filespec_data(two);
2779        if (data.status)
2780                DIFF_OPT_SET(o, CHECK_FAILED);
2781}
2782
2783struct diff_filespec *alloc_filespec(const char *path)
2784{
2785        struct diff_filespec *spec;
2786
2787        FLEXPTR_ALLOC_STR(spec, path, path);
2788        spec->count = 1;
2789        spec->is_binary = -1;
2790        return spec;
2791}
2792
2793void free_filespec(struct diff_filespec *spec)
2794{
2795        if (!--spec->count) {
2796                diff_free_filespec_data(spec);
2797                free(spec);
2798        }
2799}
2800
2801void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2802                   int oid_valid, unsigned short mode)
2803{
2804        if (mode) {
2805                spec->mode = canon_mode(mode);
2806                oidcpy(&spec->oid, oid);
2807                spec->oid_valid = oid_valid;
2808        }
2809}
2810
2811/*
2812 * Given a name and sha1 pair, if the index tells us the file in
2813 * the work tree has that object contents, return true, so that
2814 * prepare_temp_file() does not have to inflate and extract.
2815 */
2816static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2817{
2818        const struct cache_entry *ce;
2819        struct stat st;
2820        int pos, len;
2821
2822        /*
2823         * We do not read the cache ourselves here, because the
2824         * benchmark with my previous version that always reads cache
2825         * shows that it makes things worse for diff-tree comparing
2826         * two linux-2.6 kernel trees in an already checked out work
2827         * tree.  This is because most diff-tree comparisons deal with
2828         * only a small number of files, while reading the cache is
2829         * expensive for a large project, and its cost outweighs the
2830         * savings we get by not inflating the object to a temporary
2831         * file.  Practically, this code only helps when we are used
2832         * by diff-cache --cached, which does read the cache before
2833         * calling us.
2834         */
2835        if (!active_cache)
2836                return 0;
2837
2838        /* We want to avoid the working directory if our caller
2839         * doesn't need the data in a normal file, this system
2840         * is rather slow with its stat/open/mmap/close syscalls,
2841         * and the object is contained in a pack file.  The pack
2842         * is probably already open and will be faster to obtain
2843         * the data through than the working directory.  Loose
2844         * objects however would tend to be slower as they need
2845         * to be individually opened and inflated.
2846         */
2847        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2848                return 0;
2849
2850        /*
2851         * Similarly, if we'd have to convert the file contents anyway, that
2852         * makes the optimization not worthwhile.
2853         */
2854        if (!want_file && would_convert_to_git(&the_index, name))
2855                return 0;
2856
2857        len = strlen(name);
2858        pos = cache_name_pos(name, len);
2859        if (pos < 0)
2860                return 0;
2861        ce = active_cache[pos];
2862
2863        /*
2864         * This is not the sha1 we are looking for, or
2865         * unreusable because it is not a regular file.
2866         */
2867        if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2868                return 0;
2869
2870        /*
2871         * If ce is marked as "assume unchanged", there is no
2872         * guarantee that work tree matches what we are looking for.
2873         */
2874        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2875                return 0;
2876
2877        /*
2878         * If ce matches the file in the work tree, we can reuse it.
2879         */
2880        if (ce_uptodate(ce) ||
2881            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2882                return 1;
2883
2884        return 0;
2885}
2886
2887static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2888{
2889        struct strbuf buf = STRBUF_INIT;
2890        char *dirty = "";
2891
2892        /* Are we looking at the work tree? */
2893        if (s->dirty_submodule)
2894                dirty = "-dirty";
2895
2896        strbuf_addf(&buf, "Subproject commit %s%s\n",
2897                    oid_to_hex(&s->oid), dirty);
2898        s->size = buf.len;
2899        if (size_only) {
2900                s->data = NULL;
2901                strbuf_release(&buf);
2902        } else {
2903                s->data = strbuf_detach(&buf, NULL);
2904                s->should_free = 1;
2905        }
2906        return 0;
2907}
2908
2909/*
2910 * While doing rename detection and pickaxe operation, we may need to
2911 * grab the data for the blob (or file) for our own in-core comparison.
2912 * diff_filespec has data and size fields for this purpose.
2913 */
2914int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2915{
2916        int size_only = flags & CHECK_SIZE_ONLY;
2917        int err = 0;
2918        /*
2919         * demote FAIL to WARN to allow inspecting the situation
2920         * instead of refusing.
2921         */
2922        enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2923                                    ? SAFE_CRLF_WARN
2924                                    : safe_crlf);
2925
2926        if (!DIFF_FILE_VALID(s))
2927                die("internal error: asking to populate invalid file.");
2928        if (S_ISDIR(s->mode))
2929                return -1;
2930
2931        if (s->data)
2932                return 0;
2933
2934        if (size_only && 0 < s->size)
2935                return 0;
2936
2937        if (S_ISGITLINK(s->mode))
2938                return diff_populate_gitlink(s, size_only);
2939
2940        if (!s->oid_valid ||
2941            reuse_worktree_file(s->path, &s->oid, 0)) {
2942                struct strbuf buf = STRBUF_INIT;
2943                struct stat st;
2944                int fd;
2945
2946                if (lstat(s->path, &st) < 0) {
2947                        if (errno == ENOENT) {
2948                        err_empty:
2949                                err = -1;
2950                        empty:
2951                                s->data = (char *)"";
2952                                s->size = 0;
2953                                return err;
2954                        }
2955                }
2956                s->size = xsize_t(st.st_size);
2957                if (!s->size)
2958                        goto empty;
2959                if (S_ISLNK(st.st_mode)) {
2960                        struct strbuf sb = STRBUF_INIT;
2961
2962                        if (strbuf_readlink(&sb, s->path, s->size))
2963                                goto err_empty;
2964                        s->size = sb.len;
2965                        s->data = strbuf_detach(&sb, NULL);
2966                        s->should_free = 1;
2967                        return 0;
2968                }
2969
2970                /*
2971                 * Even if the caller would be happy with getting
2972                 * only the size, we cannot return early at this
2973                 * point if the path requires us to run the content
2974                 * conversion.
2975                 */
2976                if (size_only && !would_convert_to_git(&the_index, s->path))
2977                        return 0;
2978
2979                /*
2980                 * Note: this check uses xsize_t(st.st_size) that may
2981                 * not be the true size of the blob after it goes
2982                 * through convert_to_git().  This may not strictly be
2983                 * correct, but the whole point of big_file_threshold
2984                 * and is_binary check being that we want to avoid
2985                 * opening the file and inspecting the contents, this
2986                 * is probably fine.
2987                 */
2988                if ((flags & CHECK_BINARY) &&
2989                    s->size > big_file_threshold && s->is_binary == -1) {
2990                        s->is_binary = 1;
2991                        return 0;
2992                }
2993                fd = open(s->path, O_RDONLY);
2994                if (fd < 0)
2995                        goto err_empty;
2996                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2997                close(fd);
2998                s->should_munmap = 1;
2999
3000                /*
3001                 * Convert from working tree format to canonical git format
3002                 */
3003                if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
3004                        size_t size = 0;
3005                        munmap(s->data, s->size);
3006                        s->should_munmap = 0;
3007                        s->data = strbuf_detach(&buf, &size);
3008                        s->size = size;
3009                        s->should_free = 1;
3010                }
3011        }
3012        else {
3013                enum object_type type;
3014                if (size_only || (flags & CHECK_BINARY)) {
3015                        type = sha1_object_info(s->oid.hash, &s->size);
3016                        if (type < 0)
3017                                die("unable to read %s",
3018                                    oid_to_hex(&s->oid));
3019                        if (size_only)
3020                                return 0;
3021                        if (s->size > big_file_threshold && s->is_binary == -1) {
3022                                s->is_binary = 1;
3023                                return 0;
3024                        }
3025                }
3026                s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3027                if (!s->data)
3028                        die("unable to read %s", oid_to_hex(&s->oid));
3029                s->should_free = 1;
3030        }
3031        return 0;
3032}
3033
3034void diff_free_filespec_blob(struct diff_filespec *s)
3035{
3036        if (s->should_free)
3037                free(s->data);
3038        else if (s->should_munmap)
3039                munmap(s->data, s->size);
3040
3041        if (s->should_free || s->should_munmap) {
3042                s->should_free = s->should_munmap = 0;
3043                s->data = NULL;
3044        }
3045}
3046
3047void diff_free_filespec_data(struct diff_filespec *s)
3048{
3049        diff_free_filespec_blob(s);
3050        FREE_AND_NULL(s->cnt_data);
3051}
3052
3053static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3054                           void *blob,
3055                           unsigned long size,
3056                           const struct object_id *oid,
3057                           int mode)
3058{
3059        int fd;
3060        struct strbuf buf = STRBUF_INIT;
3061        struct strbuf template = STRBUF_INIT;
3062        char *path_dup = xstrdup(path);
3063        const char *base = basename(path_dup);
3064
3065        /* Generate "XXXXXX_basename.ext" */
3066        strbuf_addstr(&template, "XXXXXX_");
3067        strbuf_addstr(&template, base);
3068
3069        fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3070        if (fd < 0)
3071                die_errno("unable to create temp-file");
3072        if (convert_to_working_tree(path,
3073                        (const char *)blob, (size_t)size, &buf)) {
3074                blob = buf.buf;
3075                size = buf.len;
3076        }
3077        if (write_in_full(fd, blob, size) != size)
3078                die_errno("unable to write temp-file");
3079        close_tempfile(&temp->tempfile);
3080        temp->name = get_tempfile_path(&temp->tempfile);
3081        oid_to_hex_r(temp->hex, oid);
3082        xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3083        strbuf_release(&buf);
3084        strbuf_release(&template);
3085        free(path_dup);
3086}
3087
3088static struct diff_tempfile *prepare_temp_file(const char *name,
3089                struct diff_filespec *one)
3090{
3091        struct diff_tempfile *temp = claim_diff_tempfile();
3092
3093        if (!DIFF_FILE_VALID(one)) {
3094        not_a_valid_file:
3095                /* A '-' entry produces this for file-2, and
3096                 * a '+' entry produces this for file-1.
3097                 */
3098                temp->name = "/dev/null";
3099                xsnprintf(temp->hex, sizeof(temp->hex), ".");
3100                xsnprintf(temp->mode, sizeof(temp->mode), ".");
3101                return temp;
3102        }
3103
3104        if (!S_ISGITLINK(one->mode) &&
3105            (!one->oid_valid ||
3106             reuse_worktree_file(name, &one->oid, 1))) {
3107                struct stat st;
3108                if (lstat(name, &st) < 0) {
3109                        if (errno == ENOENT)
3110                                goto not_a_valid_file;
3111                        die_errno("stat(%s)", name);
3112                }
3113                if (S_ISLNK(st.st_mode)) {
3114                        struct strbuf sb = STRBUF_INIT;
3115                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
3116                                die_errno("readlink(%s)", name);
3117                        prep_temp_blob(name, temp, sb.buf, sb.len,
3118                                       (one->oid_valid ?
3119                                        &one->oid : &null_oid),
3120                                       (one->oid_valid ?
3121                                        one->mode : S_IFLNK));
3122                        strbuf_release(&sb);
3123                }
3124                else {
3125                        /* we can borrow from the file in the work tree */
3126                        temp->name = name;
3127                        if (!one->oid_valid)
3128                                oid_to_hex_r(temp->hex, &null_oid);
3129                        else
3130                                oid_to_hex_r(temp->hex, &one->oid);
3131                        /* Even though we may sometimes borrow the
3132                         * contents from the work tree, we always want
3133                         * one->mode.  mode is trustworthy even when
3134                         * !(one->oid_valid), as long as
3135                         * DIFF_FILE_VALID(one).
3136                         */
3137                        xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3138                }
3139                return temp;
3140        }
3141        else {
3142                if (diff_populate_filespec(one, 0))
3143                        die("cannot read data blob for %s", one->path);
3144                prep_temp_blob(name, temp, one->data, one->size,
3145                               &one->oid, one->mode);
3146        }
3147        return temp;
3148}
3149
3150static void add_external_diff_name(struct argv_array *argv,
3151                                   const char *name,
3152                                   struct diff_filespec *df)
3153{
3154        struct diff_tempfile *temp = prepare_temp_file(name, df);
3155        argv_array_push(argv, temp->name);
3156        argv_array_push(argv, temp->hex);
3157        argv_array_push(argv, temp->mode);
3158}
3159
3160/* An external diff command takes:
3161 *
3162 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3163 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
3164 *
3165 */
3166static void run_external_diff(const char *pgm,
3167                              const char *name,
3168                              const char *other,
3169                              struct diff_filespec *one,
3170                              struct diff_filespec *two,
3171                              const char *xfrm_msg,
3172                              int complete_rewrite,
3173                              struct diff_options *o)
3174{
3175        struct argv_array argv = ARGV_ARRAY_INIT;
3176        struct argv_array env = ARGV_ARRAY_INIT;
3177        struct diff_queue_struct *q = &diff_queued_diff;
3178
3179        argv_array_push(&argv, pgm);
3180        argv_array_push(&argv, name);
3181
3182        if (one && two) {
3183                add_external_diff_name(&argv, name, one);
3184                if (!other)
3185                        add_external_diff_name(&argv, name, two);
3186                else {
3187                        add_external_diff_name(&argv, other, two);
3188                        argv_array_push(&argv, other);
3189                        argv_array_push(&argv, xfrm_msg);
3190                }
3191        }
3192
3193        argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3194        argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3195
3196        if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3197                die(_("external diff died, stopping at %s"), name);
3198
3199        remove_tempfile();
3200        argv_array_clear(&argv);
3201        argv_array_clear(&env);
3202}
3203
3204static int similarity_index(struct diff_filepair *p)
3205{
3206        return p->score * 100 / MAX_SCORE;
3207}
3208
3209static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3210{
3211        if (startup_info->have_repository)
3212                return find_unique_abbrev(oid->hash, abbrev);
3213        else {
3214                char *hex = oid_to_hex(oid);
3215                if (abbrev < 0)
3216                        abbrev = FALLBACK_DEFAULT_ABBREV;
3217                if (abbrev > GIT_SHA1_HEXSZ)
3218                        die("BUG: oid abbreviation out of range: %d", abbrev);
3219                if (abbrev)
3220                        hex[abbrev] = '\0';
3221                return hex;
3222        }
3223}
3224
3225static void fill_metainfo(struct strbuf *msg,
3226                          const char *name,
3227                          const char *other,
3228                          struct diff_filespec *one,
3229                          struct diff_filespec *two,
3230                          struct diff_options *o,
3231                          struct diff_filepair *p,
3232                          int *must_show_header,
3233                          int use_color)
3234{
3235        const char *set = diff_get_color(use_color, DIFF_METAINFO);
3236        const char *reset = diff_get_color(use_color, DIFF_RESET);
3237        const char *line_prefix = diff_line_prefix(o);
3238
3239        *must_show_header = 1;
3240        strbuf_init(msg, PATH_MAX * 2 + 300);
3241        switch (p->status) {
3242        case DIFF_STATUS_COPIED:
3243                strbuf_addf(msg, "%s%ssimilarity index %d%%",
3244                            line_prefix, set, similarity_index(p));
3245                strbuf_addf(msg, "%s\n%s%scopy from ",
3246                            reset,  line_prefix, set);
3247                quote_c_style(name, msg, NULL, 0);
3248                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3249                quote_c_style(other, msg, NULL, 0);
3250                strbuf_addf(msg, "%s\n", reset);
3251                break;
3252        case DIFF_STATUS_RENAMED:
3253                strbuf_addf(msg, "%s%ssimilarity index %d%%",
3254                            line_prefix, set, similarity_index(p));
3255                strbuf_addf(msg, "%s\n%s%srename from ",
3256                            reset, line_prefix, set);
3257                quote_c_style(name, msg, NULL, 0);
3258                strbuf_addf(msg, "%s\n%s%srename to ",
3259                            reset, line_prefix, set);
3260                quote_c_style(other, msg, NULL, 0);
3261                strbuf_addf(msg, "%s\n", reset);
3262                break;
3263        case DIFF_STATUS_MODIFIED:
3264                if (p->score) {
3265                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3266                                    line_prefix,
3267                                    set, similarity_index(p), reset);
3268                        break;
3269                }
3270                /* fallthru */
3271        default:
3272                *must_show_header = 0;
3273        }
3274        if (one && two && oidcmp(&one->oid, &two->oid)) {
3275                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3276
3277                if (DIFF_OPT_TST(o, BINARY)) {
3278                        mmfile_t mf;
3279                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3280                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3281                                abbrev = 40;
3282                }
3283                strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3284                            diff_abbrev_oid(&one->oid, abbrev),
3285                            diff_abbrev_oid(&two->oid, abbrev));
3286                if (one->mode == two->mode)
3287                        strbuf_addf(msg, " %06o", one->mode);
3288                strbuf_addf(msg, "%s\n", reset);
3289        }
3290}
3291
3292static void run_diff_cmd(const char *pgm,
3293                         const char *name,
3294                         const char *other,
3295                         const char *attr_path,
3296                         struct diff_filespec *one,
3297                         struct diff_filespec *two,
3298                         struct strbuf *msg,
3299                         struct diff_options *o,
3300                         struct diff_filepair *p)
3301{
3302        const char *xfrm_msg = NULL;
3303        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3304        int must_show_header = 0;
3305
3306
3307        if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3308                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3309                if (drv && drv->external)
3310                        pgm = drv->external;
3311        }
3312
3313        if (msg) {
3314                /*
3315                 * don't use colors when the header is intended for an
3316                 * external diff driver
3317                 */
3318                fill_metainfo(msg, name, other, one, two, o, p,
3319                              &must_show_header,
3320                              want_color(o->use_color) && !pgm);
3321                xfrm_msg = msg->len ? msg->buf : NULL;
3322        }
3323
3324        if (pgm) {
3325                run_external_diff(pgm, name, other, one, two, xfrm_msg,
3326                                  complete_rewrite, o);
3327                return;
3328        }
3329        if (one && two)
3330                builtin_diff(name, other ? other : name,
3331                             one, two, xfrm_msg, must_show_header,
3332                             o, complete_rewrite);
3333        else
3334                fprintf(o->file, "* Unmerged path %s\n", name);
3335}
3336
3337static void diff_fill_oid_info(struct diff_filespec *one)
3338{
3339        if (DIFF_FILE_VALID(one)) {
3340                if (!one->oid_valid) {
3341                        struct stat st;
3342                        if (one->is_stdin) {
3343                                oidclr(&one->oid);
3344                                return;
3345                        }
3346                        if (lstat(one->path, &st) < 0)
3347                                die_errno("stat '%s'", one->path);
3348                        if (index_path(one->oid.hash, one->path, &st, 0))
3349                                die("cannot hash %s", one->path);
3350                }
3351        }
3352        else
3353                oidclr(&one->oid);
3354}
3355
3356static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3357{
3358        /* Strip the prefix but do not molest /dev/null and absolute paths */
3359        if (*namep && **namep != '/') {
3360                *namep += prefix_length;
3361                if (**namep == '/')
3362                        ++*namep;
3363        }
3364        if (*otherp && **otherp != '/') {
3365                *otherp += prefix_length;
3366                if (**otherp == '/')
3367                        ++*otherp;
3368        }
3369}
3370
3371static void run_diff(struct diff_filepair *p, struct diff_options *o)
3372{
3373        const char *pgm = external_diff();
3374        struct strbuf msg;
3375        struct diff_filespec *one = p->one;
3376        struct diff_filespec *two = p->two;
3377        const char *name;
3378        const char *other;
3379        const char *attr_path;
3380
3381        name  = one->path;
3382        other = (strcmp(name, two->path) ? two->path : NULL);
3383        attr_path = name;
3384        if (o->prefix_length)
3385                strip_prefix(o->prefix_length, &name, &other);
3386
3387        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3388                pgm = NULL;
3389
3390        if (DIFF_PAIR_UNMERGED(p)) {
3391                run_diff_cmd(pgm, name, NULL, attr_path,
3392                             NULL, NULL, NULL, o, p);
3393                return;
3394        }
3395
3396        diff_fill_oid_info(one);
3397        diff_fill_oid_info(two);
3398
3399        if (!pgm &&
3400            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3401            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3402                /*
3403                 * a filepair that changes between file and symlink
3404                 * needs to be split into deletion and creation.
3405                 */
3406                struct diff_filespec *null = alloc_filespec(two->path);
3407                run_diff_cmd(NULL, name, other, attr_path,
3408                             one, null, &msg, o, p);
3409                free(null);
3410                strbuf_release(&msg);
3411
3412                null = alloc_filespec(one->path);
3413                run_diff_cmd(NULL, name, other, attr_path,
3414                             null, two, &msg, o, p);
3415                free(null);
3416        }
3417        else
3418                run_diff_cmd(pgm, name, other, attr_path,
3419                             one, two, &msg, o, p);
3420
3421        strbuf_release(&msg);
3422}
3423
3424static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3425                         struct diffstat_t *diffstat)
3426{
3427        const char *name;
3428        const char *other;
3429
3430        if (DIFF_PAIR_UNMERGED(p)) {
3431                /* unmerged */
3432                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3433                return;
3434        }
3435
3436        name = p->one->path;
3437        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3438
3439        if (o->prefix_length)
3440                strip_prefix(o->prefix_length, &name, &other);
3441
3442        diff_fill_oid_info(p->one);
3443        diff_fill_oid_info(p->two);
3444
3445        builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3446}
3447
3448static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3449{
3450        const char *name;
3451        const char *other;
3452        const char *attr_path;
3453
3454        if (DIFF_PAIR_UNMERGED(p)) {
3455                /* unmerged */
3456                return;
3457        }
3458
3459        name = p->one->path;
3460        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3461        attr_path = other ? other : name;
3462
3463        if (o->prefix_length)
3464                strip_prefix(o->prefix_length, &name, &other);
3465
3466        diff_fill_oid_info(p->one);
3467        diff_fill_oid_info(p->two);
3468
3469        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3470}
3471
3472void diff_setup(struct diff_options *options)
3473{
3474        memcpy(options, &default_diff_options, sizeof(*options));
3475
3476        options->file = stdout;
3477
3478        options->abbrev = DEFAULT_ABBREV;
3479        options->line_termination = '\n';
3480        options->break_opt = -1;
3481        options->rename_limit = -1;
3482        options->dirstat_permille = diff_dirstat_permille_default;
3483        options->context = diff_context_default;
3484        options->interhunkcontext = diff_interhunk_context_default;
3485        options->ws_error_highlight = ws_error_highlight_default;
3486        DIFF_OPT_SET(options, RENAME_EMPTY);
3487
3488        /* pathchange left =NULL by default */
3489        options->change = diff_change;
3490        options->add_remove = diff_addremove;
3491        options->use_color = diff_use_color_default;
3492        options->detect_rename = diff_detect_rename_default;
3493        options->xdl_opts |= diff_algorithm;
3494        if (diff_indent_heuristic)
3495                DIFF_XDL_SET(options, INDENT_HEURISTIC);
3496
3497        options->orderfile = diff_order_file_cfg;
3498
3499        if (diff_no_prefix) {
3500                options->a_prefix = options->b_prefix = "";
3501        } else if (!diff_mnemonic_prefix) {
3502                options->a_prefix = "a/";
3503                options->b_prefix = "b/";
3504        }
3505}
3506
3507void diff_setup_done(struct diff_options *options)
3508{
3509        int count = 0;
3510
3511        if (options->set_default)
3512                options->set_default(options);
3513
3514        if (options->output_format & DIFF_FORMAT_NAME)
3515                count++;
3516        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3517                count++;
3518        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3519                count++;
3520        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3521                count++;
3522        if (count > 1)
3523                die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3524
3525        /*
3526         * Most of the time we can say "there are changes"
3527         * only by checking if there are changed paths, but
3528         * --ignore-whitespace* options force us to look
3529         * inside contents.
3530         */
3531
3532        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3533            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3534            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3535                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3536        else
3537                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3538
3539        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3540                options->detect_rename = DIFF_DETECT_COPY;
3541
3542        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3543                options->prefix = NULL;
3544        if (options->prefix)
3545                options->prefix_length = strlen(options->prefix);
3546        else
3547                options->prefix_length = 0;
3548
3549        if (options->output_format & (DIFF_FORMAT_NAME |
3550                                      DIFF_FORMAT_NAME_STATUS |
3551                                      DIFF_FORMAT_CHECKDIFF |
3552                                      DIFF_FORMAT_NO_OUTPUT))
3553                options->output_format &= ~(DIFF_FORMAT_RAW |
3554                                            DIFF_FORMAT_NUMSTAT |
3555                                            DIFF_FORMAT_DIFFSTAT |
3556                                            DIFF_FORMAT_SHORTSTAT |
3557                                            DIFF_FORMAT_DIRSTAT |
3558                                            DIFF_FORMAT_SUMMARY |
3559                                            DIFF_FORMAT_PATCH);
3560
3561        /*
3562         * These cases always need recursive; we do not drop caller-supplied
3563         * recursive bits for other formats here.
3564         */
3565        if (options->output_format & (DIFF_FORMAT_PATCH |
3566                                      DIFF_FORMAT_NUMSTAT |
3567                                      DIFF_FORMAT_DIFFSTAT |
3568                                      DIFF_FORMAT_SHORTSTAT |
3569                                      DIFF_FORMAT_DIRSTAT |
3570                                      DIFF_FORMAT_SUMMARY |
3571                                      DIFF_FORMAT_CHECKDIFF))
3572                DIFF_OPT_SET(options, RECURSIVE);
3573        /*
3574         * Also pickaxe would not work very well if you do not say recursive
3575         */
3576        if (options->pickaxe)
3577                DIFF_OPT_SET(options, RECURSIVE);
3578        /*
3579         * When patches are generated, submodules diffed against the work tree
3580         * must be checked for dirtiness too so it can be shown in the output
3581         */
3582        if (options->output_format & DIFF_FORMAT_PATCH)
3583                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3584
3585        if (options->detect_rename && options->rename_limit < 0)
3586                options->rename_limit = diff_rename_limit_default;
3587        if (options->setup & DIFF_SETUP_USE_CACHE) {
3588                if (!active_cache)
3589                        /* read-cache does not die even when it fails
3590                         * so it is safe for us to do this here.  Also
3591                         * it does not smudge active_cache or active_nr
3592                         * when it fails, so we do not have to worry about
3593                         * cleaning it up ourselves either.
3594                         */
3595                        read_cache();
3596        }
3597        if (40 < options->abbrev)
3598                options->abbrev = 40; /* full */
3599
3600        /*
3601         * It does not make sense to show the first hit we happened
3602         * to have found.  It does not make sense not to return with
3603         * exit code in such a case either.
3604         */
3605        if (DIFF_OPT_TST(options, QUICK)) {
3606                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3607                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3608        }
3609
3610        options->diff_path_counter = 0;
3611
3612        if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3613                die(_("--follow requires exactly one pathspec"));
3614}
3615
3616static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3617{
3618        char c, *eq;
3619        int len;
3620
3621        if (*arg != '-')
3622                return 0;
3623        c = *++arg;
3624        if (!c)
3625                return 0;
3626        if (c == arg_short) {
3627                c = *++arg;
3628                if (!c)
3629                        return 1;
3630                if (val && isdigit(c)) {
3631                        char *end;
3632                        int n = strtoul(arg, &end, 10);
3633                        if (*end)
3634                                return 0;
3635                        *val = n;
3636                        return 1;
3637                }
3638                return 0;
3639        }
3640        if (c != '-')
3641                return 0;
3642        arg++;
3643        eq = strchrnul(arg, '=');
3644        len = eq - arg;
3645        if (!len || strncmp(arg, arg_long, len))
3646                return 0;
3647        if (*eq) {
3648                int n;
3649                char *end;
3650                if (!isdigit(*++eq))
3651                        return 0;
3652                n = strtoul(eq, &end, 10);
3653                if (*end)
3654                        return 0;
3655                *val = n;
3656        }
3657        return 1;
3658}
3659
3660static int diff_scoreopt_parse(const char *opt);
3661
3662static inline int short_opt(char opt, const char **argv,
3663                            const char **optarg)
3664{
3665        const char *arg = argv[0];
3666        if (arg[0] != '-' || arg[1] != opt)
3667                return 0;
3668        if (arg[2] != '\0') {
3669                *optarg = arg + 2;
3670                return 1;
3671        }
3672        if (!argv[1])
3673                die("Option '%c' requires a value", opt);
3674        *optarg = argv[1];
3675        return 2;
3676}
3677
3678int parse_long_opt(const char *opt, const char **argv,
3679                   const char **optarg)
3680{
3681        const char *arg = argv[0];
3682        if (!skip_prefix(arg, "--", &arg))
3683                return 0;
3684        if (!skip_prefix(arg, opt, &arg))
3685                return 0;
3686        if (*arg == '=') { /* stuck form: --option=value */
3687                *optarg = arg + 1;
3688                return 1;
3689        }
3690        if (*arg != '\0')
3691                return 0;
3692        /* separate form: --option value */
3693        if (!argv[1])
3694                die("Option '--%s' requires a value", opt);
3695        *optarg = argv[1];
3696        return 2;
3697}
3698
3699static int stat_opt(struct diff_options *options, const char **av)
3700{
3701        const char *arg = av[0];
3702        char *end;
3703        int width = options->stat_width;
3704        int name_width = options->stat_name_width;
3705        int graph_width = options->stat_graph_width;
3706        int count = options->stat_count;
3707        int argcount = 1;
3708
3709        if (!skip_prefix(arg, "--stat", &arg))
3710                die("BUG: stat option does not begin with --stat: %s", arg);
3711        end = (char *)arg;
3712
3713        switch (*arg) {
3714        case '-':
3715                if (skip_prefix(arg, "-width", &arg)) {
3716                        if (*arg == '=')
3717                                width = strtoul(arg + 1, &end, 10);
3718                        else if (!*arg && !av[1])
3719                                die_want_option("--stat-width");
3720                        else if (!*arg) {
3721                                width = strtoul(av[1], &end, 10);
3722                                argcount = 2;
3723                        }
3724                } else if (skip_prefix(arg, "-name-width", &arg)) {
3725                        if (*arg == '=')
3726                                name_width = strtoul(arg + 1, &end, 10);
3727                        else if (!*arg && !av[1])
3728                                die_want_option("--stat-name-width");
3729                        else if (!*arg) {
3730                                name_width = strtoul(av[1], &end, 10);
3731                                argcount = 2;
3732                        }
3733                } else if (skip_prefix(arg, "-graph-width", &arg)) {
3734                        if (*arg == '=')
3735                                graph_width = strtoul(arg + 1, &end, 10);
3736                        else if (!*arg && !av[1])
3737                                die_want_option("--stat-graph-width");
3738                        else if (!*arg) {
3739                                graph_width = strtoul(av[1], &end, 10);
3740                                argcount = 2;
3741                        }
3742                } else if (skip_prefix(arg, "-count", &arg)) {
3743                        if (*arg == '=')
3744                                count = strtoul(arg + 1, &end, 10);
3745                        else if (!*arg && !av[1])
3746                                die_want_option("--stat-count");
3747                        else if (!*arg) {
3748                                count = strtoul(av[1], &end, 10);
3749                                argcount = 2;
3750                        }
3751                }
3752                break;
3753        case '=':
3754                width = strtoul(arg+1, &end, 10);
3755                if (*end == ',')
3756                        name_width = strtoul(end+1, &end, 10);
3757                if (*end == ',')
3758                        count = strtoul(end+1, &end, 10);
3759        }
3760
3761        /* Important! This checks all the error cases! */
3762        if (*end)
3763                return 0;
3764        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3765        options->stat_name_width = name_width;
3766        options->stat_graph_width = graph_width;
3767        options->stat_width = width;
3768        options->stat_count = count;
3769        return argcount;
3770}
3771
3772static int parse_dirstat_opt(struct diff_options *options, const char *params)
3773{
3774        struct strbuf errmsg = STRBUF_INIT;
3775        if (parse_dirstat_params(options, params, &errmsg))
3776                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3777                    errmsg.buf);
3778        strbuf_release(&errmsg);
3779        /*
3780         * The caller knows a dirstat-related option is given from the command
3781         * line; allow it to say "return this_function();"
3782         */
3783        options->output_format |= DIFF_FORMAT_DIRSTAT;
3784        return 1;
3785}
3786
3787static int parse_submodule_opt(struct diff_options *options, const char *value)
3788{
3789        if (parse_submodule_params(options, value))
3790                die(_("Failed to parse --submodule option parameter: '%s'"),
3791                        value);
3792        return 1;
3793}
3794
3795static const char diff_status_letters[] = {
3796        DIFF_STATUS_ADDED,
3797        DIFF_STATUS_COPIED,
3798        DIFF_STATUS_DELETED,
3799        DIFF_STATUS_MODIFIED,
3800        DIFF_STATUS_RENAMED,
3801        DIFF_STATUS_TYPE_CHANGED,
3802        DIFF_STATUS_UNKNOWN,
3803        DIFF_STATUS_UNMERGED,
3804        DIFF_STATUS_FILTER_AON,
3805        DIFF_STATUS_FILTER_BROKEN,
3806        '\0',
3807};
3808
3809static unsigned int filter_bit['Z' + 1];
3810
3811static void prepare_filter_bits(void)
3812{
3813        int i;
3814
3815        if (!filter_bit[DIFF_STATUS_ADDED]) {
3816                for (i = 0; diff_status_letters[i]; i++)
3817                        filter_bit[(int) diff_status_letters[i]] = (1 << i);
3818        }
3819}
3820
3821static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3822{
3823        return opt->filter & filter_bit[(int) status];
3824}
3825
3826static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3827{
3828        int i, optch;
3829
3830        prepare_filter_bits();
3831
3832        /*
3833         * If there is a negation e.g. 'd' in the input, and we haven't
3834         * initialized the filter field with another --diff-filter, start
3835         * from full set of bits, except for AON.
3836         */
3837        if (!opt->filter) {
3838                for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3839                        if (optch < 'a' || 'z' < optch)
3840                                continue;
3841                        opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3842                        opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3843                        break;
3844                }
3845        }
3846
3847        for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3848                unsigned int bit;
3849                int negate;
3850
3851                if ('a' <= optch && optch <= 'z') {
3852                        negate = 1;
3853                        optch = toupper(optch);
3854                } else {
3855                        negate = 0;
3856                }
3857
3858                bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3859                if (!bit)
3860                        return optarg[i];
3861                if (negate)
3862                        opt->filter &= ~bit;
3863                else
3864                        opt->filter |= bit;
3865        }
3866        return 0;
3867}
3868
3869static void enable_patch_output(int *fmt) {
3870        *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3871        *fmt |= DIFF_FORMAT_PATCH;
3872}
3873
3874static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3875{
3876        int val = parse_ws_error_highlight(arg);
3877
3878        if (val < 0) {
3879                error("unknown value after ws-error-highlight=%.*s",
3880                      -1 - val, arg);
3881                return 0;
3882        }
3883        opt->ws_error_highlight = val;
3884        return 1;
3885}
3886
3887int diff_opt_parse(struct diff_options *options,
3888                   const char **av, int ac, const char *prefix)
3889{
3890        const char *arg = av[0];
3891        const char *optarg;
3892        int argcount;
3893
3894        if (!prefix)
3895                prefix = "";
3896
3897        /* Output format options */
3898        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3899            || opt_arg(arg, 'U', "unified", &options->context))
3900                enable_patch_output(&options->output_format);
3901        else if (!strcmp(arg, "--raw"))
3902                options->output_format |= DIFF_FORMAT_RAW;
3903        else if (!strcmp(arg, "--patch-with-raw")) {
3904                enable_patch_output(&options->output_format);
3905                options->output_format |= DIFF_FORMAT_RAW;
3906        } else if (!strcmp(arg, "--numstat"))
3907                options->output_format |= DIFF_FORMAT_NUMSTAT;
3908        else if (!strcmp(arg, "--shortstat"))
3909                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3910        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3911                return parse_dirstat_opt(options, "");
3912        else if (skip_prefix(arg, "-X", &arg))
3913                return parse_dirstat_opt(options, arg);
3914        else if (skip_prefix(arg, "--dirstat=", &arg))
3915                return parse_dirstat_opt(options, arg);
3916        else if (!strcmp(arg, "--cumulative"))
3917                return parse_dirstat_opt(options, "cumulative");
3918        else if (!strcmp(arg, "--dirstat-by-file"))
3919                return parse_dirstat_opt(options, "files");
3920        else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3921                parse_dirstat_opt(options, "files");
3922                return parse_dirstat_opt(options, arg);
3923        }
3924        else if (!strcmp(arg, "--check"))
3925                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3926        else if (!strcmp(arg, "--summary"))
3927                options->output_format |= DIFF_FORMAT_SUMMARY;
3928        else if (!strcmp(arg, "--patch-with-stat")) {
3929                enable_patch_output(&options->output_format);
3930                options->output_format |= DIFF_FORMAT_DIFFSTAT;
3931        } else if (!strcmp(arg, "--name-only"))
3932                options->output_format |= DIFF_FORMAT_NAME;
3933        else if (!strcmp(arg, "--name-status"))
3934                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3935        else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3936                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3937        else if (starts_with(arg, "--stat"))
3938                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3939                return stat_opt(options, av);
3940
3941        /* renames options */
3942        else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3943                 !strcmp(arg, "--break-rewrites")) {
3944                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3945                        return error("invalid argument to -B: %s", arg+2);
3946        }
3947        else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3948                 !strcmp(arg, "--find-renames")) {
3949                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3950                        return error("invalid argument to -M: %s", arg+2);
3951                options->detect_rename = DIFF_DETECT_RENAME;
3952        }
3953        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3954                options->irreversible_delete = 1;
3955        }
3956        else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3957                 !strcmp(arg, "--find-copies")) {
3958                if (options->detect_rename == DIFF_DETECT_COPY)
3959                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3960                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3961                        return error("invalid argument to -C: %s", arg+2);
3962                options->detect_rename = DIFF_DETECT_COPY;
3963        }
3964        else if (!strcmp(arg, "--no-renames"))
3965                options->detect_rename = 0;
3966        else if (!strcmp(arg, "--rename-empty"))
3967                DIFF_OPT_SET(options, RENAME_EMPTY);
3968        else if (!strcmp(arg, "--no-rename-empty"))
3969                DIFF_OPT_CLR(options, RENAME_EMPTY);
3970        else if (!strcmp(arg, "--relative"))
3971                DIFF_OPT_SET(options, RELATIVE_NAME);
3972        else if (skip_prefix(arg, "--relative=", &arg)) {
3973                DIFF_OPT_SET(options, RELATIVE_NAME);
3974                options->prefix = arg;
3975        }
3976
3977        /* xdiff options */
3978        else if (!strcmp(arg, "--minimal"))
3979                DIFF_XDL_SET(options, NEED_MINIMAL);
3980        else if (!strcmp(arg, "--no-minimal"))
3981                DIFF_XDL_CLR(options, NEED_MINIMAL);
3982        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3983                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3984        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3985                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3986        else if (!strcmp(arg, "--ignore-space-at-eol"))
3987                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3988        else if (!strcmp(arg, "--ignore-blank-lines"))
3989                DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3990        else if (!strcmp(arg, "--indent-heuristic"))
3991                DIFF_XDL_SET(options, INDENT_HEURISTIC);
3992        else if (!strcmp(arg, "--no-indent-heuristic"))
3993                DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3994        else if (!strcmp(arg, "--patience"))
3995                options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3996        else if (!strcmp(arg, "--histogram"))
3997                options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3998        else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3999                long value = parse_algorithm_value(optarg);
4000                if (value < 0)
4001                        return error("option diff-algorithm accepts \"myers\", "
4002                                     "\"minimal\", \"patience\" and \"histogram\"");
4003                /* clear out previous settings */
4004                DIFF_XDL_CLR(options, NEED_MINIMAL);
4005                options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
4006                options->xdl_opts |= value;
4007                return argcount;
4008        }
4009
4010        /* flags options */
4011        else if (!strcmp(arg, "--binary")) {
4012                enable_patch_output(&options->output_format);
4013                DIFF_OPT_SET(options, BINARY);
4014        }
4015        else if (!strcmp(arg, "--full-index"))
4016                DIFF_OPT_SET(options, FULL_INDEX);
4017        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
4018                DIFF_OPT_SET(options, TEXT);
4019        else if (!strcmp(arg, "-R"))
4020                DIFF_OPT_SET(options, REVERSE_DIFF);
4021        else if (!strcmp(arg, "--find-copies-harder"))
4022                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
4023        else if (!strcmp(arg, "--follow"))
4024                DIFF_OPT_SET(options, FOLLOW_RENAMES);
4025        else if (!strcmp(arg, "--no-follow")) {
4026                DIFF_OPT_CLR(options, FOLLOW_RENAMES);
4027                DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
4028        } else if (!strcmp(arg, "--color"))
4029                options->use_color = 1;
4030        else if (skip_prefix(arg, "--color=", &arg)) {
4031                int value = git_config_colorbool(NULL, arg);
4032                if (value < 0)
4033                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
4034                options->use_color = value;
4035        }
4036        else if (!strcmp(arg, "--no-color"))
4037                options->use_color = 0;
4038        else if (!strcmp(arg, "--color-words")) {
4039                options->use_color = 1;
4040                options->word_diff = DIFF_WORDS_COLOR;
4041        }
4042        else if (skip_prefix(arg, "--color-words=", &arg)) {
4043                options->use_color = 1;
4044                options->word_diff = DIFF_WORDS_COLOR;
4045                options->word_regex = arg;
4046        }
4047        else if (!strcmp(arg, "--word-diff")) {
4048                if (options->word_diff == DIFF_WORDS_NONE)
4049                        options->word_diff = DIFF_WORDS_PLAIN;
4050        }
4051        else if (skip_prefix(arg, "--word-diff=", &arg)) {
4052                if (!strcmp(arg, "plain"))
4053                        options->word_diff = DIFF_WORDS_PLAIN;
4054                else if (!strcmp(arg, "color")) {
4055                        options->use_color = 1;
4056                        options->word_diff = DIFF_WORDS_COLOR;
4057                }
4058                else if (!strcmp(arg, "porcelain"))
4059                        options->word_diff = DIFF_WORDS_PORCELAIN;
4060                else if (!strcmp(arg, "none"))
4061                        options->word_diff = DIFF_WORDS_NONE;
4062                else
4063                        die("bad --word-diff argument: %s", arg);
4064        }
4065        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
4066                if (options->word_diff == DIFF_WORDS_NONE)
4067                        options->word_diff = DIFF_WORDS_PLAIN;
4068                options->word_regex = optarg;
4069                return argcount;
4070        }
4071        else if (!strcmp(arg, "--exit-code"))
4072                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4073        else if (!strcmp(arg, "--quiet"))
4074                DIFF_OPT_SET(options, QUICK);
4075        else if (!strcmp(arg, "--ext-diff"))
4076                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4077        else if (!strcmp(arg, "--no-ext-diff"))
4078                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4079        else if (!strcmp(arg, "--textconv"))
4080                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4081        else if (!strcmp(arg, "--no-textconv"))
4082                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4083        else if (!strcmp(arg, "--ignore-submodules")) {
4084                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4085                handle_ignore_submodules_arg(options, "all");
4086        } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4087                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4088                handle_ignore_submodules_arg(options, arg);
4089        } else if (!strcmp(arg, "--submodule"))
4090                options->submodule_format = DIFF_SUBMODULE_LOG;
4091        else if (skip_prefix(arg, "--submodule=", &arg))
4092                return parse_submodule_opt(options, arg);
4093        else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4094                return parse_ws_error_highlight_opt(options, arg);
4095        else if (!strcmp(arg, "--ita-invisible-in-index"))
4096                options->ita_invisible_in_index = 1;
4097        else if (!strcmp(arg, "--ita-visible-in-index"))
4098                options->ita_invisible_in_index = 0;
4099
4100        /* misc options */
4101        else if (!strcmp(arg, "-z"))
4102                options->line_termination = 0;
4103        else if ((argcount = short_opt('l', av, &optarg))) {
4104                options->rename_limit = strtoul(optarg, NULL, 10);
4105                return argcount;
4106        }
4107        else if ((argcount = short_opt('S', av, &optarg))) {
4108                options->pickaxe = optarg;
4109                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4110                return argcount;
4111        } else if ((argcount = short_opt('G', av, &optarg))) {
4112                options->pickaxe = optarg;
4113                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4114                return argcount;
4115        }
4116        else if (!strcmp(arg, "--pickaxe-all"))
4117                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4118        else if (!strcmp(arg, "--pickaxe-regex"))
4119                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4120        else if ((argcount = short_opt('O', av, &optarg))) {
4121                options->orderfile = prefix_filename(prefix, optarg);
4122                return argcount;
4123        }
4124        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4125                int offending = parse_diff_filter_opt(optarg, options);
4126                if (offending)
4127                        die("unknown change class '%c' in --diff-filter=%s",
4128                            offending, optarg);
4129                return argcount;
4130        }
4131        else if (!strcmp(arg, "--no-abbrev"))
4132                options->abbrev = 0;
4133        else if (!strcmp(arg, "--abbrev"))
4134                options->abbrev = DEFAULT_ABBREV;
4135        else if (skip_prefix(arg, "--abbrev=", &arg)) {
4136                options->abbrev = strtoul(arg, NULL, 10);
4137                if (options->abbrev < MINIMUM_ABBREV)
4138                        options->abbrev = MINIMUM_ABBREV;
4139                else if (40 < options->abbrev)
4140                        options->abbrev = 40;
4141        }
4142        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4143                options->a_prefix = optarg;
4144                return argcount;
4145        }
4146        else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4147                options->line_prefix = optarg;
4148                options->line_prefix_length = strlen(options->line_prefix);
4149                graph_setup_line_prefix(options);
4150                return argcount;
4151        }
4152        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4153                options->b_prefix = optarg;
4154                return argcount;
4155        }
4156        else if (!strcmp(arg, "--no-prefix"))
4157                options->a_prefix = options->b_prefix = "";
4158        else if (opt_arg(arg, '\0', "inter-hunk-context",
4159                         &options->interhunkcontext))
4160                ;
4161        else if (!strcmp(arg, "-W"))
4162                DIFF_OPT_SET(options, FUNCCONTEXT);
4163        else if (!strcmp(arg, "--function-context"))
4164                DIFF_OPT_SET(options, FUNCCONTEXT);
4165        else if (!strcmp(arg, "--no-function-context"))
4166                DIFF_OPT_CLR(options, FUNCCONTEXT);
4167        else if ((argcount = parse_long_opt("output", av, &optarg))) {
4168                char *path = prefix_filename(prefix, optarg);
4169                options->file = xfopen(path, "w");
4170                options->close_file = 1;
4171                if (options->use_color != GIT_COLOR_ALWAYS)
4172                        options->use_color = GIT_COLOR_NEVER;
4173                free(path);
4174                return argcount;
4175        } else
4176                return 0;
4177        return 1;
4178}
4179
4180int parse_rename_score(const char **cp_p)
4181{
4182        unsigned long num, scale;
4183        int ch, dot;
4184        const char *cp = *cp_p;
4185
4186        num = 0;
4187        scale = 1;
4188        dot = 0;
4189        for (;;) {
4190                ch = *cp;
4191                if ( !dot && ch == '.' ) {
4192                        scale = 1;
4193                        dot = 1;
4194                } else if ( ch == '%' ) {
4195                        scale = dot ? scale*100 : 100;
4196                        cp++;   /* % is always at the end */
4197                        break;
4198                } else if ( ch >= '0' && ch <= '9' ) {
4199                        if ( scale < 100000 ) {
4200                                scale *= 10;
4201                                num = (num*10) + (ch-'0');
4202                        }
4203                } else {
4204                        break;
4205                }
4206                cp++;
4207        }
4208        *cp_p = cp;
4209
4210        /* user says num divided by scale and we say internally that
4211         * is MAX_SCORE * num / scale.
4212         */
4213        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4214}
4215
4216static int diff_scoreopt_parse(const char *opt)
4217{
4218        int opt1, opt2, cmd;
4219
4220        if (*opt++ != '-')
4221                return -1;
4222        cmd = *opt++;
4223        if (cmd == '-') {
4224                /* convert the long-form arguments into short-form versions */
4225                if (skip_prefix(opt, "break-rewrites", &opt)) {
4226                        if (*opt == 0 || *opt++ == '=')
4227                                cmd = 'B';
4228                } else if (skip_prefix(opt, "find-copies", &opt)) {
4229                        if (*opt == 0 || *opt++ == '=')
4230                                cmd = 'C';
4231                } else if (skip_prefix(opt, "find-renames", &opt)) {
4232                        if (*opt == 0 || *opt++ == '=')
4233                                cmd = 'M';
4234                }
4235        }
4236        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4237                return -1; /* that is not a -M, -C, or -B option */
4238
4239        opt1 = parse_rename_score(&opt);
4240        if (cmd != 'B')
4241                opt2 = 0;
4242        else {
4243                if (*opt == 0)
4244                        opt2 = 0;
4245                else if (*opt != '/')
4246                        return -1; /* we expect -B80/99 or -B80 */
4247                else {
4248                        opt++;
4249                        opt2 = parse_rename_score(&opt);
4250                }
4251        }
4252        if (*opt != 0)
4253                return -1;
4254        return opt1 | (opt2 << 16);
4255}
4256
4257struct diff_queue_struct diff_queued_diff;
4258
4259void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4260{
4261        ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4262        queue->queue[queue->nr++] = dp;
4263}
4264
4265struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4266                                 struct diff_filespec *one,
4267                                 struct diff_filespec *two)
4268{
4269        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4270        dp->one = one;
4271        dp->two = two;
4272        if (queue)
4273                diff_q(queue, dp);
4274        return dp;
4275}
4276
4277void diff_free_filepair(struct diff_filepair *p)
4278{
4279        free_filespec(p->one);
4280        free_filespec(p->two);
4281        free(p);
4282}
4283
4284const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4285{
4286        int abblen;
4287        const char *abbrev;
4288
4289        if (len == GIT_SHA1_HEXSZ)
4290                return oid_to_hex(oid);
4291
4292        abbrev = diff_abbrev_oid(oid, len);
4293        abblen = strlen(abbrev);
4294
4295        /*
4296         * In well-behaved cases, where the abbbreviated result is the
4297         * same as the requested length, append three dots after the
4298         * abbreviation (hence the whole logic is limited to the case
4299         * where abblen < 37); when the actual abbreviated result is a
4300         * bit longer than the requested length, we reduce the number
4301         * of dots so that they match the well-behaved ones.  However,
4302         * if the actual abbreviation is longer than the requested
4303         * length by more than three, we give up on aligning, and add
4304         * three dots anyway, to indicate that the output is not the
4305         * full object name.  Yes, this may be suboptimal, but this
4306         * appears only in "diff --raw --abbrev" output and it is not
4307         * worth the effort to change it now.  Note that this would
4308         * likely to work fine when the automatic sizing of default
4309         * abbreviation length is used--we would be fed -1 in "len" in
4310         * that case, and will end up always appending three-dots, but
4311         * the automatic sizing is supposed to give abblen that ensures
4312         * uniqueness across all objects (statistically speaking).
4313         */
4314        if (abblen < GIT_SHA1_HEXSZ - 3) {
4315                static char hex[GIT_MAX_HEXSZ + 1];
4316                if (len < abblen && abblen <= len + 2)
4317                        xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4318                else
4319                        xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4320                return hex;
4321        }
4322
4323        return oid_to_hex(oid);
4324}
4325
4326static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4327{
4328        int line_termination = opt->line_termination;
4329        int inter_name_termination = line_termination ? '\t' : '\0';
4330
4331        fprintf(opt->file, "%s", diff_line_prefix(opt));
4332        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4333                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4334                        diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4335                fprintf(opt->file, "%s ",
4336                        diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4337        }
4338        if (p->score) {
4339                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4340                        inter_name_termination);
4341        } else {
4342                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4343        }
4344
4345        if (p->status == DIFF_STATUS_COPIED ||
4346            p->status == DIFF_STATUS_RENAMED) {
4347                const char *name_a, *name_b;
4348                name_a = p->one->path;
4349                name_b = p->two->path;
4350                strip_prefix(opt->prefix_length, &name_a, &name_b);
4351                write_name_quoted(name_a, opt->file, inter_name_termination);
4352                write_name_quoted(name_b, opt->file, line_termination);
4353        } else {
4354                const char *name_a, *name_b;
4355                name_a = p->one->mode ? p->one->path : p->two->path;
4356                name_b = NULL;
4357                strip_prefix(opt->prefix_length, &name_a, &name_b);
4358                write_name_quoted(name_a, opt->file, line_termination);
4359        }
4360}
4361
4362int diff_unmodified_pair(struct diff_filepair *p)
4363{
4364        /* This function is written stricter than necessary to support
4365         * the currently implemented transformers, but the idea is to
4366         * let transformers to produce diff_filepairs any way they want,
4367         * and filter and clean them up here before producing the output.
4368         */
4369        struct diff_filespec *one = p->one, *two = p->two;
4370
4371        if (DIFF_PAIR_UNMERGED(p))
4372                return 0; /* unmerged is interesting */
4373
4374        /* deletion, addition, mode or type change
4375         * and rename are all interesting.
4376         */
4377        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4378            DIFF_PAIR_MODE_CHANGED(p) ||
4379            strcmp(one->path, two->path))
4380                return 0;
4381
4382        /* both are valid and point at the same path.  that is, we are
4383         * dealing with a change.
4384         */
4385        if (one->oid_valid && two->oid_valid &&
4386            !oidcmp(&one->oid, &two->oid) &&
4387            !one->dirty_submodule && !two->dirty_submodule)
4388                return 1; /* no change */
4389        if (!one->oid_valid && !two->oid_valid)
4390                return 1; /* both look at the same file on the filesystem. */
4391        return 0;
4392}
4393
4394static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4395{
4396        if (diff_unmodified_pair(p))
4397                return;
4398
4399        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4400            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4401                return; /* no tree diffs in patch format */
4402
4403        run_diff(p, o);
4404}
4405
4406static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4407                            struct diffstat_t *diffstat)
4408{
4409        if (diff_unmodified_pair(p))
4410                return;
4411
4412        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4413            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4414                return; /* no useful stat for tree diffs */
4415
4416        run_diffstat(p, o, diffstat);
4417}
4418
4419static void diff_flush_checkdiff(struct diff_filepair *p,
4420                struct diff_options *o)
4421{
4422        if (diff_unmodified_pair(p))
4423                return;
4424
4425        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4426            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4427                return; /* nothing to check in tree diffs */
4428
4429        run_checkdiff(p, o);
4430}
4431
4432int diff_queue_is_empty(void)
4433{
4434        struct diff_queue_struct *q = &diff_queued_diff;
4435        int i;
4436        for (i = 0; i < q->nr; i++)
4437                if (!diff_unmodified_pair(q->queue[i]))
4438                        return 0;
4439        return 1;
4440}
4441
4442#if DIFF_DEBUG
4443void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4444{
4445        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4446                x, one ? one : "",
4447                s->path,
4448                DIFF_FILE_VALID(s) ? "valid" : "invalid",
4449                s->mode,
4450                s->oid_valid ? oid_to_hex(&s->oid) : "");
4451        fprintf(stderr, "queue[%d] %s size %lu\n",
4452                x, one ? one : "",
4453                s->size);
4454}
4455
4456void diff_debug_filepair(const struct diff_filepair *p, int i)
4457{
4458        diff_debug_filespec(p->one, i, "one");
4459        diff_debug_filespec(p->two, i, "two");
4460        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4461                p->score, p->status ? p->status : '?',
4462                p->one->rename_used, p->broken_pair);
4463}
4464
4465void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4466{
4467        int i;
4468        if (msg)
4469                fprintf(stderr, "%s\n", msg);
4470        fprintf(stderr, "q->nr = %d\n", q->nr);
4471        for (i = 0; i < q->nr; i++) {
4472                struct diff_filepair *p = q->queue[i];
4473                diff_debug_filepair(p, i);
4474        }
4475}
4476#endif
4477
4478static void diff_resolve_rename_copy(void)
4479{
4480        int i;
4481        struct diff_filepair *p;
4482        struct diff_queue_struct *q = &diff_queued_diff;
4483
4484        diff_debug_queue("resolve-rename-copy", q);
4485
4486        for (i = 0; i < q->nr; i++) {
4487                p = q->queue[i];
4488                p->status = 0; /* undecided */
4489                if (DIFF_PAIR_UNMERGED(p))
4490                        p->status = DIFF_STATUS_UNMERGED;
4491                else if (!DIFF_FILE_VALID(p->one))
4492                        p->status = DIFF_STATUS_ADDED;
4493                else if (!DIFF_FILE_VALID(p->two))
4494                        p->status = DIFF_STATUS_DELETED;
4495                else if (DIFF_PAIR_TYPE_CHANGED(p))
4496                        p->status = DIFF_STATUS_TYPE_CHANGED;
4497
4498                /* from this point on, we are dealing with a pair
4499                 * whose both sides are valid and of the same type, i.e.
4500                 * either in-place edit or rename/copy edit.
4501                 */
4502                else if (DIFF_PAIR_RENAME(p)) {
4503                        /*
4504                         * A rename might have re-connected a broken
4505                         * pair up, causing the pathnames to be the
4506                         * same again. If so, that's not a rename at
4507                         * all, just a modification..
4508                         *
4509                         * Otherwise, see if this source was used for
4510                         * multiple renames, in which case we decrement
4511                         * the count, and call it a copy.
4512                         */
4513                        if (!strcmp(p->one->path, p->two->path))
4514                                p->status = DIFF_STATUS_MODIFIED;
4515                        else if (--p->one->rename_used > 0)
4516                                p->status = DIFF_STATUS_COPIED;
4517                        else
4518                                p->status = DIFF_STATUS_RENAMED;
4519                }
4520                else if (oidcmp(&p->one->oid, &p->two->oid) ||
4521                         p->one->mode != p->two->mode ||
4522                         p->one->dirty_submodule ||
4523                         p->two->dirty_submodule ||
4524                         is_null_oid(&p->one->oid))
4525                        p->status = DIFF_STATUS_MODIFIED;
4526                else {
4527                        /* This is a "no-change" entry and should not
4528                         * happen anymore, but prepare for broken callers.
4529                         */
4530                        error("feeding unmodified %s to diffcore",
4531                              p->one->path);
4532                        p->status = DIFF_STATUS_UNKNOWN;
4533                }
4534        }
4535        diff_debug_queue("resolve-rename-copy done", q);
4536}
4537
4538static int check_pair_status(struct diff_filepair *p)
4539{
4540        switch (p->status) {
4541        case DIFF_STATUS_UNKNOWN:
4542                return 0;
4543        case 0:
4544                die("internal error in diff-resolve-rename-copy");
4545        default:
4546                return 1;
4547        }
4548}
4549
4550static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4551{
4552        int fmt = opt->output_format;
4553
4554        if (fmt & DIFF_FORMAT_CHECKDIFF)
4555                diff_flush_checkdiff(p, opt);
4556        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4557                diff_flush_raw(p, opt);
4558        else if (fmt & DIFF_FORMAT_NAME) {
4559                const char *name_a, *name_b;
4560                name_a = p->two->path;
4561                name_b = NULL;
4562                strip_prefix(opt->prefix_length, &name_a, &name_b);
4563                fprintf(opt->file, "%s", diff_line_prefix(opt));
4564                write_name_quoted(name_a, opt->file, opt->line_termination);
4565        }
4566}
4567
4568static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4569{
4570        if (fs->mode)
4571                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4572        else
4573                fprintf(file, " %s ", newdelete);
4574        write_name_quoted(fs->path, file, '\n');
4575}
4576
4577
4578static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4579                const char *line_prefix)
4580{
4581        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4582                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4583                        p->two->mode, show_name ? ' ' : '\n');
4584                if (show_name) {
4585                        write_name_quoted(p->two->path, file, '\n');
4586                }
4587        }
4588}
4589
4590static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4591                        const char *line_prefix)
4592{
4593        char *names = pprint_rename(p->one->path, p->two->path);
4594
4595        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4596        free(names);
4597        show_mode_change(file, p, 0, line_prefix);
4598}
4599
4600static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4601{
4602        FILE *file = opt->file;
4603        const char *line_prefix = diff_line_prefix(opt);
4604
4605        switch(p->status) {
4606        case DIFF_STATUS_DELETED:
4607                fputs(line_prefix, file);
4608                show_file_mode_name(file, "delete", p->one);
4609                break;
4610        case DIFF_STATUS_ADDED:
4611                fputs(line_prefix, file);
4612                show_file_mode_name(file, "create", p->two);
4613                break;
4614        case DIFF_STATUS_COPIED:
4615                fputs(line_prefix, file);
4616                show_rename_copy(file, "copy", p, line_prefix);
4617                break;
4618        case DIFF_STATUS_RENAMED:
4619                fputs(line_prefix, file);
4620                show_rename_copy(file, "rename", p, line_prefix);
4621                break;
4622        default:
4623                if (p->score) {
4624                        fprintf(file, "%s rewrite ", line_prefix);
4625                        write_name_quoted(p->two->path, file, ' ');
4626                        fprintf(file, "(%d%%)\n", similarity_index(p));
4627                }
4628                show_mode_change(file, p, !p->score, line_prefix);
4629                break;
4630        }
4631}
4632
4633struct patch_id_t {
4634        git_SHA_CTX *ctx;
4635        int patchlen;
4636};
4637
4638static int remove_space(char *line, int len)
4639{
4640        int i;
4641        char *dst = line;
4642        unsigned char c;
4643
4644        for (i = 0; i < len; i++)
4645                if (!isspace((c = line[i])))
4646                        *dst++ = c;
4647
4648        return dst - line;
4649}
4650
4651static void patch_id_consume(void *priv, char *line, unsigned long len)
4652{
4653        struct patch_id_t *data = priv;
4654        int new_len;
4655
4656        /* Ignore line numbers when computing the SHA1 of the patch */
4657        if (starts_with(line, "@@ -"))
4658                return;
4659
4660        new_len = remove_space(line, len);
4661
4662        git_SHA1_Update(data->ctx, line, new_len);
4663        data->patchlen += new_len;
4664}
4665
4666static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4667{
4668        git_SHA1_Update(ctx, str, strlen(str));
4669}
4670
4671static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4672{
4673        /* large enough for 2^32 in octal */
4674        char buf[12];
4675        int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4676        git_SHA1_Update(ctx, buf, len);
4677}
4678
4679/* returns 0 upon success, and writes result into sha1 */
4680static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4681{
4682        struct diff_queue_struct *q = &diff_queued_diff;
4683        int i;
4684        git_SHA_CTX ctx;
4685        struct patch_id_t data;
4686
4687        git_SHA1_Init(&ctx);
4688        memset(&data, 0, sizeof(struct patch_id_t));
4689        data.ctx = &ctx;
4690
4691        for (i = 0; i < q->nr; i++) {
4692                xpparam_t xpp;
4693                xdemitconf_t xecfg;
4694                mmfile_t mf1, mf2;
4695                struct diff_filepair *p = q->queue[i];
4696                int len1, len2;
4697
4698                memset(&xpp, 0, sizeof(xpp));
4699                memset(&xecfg, 0, sizeof(xecfg));
4700                if (p->status == 0)
4701                        return error("internal diff status error");
4702                if (p->status == DIFF_STATUS_UNKNOWN)
4703                        continue;
4704                if (diff_unmodified_pair(p))
4705                        continue;
4706                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4707                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4708                        continue;
4709                if (DIFF_PAIR_UNMERGED(p))
4710                        continue;
4711
4712                diff_fill_oid_info(p->one);
4713                diff_fill_oid_info(p->two);
4714
4715                len1 = remove_space(p->one->path, strlen(p->one->path));
4716                len2 = remove_space(p->two->path, strlen(p->two->path));
4717                patch_id_add_string(&ctx, "diff--git");
4718                patch_id_add_string(&ctx, "a/");
4719                git_SHA1_Update(&ctx, p->one->path, len1);
4720                patch_id_add_string(&ctx, "b/");
4721                git_SHA1_Update(&ctx, p->two->path, len2);
4722
4723                if (p->one->mode == 0) {
4724                        patch_id_add_string(&ctx, "newfilemode");
4725                        patch_id_add_mode(&ctx, p->two->mode);
4726                        patch_id_add_string(&ctx, "---/dev/null");
4727                        patch_id_add_string(&ctx, "+++b/");
4728                        git_SHA1_Update(&ctx, p->two->path, len2);
4729                } else if (p->two->mode == 0) {
4730                        patch_id_add_string(&ctx, "deletedfilemode");
4731                        patch_id_add_mode(&ctx, p->one->mode);
4732                        patch_id_add_string(&ctx, "---a/");
4733                        git_SHA1_Update(&ctx, p->one->path, len1);
4734                        patch_id_add_string(&ctx, "+++/dev/null");
4735                } else {
4736                        patch_id_add_string(&ctx, "---a/");
4737                        git_SHA1_Update(&ctx, p->one->path, len1);
4738                        patch_id_add_string(&ctx, "+++b/");
4739                        git_SHA1_Update(&ctx, p->two->path, len2);
4740                }
4741
4742                if (diff_header_only)
4743                        continue;
4744
4745                if (fill_mmfile(&mf1, p->one) < 0 ||
4746                    fill_mmfile(&mf2, p->two) < 0)
4747                        return error("unable to read files to diff");
4748
4749                if (diff_filespec_is_binary(p->one) ||
4750                    diff_filespec_is_binary(p->two)) {
4751                        git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4752                                        GIT_SHA1_HEXSZ);
4753                        git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4754                                        GIT_SHA1_HEXSZ);
4755                        continue;
4756                }
4757
4758                xpp.flags = 0;
4759                xecfg.ctxlen = 3;
4760                xecfg.flags = 0;
4761                if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4762                                  &xpp, &xecfg))
4763                        return error("unable to generate patch-id diff for %s",
4764                                     p->one->path);
4765        }
4766
4767        git_SHA1_Final(oid->hash, &ctx);
4768        return 0;
4769}
4770
4771int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4772{
4773        struct diff_queue_struct *q = &diff_queued_diff;
4774        int i;
4775        int result = diff_get_patch_id(options, oid, diff_header_only);
4776
4777        for (i = 0; i < q->nr; i++)
4778                diff_free_filepair(q->queue[i]);
4779
4780        free(q->queue);
4781        DIFF_QUEUE_CLEAR(q);
4782
4783        return result;
4784}
4785
4786static int is_summary_empty(const struct diff_queue_struct *q)
4787{
4788        int i;
4789
4790        for (i = 0; i < q->nr; i++) {
4791                const struct diff_filepair *p = q->queue[i];
4792
4793                switch (p->status) {
4794                case DIFF_STATUS_DELETED:
4795                case DIFF_STATUS_ADDED:
4796                case DIFF_STATUS_COPIED:
4797                case DIFF_STATUS_RENAMED:
4798                        return 0;
4799                default:
4800                        if (p->score)
4801                                return 0;
4802                        if (p->one->mode && p->two->mode &&
4803                            p->one->mode != p->two->mode)
4804                                return 0;
4805                        break;
4806                }
4807        }
4808        return 1;
4809}
4810
4811static const char rename_limit_warning[] =
4812N_("inexact rename detection was skipped due to too many files.");
4813
4814static const char degrade_cc_to_c_warning[] =
4815N_("only found copies from modified paths due to too many files.");
4816
4817static const char rename_limit_advice[] =
4818N_("you may want to set your %s variable to at least "
4819   "%d and retry the command.");
4820
4821void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4822{
4823        if (degraded_cc)
4824                warning(_(degrade_cc_to_c_warning));
4825        else if (needed)
4826                warning(_(rename_limit_warning));
4827        else
4828                return;
4829        if (0 < needed && needed < 32767)
4830                warning(_(rename_limit_advice), varname, needed);
4831}
4832
4833static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4834{
4835        int i;
4836        struct diff_queue_struct *q = &diff_queued_diff;
4837
4838        if (WSEH_NEW & WS_RULE_MASK)
4839                die("BUG: WS rules bit mask overlaps with diff symbol flags");
4840
4841        for (i = 0; i < q->nr; i++) {
4842                struct diff_filepair *p = q->queue[i];
4843                if (check_pair_status(p))
4844                        diff_flush_patch(p, o);
4845        }
4846}
4847
4848void diff_flush(struct diff_options *options)
4849{
4850        struct diff_queue_struct *q = &diff_queued_diff;
4851        int i, output_format = options->output_format;
4852        int separator = 0;
4853        int dirstat_by_line = 0;
4854
4855        /*
4856         * Order: raw, stat, summary, patch
4857         * or:    name/name-status/checkdiff (other bits clear)
4858         */
4859        if (!q->nr)
4860                goto free_queue;
4861
4862        if (output_format & (DIFF_FORMAT_RAW |
4863                             DIFF_FORMAT_NAME |
4864                             DIFF_FORMAT_NAME_STATUS |
4865                             DIFF_FORMAT_CHECKDIFF)) {
4866                for (i = 0; i < q->nr; i++) {
4867                        struct diff_filepair *p = q->queue[i];
4868                        if (check_pair_status(p))
4869                                flush_one_pair(p, options);
4870                }
4871                separator++;
4872        }
4873
4874        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4875                dirstat_by_line = 1;
4876
4877        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4878            dirstat_by_line) {
4879                struct diffstat_t diffstat;
4880
4881                memset(&diffstat, 0, sizeof(struct diffstat_t));
4882                for (i = 0; i < q->nr; i++) {
4883                        struct diff_filepair *p = q->queue[i];
4884                        if (check_pair_status(p))
4885                                diff_flush_stat(p, options, &diffstat);
4886                }
4887                if (output_format & DIFF_FORMAT_NUMSTAT)
4888                        show_numstat(&diffstat, options);
4889                if (output_format & DIFF_FORMAT_DIFFSTAT)
4890                        show_stats(&diffstat, options);
4891                if (output_format & DIFF_FORMAT_SHORTSTAT)
4892                        show_shortstats(&diffstat, options);
4893                if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4894                        show_dirstat_by_line(&diffstat, options);
4895                free_diffstat_info(&diffstat);
4896                separator++;
4897        }
4898        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4899                show_dirstat(options);
4900
4901        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4902                for (i = 0; i < q->nr; i++) {
4903                        diff_summary(options, q->queue[i]);
4904                }
4905                separator++;
4906        }
4907
4908        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4909            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4910            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4911                /*
4912                 * run diff_flush_patch for the exit status. setting
4913                 * options->file to /dev/null should be safe, because we
4914                 * aren't supposed to produce any output anyway.
4915                 */
4916                if (options->close_file)
4917                        fclose(options->file);
4918                options->file = xfopen("/dev/null", "w");
4919                options->close_file = 1;
4920                for (i = 0; i < q->nr; i++) {
4921                        struct diff_filepair *p = q->queue[i];
4922                        if (check_pair_status(p))
4923                                diff_flush_patch(p, options);
4924                        if (options->found_changes)
4925                                break;
4926                }
4927        }
4928
4929        if (output_format & DIFF_FORMAT_PATCH) {
4930                if (separator) {
4931                        emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0, 0);
4932                        if (options->stat_sep) {
4933                                /* attach patch instead of inline */
4934                                fputs(options->stat_sep, options->file);
4935                        }
4936                }
4937
4938                diff_flush_patch_all_file_pairs(options);
4939        }
4940
4941        if (output_format & DIFF_FORMAT_CALLBACK)
4942                options->format_callback(q, options, options->format_callback_data);
4943
4944        for (i = 0; i < q->nr; i++)
4945                diff_free_filepair(q->queue[i]);
4946free_queue:
4947        free(q->queue);
4948        DIFF_QUEUE_CLEAR(q);
4949        if (options->close_file)
4950                fclose(options->file);
4951
4952        /*
4953         * Report the content-level differences with HAS_CHANGES;
4954         * diff_addremove/diff_change does not set the bit when
4955         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4956         */
4957        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4958                if (options->found_changes)
4959                        DIFF_OPT_SET(options, HAS_CHANGES);
4960                else
4961                        DIFF_OPT_CLR(options, HAS_CHANGES);
4962        }
4963}
4964
4965static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4966{
4967        return (((p->status == DIFF_STATUS_MODIFIED) &&
4968                 ((p->score &&
4969                   filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4970                  (!p->score &&
4971                   filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4972                ((p->status != DIFF_STATUS_MODIFIED) &&
4973                 filter_bit_tst(p->status, options)));
4974}
4975
4976static void diffcore_apply_filter(struct diff_options *options)
4977{
4978        int i;
4979        struct diff_queue_struct *q = &diff_queued_diff;
4980        struct diff_queue_struct outq;
4981
4982        DIFF_QUEUE_CLEAR(&outq);
4983
4984        if (!options->filter)
4985                return;
4986
4987        if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4988                int found;
4989                for (i = found = 0; !found && i < q->nr; i++) {
4990                        if (match_filter(options, q->queue[i]))
4991                                found++;
4992                }
4993                if (found)
4994                        return;
4995
4996                /* otherwise we will clear the whole queue
4997                 * by copying the empty outq at the end of this
4998                 * function, but first clear the current entries
4999                 * in the queue.
5000                 */
5001                for (i = 0; i < q->nr; i++)
5002                        diff_free_filepair(q->queue[i]);
5003        }
5004        else {
5005                /* Only the matching ones */
5006                for (i = 0; i < q->nr; i++) {
5007                        struct diff_filepair *p = q->queue[i];
5008                        if (match_filter(options, p))
5009                                diff_q(&outq, p);
5010                        else
5011                                diff_free_filepair(p);
5012                }
5013        }
5014        free(q->queue);
5015        *q = outq;
5016}
5017
5018/* Check whether two filespecs with the same mode and size are identical */
5019static int diff_filespec_is_identical(struct diff_filespec *one,
5020                                      struct diff_filespec *two)
5021{
5022        if (S_ISGITLINK(one->mode))
5023                return 0;
5024        if (diff_populate_filespec(one, 0))
5025                return 0;
5026        if (diff_populate_filespec(two, 0))
5027                return 0;
5028        return !memcmp(one->data, two->data, one->size);
5029}
5030
5031static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
5032{
5033        if (p->done_skip_stat_unmatch)
5034                return p->skip_stat_unmatch_result;
5035
5036        p->done_skip_stat_unmatch = 1;
5037        p->skip_stat_unmatch_result = 0;
5038        /*
5039         * 1. Entries that come from stat info dirtiness
5040         *    always have both sides (iow, not create/delete),
5041         *    one side of the object name is unknown, with
5042         *    the same mode and size.  Keep the ones that
5043         *    do not match these criteria.  They have real
5044         *    differences.
5045         *
5046         * 2. At this point, the file is known to be modified,
5047         *    with the same mode and size, and the object
5048         *    name of one side is unknown.  Need to inspect
5049         *    the identical contents.
5050         */
5051        if (!DIFF_FILE_VALID(p->one) || /* (1) */
5052            !DIFF_FILE_VALID(p->two) ||
5053            (p->one->oid_valid && p->two->oid_valid) ||
5054            (p->one->mode != p->two->mode) ||
5055            diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
5056            diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
5057            (p->one->size != p->two->size) ||
5058            !diff_filespec_is_identical(p->one, p->two)) /* (2) */
5059                p->skip_stat_unmatch_result = 1;
5060        return p->skip_stat_unmatch_result;
5061}
5062
5063static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
5064{
5065        int i;
5066        struct diff_queue_struct *q = &diff_queued_diff;
5067        struct diff_queue_struct outq;
5068        DIFF_QUEUE_CLEAR(&outq);
5069
5070        for (i = 0; i < q->nr; i++) {
5071                struct diff_filepair *p = q->queue[i];
5072
5073                if (diff_filespec_check_stat_unmatch(p))
5074                        diff_q(&outq, p);
5075                else {
5076                        /*
5077                         * The caller can subtract 1 from skip_stat_unmatch
5078                         * to determine how many paths were dirty only
5079                         * due to stat info mismatch.
5080                         */
5081                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5082                                diffopt->skip_stat_unmatch++;
5083                        diff_free_filepair(p);
5084                }
5085        }
5086        free(q->queue);
5087        *q = outq;
5088}
5089
5090static int diffnamecmp(const void *a_, const void *b_)
5091{
5092        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5093        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5094        const char *name_a, *name_b;
5095
5096        name_a = a->one ? a->one->path : a->two->path;
5097        name_b = b->one ? b->one->path : b->two->path;
5098        return strcmp(name_a, name_b);
5099}
5100
5101void diffcore_fix_diff_index(struct diff_options *options)
5102{
5103        struct diff_queue_struct *q = &diff_queued_diff;
5104        QSORT(q->queue, q->nr, diffnamecmp);
5105}
5106
5107void diffcore_std(struct diff_options *options)
5108{
5109        /* NOTE please keep the following in sync with diff_tree_combined() */
5110        if (options->skip_stat_unmatch)
5111                diffcore_skip_stat_unmatch(options);
5112        if (!options->found_follow) {
5113                /* See try_to_follow_renames() in tree-diff.c */
5114                if (options->break_opt != -1)
5115                        diffcore_break(options->break_opt);
5116                if (options->detect_rename)
5117                        diffcore_rename(options);
5118                if (options->break_opt != -1)
5119                        diffcore_merge_broken();
5120        }
5121        if (options->pickaxe)
5122                diffcore_pickaxe(options);
5123        if (options->orderfile)
5124                diffcore_order(options->orderfile);
5125        if (!options->found_follow)
5126                /* See try_to_follow_renames() in tree-diff.c */
5127                diff_resolve_rename_copy();
5128        diffcore_apply_filter(options);
5129
5130        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5131                DIFF_OPT_SET(options, HAS_CHANGES);
5132        else
5133                DIFF_OPT_CLR(options, HAS_CHANGES);
5134
5135        options->found_follow = 0;
5136}
5137
5138int diff_result_code(struct diff_options *opt, int status)
5139{
5140        int result = 0;
5141
5142        diff_warn_rename_limit("diff.renameLimit",
5143                               opt->needed_rename_limit,
5144                               opt->degraded_cc_to_c);
5145        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5146            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5147                return status;
5148        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5149            DIFF_OPT_TST(opt, HAS_CHANGES))
5150                result |= 01;
5151        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5152            DIFF_OPT_TST(opt, CHECK_FAILED))
5153                result |= 02;
5154        return result;
5155}
5156
5157int diff_can_quit_early(struct diff_options *opt)
5158{
5159        return (DIFF_OPT_TST(opt, QUICK) &&
5160                !opt->filter &&
5161                DIFF_OPT_TST(opt, HAS_CHANGES));
5162}
5163
5164/*
5165 * Shall changes to this submodule be ignored?
5166 *
5167 * Submodule changes can be configured to be ignored separately for each path,
5168 * but that configuration can be overridden from the command line.
5169 */
5170static int is_submodule_ignored(const char *path, struct diff_options *options)
5171{
5172        int ignored = 0;
5173        unsigned orig_flags = options->flags;
5174        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5175                set_diffopt_flags_from_submodule_config(options, path);
5176        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5177                ignored = 1;
5178        options->flags = orig_flags;
5179        return ignored;
5180}
5181
5182void diff_addremove(struct diff_options *options,
5183                    int addremove, unsigned mode,
5184                    const struct object_id *oid,
5185                    int oid_valid,
5186                    const char *concatpath, unsigned dirty_submodule)
5187{
5188        struct diff_filespec *one, *two;
5189
5190        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5191                return;
5192
5193        /* This may look odd, but it is a preparation for
5194         * feeding "there are unchanged files which should
5195         * not produce diffs, but when you are doing copy
5196         * detection you would need them, so here they are"
5197         * entries to the diff-core.  They will be prefixed
5198         * with something like '=' or '*' (I haven't decided
5199         * which but should not make any difference).
5200         * Feeding the same new and old to diff_change()
5201         * also has the same effect.
5202         * Before the final output happens, they are pruned after
5203         * merged into rename/copy pairs as appropriate.
5204         */
5205        if (DIFF_OPT_TST(options, REVERSE_DIFF))
5206                addremove = (addremove == '+' ? '-' :
5207                             addremove == '-' ? '+' : addremove);
5208
5209        if (options->prefix &&
5210            strncmp(concatpath, options->prefix, options->prefix_length))
5211                return;
5212
5213        one = alloc_filespec(concatpath);
5214        two = alloc_filespec(concatpath);
5215
5216        if (addremove != '+')
5217                fill_filespec(one, oid, oid_valid, mode);
5218        if (addremove != '-') {
5219                fill_filespec(two, oid, oid_valid, mode);
5220                two->dirty_submodule = dirty_submodule;
5221        }
5222
5223        diff_queue(&diff_queued_diff, one, two);
5224        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5225                DIFF_OPT_SET(options, HAS_CHANGES);
5226}
5227
5228void diff_change(struct diff_options *options,
5229                 unsigned old_mode, unsigned new_mode,
5230                 const struct object_id *old_oid,
5231                 const struct object_id *new_oid,
5232                 int old_oid_valid, int new_oid_valid,
5233                 const char *concatpath,
5234                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5235{
5236        struct diff_filespec *one, *two;
5237        struct diff_filepair *p;
5238
5239        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5240            is_submodule_ignored(concatpath, options))
5241                return;
5242
5243        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5244                SWAP(old_mode, new_mode);
5245                SWAP(old_oid, new_oid);
5246                SWAP(old_oid_valid, new_oid_valid);
5247                SWAP(old_dirty_submodule, new_dirty_submodule);
5248        }
5249
5250        if (options->prefix &&
5251            strncmp(concatpath, options->prefix, options->prefix_length))
5252                return;
5253
5254        one = alloc_filespec(concatpath);
5255        two = alloc_filespec(concatpath);
5256        fill_filespec(one, old_oid, old_oid_valid, old_mode);
5257        fill_filespec(two, new_oid, new_oid_valid, new_mode);
5258        one->dirty_submodule = old_dirty_submodule;
5259        two->dirty_submodule = new_dirty_submodule;
5260        p = diff_queue(&diff_queued_diff, one, two);
5261
5262        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5263                return;
5264
5265        if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5266            !diff_filespec_check_stat_unmatch(p))
5267                return;
5268
5269        DIFF_OPT_SET(options, HAS_CHANGES);
5270}
5271
5272struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5273{
5274        struct diff_filepair *pair;
5275        struct diff_filespec *one, *two;
5276
5277        if (options->prefix &&
5278            strncmp(path, options->prefix, options->prefix_length))
5279                return NULL;
5280
5281        one = alloc_filespec(path);
5282        two = alloc_filespec(path);
5283        pair = diff_queue(&diff_queued_diff, one, two);
5284        pair->is_unmerged = 1;
5285        return pair;
5286}
5287
5288static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5289                size_t *outsize)
5290{
5291        struct diff_tempfile *temp;
5292        const char *argv[3];
5293        const char **arg = argv;
5294        struct child_process child = CHILD_PROCESS_INIT;
5295        struct strbuf buf = STRBUF_INIT;
5296        int err = 0;
5297
5298        temp = prepare_temp_file(spec->path, spec);
5299        *arg++ = pgm;
5300        *arg++ = temp->name;
5301        *arg = NULL;
5302
5303        child.use_shell = 1;
5304        child.argv = argv;
5305        child.out = -1;
5306        if (start_command(&child)) {
5307                remove_tempfile();
5308                return NULL;
5309        }
5310
5311        if (strbuf_read(&buf, child.out, 0) < 0)
5312                err = error("error reading from textconv command '%s'", pgm);
5313        close(child.out);
5314
5315        if (finish_command(&child) || err) {
5316                strbuf_release(&buf);
5317                remove_tempfile();
5318                return NULL;
5319        }
5320        remove_tempfile();
5321
5322        return strbuf_detach(&buf, outsize);
5323}
5324
5325size_t fill_textconv(struct userdiff_driver *driver,
5326                     struct diff_filespec *df,
5327                     char **outbuf)
5328{
5329        size_t size;
5330
5331        if (!driver) {
5332                if (!DIFF_FILE_VALID(df)) {
5333                        *outbuf = "";
5334                        return 0;
5335                }
5336                if (diff_populate_filespec(df, 0))
5337                        die("unable to read files to diff");
5338                *outbuf = df->data;
5339                return df->size;
5340        }
5341
5342        if (!driver->textconv)
5343                die("BUG: fill_textconv called with non-textconv driver");
5344
5345        if (driver->textconv_cache && df->oid_valid) {
5346                *outbuf = notes_cache_get(driver->textconv_cache,
5347                                          &df->oid,
5348                                          &size);
5349                if (*outbuf)
5350                        return size;
5351        }
5352
5353        *outbuf = run_textconv(driver->textconv, df, &size);
5354        if (!*outbuf)
5355                die("unable to read files to diff");
5356
5357        if (driver->textconv_cache && df->oid_valid) {
5358                /* ignore errors, as we might be in a readonly repository */
5359                notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5360                                size);
5361                /*
5362                 * we could save up changes and flush them all at the end,
5363                 * but we would need an extra call after all diffing is done.
5364                 * Since generating a cache entry is the slow path anyway,
5365                 * this extra overhead probably isn't a big deal.
5366                 */
5367                notes_cache_write(driver->textconv_cache);
5368        }
5369
5370        return size;
5371}
5372
5373int textconv_object(const char *path,
5374                    unsigned mode,
5375                    const struct object_id *oid,
5376                    int oid_valid,
5377                    char **buf,
5378                    unsigned long *buf_size)
5379{
5380        struct diff_filespec *df;
5381        struct userdiff_driver *textconv;
5382
5383        df = alloc_filespec(path);
5384        fill_filespec(df, oid, oid_valid, mode);
5385        textconv = get_textconv(df);
5386        if (!textconv) {
5387                free_filespec(df);
5388                return 0;
5389        }
5390
5391        *buf_size = fill_textconv(textconv, df, buf);
5392        free_filespec(df);
5393        return 1;
5394}
5395
5396void setup_diff_pager(struct diff_options *opt)
5397{
5398        /*
5399         * If the user asked for our exit code, then either they want --quiet
5400         * or --exit-code. We should definitely not bother with a pager in the
5401         * former case, as we will generate no output. Since we still properly
5402         * report our exit code even when a pager is run, we _could_ run a
5403         * pager with --exit-code. But since we have not done so historically,
5404         * and because it is easy to find people oneline advising "git diff
5405         * --exit-code" in hooks and other scripts, we do not do so.
5406         */
5407        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5408            check_pager_config("diff") != 0)
5409                setup_pager();
5410}