534c12e28ea8b077c9fa305a634f7a73cf30589d
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "tempfile.h"
   6#include "quote.h"
   7#include "diff.h"
   8#include "diffcore.h"
   9#include "delta.h"
  10#include "xdiff-interface.h"
  11#include "color.h"
  12#include "attr.h"
  13#include "run-command.h"
  14#include "utf8.h"
  15#include "userdiff.h"
  16#include "submodule-config.h"
  17#include "submodule.h"
  18#include "ll-merge.h"
  19#include "string-list.h"
  20#include "argv-array.h"
  21
  22#ifdef NO_FAST_WORKING_DIRECTORY
  23#define FAST_WORKING_DIRECTORY 0
  24#else
  25#define FAST_WORKING_DIRECTORY 1
  26#endif
  27
  28static int diff_detect_rename_default;
  29static int diff_compaction_heuristic; /* experimental */
  30static int diff_rename_limit_default = 400;
  31static int diff_suppress_blank_empty;
  32static int diff_use_color_default = -1;
  33static int diff_context_default = 3;
  34static const char *diff_word_regex_cfg;
  35static const char *external_diff_cmd_cfg;
  36static const char *diff_order_file_cfg;
  37int diff_auto_refresh_index = 1;
  38static int diff_mnemonic_prefix;
  39static int diff_no_prefix;
  40static int diff_stat_graph_width;
  41static int diff_dirstat_permille_default = 30;
  42static struct diff_options default_diff_options;
  43static long diff_algorithm;
  44
  45static char diff_colors[][COLOR_MAXLEN] = {
  46        GIT_COLOR_RESET,
  47        GIT_COLOR_NORMAL,       /* CONTEXT */
  48        GIT_COLOR_BOLD,         /* METAINFO */
  49        GIT_COLOR_CYAN,         /* FRAGINFO */
  50        GIT_COLOR_RED,          /* OLD */
  51        GIT_COLOR_GREEN,        /* NEW */
  52        GIT_COLOR_YELLOW,       /* COMMIT */
  53        GIT_COLOR_BG_RED,       /* WHITESPACE */
  54        GIT_COLOR_NORMAL,       /* FUNCINFO */
  55};
  56
  57static int parse_diff_color_slot(const char *var)
  58{
  59        if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
  60                return DIFF_CONTEXT;
  61        if (!strcasecmp(var, "meta"))
  62                return DIFF_METAINFO;
  63        if (!strcasecmp(var, "frag"))
  64                return DIFF_FRAGINFO;
  65        if (!strcasecmp(var, "old"))
  66                return DIFF_FILE_OLD;
  67        if (!strcasecmp(var, "new"))
  68                return DIFF_FILE_NEW;
  69        if (!strcasecmp(var, "commit"))
  70                return DIFF_COMMIT;
  71        if (!strcasecmp(var, "whitespace"))
  72                return DIFF_WHITESPACE;
  73        if (!strcasecmp(var, "func"))
  74                return DIFF_FUNCINFO;
  75        return -1;
  76}
  77
  78static int parse_dirstat_params(struct diff_options *options, const char *params_string,
  79                                struct strbuf *errmsg)
  80{
  81        char *params_copy = xstrdup(params_string);
  82        struct string_list params = STRING_LIST_INIT_NODUP;
  83        int ret = 0;
  84        int i;
  85
  86        if (*params_copy)
  87                string_list_split_in_place(&params, params_copy, ',', -1);
  88        for (i = 0; i < params.nr; i++) {
  89                const char *p = params.items[i].string;
  90                if (!strcmp(p, "changes")) {
  91                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  92                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  93                } else if (!strcmp(p, "lines")) {
  94                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  95                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  96                } else if (!strcmp(p, "files")) {
  97                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  98                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  99                } else if (!strcmp(p, "noncumulative")) {
 100                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
 101                } else if (!strcmp(p, "cumulative")) {
 102                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
 103                } else if (isdigit(*p)) {
 104                        char *end;
 105                        int permille = strtoul(p, &end, 10) * 10;
 106                        if (*end == '.' && isdigit(*++end)) {
 107                                /* only use first digit */
 108                                permille += *end - '0';
 109                                /* .. and ignore any further digits */
 110                                while (isdigit(*++end))
 111                                        ; /* nothing */
 112                        }
 113                        if (!*end)
 114                                options->dirstat_permille = permille;
 115                        else {
 116                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%s'\n"),
 117                                            p);
 118                                ret++;
 119                        }
 120                } else {
 121                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%s'\n"), p);
 122                        ret++;
 123                }
 124
 125        }
 126        string_list_clear(&params, 0);
 127        free(params_copy);
 128        return ret;
 129}
 130
 131static int parse_submodule_params(struct diff_options *options, const char *value)
 132{
 133        if (!strcmp(value, "log"))
 134                DIFF_OPT_SET(options, SUBMODULE_LOG);
 135        else if (!strcmp(value, "short"))
 136                DIFF_OPT_CLR(options, SUBMODULE_LOG);
 137        else
 138                return -1;
 139        return 0;
 140}
 141
 142static int git_config_rename(const char *var, const char *value)
 143{
 144        if (!value)
 145                return DIFF_DETECT_RENAME;
 146        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 147                return  DIFF_DETECT_COPY;
 148        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 149}
 150
 151long parse_algorithm_value(const char *value)
 152{
 153        if (!value)
 154                return -1;
 155        else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
 156                return 0;
 157        else if (!strcasecmp(value, "minimal"))
 158                return XDF_NEED_MINIMAL;
 159        else if (!strcasecmp(value, "patience"))
 160                return XDF_PATIENCE_DIFF;
 161        else if (!strcasecmp(value, "histogram"))
 162                return XDF_HISTOGRAM_DIFF;
 163        return -1;
 164}
 165
 166/*
 167 * These are to give UI layer defaults.
 168 * The core-level commands such as git-diff-files should
 169 * never be affected by the setting of diff.renames
 170 * the user happens to have in the configuration file.
 171 */
 172void init_diff_ui_defaults(void)
 173{
 174        diff_detect_rename_default = 1;
 175}
 176
 177int git_diff_ui_config(const char *var, const char *value, void *cb)
 178{
 179        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 180                diff_use_color_default = git_config_colorbool(var, value);
 181                return 0;
 182        }
 183        if (!strcmp(var, "diff.context")) {
 184                diff_context_default = git_config_int(var, value);
 185                if (diff_context_default < 0)
 186                        return -1;
 187                return 0;
 188        }
 189        if (!strcmp(var, "diff.renames")) {
 190                diff_detect_rename_default = git_config_rename(var, value);
 191                return 0;
 192        }
 193        if (!strcmp(var, "diff.compactionheuristic")) {
 194                diff_compaction_heuristic = git_config_bool(var, value);
 195                return 0;
 196        }
 197        if (!strcmp(var, "diff.autorefreshindex")) {
 198                diff_auto_refresh_index = git_config_bool(var, value);
 199                return 0;
 200        }
 201        if (!strcmp(var, "diff.mnemonicprefix")) {
 202                diff_mnemonic_prefix = git_config_bool(var, value);
 203                return 0;
 204        }
 205        if (!strcmp(var, "diff.noprefix")) {
 206                diff_no_prefix = git_config_bool(var, value);
 207                return 0;
 208        }
 209        if (!strcmp(var, "diff.statgraphwidth")) {
 210                diff_stat_graph_width = git_config_int(var, value);
 211                return 0;
 212        }
 213        if (!strcmp(var, "diff.external"))
 214                return git_config_string(&external_diff_cmd_cfg, var, value);
 215        if (!strcmp(var, "diff.wordregex"))
 216                return git_config_string(&diff_word_regex_cfg, var, value);
 217        if (!strcmp(var, "diff.orderfile"))
 218                return git_config_pathname(&diff_order_file_cfg, var, value);
 219
 220        if (!strcmp(var, "diff.ignoresubmodules"))
 221                handle_ignore_submodules_arg(&default_diff_options, value);
 222
 223        if (!strcmp(var, "diff.submodule")) {
 224                if (parse_submodule_params(&default_diff_options, value))
 225                        warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
 226                                value);
 227                return 0;
 228        }
 229
 230        if (!strcmp(var, "diff.algorithm")) {
 231                diff_algorithm = parse_algorithm_value(value);
 232                if (diff_algorithm < 0)
 233                        return -1;
 234                return 0;
 235        }
 236
 237        if (git_color_config(var, value, cb) < 0)
 238                return -1;
 239
 240        return git_diff_basic_config(var, value, cb);
 241}
 242
 243int git_diff_basic_config(const char *var, const char *value, void *cb)
 244{
 245        const char *name;
 246
 247        if (!strcmp(var, "diff.renamelimit")) {
 248                diff_rename_limit_default = git_config_int(var, value);
 249                return 0;
 250        }
 251
 252        if (userdiff_config(var, value) < 0)
 253                return -1;
 254
 255        if (skip_prefix(var, "diff.color.", &name) ||
 256            skip_prefix(var, "color.diff.", &name)) {
 257                int slot = parse_diff_color_slot(name);
 258                if (slot < 0)
 259                        return 0;
 260                if (!value)
 261                        return config_error_nonbool(var);
 262                return color_parse(value, diff_colors[slot]);
 263        }
 264
 265        /* like GNU diff's --suppress-blank-empty option  */
 266        if (!strcmp(var, "diff.suppressblankempty") ||
 267                        /* for backwards compatibility */
 268                        !strcmp(var, "diff.suppress-blank-empty")) {
 269                diff_suppress_blank_empty = git_config_bool(var, value);
 270                return 0;
 271        }
 272
 273        if (!strcmp(var, "diff.dirstat")) {
 274                struct strbuf errmsg = STRBUF_INIT;
 275                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 276                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 277                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 278                                errmsg.buf);
 279                strbuf_release(&errmsg);
 280                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 281                return 0;
 282        }
 283
 284        if (starts_with(var, "submodule."))
 285                return parse_submodule_config_option(var, value);
 286
 287        return git_default_config(var, value, cb);
 288}
 289
 290static char *quote_two(const char *one, const char *two)
 291{
 292        int need_one = quote_c_style(one, NULL, NULL, 1);
 293        int need_two = quote_c_style(two, NULL, NULL, 1);
 294        struct strbuf res = STRBUF_INIT;
 295
 296        if (need_one + need_two) {
 297                strbuf_addch(&res, '"');
 298                quote_c_style(one, &res, NULL, 1);
 299                quote_c_style(two, &res, NULL, 1);
 300                strbuf_addch(&res, '"');
 301        } else {
 302                strbuf_addstr(&res, one);
 303                strbuf_addstr(&res, two);
 304        }
 305        return strbuf_detach(&res, NULL);
 306}
 307
 308static const char *external_diff(void)
 309{
 310        static const char *external_diff_cmd = NULL;
 311        static int done_preparing = 0;
 312
 313        if (done_preparing)
 314                return external_diff_cmd;
 315        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 316        if (!external_diff_cmd)
 317                external_diff_cmd = external_diff_cmd_cfg;
 318        done_preparing = 1;
 319        return external_diff_cmd;
 320}
 321
 322/*
 323 * Keep track of files used for diffing. Sometimes such an entry
 324 * refers to a temporary file, sometimes to an existing file, and
 325 * sometimes to "/dev/null".
 326 */
 327static struct diff_tempfile {
 328        /*
 329         * filename external diff should read from, or NULL if this
 330         * entry is currently not in use:
 331         */
 332        const char *name;
 333
 334        char hex[GIT_SHA1_HEXSZ + 1];
 335        char mode[10];
 336
 337        /*
 338         * If this diff_tempfile instance refers to a temporary file,
 339         * this tempfile object is used to manage its lifetime.
 340         */
 341        struct tempfile tempfile;
 342} diff_temp[2];
 343
 344typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 345
 346struct emit_callback {
 347        int color_diff;
 348        unsigned ws_rule;
 349        int blank_at_eof_in_preimage;
 350        int blank_at_eof_in_postimage;
 351        int lno_in_preimage;
 352        int lno_in_postimage;
 353        sane_truncate_fn truncate;
 354        const char **label_path;
 355        struct diff_words_data *diff_words;
 356        struct diff_options *opt;
 357        int *found_changesp;
 358        struct strbuf *header;
 359};
 360
 361static int count_lines(const char *data, int size)
 362{
 363        int count, ch, completely_empty = 1, nl_just_seen = 0;
 364        count = 0;
 365        while (0 < size--) {
 366                ch = *data++;
 367                if (ch == '\n') {
 368                        count++;
 369                        nl_just_seen = 1;
 370                        completely_empty = 0;
 371                }
 372                else {
 373                        nl_just_seen = 0;
 374                        completely_empty = 0;
 375                }
 376        }
 377        if (completely_empty)
 378                return 0;
 379        if (!nl_just_seen)
 380                count++; /* no trailing newline */
 381        return count;
 382}
 383
 384static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 385{
 386        if (!DIFF_FILE_VALID(one)) {
 387                mf->ptr = (char *)""; /* does not matter */
 388                mf->size = 0;
 389                return 0;
 390        }
 391        else if (diff_populate_filespec(one, 0))
 392                return -1;
 393
 394        mf->ptr = one->data;
 395        mf->size = one->size;
 396        return 0;
 397}
 398
 399/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 400static unsigned long diff_filespec_size(struct diff_filespec *one)
 401{
 402        if (!DIFF_FILE_VALID(one))
 403                return 0;
 404        diff_populate_filespec(one, CHECK_SIZE_ONLY);
 405        return one->size;
 406}
 407
 408static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 409{
 410        char *ptr = mf->ptr;
 411        long size = mf->size;
 412        int cnt = 0;
 413
 414        if (!size)
 415                return cnt;
 416        ptr += size - 1; /* pointing at the very end */
 417        if (*ptr != '\n')
 418                ; /* incomplete line */
 419        else
 420                ptr--; /* skip the last LF */
 421        while (mf->ptr < ptr) {
 422                char *prev_eol;
 423                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 424                        if (*prev_eol == '\n')
 425                                break;
 426                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 427                        break;
 428                cnt++;
 429                ptr = prev_eol - 1;
 430        }
 431        return cnt;
 432}
 433
 434static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 435                               struct emit_callback *ecbdata)
 436{
 437        int l1, l2, at;
 438        unsigned ws_rule = ecbdata->ws_rule;
 439        l1 = count_trailing_blank(mf1, ws_rule);
 440        l2 = count_trailing_blank(mf2, ws_rule);
 441        if (l2 <= l1) {
 442                ecbdata->blank_at_eof_in_preimage = 0;
 443                ecbdata->blank_at_eof_in_postimage = 0;
 444                return;
 445        }
 446        at = count_lines(mf1->ptr, mf1->size);
 447        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 448
 449        at = count_lines(mf2->ptr, mf2->size);
 450        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 451}
 452
 453static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 454                        int first, const char *line, int len)
 455{
 456        int has_trailing_newline, has_trailing_carriage_return;
 457        int nofirst;
 458        FILE *file = o->file;
 459
 460        fputs(diff_line_prefix(o), file);
 461
 462        if (len == 0) {
 463                has_trailing_newline = (first == '\n');
 464                has_trailing_carriage_return = (!has_trailing_newline &&
 465                                                (first == '\r'));
 466                nofirst = has_trailing_newline || has_trailing_carriage_return;
 467        } else {
 468                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 469                if (has_trailing_newline)
 470                        len--;
 471                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 472                if (has_trailing_carriage_return)
 473                        len--;
 474                nofirst = 0;
 475        }
 476
 477        if (len || !nofirst) {
 478                fputs(set, file);
 479                if (!nofirst)
 480                        fputc(first, file);
 481                fwrite(line, len, 1, file);
 482                fputs(reset, file);
 483        }
 484        if (has_trailing_carriage_return)
 485                fputc('\r', file);
 486        if (has_trailing_newline)
 487                fputc('\n', file);
 488}
 489
 490static void emit_line(struct diff_options *o, const char *set, const char *reset,
 491                      const char *line, int len)
 492{
 493        emit_line_0(o, set, reset, line[0], line+1, len-1);
 494}
 495
 496static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 497{
 498        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 499              ecbdata->blank_at_eof_in_preimage &&
 500              ecbdata->blank_at_eof_in_postimage &&
 501              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 502              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 503                return 0;
 504        return ws_blank_line(line, len, ecbdata->ws_rule);
 505}
 506
 507static void emit_line_checked(const char *reset,
 508                              struct emit_callback *ecbdata,
 509                              const char *line, int len,
 510                              enum color_diff color,
 511                              unsigned ws_error_highlight,
 512                              char sign)
 513{
 514        const char *set = diff_get_color(ecbdata->color_diff, color);
 515        const char *ws = NULL;
 516
 517        if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
 518                ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 519                if (!*ws)
 520                        ws = NULL;
 521        }
 522
 523        if (!ws)
 524                emit_line_0(ecbdata->opt, set, reset, sign, line, len);
 525        else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
 526                /* Blank line at EOF - paint '+' as well */
 527                emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
 528        else {
 529                /* Emit just the prefix, then the rest. */
 530                emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
 531                ws_check_emit(line, len, ecbdata->ws_rule,
 532                              ecbdata->opt->file, set, reset, ws);
 533        }
 534}
 535
 536static void emit_add_line(const char *reset,
 537                          struct emit_callback *ecbdata,
 538                          const char *line, int len)
 539{
 540        emit_line_checked(reset, ecbdata, line, len,
 541                          DIFF_FILE_NEW, WSEH_NEW, '+');
 542}
 543
 544static void emit_del_line(const char *reset,
 545                          struct emit_callback *ecbdata,
 546                          const char *line, int len)
 547{
 548        emit_line_checked(reset, ecbdata, line, len,
 549                          DIFF_FILE_OLD, WSEH_OLD, '-');
 550}
 551
 552static void emit_context_line(const char *reset,
 553                              struct emit_callback *ecbdata,
 554                              const char *line, int len)
 555{
 556        emit_line_checked(reset, ecbdata, line, len,
 557                          DIFF_CONTEXT, WSEH_CONTEXT, ' ');
 558}
 559
 560static void emit_hunk_header(struct emit_callback *ecbdata,
 561                             const char *line, int len)
 562{
 563        const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
 564        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 565        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 566        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 567        static const char atat[2] = { '@', '@' };
 568        const char *cp, *ep;
 569        struct strbuf msgbuf = STRBUF_INIT;
 570        int org_len = len;
 571        int i = 1;
 572
 573        /*
 574         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 575         * it always is at least 10 bytes long.
 576         */
 577        if (len < 10 ||
 578            memcmp(line, atat, 2) ||
 579            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 580                emit_line(ecbdata->opt, context, reset, line, len);
 581                return;
 582        }
 583        ep += 2; /* skip over @@ */
 584
 585        /* The hunk header in fraginfo color */
 586        strbuf_addstr(&msgbuf, frag);
 587        strbuf_add(&msgbuf, line, ep - line);
 588        strbuf_addstr(&msgbuf, reset);
 589
 590        /*
 591         * trailing "\r\n"
 592         */
 593        for ( ; i < 3; i++)
 594                if (line[len - i] == '\r' || line[len - i] == '\n')
 595                        len--;
 596
 597        /* blank before the func header */
 598        for (cp = ep; ep - line < len; ep++)
 599                if (*ep != ' ' && *ep != '\t')
 600                        break;
 601        if (ep != cp) {
 602                strbuf_addstr(&msgbuf, context);
 603                strbuf_add(&msgbuf, cp, ep - cp);
 604                strbuf_addstr(&msgbuf, reset);
 605        }
 606
 607        if (ep < line + len) {
 608                strbuf_addstr(&msgbuf, func);
 609                strbuf_add(&msgbuf, ep, line + len - ep);
 610                strbuf_addstr(&msgbuf, reset);
 611        }
 612
 613        strbuf_add(&msgbuf, line + len, org_len - len);
 614        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 615        strbuf_release(&msgbuf);
 616}
 617
 618static struct diff_tempfile *claim_diff_tempfile(void) {
 619        int i;
 620        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 621                if (!diff_temp[i].name)
 622                        return diff_temp + i;
 623        die("BUG: diff is failing to clean up its tempfiles");
 624}
 625
 626static void remove_tempfile(void)
 627{
 628        int i;
 629        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 630                if (is_tempfile_active(&diff_temp[i].tempfile))
 631                        delete_tempfile(&diff_temp[i].tempfile);
 632                diff_temp[i].name = NULL;
 633        }
 634}
 635
 636static void print_line_count(FILE *file, int count)
 637{
 638        switch (count) {
 639        case 0:
 640                fprintf(file, "0,0");
 641                break;
 642        case 1:
 643                fprintf(file, "1");
 644                break;
 645        default:
 646                fprintf(file, "1,%d", count);
 647                break;
 648        }
 649}
 650
 651static void emit_rewrite_lines(struct emit_callback *ecb,
 652                               int prefix, const char *data, int size)
 653{
 654        const char *endp = NULL;
 655        static const char *nneof = " No newline at end of file\n";
 656        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 657
 658        while (0 < size) {
 659                int len;
 660
 661                endp = memchr(data, '\n', size);
 662                len = endp ? (endp - data + 1) : size;
 663                if (prefix != '+') {
 664                        ecb->lno_in_preimage++;
 665                        emit_del_line(reset, ecb, data, len);
 666                } else {
 667                        ecb->lno_in_postimage++;
 668                        emit_add_line(reset, ecb, data, len);
 669                }
 670                size -= len;
 671                data += len;
 672        }
 673        if (!endp) {
 674                const char *context = diff_get_color(ecb->color_diff,
 675                                                     DIFF_CONTEXT);
 676                putc('\n', ecb->opt->file);
 677                emit_line_0(ecb->opt, context, reset, '\\',
 678                            nneof, strlen(nneof));
 679        }
 680}
 681
 682static void emit_rewrite_diff(const char *name_a,
 683                              const char *name_b,
 684                              struct diff_filespec *one,
 685                              struct diff_filespec *two,
 686                              struct userdiff_driver *textconv_one,
 687                              struct userdiff_driver *textconv_two,
 688                              struct diff_options *o)
 689{
 690        int lc_a, lc_b;
 691        const char *name_a_tab, *name_b_tab;
 692        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 693        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 694        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 695        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 696        const char *a_prefix, *b_prefix;
 697        char *data_one, *data_two;
 698        size_t size_one, size_two;
 699        struct emit_callback ecbdata;
 700        const char *line_prefix = diff_line_prefix(o);
 701
 702        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 703                a_prefix = o->b_prefix;
 704                b_prefix = o->a_prefix;
 705        } else {
 706                a_prefix = o->a_prefix;
 707                b_prefix = o->b_prefix;
 708        }
 709
 710        name_a += (*name_a == '/');
 711        name_b += (*name_b == '/');
 712        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 713        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 714
 715        strbuf_reset(&a_name);
 716        strbuf_reset(&b_name);
 717        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 718        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 719
 720        size_one = fill_textconv(textconv_one, one, &data_one);
 721        size_two = fill_textconv(textconv_two, two, &data_two);
 722
 723        memset(&ecbdata, 0, sizeof(ecbdata));
 724        ecbdata.color_diff = want_color(o->use_color);
 725        ecbdata.found_changesp = &o->found_changes;
 726        ecbdata.ws_rule = whitespace_rule(name_b);
 727        ecbdata.opt = o;
 728        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 729                mmfile_t mf1, mf2;
 730                mf1.ptr = (char *)data_one;
 731                mf2.ptr = (char *)data_two;
 732                mf1.size = size_one;
 733                mf2.size = size_two;
 734                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 735        }
 736        ecbdata.lno_in_preimage = 1;
 737        ecbdata.lno_in_postimage = 1;
 738
 739        lc_a = count_lines(data_one, size_one);
 740        lc_b = count_lines(data_two, size_two);
 741        fprintf(o->file,
 742                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 743                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 744                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 745                line_prefix, fraginfo);
 746        if (!o->irreversible_delete)
 747                print_line_count(o->file, lc_a);
 748        else
 749                fprintf(o->file, "?,?");
 750        fprintf(o->file, " +");
 751        print_line_count(o->file, lc_b);
 752        fprintf(o->file, " @@%s\n", reset);
 753        if (lc_a && !o->irreversible_delete)
 754                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 755        if (lc_b)
 756                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 757        if (textconv_one)
 758                free((char *)data_one);
 759        if (textconv_two)
 760                free((char *)data_two);
 761}
 762
 763struct diff_words_buffer {
 764        mmfile_t text;
 765        long alloc;
 766        struct diff_words_orig {
 767                const char *begin, *end;
 768        } *orig;
 769        int orig_nr, orig_alloc;
 770};
 771
 772static void diff_words_append(char *line, unsigned long len,
 773                struct diff_words_buffer *buffer)
 774{
 775        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 776        line++;
 777        len--;
 778        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 779        buffer->text.size += len;
 780        buffer->text.ptr[buffer->text.size] = '\0';
 781}
 782
 783struct diff_words_style_elem {
 784        const char *prefix;
 785        const char *suffix;
 786        const char *color; /* NULL; filled in by the setup code if
 787                            * color is enabled */
 788};
 789
 790struct diff_words_style {
 791        enum diff_words_type type;
 792        struct diff_words_style_elem new, old, ctx;
 793        const char *newline;
 794};
 795
 796static struct diff_words_style diff_words_styles[] = {
 797        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 798        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 799        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 800};
 801
 802struct diff_words_data {
 803        struct diff_words_buffer minus, plus;
 804        const char *current_plus;
 805        int last_minus;
 806        struct diff_options *opt;
 807        regex_t *word_regex;
 808        enum diff_words_type type;
 809        struct diff_words_style *style;
 810};
 811
 812static int fn_out_diff_words_write_helper(FILE *fp,
 813                                          struct diff_words_style_elem *st_el,
 814                                          const char *newline,
 815                                          size_t count, const char *buf,
 816                                          const char *line_prefix)
 817{
 818        int print = 0;
 819
 820        while (count) {
 821                char *p = memchr(buf, '\n', count);
 822                if (print)
 823                        fputs(line_prefix, fp);
 824                if (p != buf) {
 825                        if (st_el->color && fputs(st_el->color, fp) < 0)
 826                                return -1;
 827                        if (fputs(st_el->prefix, fp) < 0 ||
 828                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 829                            fputs(st_el->suffix, fp) < 0)
 830                                return -1;
 831                        if (st_el->color && *st_el->color
 832                            && fputs(GIT_COLOR_RESET, fp) < 0)
 833                                return -1;
 834                }
 835                if (!p)
 836                        return 0;
 837                if (fputs(newline, fp) < 0)
 838                        return -1;
 839                count -= p + 1 - buf;
 840                buf = p + 1;
 841                print = 1;
 842        }
 843        return 0;
 844}
 845
 846/*
 847 * '--color-words' algorithm can be described as:
 848 *
 849 *   1. collect a the minus/plus lines of a diff hunk, divided into
 850 *      minus-lines and plus-lines;
 851 *
 852 *   2. break both minus-lines and plus-lines into words and
 853 *      place them into two mmfile_t with one word for each line;
 854 *
 855 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 856 *
 857 * And for the common parts of the both file, we output the plus side text.
 858 * diff_words->current_plus is used to trace the current position of the plus file
 859 * which printed. diff_words->last_minus is used to trace the last minus word
 860 * printed.
 861 *
 862 * For '--graph' to work with '--color-words', we need to output the graph prefix
 863 * on each line of color words output. Generally, there are two conditions on
 864 * which we should output the prefix.
 865 *
 866 *   1. diff_words->last_minus == 0 &&
 867 *      diff_words->current_plus == diff_words->plus.text.ptr
 868 *
 869 *      that is: the plus text must start as a new line, and if there is no minus
 870 *      word printed, a graph prefix must be printed.
 871 *
 872 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 873 *      *(diff_words->current_plus - 1) == '\n'
 874 *
 875 *      that is: a graph prefix must be printed following a '\n'
 876 */
 877static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 878{
 879        if ((diff_words->last_minus == 0 &&
 880                diff_words->current_plus == diff_words->plus.text.ptr) ||
 881                (diff_words->current_plus > diff_words->plus.text.ptr &&
 882                *(diff_words->current_plus - 1) == '\n')) {
 883                return 1;
 884        } else {
 885                return 0;
 886        }
 887}
 888
 889static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 890{
 891        struct diff_words_data *diff_words = priv;
 892        struct diff_words_style *style = diff_words->style;
 893        int minus_first, minus_len, plus_first, plus_len;
 894        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 895        struct diff_options *opt = diff_words->opt;
 896        const char *line_prefix;
 897
 898        if (line[0] != '@' || parse_hunk_header(line, len,
 899                        &minus_first, &minus_len, &plus_first, &plus_len))
 900                return;
 901
 902        assert(opt);
 903        line_prefix = diff_line_prefix(opt);
 904
 905        /* POSIX requires that first be decremented by one if len == 0... */
 906        if (minus_len) {
 907                minus_begin = diff_words->minus.orig[minus_first].begin;
 908                minus_end =
 909                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 910        } else
 911                minus_begin = minus_end =
 912                        diff_words->minus.orig[minus_first].end;
 913
 914        if (plus_len) {
 915                plus_begin = diff_words->plus.orig[plus_first].begin;
 916                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 917        } else
 918                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 919
 920        if (color_words_output_graph_prefix(diff_words)) {
 921                fputs(line_prefix, diff_words->opt->file);
 922        }
 923        if (diff_words->current_plus != plus_begin) {
 924                fn_out_diff_words_write_helper(diff_words->opt->file,
 925                                &style->ctx, style->newline,
 926                                plus_begin - diff_words->current_plus,
 927                                diff_words->current_plus, line_prefix);
 928                if (*(plus_begin - 1) == '\n')
 929                        fputs(line_prefix, diff_words->opt->file);
 930        }
 931        if (minus_begin != minus_end) {
 932                fn_out_diff_words_write_helper(diff_words->opt->file,
 933                                &style->old, style->newline,
 934                                minus_end - minus_begin, minus_begin,
 935                                line_prefix);
 936        }
 937        if (plus_begin != plus_end) {
 938                fn_out_diff_words_write_helper(diff_words->opt->file,
 939                                &style->new, style->newline,
 940                                plus_end - plus_begin, plus_begin,
 941                                line_prefix);
 942        }
 943
 944        diff_words->current_plus = plus_end;
 945        diff_words->last_minus = minus_first;
 946}
 947
 948/* This function starts looking at *begin, and returns 0 iff a word was found. */
 949static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 950                int *begin, int *end)
 951{
 952        if (word_regex && *begin < buffer->size) {
 953                regmatch_t match[1];
 954                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 955                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 956                                        '\n', match[0].rm_eo - match[0].rm_so);
 957                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 958                        *begin += match[0].rm_so;
 959                        return *begin >= *end;
 960                }
 961                return -1;
 962        }
 963
 964        /* find the next word */
 965        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 966                (*begin)++;
 967        if (*begin >= buffer->size)
 968                return -1;
 969
 970        /* find the end of the word */
 971        *end = *begin + 1;
 972        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 973                (*end)++;
 974
 975        return 0;
 976}
 977
 978/*
 979 * This function splits the words in buffer->text, stores the list with
 980 * newline separator into out, and saves the offsets of the original words
 981 * in buffer->orig.
 982 */
 983static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 984                regex_t *word_regex)
 985{
 986        int i, j;
 987        long alloc = 0;
 988
 989        out->size = 0;
 990        out->ptr = NULL;
 991
 992        /* fake an empty "0th" word */
 993        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 994        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 995        buffer->orig_nr = 1;
 996
 997        for (i = 0; i < buffer->text.size; i++) {
 998                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 999                        return;
1000
1001                /* store original boundaries */
1002                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1003                                buffer->orig_alloc);
1004                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1005                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1006                buffer->orig_nr++;
1007
1008                /* store one word */
1009                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1010                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1011                out->ptr[out->size + j - i] = '\n';
1012                out->size += j - i + 1;
1013
1014                i = j - 1;
1015        }
1016}
1017
1018/* this executes the word diff on the accumulated buffers */
1019static void diff_words_show(struct diff_words_data *diff_words)
1020{
1021        xpparam_t xpp;
1022        xdemitconf_t xecfg;
1023        mmfile_t minus, plus;
1024        struct diff_words_style *style = diff_words->style;
1025
1026        struct diff_options *opt = diff_words->opt;
1027        const char *line_prefix;
1028
1029        assert(opt);
1030        line_prefix = diff_line_prefix(opt);
1031
1032        /* special case: only removal */
1033        if (!diff_words->plus.text.size) {
1034                fputs(line_prefix, diff_words->opt->file);
1035                fn_out_diff_words_write_helper(diff_words->opt->file,
1036                        &style->old, style->newline,
1037                        diff_words->minus.text.size,
1038                        diff_words->minus.text.ptr, line_prefix);
1039                diff_words->minus.text.size = 0;
1040                return;
1041        }
1042
1043        diff_words->current_plus = diff_words->plus.text.ptr;
1044        diff_words->last_minus = 0;
1045
1046        memset(&xpp, 0, sizeof(xpp));
1047        memset(&xecfg, 0, sizeof(xecfg));
1048        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1049        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1050        xpp.flags = 0;
1051        /* as only the hunk header will be parsed, we need a 0-context */
1052        xecfg.ctxlen = 0;
1053        if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1054                          &xpp, &xecfg))
1055                die("unable to generate word diff");
1056        free(minus.ptr);
1057        free(plus.ptr);
1058        if (diff_words->current_plus != diff_words->plus.text.ptr +
1059                        diff_words->plus.text.size) {
1060                if (color_words_output_graph_prefix(diff_words))
1061                        fputs(line_prefix, diff_words->opt->file);
1062                fn_out_diff_words_write_helper(diff_words->opt->file,
1063                        &style->ctx, style->newline,
1064                        diff_words->plus.text.ptr + diff_words->plus.text.size
1065                        - diff_words->current_plus, diff_words->current_plus,
1066                        line_prefix);
1067        }
1068        diff_words->minus.text.size = diff_words->plus.text.size = 0;
1069}
1070
1071/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1072static void diff_words_flush(struct emit_callback *ecbdata)
1073{
1074        if (ecbdata->diff_words->minus.text.size ||
1075            ecbdata->diff_words->plus.text.size)
1076                diff_words_show(ecbdata->diff_words);
1077}
1078
1079static void diff_filespec_load_driver(struct diff_filespec *one)
1080{
1081        /* Use already-loaded driver */
1082        if (one->driver)
1083                return;
1084
1085        if (S_ISREG(one->mode))
1086                one->driver = userdiff_find_by_path(one->path);
1087
1088        /* Fallback to default settings */
1089        if (!one->driver)
1090                one->driver = userdiff_find_by_name("default");
1091}
1092
1093static const char *userdiff_word_regex(struct diff_filespec *one)
1094{
1095        diff_filespec_load_driver(one);
1096        return one->driver->word_regex;
1097}
1098
1099static void init_diff_words_data(struct emit_callback *ecbdata,
1100                                 struct diff_options *orig_opts,
1101                                 struct diff_filespec *one,
1102                                 struct diff_filespec *two)
1103{
1104        int i;
1105        struct diff_options *o = xmalloc(sizeof(struct diff_options));
1106        memcpy(o, orig_opts, sizeof(struct diff_options));
1107
1108        ecbdata->diff_words =
1109                xcalloc(1, sizeof(struct diff_words_data));
1110        ecbdata->diff_words->type = o->word_diff;
1111        ecbdata->diff_words->opt = o;
1112        if (!o->word_regex)
1113                o->word_regex = userdiff_word_regex(one);
1114        if (!o->word_regex)
1115                o->word_regex = userdiff_word_regex(two);
1116        if (!o->word_regex)
1117                o->word_regex = diff_word_regex_cfg;
1118        if (o->word_regex) {
1119                ecbdata->diff_words->word_regex = (regex_t *)
1120                        xmalloc(sizeof(regex_t));
1121                if (regcomp(ecbdata->diff_words->word_regex,
1122                            o->word_regex,
1123                            REG_EXTENDED | REG_NEWLINE))
1124                        die ("Invalid regular expression: %s",
1125                             o->word_regex);
1126        }
1127        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1128                if (o->word_diff == diff_words_styles[i].type) {
1129                        ecbdata->diff_words->style =
1130                                &diff_words_styles[i];
1131                        break;
1132                }
1133        }
1134        if (want_color(o->use_color)) {
1135                struct diff_words_style *st = ecbdata->diff_words->style;
1136                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1137                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1138                st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1139        }
1140}
1141
1142static void free_diff_words_data(struct emit_callback *ecbdata)
1143{
1144        if (ecbdata->diff_words) {
1145                diff_words_flush(ecbdata);
1146                free (ecbdata->diff_words->opt);
1147                free (ecbdata->diff_words->minus.text.ptr);
1148                free (ecbdata->diff_words->minus.orig);
1149                free (ecbdata->diff_words->plus.text.ptr);
1150                free (ecbdata->diff_words->plus.orig);
1151                if (ecbdata->diff_words->word_regex) {
1152                        regfree(ecbdata->diff_words->word_regex);
1153                        free(ecbdata->diff_words->word_regex);
1154                }
1155                free(ecbdata->diff_words);
1156                ecbdata->diff_words = NULL;
1157        }
1158}
1159
1160const char *diff_get_color(int diff_use_color, enum color_diff ix)
1161{
1162        if (want_color(diff_use_color))
1163                return diff_colors[ix];
1164        return "";
1165}
1166
1167const char *diff_line_prefix(struct diff_options *opt)
1168{
1169        struct strbuf *msgbuf;
1170        if (!opt->output_prefix)
1171                return "";
1172
1173        msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1174        return msgbuf->buf;
1175}
1176
1177static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1178{
1179        const char *cp;
1180        unsigned long allot;
1181        size_t l = len;
1182
1183        if (ecb->truncate)
1184                return ecb->truncate(line, len);
1185        cp = line;
1186        allot = l;
1187        while (0 < l) {
1188                (void) utf8_width(&cp, &l);
1189                if (!cp)
1190                        break; /* truncated in the middle? */
1191        }
1192        return allot - l;
1193}
1194
1195static void find_lno(const char *line, struct emit_callback *ecbdata)
1196{
1197        const char *p;
1198        ecbdata->lno_in_preimage = 0;
1199        ecbdata->lno_in_postimage = 0;
1200        p = strchr(line, '-');
1201        if (!p)
1202                return; /* cannot happen */
1203        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1204        p = strchr(p, '+');
1205        if (!p)
1206                return; /* cannot happen */
1207        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1208}
1209
1210static void fn_out_consume(void *priv, char *line, unsigned long len)
1211{
1212        struct emit_callback *ecbdata = priv;
1213        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1214        const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1215        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1216        struct diff_options *o = ecbdata->opt;
1217        const char *line_prefix = diff_line_prefix(o);
1218
1219        if (ecbdata->header) {
1220                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1221                strbuf_reset(ecbdata->header);
1222                ecbdata->header = NULL;
1223        }
1224        *(ecbdata->found_changesp) = 1;
1225
1226        if (ecbdata->label_path[0]) {
1227                const char *name_a_tab, *name_b_tab;
1228
1229                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1230                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1231
1232                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1233                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1234                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1235                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1236                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1237        }
1238
1239        if (diff_suppress_blank_empty
1240            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1241                line[0] = '\n';
1242                len = 1;
1243        }
1244
1245        if (line[0] == '@') {
1246                if (ecbdata->diff_words)
1247                        diff_words_flush(ecbdata);
1248                len = sane_truncate_line(ecbdata, line, len);
1249                find_lno(line, ecbdata);
1250                emit_hunk_header(ecbdata, line, len);
1251                if (line[len-1] != '\n')
1252                        putc('\n', ecbdata->opt->file);
1253                return;
1254        }
1255
1256        if (len < 1) {
1257                emit_line(ecbdata->opt, reset, reset, line, len);
1258                if (ecbdata->diff_words
1259                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1260                        fputs("~\n", ecbdata->opt->file);
1261                return;
1262        }
1263
1264        if (ecbdata->diff_words) {
1265                if (line[0] == '-') {
1266                        diff_words_append(line, len,
1267                                          &ecbdata->diff_words->minus);
1268                        return;
1269                } else if (line[0] == '+') {
1270                        diff_words_append(line, len,
1271                                          &ecbdata->diff_words->plus);
1272                        return;
1273                } else if (starts_with(line, "\\ ")) {
1274                        /*
1275                         * Eat the "no newline at eof" marker as if we
1276                         * saw a "+" or "-" line with nothing on it,
1277                         * and return without diff_words_flush() to
1278                         * defer processing. If this is the end of
1279                         * preimage, more "+" lines may come after it.
1280                         */
1281                        return;
1282                }
1283                diff_words_flush(ecbdata);
1284                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1285                        emit_line(ecbdata->opt, context, reset, line, len);
1286                        fputs("~\n", ecbdata->opt->file);
1287                } else {
1288                        /*
1289                         * Skip the prefix character, if any.  With
1290                         * diff_suppress_blank_empty, there may be
1291                         * none.
1292                         */
1293                        if (line[0] != '\n') {
1294                              line++;
1295                              len--;
1296                        }
1297                        emit_line(ecbdata->opt, context, reset, line, len);
1298                }
1299                return;
1300        }
1301
1302        switch (line[0]) {
1303        case '+':
1304                ecbdata->lno_in_postimage++;
1305                emit_add_line(reset, ecbdata, line + 1, len - 1);
1306                break;
1307        case '-':
1308                ecbdata->lno_in_preimage++;
1309                emit_del_line(reset, ecbdata, line + 1, len - 1);
1310                break;
1311        case ' ':
1312                ecbdata->lno_in_postimage++;
1313                ecbdata->lno_in_preimage++;
1314                emit_context_line(reset, ecbdata, line + 1, len - 1);
1315                break;
1316        default:
1317                /* incomplete line at the end */
1318                ecbdata->lno_in_preimage++;
1319                emit_line(ecbdata->opt,
1320                          diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1321                          reset, line, len);
1322                break;
1323        }
1324}
1325
1326static char *pprint_rename(const char *a, const char *b)
1327{
1328        const char *old = a;
1329        const char *new = b;
1330        struct strbuf name = STRBUF_INIT;
1331        int pfx_length, sfx_length;
1332        int pfx_adjust_for_slash;
1333        int len_a = strlen(a);
1334        int len_b = strlen(b);
1335        int a_midlen, b_midlen;
1336        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1337        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1338
1339        if (qlen_a || qlen_b) {
1340                quote_c_style(a, &name, NULL, 0);
1341                strbuf_addstr(&name, " => ");
1342                quote_c_style(b, &name, NULL, 0);
1343                return strbuf_detach(&name, NULL);
1344        }
1345
1346        /* Find common prefix */
1347        pfx_length = 0;
1348        while (*old && *new && *old == *new) {
1349                if (*old == '/')
1350                        pfx_length = old - a + 1;
1351                old++;
1352                new++;
1353        }
1354
1355        /* Find common suffix */
1356        old = a + len_a;
1357        new = b + len_b;
1358        sfx_length = 0;
1359        /*
1360         * If there is a common prefix, it must end in a slash.  In
1361         * that case we let this loop run 1 into the prefix to see the
1362         * same slash.
1363         *
1364         * If there is no common prefix, we cannot do this as it would
1365         * underrun the input strings.
1366         */
1367        pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1368        while (a + pfx_length - pfx_adjust_for_slash <= old &&
1369               b + pfx_length - pfx_adjust_for_slash <= new &&
1370               *old == *new) {
1371                if (*old == '/')
1372                        sfx_length = len_a - (old - a);
1373                old--;
1374                new--;
1375        }
1376
1377        /*
1378         * pfx{mid-a => mid-b}sfx
1379         * {pfx-a => pfx-b}sfx
1380         * pfx{sfx-a => sfx-b}
1381         * name-a => name-b
1382         */
1383        a_midlen = len_a - pfx_length - sfx_length;
1384        b_midlen = len_b - pfx_length - sfx_length;
1385        if (a_midlen < 0)
1386                a_midlen = 0;
1387        if (b_midlen < 0)
1388                b_midlen = 0;
1389
1390        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1391        if (pfx_length + sfx_length) {
1392                strbuf_add(&name, a, pfx_length);
1393                strbuf_addch(&name, '{');
1394        }
1395        strbuf_add(&name, a + pfx_length, a_midlen);
1396        strbuf_addstr(&name, " => ");
1397        strbuf_add(&name, b + pfx_length, b_midlen);
1398        if (pfx_length + sfx_length) {
1399                strbuf_addch(&name, '}');
1400                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1401        }
1402        return strbuf_detach(&name, NULL);
1403}
1404
1405struct diffstat_t {
1406        int nr;
1407        int alloc;
1408        struct diffstat_file {
1409                char *from_name;
1410                char *name;
1411                char *print_name;
1412                unsigned is_unmerged:1;
1413                unsigned is_binary:1;
1414                unsigned is_renamed:1;
1415                unsigned is_interesting:1;
1416                uintmax_t added, deleted;
1417        } **files;
1418};
1419
1420static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1421                                          const char *name_a,
1422                                          const char *name_b)
1423{
1424        struct diffstat_file *x;
1425        x = xcalloc(1, sizeof(*x));
1426        ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1427        diffstat->files[diffstat->nr++] = x;
1428        if (name_b) {
1429                x->from_name = xstrdup(name_a);
1430                x->name = xstrdup(name_b);
1431                x->is_renamed = 1;
1432        }
1433        else {
1434                x->from_name = NULL;
1435                x->name = xstrdup(name_a);
1436        }
1437        return x;
1438}
1439
1440static void diffstat_consume(void *priv, char *line, unsigned long len)
1441{
1442        struct diffstat_t *diffstat = priv;
1443        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1444
1445        if (line[0] == '+')
1446                x->added++;
1447        else if (line[0] == '-')
1448                x->deleted++;
1449}
1450
1451const char mime_boundary_leader[] = "------------";
1452
1453static int scale_linear(int it, int width, int max_change)
1454{
1455        if (!it)
1456                return 0;
1457        /*
1458         * make sure that at least one '-' or '+' is printed if
1459         * there is any change to this path. The easiest way is to
1460         * scale linearly as if the alloted width is one column shorter
1461         * than it is, and then add 1 to the result.
1462         */
1463        return 1 + (it * (width - 1) / max_change);
1464}
1465
1466static void show_name(FILE *file,
1467                      const char *prefix, const char *name, int len)
1468{
1469        fprintf(file, " %s%-*s |", prefix, len, name);
1470}
1471
1472static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1473{
1474        if (cnt <= 0)
1475                return;
1476        fprintf(file, "%s", set);
1477        while (cnt--)
1478                putc(ch, file);
1479        fprintf(file, "%s", reset);
1480}
1481
1482static void fill_print_name(struct diffstat_file *file)
1483{
1484        char *pname;
1485
1486        if (file->print_name)
1487                return;
1488
1489        if (!file->is_renamed) {
1490                struct strbuf buf = STRBUF_INIT;
1491                if (quote_c_style(file->name, &buf, NULL, 0)) {
1492                        pname = strbuf_detach(&buf, NULL);
1493                } else {
1494                        pname = file->name;
1495                        strbuf_release(&buf);
1496                }
1497        } else {
1498                pname = pprint_rename(file->from_name, file->name);
1499        }
1500        file->print_name = pname;
1501}
1502
1503int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1504{
1505        struct strbuf sb = STRBUF_INIT;
1506        int ret;
1507
1508        if (!files) {
1509                assert(insertions == 0 && deletions == 0);
1510                return fprintf(fp, "%s\n", " 0 files changed");
1511        }
1512
1513        strbuf_addf(&sb,
1514                    (files == 1) ? " %d file changed" : " %d files changed",
1515                    files);
1516
1517        /*
1518         * For binary diff, the caller may want to print "x files
1519         * changed" with insertions == 0 && deletions == 0.
1520         *
1521         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1522         * is probably less confusing (i.e skip over "2 files changed
1523         * but nothing about added/removed lines? Is this a bug in Git?").
1524         */
1525        if (insertions || deletions == 0) {
1526                strbuf_addf(&sb,
1527                            (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1528                            insertions);
1529        }
1530
1531        if (deletions || insertions == 0) {
1532                strbuf_addf(&sb,
1533                            (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1534                            deletions);
1535        }
1536        strbuf_addch(&sb, '\n');
1537        ret = fputs(sb.buf, fp);
1538        strbuf_release(&sb);
1539        return ret;
1540}
1541
1542static void show_stats(struct diffstat_t *data, struct diff_options *options)
1543{
1544        int i, len, add, del, adds = 0, dels = 0;
1545        uintmax_t max_change = 0, max_len = 0;
1546        int total_files = data->nr, count;
1547        int width, name_width, graph_width, number_width = 0, bin_width = 0;
1548        const char *reset, *add_c, *del_c;
1549        const char *line_prefix = "";
1550        int extra_shown = 0;
1551
1552        if (data->nr == 0)
1553                return;
1554
1555        line_prefix = diff_line_prefix(options);
1556        count = options->stat_count ? options->stat_count : data->nr;
1557
1558        reset = diff_get_color_opt(options, DIFF_RESET);
1559        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1560        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1561
1562        /*
1563         * Find the longest filename and max number of changes
1564         */
1565        for (i = 0; (i < count) && (i < data->nr); i++) {
1566                struct diffstat_file *file = data->files[i];
1567                uintmax_t change = file->added + file->deleted;
1568
1569                if (!file->is_interesting && (change == 0)) {
1570                        count++; /* not shown == room for one more */
1571                        continue;
1572                }
1573                fill_print_name(file);
1574                len = strlen(file->print_name);
1575                if (max_len < len)
1576                        max_len = len;
1577
1578                if (file->is_unmerged) {
1579                        /* "Unmerged" is 8 characters */
1580                        bin_width = bin_width < 8 ? 8 : bin_width;
1581                        continue;
1582                }
1583                if (file->is_binary) {
1584                        /* "Bin XXX -> YYY bytes" */
1585                        int w = 14 + decimal_width(file->added)
1586                                + decimal_width(file->deleted);
1587                        bin_width = bin_width < w ? w : bin_width;
1588                        /* Display change counts aligned with "Bin" */
1589                        number_width = 3;
1590                        continue;
1591                }
1592
1593                if (max_change < change)
1594                        max_change = change;
1595        }
1596        count = i; /* where we can stop scanning in data->files[] */
1597
1598        /*
1599         * We have width = stat_width or term_columns() columns total.
1600         * We want a maximum of min(max_len, stat_name_width) for the name part.
1601         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1602         * We also need 1 for " " and 4 + decimal_width(max_change)
1603         * for " | NNNN " and one the empty column at the end, altogether
1604         * 6 + decimal_width(max_change).
1605         *
1606         * If there's not enough space, we will use the smaller of
1607         * stat_name_width (if set) and 5/8*width for the filename,
1608         * and the rest for constant elements + graph part, but no more
1609         * than stat_graph_width for the graph part.
1610         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1611         * for the standard terminal size).
1612         *
1613         * In other words: stat_width limits the maximum width, and
1614         * stat_name_width fixes the maximum width of the filename,
1615         * and is also used to divide available columns if there
1616         * aren't enough.
1617         *
1618         * Binary files are displayed with "Bin XXX -> YYY bytes"
1619         * instead of the change count and graph. This part is treated
1620         * similarly to the graph part, except that it is not
1621         * "scaled". If total width is too small to accommodate the
1622         * guaranteed minimum width of the filename part and the
1623         * separators and this message, this message will "overflow"
1624         * making the line longer than the maximum width.
1625         */
1626
1627        if (options->stat_width == -1)
1628                width = term_columns() - options->output_prefix_length;
1629        else
1630                width = options->stat_width ? options->stat_width : 80;
1631        number_width = decimal_width(max_change) > number_width ?
1632                decimal_width(max_change) : number_width;
1633
1634        if (options->stat_graph_width == -1)
1635                options->stat_graph_width = diff_stat_graph_width;
1636
1637        /*
1638         * Guarantee 3/8*16==6 for the graph part
1639         * and 5/8*16==10 for the filename part
1640         */
1641        if (width < 16 + 6 + number_width)
1642                width = 16 + 6 + number_width;
1643
1644        /*
1645         * First assign sizes that are wanted, ignoring available width.
1646         * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1647         * starting from "XXX" should fit in graph_width.
1648         */
1649        graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1650        if (options->stat_graph_width &&
1651            options->stat_graph_width < graph_width)
1652                graph_width = options->stat_graph_width;
1653
1654        name_width = (options->stat_name_width > 0 &&
1655                      options->stat_name_width < max_len) ?
1656                options->stat_name_width : max_len;
1657
1658        /*
1659         * Adjust adjustable widths not to exceed maximum width
1660         */
1661        if (name_width + number_width + 6 + graph_width > width) {
1662                if (graph_width > width * 3/8 - number_width - 6) {
1663                        graph_width = width * 3/8 - number_width - 6;
1664                        if (graph_width < 6)
1665                                graph_width = 6;
1666                }
1667
1668                if (options->stat_graph_width &&
1669                    graph_width > options->stat_graph_width)
1670                        graph_width = options->stat_graph_width;
1671                if (name_width > width - number_width - 6 - graph_width)
1672                        name_width = width - number_width - 6 - graph_width;
1673                else
1674                        graph_width = width - number_width - 6 - name_width;
1675        }
1676
1677        /*
1678         * From here name_width is the width of the name area,
1679         * and graph_width is the width of the graph area.
1680         * max_change is used to scale graph properly.
1681         */
1682        for (i = 0; i < count; i++) {
1683                const char *prefix = "";
1684                struct diffstat_file *file = data->files[i];
1685                char *name = file->print_name;
1686                uintmax_t added = file->added;
1687                uintmax_t deleted = file->deleted;
1688                int name_len;
1689
1690                if (!file->is_interesting && (added + deleted == 0))
1691                        continue;
1692
1693                /*
1694                 * "scale" the filename
1695                 */
1696                len = name_width;
1697                name_len = strlen(name);
1698                if (name_width < name_len) {
1699                        char *slash;
1700                        prefix = "...";
1701                        len -= 3;
1702                        name += name_len - len;
1703                        slash = strchr(name, '/');
1704                        if (slash)
1705                                name = slash;
1706                }
1707
1708                if (file->is_binary) {
1709                        fprintf(options->file, "%s", line_prefix);
1710                        show_name(options->file, prefix, name, len);
1711                        fprintf(options->file, " %*s", number_width, "Bin");
1712                        if (!added && !deleted) {
1713                                putc('\n', options->file);
1714                                continue;
1715                        }
1716                        fprintf(options->file, " %s%"PRIuMAX"%s",
1717                                del_c, deleted, reset);
1718                        fprintf(options->file, " -> ");
1719                        fprintf(options->file, "%s%"PRIuMAX"%s",
1720                                add_c, added, reset);
1721                        fprintf(options->file, " bytes");
1722                        fprintf(options->file, "\n");
1723                        continue;
1724                }
1725                else if (file->is_unmerged) {
1726                        fprintf(options->file, "%s", line_prefix);
1727                        show_name(options->file, prefix, name, len);
1728                        fprintf(options->file, " Unmerged\n");
1729                        continue;
1730                }
1731
1732                /*
1733                 * scale the add/delete
1734                 */
1735                add = added;
1736                del = deleted;
1737
1738                if (graph_width <= max_change) {
1739                        int total = scale_linear(add + del, graph_width, max_change);
1740                        if (total < 2 && add && del)
1741                                /* width >= 2 due to the sanity check */
1742                                total = 2;
1743                        if (add < del) {
1744                                add = scale_linear(add, graph_width, max_change);
1745                                del = total - add;
1746                        } else {
1747                                del = scale_linear(del, graph_width, max_change);
1748                                add = total - del;
1749                        }
1750                }
1751                fprintf(options->file, "%s", line_prefix);
1752                show_name(options->file, prefix, name, len);
1753                fprintf(options->file, " %*"PRIuMAX"%s",
1754                        number_width, added + deleted,
1755                        added + deleted ? " " : "");
1756                show_graph(options->file, '+', add, add_c, reset);
1757                show_graph(options->file, '-', del, del_c, reset);
1758                fprintf(options->file, "\n");
1759        }
1760
1761        for (i = 0; i < data->nr; i++) {
1762                struct diffstat_file *file = data->files[i];
1763                uintmax_t added = file->added;
1764                uintmax_t deleted = file->deleted;
1765
1766                if (file->is_unmerged ||
1767                    (!file->is_interesting && (added + deleted == 0))) {
1768                        total_files--;
1769                        continue;
1770                }
1771
1772                if (!file->is_binary) {
1773                        adds += added;
1774                        dels += deleted;
1775                }
1776                if (i < count)
1777                        continue;
1778                if (!extra_shown)
1779                        fprintf(options->file, "%s ...\n", line_prefix);
1780                extra_shown = 1;
1781        }
1782        fprintf(options->file, "%s", line_prefix);
1783        print_stat_summary(options->file, total_files, adds, dels);
1784}
1785
1786static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1787{
1788        int i, adds = 0, dels = 0, total_files = data->nr;
1789
1790        if (data->nr == 0)
1791                return;
1792
1793        for (i = 0; i < data->nr; i++) {
1794                int added = data->files[i]->added;
1795                int deleted= data->files[i]->deleted;
1796
1797                if (data->files[i]->is_unmerged ||
1798                    (!data->files[i]->is_interesting && (added + deleted == 0))) {
1799                        total_files--;
1800                } else if (!data->files[i]->is_binary) { /* don't count bytes */
1801                        adds += added;
1802                        dels += deleted;
1803                }
1804        }
1805        fprintf(options->file, "%s", diff_line_prefix(options));
1806        print_stat_summary(options->file, total_files, adds, dels);
1807}
1808
1809static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1810{
1811        int i;
1812
1813        if (data->nr == 0)
1814                return;
1815
1816        for (i = 0; i < data->nr; i++) {
1817                struct diffstat_file *file = data->files[i];
1818
1819                fprintf(options->file, "%s", diff_line_prefix(options));
1820
1821                if (file->is_binary)
1822                        fprintf(options->file, "-\t-\t");
1823                else
1824                        fprintf(options->file,
1825                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1826                                file->added, file->deleted);
1827                if (options->line_termination) {
1828                        fill_print_name(file);
1829                        if (!file->is_renamed)
1830                                write_name_quoted(file->name, options->file,
1831                                                  options->line_termination);
1832                        else {
1833                                fputs(file->print_name, options->file);
1834                                putc(options->line_termination, options->file);
1835                        }
1836                } else {
1837                        if (file->is_renamed) {
1838                                putc('\0', options->file);
1839                                write_name_quoted(file->from_name, options->file, '\0');
1840                        }
1841                        write_name_quoted(file->name, options->file, '\0');
1842                }
1843        }
1844}
1845
1846struct dirstat_file {
1847        const char *name;
1848        unsigned long changed;
1849};
1850
1851struct dirstat_dir {
1852        struct dirstat_file *files;
1853        int alloc, nr, permille, cumulative;
1854};
1855
1856static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1857                unsigned long changed, const char *base, int baselen)
1858{
1859        unsigned long this_dir = 0;
1860        unsigned int sources = 0;
1861        const char *line_prefix = diff_line_prefix(opt);
1862
1863        while (dir->nr) {
1864                struct dirstat_file *f = dir->files;
1865                int namelen = strlen(f->name);
1866                unsigned long this;
1867                char *slash;
1868
1869                if (namelen < baselen)
1870                        break;
1871                if (memcmp(f->name, base, baselen))
1872                        break;
1873                slash = strchr(f->name + baselen, '/');
1874                if (slash) {
1875                        int newbaselen = slash + 1 - f->name;
1876                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1877                        sources++;
1878                } else {
1879                        this = f->changed;
1880                        dir->files++;
1881                        dir->nr--;
1882                        sources += 2;
1883                }
1884                this_dir += this;
1885        }
1886
1887        /*
1888         * We don't report dirstat's for
1889         *  - the top level
1890         *  - or cases where everything came from a single directory
1891         *    under this directory (sources == 1).
1892         */
1893        if (baselen && sources != 1) {
1894                if (this_dir) {
1895                        int permille = this_dir * 1000 / changed;
1896                        if (permille >= dir->permille) {
1897                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1898                                        permille / 10, permille % 10, baselen, base);
1899                                if (!dir->cumulative)
1900                                        return 0;
1901                        }
1902                }
1903        }
1904        return this_dir;
1905}
1906
1907static int dirstat_compare(const void *_a, const void *_b)
1908{
1909        const struct dirstat_file *a = _a;
1910        const struct dirstat_file *b = _b;
1911        return strcmp(a->name, b->name);
1912}
1913
1914static void show_dirstat(struct diff_options *options)
1915{
1916        int i;
1917        unsigned long changed;
1918        struct dirstat_dir dir;
1919        struct diff_queue_struct *q = &diff_queued_diff;
1920
1921        dir.files = NULL;
1922        dir.alloc = 0;
1923        dir.nr = 0;
1924        dir.permille = options->dirstat_permille;
1925        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1926
1927        changed = 0;
1928        for (i = 0; i < q->nr; i++) {
1929                struct diff_filepair *p = q->queue[i];
1930                const char *name;
1931                unsigned long copied, added, damage;
1932                int content_changed;
1933
1934                name = p->two->path ? p->two->path : p->one->path;
1935
1936                if (p->one->oid_valid && p->two->oid_valid)
1937                        content_changed = oidcmp(&p->one->oid, &p->two->oid);
1938                else
1939                        content_changed = 1;
1940
1941                if (!content_changed) {
1942                        /*
1943                         * The SHA1 has not changed, so pre-/post-content is
1944                         * identical. We can therefore skip looking at the
1945                         * file contents altogether.
1946                         */
1947                        damage = 0;
1948                        goto found_damage;
1949                }
1950
1951                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1952                        /*
1953                         * In --dirstat-by-file mode, we don't really need to
1954                         * look at the actual file contents at all.
1955                         * The fact that the SHA1 changed is enough for us to
1956                         * add this file to the list of results
1957                         * (with each file contributing equal damage).
1958                         */
1959                        damage = 1;
1960                        goto found_damage;
1961                }
1962
1963                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1964                        diff_populate_filespec(p->one, 0);
1965                        diff_populate_filespec(p->two, 0);
1966                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1967                                               &copied, &added);
1968                        diff_free_filespec_data(p->one);
1969                        diff_free_filespec_data(p->two);
1970                } else if (DIFF_FILE_VALID(p->one)) {
1971                        diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1972                        copied = added = 0;
1973                        diff_free_filespec_data(p->one);
1974                } else if (DIFF_FILE_VALID(p->two)) {
1975                        diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1976                        copied = 0;
1977                        added = p->two->size;
1978                        diff_free_filespec_data(p->two);
1979                } else
1980                        continue;
1981
1982                /*
1983                 * Original minus copied is the removed material,
1984                 * added is the new material.  They are both damages
1985                 * made to the preimage.
1986                 * If the resulting damage is zero, we know that
1987                 * diffcore_count_changes() considers the two entries to
1988                 * be identical, but since content_changed is true, we
1989                 * know that there must have been _some_ kind of change,
1990                 * so we force all entries to have damage > 0.
1991                 */
1992                damage = (p->one->size - copied) + added;
1993                if (!damage)
1994                        damage = 1;
1995
1996found_damage:
1997                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1998                dir.files[dir.nr].name = name;
1999                dir.files[dir.nr].changed = damage;
2000                changed += damage;
2001                dir.nr++;
2002        }
2003
2004        /* This can happen even with many files, if everything was renames */
2005        if (!changed)
2006                return;
2007
2008        /* Show all directories with more than x% of the changes */
2009        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2010        gather_dirstat(options, &dir, changed, "", 0);
2011}
2012
2013static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2014{
2015        int i;
2016        unsigned long changed;
2017        struct dirstat_dir dir;
2018
2019        if (data->nr == 0)
2020                return;
2021
2022        dir.files = NULL;
2023        dir.alloc = 0;
2024        dir.nr = 0;
2025        dir.permille = options->dirstat_permille;
2026        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2027
2028        changed = 0;
2029        for (i = 0; i < data->nr; i++) {
2030                struct diffstat_file *file = data->files[i];
2031                unsigned long damage = file->added + file->deleted;
2032                if (file->is_binary)
2033                        /*
2034                         * binary files counts bytes, not lines. Must find some
2035                         * way to normalize binary bytes vs. textual lines.
2036                         * The following heuristic assumes that there are 64
2037                         * bytes per "line".
2038                         * This is stupid and ugly, but very cheap...
2039                         */
2040                        damage = (damage + 63) / 64;
2041                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2042                dir.files[dir.nr].name = file->name;
2043                dir.files[dir.nr].changed = damage;
2044                changed += damage;
2045                dir.nr++;
2046        }
2047
2048        /* This can happen even with many files, if everything was renames */
2049        if (!changed)
2050                return;
2051
2052        /* Show all directories with more than x% of the changes */
2053        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2054        gather_dirstat(options, &dir, changed, "", 0);
2055}
2056
2057static void free_diffstat_info(struct diffstat_t *diffstat)
2058{
2059        int i;
2060        for (i = 0; i < diffstat->nr; i++) {
2061                struct diffstat_file *f = diffstat->files[i];
2062                if (f->name != f->print_name)
2063                        free(f->print_name);
2064                free(f->name);
2065                free(f->from_name);
2066                free(f);
2067        }
2068        free(diffstat->files);
2069}
2070
2071struct checkdiff_t {
2072        const char *filename;
2073        int lineno;
2074        int conflict_marker_size;
2075        struct diff_options *o;
2076        unsigned ws_rule;
2077        unsigned status;
2078};
2079
2080static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2081{
2082        char firstchar;
2083        int cnt;
2084
2085        if (len < marker_size + 1)
2086                return 0;
2087        firstchar = line[0];
2088        switch (firstchar) {
2089        case '=': case '>': case '<': case '|':
2090                break;
2091        default:
2092                return 0;
2093        }
2094        for (cnt = 1; cnt < marker_size; cnt++)
2095                if (line[cnt] != firstchar)
2096                        return 0;
2097        /* line[1] thru line[marker_size-1] are same as firstchar */
2098        if (len < marker_size + 1 || !isspace(line[marker_size]))
2099                return 0;
2100        return 1;
2101}
2102
2103static void checkdiff_consume(void *priv, char *line, unsigned long len)
2104{
2105        struct checkdiff_t *data = priv;
2106        int marker_size = data->conflict_marker_size;
2107        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2108        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2109        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2110        char *err;
2111        const char *line_prefix;
2112
2113        assert(data->o);
2114        line_prefix = diff_line_prefix(data->o);
2115
2116        if (line[0] == '+') {
2117                unsigned bad;
2118                data->lineno++;
2119                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2120                        data->status |= 1;
2121                        fprintf(data->o->file,
2122                                "%s%s:%d: leftover conflict marker\n",
2123                                line_prefix, data->filename, data->lineno);
2124                }
2125                bad = ws_check(line + 1, len - 1, data->ws_rule);
2126                if (!bad)
2127                        return;
2128                data->status |= bad;
2129                err = whitespace_error_string(bad);
2130                fprintf(data->o->file, "%s%s:%d: %s.\n",
2131                        line_prefix, data->filename, data->lineno, err);
2132                free(err);
2133                emit_line(data->o, set, reset, line, 1);
2134                ws_check_emit(line + 1, len - 1, data->ws_rule,
2135                              data->o->file, set, reset, ws);
2136        } else if (line[0] == ' ') {
2137                data->lineno++;
2138        } else if (line[0] == '@') {
2139                char *plus = strchr(line, '+');
2140                if (plus)
2141                        data->lineno = strtol(plus, NULL, 10) - 1;
2142                else
2143                        die("invalid diff");
2144        }
2145}
2146
2147static unsigned char *deflate_it(char *data,
2148                                 unsigned long size,
2149                                 unsigned long *result_size)
2150{
2151        int bound;
2152        unsigned char *deflated;
2153        git_zstream stream;
2154
2155        git_deflate_init(&stream, zlib_compression_level);
2156        bound = git_deflate_bound(&stream, size);
2157        deflated = xmalloc(bound);
2158        stream.next_out = deflated;
2159        stream.avail_out = bound;
2160
2161        stream.next_in = (unsigned char *)data;
2162        stream.avail_in = size;
2163        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2164                ; /* nothing */
2165        git_deflate_end(&stream);
2166        *result_size = stream.total_out;
2167        return deflated;
2168}
2169
2170static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2171                                  const char *prefix)
2172{
2173        void *cp;
2174        void *delta;
2175        void *deflated;
2176        void *data;
2177        unsigned long orig_size;
2178        unsigned long delta_size;
2179        unsigned long deflate_size;
2180        unsigned long data_size;
2181
2182        /* We could do deflated delta, or we could do just deflated two,
2183         * whichever is smaller.
2184         */
2185        delta = NULL;
2186        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2187        if (one->size && two->size) {
2188                delta = diff_delta(one->ptr, one->size,
2189                                   two->ptr, two->size,
2190                                   &delta_size, deflate_size);
2191                if (delta) {
2192                        void *to_free = delta;
2193                        orig_size = delta_size;
2194                        delta = deflate_it(delta, delta_size, &delta_size);
2195                        free(to_free);
2196                }
2197        }
2198
2199        if (delta && delta_size < deflate_size) {
2200                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2201                free(deflated);
2202                data = delta;
2203                data_size = delta_size;
2204        }
2205        else {
2206                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2207                free(delta);
2208                data = deflated;
2209                data_size = deflate_size;
2210        }
2211
2212        /* emit data encoded in base85 */
2213        cp = data;
2214        while (data_size) {
2215                int bytes = (52 < data_size) ? 52 : data_size;
2216                char line[70];
2217                data_size -= bytes;
2218                if (bytes <= 26)
2219                        line[0] = bytes + 'A' - 1;
2220                else
2221                        line[0] = bytes - 26 + 'a' - 1;
2222                encode_85(line + 1, cp, bytes);
2223                cp = (char *) cp + bytes;
2224                fprintf(file, "%s", prefix);
2225                fputs(line, file);
2226                fputc('\n', file);
2227        }
2228        fprintf(file, "%s\n", prefix);
2229        free(data);
2230}
2231
2232static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2233                             const char *prefix)
2234{
2235        fprintf(file, "%sGIT binary patch\n", prefix);
2236        emit_binary_diff_body(file, one, two, prefix);
2237        emit_binary_diff_body(file, two, one, prefix);
2238}
2239
2240int diff_filespec_is_binary(struct diff_filespec *one)
2241{
2242        if (one->is_binary == -1) {
2243                diff_filespec_load_driver(one);
2244                if (one->driver->binary != -1)
2245                        one->is_binary = one->driver->binary;
2246                else {
2247                        if (!one->data && DIFF_FILE_VALID(one))
2248                                diff_populate_filespec(one, CHECK_BINARY);
2249                        if (one->is_binary == -1 && one->data)
2250                                one->is_binary = buffer_is_binary(one->data,
2251                                                one->size);
2252                        if (one->is_binary == -1)
2253                                one->is_binary = 0;
2254                }
2255        }
2256        return one->is_binary;
2257}
2258
2259static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2260{
2261        diff_filespec_load_driver(one);
2262        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2263}
2264
2265void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2266{
2267        if (!options->a_prefix)
2268                options->a_prefix = a;
2269        if (!options->b_prefix)
2270                options->b_prefix = b;
2271}
2272
2273struct userdiff_driver *get_textconv(struct diff_filespec *one)
2274{
2275        if (!DIFF_FILE_VALID(one))
2276                return NULL;
2277
2278        diff_filespec_load_driver(one);
2279        return userdiff_get_textconv(one->driver);
2280}
2281
2282static void builtin_diff(const char *name_a,
2283                         const char *name_b,
2284                         struct diff_filespec *one,
2285                         struct diff_filespec *two,
2286                         const char *xfrm_msg,
2287                         int must_show_header,
2288                         struct diff_options *o,
2289                         int complete_rewrite)
2290{
2291        mmfile_t mf1, mf2;
2292        const char *lbl[2];
2293        char *a_one, *b_two;
2294        const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2295        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2296        const char *a_prefix, *b_prefix;
2297        struct userdiff_driver *textconv_one = NULL;
2298        struct userdiff_driver *textconv_two = NULL;
2299        struct strbuf header = STRBUF_INIT;
2300        const char *line_prefix = diff_line_prefix(o);
2301
2302        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2303                        (!one->mode || S_ISGITLINK(one->mode)) &&
2304                        (!two->mode || S_ISGITLINK(two->mode))) {
2305                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2306                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2307                show_submodule_summary(o->file, one->path ? one->path : two->path,
2308                                line_prefix,
2309                                one->oid.hash, two->oid.hash,
2310                                two->dirty_submodule,
2311                                meta, del, add, reset);
2312                return;
2313        }
2314
2315        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2316                textconv_one = get_textconv(one);
2317                textconv_two = get_textconv(two);
2318        }
2319
2320        diff_set_mnemonic_prefix(o, "a/", "b/");
2321        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2322                a_prefix = o->b_prefix;
2323                b_prefix = o->a_prefix;
2324        } else {
2325                a_prefix = o->a_prefix;
2326                b_prefix = o->b_prefix;
2327        }
2328
2329        /* Never use a non-valid filename anywhere if at all possible */
2330        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2331        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2332
2333        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2334        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2335        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2336        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2337        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2338        if (lbl[0][0] == '/') {
2339                /* /dev/null */
2340                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2341                if (xfrm_msg)
2342                        strbuf_addstr(&header, xfrm_msg);
2343                must_show_header = 1;
2344        }
2345        else if (lbl[1][0] == '/') {
2346                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2347                if (xfrm_msg)
2348                        strbuf_addstr(&header, xfrm_msg);
2349                must_show_header = 1;
2350        }
2351        else {
2352                if (one->mode != two->mode) {
2353                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2354                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2355                        must_show_header = 1;
2356                }
2357                if (xfrm_msg)
2358                        strbuf_addstr(&header, xfrm_msg);
2359
2360                /*
2361                 * we do not run diff between different kind
2362                 * of objects.
2363                 */
2364                if ((one->mode ^ two->mode) & S_IFMT)
2365                        goto free_ab_and_return;
2366                if (complete_rewrite &&
2367                    (textconv_one || !diff_filespec_is_binary(one)) &&
2368                    (textconv_two || !diff_filespec_is_binary(two))) {
2369                        fprintf(o->file, "%s", header.buf);
2370                        strbuf_reset(&header);
2371                        emit_rewrite_diff(name_a, name_b, one, two,
2372                                                textconv_one, textconv_two, o);
2373                        o->found_changes = 1;
2374                        goto free_ab_and_return;
2375                }
2376        }
2377
2378        if (o->irreversible_delete && lbl[1][0] == '/') {
2379                fprintf(o->file, "%s", header.buf);
2380                strbuf_reset(&header);
2381                goto free_ab_and_return;
2382        } else if (!DIFF_OPT_TST(o, TEXT) &&
2383            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2384              (!textconv_two && diff_filespec_is_binary(two)) )) {
2385                if (!one->data && !two->data &&
2386                    S_ISREG(one->mode) && S_ISREG(two->mode) &&
2387                    !DIFF_OPT_TST(o, BINARY)) {
2388                        if (!oidcmp(&one->oid, &two->oid)) {
2389                                if (must_show_header)
2390                                        fprintf(o->file, "%s", header.buf);
2391                                goto free_ab_and_return;
2392                        }
2393                        fprintf(o->file, "%s", header.buf);
2394                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2395                                line_prefix, lbl[0], lbl[1]);
2396                        goto free_ab_and_return;
2397                }
2398                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2399                        die("unable to read files to diff");
2400                /* Quite common confusing case */
2401                if (mf1.size == mf2.size &&
2402                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2403                        if (must_show_header)
2404                                fprintf(o->file, "%s", header.buf);
2405                        goto free_ab_and_return;
2406                }
2407                fprintf(o->file, "%s", header.buf);
2408                strbuf_reset(&header);
2409                if (DIFF_OPT_TST(o, BINARY))
2410                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2411                else
2412                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2413                                line_prefix, lbl[0], lbl[1]);
2414                o->found_changes = 1;
2415        } else {
2416                /* Crazy xdl interfaces.. */
2417                const char *diffopts = getenv("GIT_DIFF_OPTS");
2418                const char *v;
2419                xpparam_t xpp;
2420                xdemitconf_t xecfg;
2421                struct emit_callback ecbdata;
2422                const struct userdiff_funcname *pe;
2423
2424                if (must_show_header) {
2425                        fprintf(o->file, "%s", header.buf);
2426                        strbuf_reset(&header);
2427                }
2428
2429                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2430                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2431
2432                pe = diff_funcname_pattern(one);
2433                if (!pe)
2434                        pe = diff_funcname_pattern(two);
2435
2436                memset(&xpp, 0, sizeof(xpp));
2437                memset(&xecfg, 0, sizeof(xecfg));
2438                memset(&ecbdata, 0, sizeof(ecbdata));
2439                ecbdata.label_path = lbl;
2440                ecbdata.color_diff = want_color(o->use_color);
2441                ecbdata.found_changesp = &o->found_changes;
2442                ecbdata.ws_rule = whitespace_rule(name_b);
2443                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2444                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2445                ecbdata.opt = o;
2446                ecbdata.header = header.len ? &header : NULL;
2447                xpp.flags = o->xdl_opts;
2448                xecfg.ctxlen = o->context;
2449                xecfg.interhunkctxlen = o->interhunkcontext;
2450                xecfg.flags = XDL_EMIT_FUNCNAMES;
2451                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2452                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2453                if (pe)
2454                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2455                if (!diffopts)
2456                        ;
2457                else if (skip_prefix(diffopts, "--unified=", &v))
2458                        xecfg.ctxlen = strtoul(v, NULL, 10);
2459                else if (skip_prefix(diffopts, "-u", &v))
2460                        xecfg.ctxlen = strtoul(v, NULL, 10);
2461                if (o->word_diff)
2462                        init_diff_words_data(&ecbdata, o, one, two);
2463                if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2464                                  &xpp, &xecfg))
2465                        die("unable to generate diff for %s", one->path);
2466                if (o->word_diff)
2467                        free_diff_words_data(&ecbdata);
2468                if (textconv_one)
2469                        free(mf1.ptr);
2470                if (textconv_two)
2471                        free(mf2.ptr);
2472                xdiff_clear_find_func(&xecfg);
2473        }
2474
2475 free_ab_and_return:
2476        strbuf_release(&header);
2477        diff_free_filespec_data(one);
2478        diff_free_filespec_data(two);
2479        free(a_one);
2480        free(b_two);
2481        return;
2482}
2483
2484static void builtin_diffstat(const char *name_a, const char *name_b,
2485                             struct diff_filespec *one,
2486                             struct diff_filespec *two,
2487                             struct diffstat_t *diffstat,
2488                             struct diff_options *o,
2489                             struct diff_filepair *p)
2490{
2491        mmfile_t mf1, mf2;
2492        struct diffstat_file *data;
2493        int same_contents;
2494        int complete_rewrite = 0;
2495
2496        if (!DIFF_PAIR_UNMERGED(p)) {
2497                if (p->status == DIFF_STATUS_MODIFIED && p->score)
2498                        complete_rewrite = 1;
2499        }
2500
2501        data = diffstat_add(diffstat, name_a, name_b);
2502        data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2503
2504        if (!one || !two) {
2505                data->is_unmerged = 1;
2506                return;
2507        }
2508
2509        same_contents = !oidcmp(&one->oid, &two->oid);
2510
2511        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2512                data->is_binary = 1;
2513                if (same_contents) {
2514                        data->added = 0;
2515                        data->deleted = 0;
2516                } else {
2517                        data->added = diff_filespec_size(two);
2518                        data->deleted = diff_filespec_size(one);
2519                }
2520        }
2521
2522        else if (complete_rewrite) {
2523                diff_populate_filespec(one, 0);
2524                diff_populate_filespec(two, 0);
2525                data->deleted = count_lines(one->data, one->size);
2526                data->added = count_lines(two->data, two->size);
2527        }
2528
2529        else if (!same_contents) {
2530                /* Crazy xdl interfaces.. */
2531                xpparam_t xpp;
2532                xdemitconf_t xecfg;
2533
2534                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2535                        die("unable to read files to diff");
2536
2537                memset(&xpp, 0, sizeof(xpp));
2538                memset(&xecfg, 0, sizeof(xecfg));
2539                xpp.flags = o->xdl_opts;
2540                xecfg.ctxlen = o->context;
2541                xecfg.interhunkctxlen = o->interhunkcontext;
2542                if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2543                                  &xpp, &xecfg))
2544                        die("unable to generate diffstat for %s", one->path);
2545        }
2546
2547        diff_free_filespec_data(one);
2548        diff_free_filespec_data(two);
2549}
2550
2551static void builtin_checkdiff(const char *name_a, const char *name_b,
2552                              const char *attr_path,
2553                              struct diff_filespec *one,
2554                              struct diff_filespec *two,
2555                              struct diff_options *o)
2556{
2557        mmfile_t mf1, mf2;
2558        struct checkdiff_t data;
2559
2560        if (!two)
2561                return;
2562
2563        memset(&data, 0, sizeof(data));
2564        data.filename = name_b ? name_b : name_a;
2565        data.lineno = 0;
2566        data.o = o;
2567        data.ws_rule = whitespace_rule(attr_path);
2568        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2569
2570        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2571                die("unable to read files to diff");
2572
2573        /*
2574         * All the other codepaths check both sides, but not checking
2575         * the "old" side here is deliberate.  We are checking the newly
2576         * introduced changes, and as long as the "new" side is text, we
2577         * can and should check what it introduces.
2578         */
2579        if (diff_filespec_is_binary(two))
2580                goto free_and_return;
2581        else {
2582                /* Crazy xdl interfaces.. */
2583                xpparam_t xpp;
2584                xdemitconf_t xecfg;
2585
2586                memset(&xpp, 0, sizeof(xpp));
2587                memset(&xecfg, 0, sizeof(xecfg));
2588                xecfg.ctxlen = 1; /* at least one context line */
2589                xpp.flags = 0;
2590                if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2591                                  &xpp, &xecfg))
2592                        die("unable to generate checkdiff for %s", one->path);
2593
2594                if (data.ws_rule & WS_BLANK_AT_EOF) {
2595                        struct emit_callback ecbdata;
2596                        int blank_at_eof;
2597
2598                        ecbdata.ws_rule = data.ws_rule;
2599                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2600                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2601
2602                        if (blank_at_eof) {
2603                                static char *err;
2604                                if (!err)
2605                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2606                                fprintf(o->file, "%s:%d: %s.\n",
2607                                        data.filename, blank_at_eof, err);
2608                                data.status = 1; /* report errors */
2609                        }
2610                }
2611        }
2612 free_and_return:
2613        diff_free_filespec_data(one);
2614        diff_free_filespec_data(two);
2615        if (data.status)
2616                DIFF_OPT_SET(o, CHECK_FAILED);
2617}
2618
2619struct diff_filespec *alloc_filespec(const char *path)
2620{
2621        struct diff_filespec *spec;
2622
2623        FLEXPTR_ALLOC_STR(spec, path, path);
2624        spec->count = 1;
2625        spec->is_binary = -1;
2626        return spec;
2627}
2628
2629void free_filespec(struct diff_filespec *spec)
2630{
2631        if (!--spec->count) {
2632                diff_free_filespec_data(spec);
2633                free(spec);
2634        }
2635}
2636
2637void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2638                   int sha1_valid, unsigned short mode)
2639{
2640        if (mode) {
2641                spec->mode = canon_mode(mode);
2642                hashcpy(spec->oid.hash, sha1);
2643                spec->oid_valid = sha1_valid;
2644        }
2645}
2646
2647/*
2648 * Given a name and sha1 pair, if the index tells us the file in
2649 * the work tree has that object contents, return true, so that
2650 * prepare_temp_file() does not have to inflate and extract.
2651 */
2652static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2653{
2654        const struct cache_entry *ce;
2655        struct stat st;
2656        int pos, len;
2657
2658        /*
2659         * We do not read the cache ourselves here, because the
2660         * benchmark with my previous version that always reads cache
2661         * shows that it makes things worse for diff-tree comparing
2662         * two linux-2.6 kernel trees in an already checked out work
2663         * tree.  This is because most diff-tree comparisons deal with
2664         * only a small number of files, while reading the cache is
2665         * expensive for a large project, and its cost outweighs the
2666         * savings we get by not inflating the object to a temporary
2667         * file.  Practically, this code only helps when we are used
2668         * by diff-cache --cached, which does read the cache before
2669         * calling us.
2670         */
2671        if (!active_cache)
2672                return 0;
2673
2674        /* We want to avoid the working directory if our caller
2675         * doesn't need the data in a normal file, this system
2676         * is rather slow with its stat/open/mmap/close syscalls,
2677         * and the object is contained in a pack file.  The pack
2678         * is probably already open and will be faster to obtain
2679         * the data through than the working directory.  Loose
2680         * objects however would tend to be slower as they need
2681         * to be individually opened and inflated.
2682         */
2683        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2684                return 0;
2685
2686        /*
2687         * Similarly, if we'd have to convert the file contents anyway, that
2688         * makes the optimization not worthwhile.
2689         */
2690        if (!want_file && would_convert_to_git(name))
2691                return 0;
2692
2693        len = strlen(name);
2694        pos = cache_name_pos(name, len);
2695        if (pos < 0)
2696                return 0;
2697        ce = active_cache[pos];
2698
2699        /*
2700         * This is not the sha1 we are looking for, or
2701         * unreusable because it is not a regular file.
2702         */
2703        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2704                return 0;
2705
2706        /*
2707         * If ce is marked as "assume unchanged", there is no
2708         * guarantee that work tree matches what we are looking for.
2709         */
2710        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2711                return 0;
2712
2713        /*
2714         * If ce matches the file in the work tree, we can reuse it.
2715         */
2716        if (ce_uptodate(ce) ||
2717            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2718                return 1;
2719
2720        return 0;
2721}
2722
2723static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2724{
2725        struct strbuf buf = STRBUF_INIT;
2726        char *dirty = "";
2727
2728        /* Are we looking at the work tree? */
2729        if (s->dirty_submodule)
2730                dirty = "-dirty";
2731
2732        strbuf_addf(&buf, "Subproject commit %s%s\n",
2733                    oid_to_hex(&s->oid), dirty);
2734        s->size = buf.len;
2735        if (size_only) {
2736                s->data = NULL;
2737                strbuf_release(&buf);
2738        } else {
2739                s->data = strbuf_detach(&buf, NULL);
2740                s->should_free = 1;
2741        }
2742        return 0;
2743}
2744
2745/*
2746 * While doing rename detection and pickaxe operation, we may need to
2747 * grab the data for the blob (or file) for our own in-core comparison.
2748 * diff_filespec has data and size fields for this purpose.
2749 */
2750int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2751{
2752        int size_only = flags & CHECK_SIZE_ONLY;
2753        int err = 0;
2754        /*
2755         * demote FAIL to WARN to allow inspecting the situation
2756         * instead of refusing.
2757         */
2758        enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2759                                    ? SAFE_CRLF_WARN
2760                                    : safe_crlf);
2761
2762        if (!DIFF_FILE_VALID(s))
2763                die("internal error: asking to populate invalid file.");
2764        if (S_ISDIR(s->mode))
2765                return -1;
2766
2767        if (s->data)
2768                return 0;
2769
2770        if (size_only && 0 < s->size)
2771                return 0;
2772
2773        if (S_ISGITLINK(s->mode))
2774                return diff_populate_gitlink(s, size_only);
2775
2776        if (!s->oid_valid ||
2777            reuse_worktree_file(s->path, s->oid.hash, 0)) {
2778                struct strbuf buf = STRBUF_INIT;
2779                struct stat st;
2780                int fd;
2781
2782                if (lstat(s->path, &st) < 0) {
2783                        if (errno == ENOENT) {
2784                        err_empty:
2785                                err = -1;
2786                        empty:
2787                                s->data = (char *)"";
2788                                s->size = 0;
2789                                return err;
2790                        }
2791                }
2792                s->size = xsize_t(st.st_size);
2793                if (!s->size)
2794                        goto empty;
2795                if (S_ISLNK(st.st_mode)) {
2796                        struct strbuf sb = STRBUF_INIT;
2797
2798                        if (strbuf_readlink(&sb, s->path, s->size))
2799                                goto err_empty;
2800                        s->size = sb.len;
2801                        s->data = strbuf_detach(&sb, NULL);
2802                        s->should_free = 1;
2803                        return 0;
2804                }
2805                if (size_only)
2806                        return 0;
2807                if ((flags & CHECK_BINARY) &&
2808                    s->size > big_file_threshold && s->is_binary == -1) {
2809                        s->is_binary = 1;
2810                        return 0;
2811                }
2812                fd = open(s->path, O_RDONLY);
2813                if (fd < 0)
2814                        goto err_empty;
2815                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2816                close(fd);
2817                s->should_munmap = 1;
2818
2819                /*
2820                 * Convert from working tree format to canonical git format
2821                 */
2822                if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2823                        size_t size = 0;
2824                        munmap(s->data, s->size);
2825                        s->should_munmap = 0;
2826                        s->data = strbuf_detach(&buf, &size);
2827                        s->size = size;
2828                        s->should_free = 1;
2829                }
2830        }
2831        else {
2832                enum object_type type;
2833                if (size_only || (flags & CHECK_BINARY)) {
2834                        type = sha1_object_info(s->oid.hash, &s->size);
2835                        if (type < 0)
2836                                die("unable to read %s",
2837                                    oid_to_hex(&s->oid));
2838                        if (size_only)
2839                                return 0;
2840                        if (s->size > big_file_threshold && s->is_binary == -1) {
2841                                s->is_binary = 1;
2842                                return 0;
2843                        }
2844                }
2845                s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2846                if (!s->data)
2847                        die("unable to read %s", oid_to_hex(&s->oid));
2848                s->should_free = 1;
2849        }
2850        return 0;
2851}
2852
2853void diff_free_filespec_blob(struct diff_filespec *s)
2854{
2855        if (s->should_free)
2856                free(s->data);
2857        else if (s->should_munmap)
2858                munmap(s->data, s->size);
2859
2860        if (s->should_free || s->should_munmap) {
2861                s->should_free = s->should_munmap = 0;
2862                s->data = NULL;
2863        }
2864}
2865
2866void diff_free_filespec_data(struct diff_filespec *s)
2867{
2868        diff_free_filespec_blob(s);
2869        free(s->cnt_data);
2870        s->cnt_data = NULL;
2871}
2872
2873static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2874                           void *blob,
2875                           unsigned long size,
2876                           const struct object_id *oid,
2877                           int mode)
2878{
2879        int fd;
2880        struct strbuf buf = STRBUF_INIT;
2881        struct strbuf template = STRBUF_INIT;
2882        char *path_dup = xstrdup(path);
2883        const char *base = basename(path_dup);
2884
2885        /* Generate "XXXXXX_basename.ext" */
2886        strbuf_addstr(&template, "XXXXXX_");
2887        strbuf_addstr(&template, base);
2888
2889        fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2890        if (fd < 0)
2891                die_errno("unable to create temp-file");
2892        if (convert_to_working_tree(path,
2893                        (const char *)blob, (size_t)size, &buf)) {
2894                blob = buf.buf;
2895                size = buf.len;
2896        }
2897        if (write_in_full(fd, blob, size) != size)
2898                die_errno("unable to write temp-file");
2899        close_tempfile(&temp->tempfile);
2900        temp->name = get_tempfile_path(&temp->tempfile);
2901        oid_to_hex_r(temp->hex, oid);
2902        xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
2903        strbuf_release(&buf);
2904        strbuf_release(&template);
2905        free(path_dup);
2906}
2907
2908static struct diff_tempfile *prepare_temp_file(const char *name,
2909                struct diff_filespec *one)
2910{
2911        struct diff_tempfile *temp = claim_diff_tempfile();
2912
2913        if (!DIFF_FILE_VALID(one)) {
2914        not_a_valid_file:
2915                /* A '-' entry produces this for file-2, and
2916                 * a '+' entry produces this for file-1.
2917                 */
2918                temp->name = "/dev/null";
2919                xsnprintf(temp->hex, sizeof(temp->hex), ".");
2920                xsnprintf(temp->mode, sizeof(temp->mode), ".");
2921                return temp;
2922        }
2923
2924        if (!S_ISGITLINK(one->mode) &&
2925            (!one->oid_valid ||
2926             reuse_worktree_file(name, one->oid.hash, 1))) {
2927                struct stat st;
2928                if (lstat(name, &st) < 0) {
2929                        if (errno == ENOENT)
2930                                goto not_a_valid_file;
2931                        die_errno("stat(%s)", name);
2932                }
2933                if (S_ISLNK(st.st_mode)) {
2934                        struct strbuf sb = STRBUF_INIT;
2935                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2936                                die_errno("readlink(%s)", name);
2937                        prep_temp_blob(name, temp, sb.buf, sb.len,
2938                                       (one->oid_valid ?
2939                                        &one->oid : &null_oid),
2940                                       (one->oid_valid ?
2941                                        one->mode : S_IFLNK));
2942                        strbuf_release(&sb);
2943                }
2944                else {
2945                        /* we can borrow from the file in the work tree */
2946                        temp->name = name;
2947                        if (!one->oid_valid)
2948                                sha1_to_hex_r(temp->hex, null_sha1);
2949                        else
2950                                sha1_to_hex_r(temp->hex, one->oid.hash);
2951                        /* Even though we may sometimes borrow the
2952                         * contents from the work tree, we always want
2953                         * one->mode.  mode is trustworthy even when
2954                         * !(one->sha1_valid), as long as
2955                         * DIFF_FILE_VALID(one).
2956                         */
2957                        xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
2958                }
2959                return temp;
2960        }
2961        else {
2962                if (diff_populate_filespec(one, 0))
2963                        die("cannot read data blob for %s", one->path);
2964                prep_temp_blob(name, temp, one->data, one->size,
2965                               &one->oid, one->mode);
2966        }
2967        return temp;
2968}
2969
2970static void add_external_diff_name(struct argv_array *argv,
2971                                   const char *name,
2972                                   struct diff_filespec *df)
2973{
2974        struct diff_tempfile *temp = prepare_temp_file(name, df);
2975        argv_array_push(argv, temp->name);
2976        argv_array_push(argv, temp->hex);
2977        argv_array_push(argv, temp->mode);
2978}
2979
2980/* An external diff command takes:
2981 *
2982 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2983 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2984 *
2985 */
2986static void run_external_diff(const char *pgm,
2987                              const char *name,
2988                              const char *other,
2989                              struct diff_filespec *one,
2990                              struct diff_filespec *two,
2991                              const char *xfrm_msg,
2992                              int complete_rewrite,
2993                              struct diff_options *o)
2994{
2995        struct argv_array argv = ARGV_ARRAY_INIT;
2996        struct argv_array env = ARGV_ARRAY_INIT;
2997        struct diff_queue_struct *q = &diff_queued_diff;
2998
2999        argv_array_push(&argv, pgm);
3000        argv_array_push(&argv, name);
3001
3002        if (one && two) {
3003                add_external_diff_name(&argv, name, one);
3004                if (!other)
3005                        add_external_diff_name(&argv, name, two);
3006                else {
3007                        add_external_diff_name(&argv, other, two);
3008                        argv_array_push(&argv, other);
3009                        argv_array_push(&argv, xfrm_msg);
3010                }
3011        }
3012
3013        argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3014        argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3015
3016        if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3017                die(_("external diff died, stopping at %s"), name);
3018
3019        remove_tempfile();
3020        argv_array_clear(&argv);
3021        argv_array_clear(&env);
3022}
3023
3024static int similarity_index(struct diff_filepair *p)
3025{
3026        return p->score * 100 / MAX_SCORE;
3027}
3028
3029static void fill_metainfo(struct strbuf *msg,
3030                          const char *name,
3031                          const char *other,
3032                          struct diff_filespec *one,
3033                          struct diff_filespec *two,
3034                          struct diff_options *o,
3035                          struct diff_filepair *p,
3036                          int *must_show_header,
3037                          int use_color)
3038{
3039        const char *set = diff_get_color(use_color, DIFF_METAINFO);
3040        const char *reset = diff_get_color(use_color, DIFF_RESET);
3041        const char *line_prefix = diff_line_prefix(o);
3042
3043        *must_show_header = 1;
3044        strbuf_init(msg, PATH_MAX * 2 + 300);
3045        switch (p->status) {
3046        case DIFF_STATUS_COPIED:
3047                strbuf_addf(msg, "%s%ssimilarity index %d%%",
3048                            line_prefix, set, similarity_index(p));
3049                strbuf_addf(msg, "%s\n%s%scopy from ",
3050                            reset,  line_prefix, set);
3051                quote_c_style(name, msg, NULL, 0);
3052                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3053                quote_c_style(other, msg, NULL, 0);
3054                strbuf_addf(msg, "%s\n", reset);
3055                break;
3056        case DIFF_STATUS_RENAMED:
3057                strbuf_addf(msg, "%s%ssimilarity index %d%%",
3058                            line_prefix, set, similarity_index(p));
3059                strbuf_addf(msg, "%s\n%s%srename from ",
3060                            reset, line_prefix, set);
3061                quote_c_style(name, msg, NULL, 0);
3062                strbuf_addf(msg, "%s\n%s%srename to ",
3063                            reset, line_prefix, set);
3064                quote_c_style(other, msg, NULL, 0);
3065                strbuf_addf(msg, "%s\n", reset);
3066                break;
3067        case DIFF_STATUS_MODIFIED:
3068                if (p->score) {
3069                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3070                                    line_prefix,
3071                                    set, similarity_index(p), reset);
3072                        break;
3073                }
3074                /* fallthru */
3075        default:
3076                *must_show_header = 0;
3077        }
3078        if (one && two && oidcmp(&one->oid, &two->oid)) {
3079                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3080
3081                if (DIFF_OPT_TST(o, BINARY)) {
3082                        mmfile_t mf;
3083                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3084                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3085                                abbrev = 40;
3086                }
3087                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3088                            find_unique_abbrev(one->oid.hash, abbrev));
3089                strbuf_addstr(msg, find_unique_abbrev(two->oid.hash, abbrev));
3090                if (one->mode == two->mode)
3091                        strbuf_addf(msg, " %06o", one->mode);
3092                strbuf_addf(msg, "%s\n", reset);
3093        }
3094}
3095
3096static void run_diff_cmd(const char *pgm,
3097                         const char *name,
3098                         const char *other,
3099                         const char *attr_path,
3100                         struct diff_filespec *one,
3101                         struct diff_filespec *two,
3102                         struct strbuf *msg,
3103                         struct diff_options *o,
3104                         struct diff_filepair *p)
3105{
3106        const char *xfrm_msg = NULL;
3107        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3108        int must_show_header = 0;
3109
3110
3111        if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3112                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3113                if (drv && drv->external)
3114                        pgm = drv->external;
3115        }
3116
3117        if (msg) {
3118                /*
3119                 * don't use colors when the header is intended for an
3120                 * external diff driver
3121                 */
3122                fill_metainfo(msg, name, other, one, two, o, p,
3123                              &must_show_header,
3124                              want_color(o->use_color) && !pgm);
3125                xfrm_msg = msg->len ? msg->buf : NULL;
3126        }
3127
3128        if (pgm) {
3129                run_external_diff(pgm, name, other, one, two, xfrm_msg,
3130                                  complete_rewrite, o);
3131                return;
3132        }
3133        if (one && two)
3134                builtin_diff(name, other ? other : name,
3135                             one, two, xfrm_msg, must_show_header,
3136                             o, complete_rewrite);
3137        else
3138                fprintf(o->file, "* Unmerged path %s\n", name);
3139}
3140
3141static void diff_fill_sha1_info(struct diff_filespec *one)
3142{
3143        if (DIFF_FILE_VALID(one)) {
3144                if (!one->oid_valid) {
3145                        struct stat st;
3146                        if (one->is_stdin) {
3147                                oidclr(&one->oid);
3148                                return;
3149                        }
3150                        if (lstat(one->path, &st) < 0)
3151                                die_errno("stat '%s'", one->path);
3152                        if (index_path(one->oid.hash, one->path, &st, 0))
3153                                die("cannot hash %s", one->path);
3154                }
3155        }
3156        else
3157                oidclr(&one->oid);
3158}
3159
3160static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3161{
3162        /* Strip the prefix but do not molest /dev/null and absolute paths */
3163        if (*namep && **namep != '/') {
3164                *namep += prefix_length;
3165                if (**namep == '/')
3166                        ++*namep;
3167        }
3168        if (*otherp && **otherp != '/') {
3169                *otherp += prefix_length;
3170                if (**otherp == '/')
3171                        ++*otherp;
3172        }
3173}
3174
3175static void run_diff(struct diff_filepair *p, struct diff_options *o)
3176{
3177        const char *pgm = external_diff();
3178        struct strbuf msg;
3179        struct diff_filespec *one = p->one;
3180        struct diff_filespec *two = p->two;
3181        const char *name;
3182        const char *other;
3183        const char *attr_path;
3184
3185        name  = p->one->path;
3186        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3187        attr_path = name;
3188        if (o->prefix_length)
3189                strip_prefix(o->prefix_length, &name, &other);
3190
3191        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3192                pgm = NULL;
3193
3194        if (DIFF_PAIR_UNMERGED(p)) {
3195                run_diff_cmd(pgm, name, NULL, attr_path,
3196                             NULL, NULL, NULL, o, p);
3197                return;
3198        }
3199
3200        diff_fill_sha1_info(one);
3201        diff_fill_sha1_info(two);
3202
3203        if (!pgm &&
3204            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3205            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3206                /*
3207                 * a filepair that changes between file and symlink
3208                 * needs to be split into deletion and creation.
3209                 */
3210                struct diff_filespec *null = alloc_filespec(two->path);
3211                run_diff_cmd(NULL, name, other, attr_path,
3212                             one, null, &msg, o, p);
3213                free(null);
3214                strbuf_release(&msg);
3215
3216                null = alloc_filespec(one->path);
3217                run_diff_cmd(NULL, name, other, attr_path,
3218                             null, two, &msg, o, p);
3219                free(null);
3220        }
3221        else
3222                run_diff_cmd(pgm, name, other, attr_path,
3223                             one, two, &msg, o, p);
3224
3225        strbuf_release(&msg);
3226}
3227
3228static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3229                         struct diffstat_t *diffstat)
3230{
3231        const char *name;
3232        const char *other;
3233
3234        if (DIFF_PAIR_UNMERGED(p)) {
3235                /* unmerged */
3236                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3237                return;
3238        }
3239
3240        name = p->one->path;
3241        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3242
3243        if (o->prefix_length)
3244                strip_prefix(o->prefix_length, &name, &other);
3245
3246        diff_fill_sha1_info(p->one);
3247        diff_fill_sha1_info(p->two);
3248
3249        builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3250}
3251
3252static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3253{
3254        const char *name;
3255        const char *other;
3256        const char *attr_path;
3257
3258        if (DIFF_PAIR_UNMERGED(p)) {
3259                /* unmerged */
3260                return;
3261        }
3262
3263        name = p->one->path;
3264        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3265        attr_path = other ? other : name;
3266
3267        if (o->prefix_length)
3268                strip_prefix(o->prefix_length, &name, &other);
3269
3270        diff_fill_sha1_info(p->one);
3271        diff_fill_sha1_info(p->two);
3272
3273        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3274}
3275
3276void diff_setup(struct diff_options *options)
3277{
3278        memcpy(options, &default_diff_options, sizeof(*options));
3279
3280        options->file = stdout;
3281
3282        options->line_termination = '\n';
3283        options->break_opt = -1;
3284        options->rename_limit = -1;
3285        options->dirstat_permille = diff_dirstat_permille_default;
3286        options->context = diff_context_default;
3287        options->ws_error_highlight = WSEH_NEW;
3288        DIFF_OPT_SET(options, RENAME_EMPTY);
3289
3290        /* pathchange left =NULL by default */
3291        options->change = diff_change;
3292        options->add_remove = diff_addremove;
3293        options->use_color = diff_use_color_default;
3294        options->detect_rename = diff_detect_rename_default;
3295        options->xdl_opts |= diff_algorithm;
3296        if (diff_compaction_heuristic)
3297                DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3298
3299        options->orderfile = diff_order_file_cfg;
3300
3301        if (diff_no_prefix) {
3302                options->a_prefix = options->b_prefix = "";
3303        } else if (!diff_mnemonic_prefix) {
3304                options->a_prefix = "a/";
3305                options->b_prefix = "b/";
3306        }
3307}
3308
3309void diff_setup_done(struct diff_options *options)
3310{
3311        int count = 0;
3312
3313        if (options->set_default)
3314                options->set_default(options);
3315
3316        if (options->output_format & DIFF_FORMAT_NAME)
3317                count++;
3318        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3319                count++;
3320        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3321                count++;
3322        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3323                count++;
3324        if (count > 1)
3325                die("--name-only, --name-status, --check and -s are mutually exclusive");
3326
3327        /*
3328         * Most of the time we can say "there are changes"
3329         * only by checking if there are changed paths, but
3330         * --ignore-whitespace* options force us to look
3331         * inside contents.
3332         */
3333
3334        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3335            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3336            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3337                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3338        else
3339                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3340
3341        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3342                options->detect_rename = DIFF_DETECT_COPY;
3343
3344        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3345                options->prefix = NULL;
3346        if (options->prefix)
3347                options->prefix_length = strlen(options->prefix);
3348        else
3349                options->prefix_length = 0;
3350
3351        if (options->output_format & (DIFF_FORMAT_NAME |
3352                                      DIFF_FORMAT_NAME_STATUS |
3353                                      DIFF_FORMAT_CHECKDIFF |
3354                                      DIFF_FORMAT_NO_OUTPUT))
3355                options->output_format &= ~(DIFF_FORMAT_RAW |
3356                                            DIFF_FORMAT_NUMSTAT |
3357                                            DIFF_FORMAT_DIFFSTAT |
3358                                            DIFF_FORMAT_SHORTSTAT |
3359                                            DIFF_FORMAT_DIRSTAT |
3360                                            DIFF_FORMAT_SUMMARY |
3361                                            DIFF_FORMAT_PATCH);
3362
3363        /*
3364         * These cases always need recursive; we do not drop caller-supplied
3365         * recursive bits for other formats here.
3366         */
3367        if (options->output_format & (DIFF_FORMAT_PATCH |
3368                                      DIFF_FORMAT_NUMSTAT |
3369                                      DIFF_FORMAT_DIFFSTAT |
3370                                      DIFF_FORMAT_SHORTSTAT |
3371                                      DIFF_FORMAT_DIRSTAT |
3372                                      DIFF_FORMAT_SUMMARY |
3373                                      DIFF_FORMAT_CHECKDIFF))
3374                DIFF_OPT_SET(options, RECURSIVE);
3375        /*
3376         * Also pickaxe would not work very well if you do not say recursive
3377         */
3378        if (options->pickaxe)
3379                DIFF_OPT_SET(options, RECURSIVE);
3380        /*
3381         * When patches are generated, submodules diffed against the work tree
3382         * must be checked for dirtiness too so it can be shown in the output
3383         */
3384        if (options->output_format & DIFF_FORMAT_PATCH)
3385                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3386
3387        if (options->detect_rename && options->rename_limit < 0)
3388                options->rename_limit = diff_rename_limit_default;
3389        if (options->setup & DIFF_SETUP_USE_CACHE) {
3390                if (!active_cache)
3391                        /* read-cache does not die even when it fails
3392                         * so it is safe for us to do this here.  Also
3393                         * it does not smudge active_cache or active_nr
3394                         * when it fails, so we do not have to worry about
3395                         * cleaning it up ourselves either.
3396                         */
3397                        read_cache();
3398        }
3399        if (options->abbrev <= 0 || 40 < options->abbrev)
3400                options->abbrev = 40; /* full */
3401
3402        /*
3403         * It does not make sense to show the first hit we happened
3404         * to have found.  It does not make sense not to return with
3405         * exit code in such a case either.
3406         */
3407        if (DIFF_OPT_TST(options, QUICK)) {
3408                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3409                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3410        }
3411
3412        options->diff_path_counter = 0;
3413
3414        if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3415                die(_("--follow requires exactly one pathspec"));
3416}
3417
3418static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3419{
3420        char c, *eq;
3421        int len;
3422
3423        if (*arg != '-')
3424                return 0;
3425        c = *++arg;
3426        if (!c)
3427                return 0;
3428        if (c == arg_short) {
3429                c = *++arg;
3430                if (!c)
3431                        return 1;
3432                if (val && isdigit(c)) {
3433                        char *end;
3434                        int n = strtoul(arg, &end, 10);
3435                        if (*end)
3436                                return 0;
3437                        *val = n;
3438                        return 1;
3439                }
3440                return 0;
3441        }
3442        if (c != '-')
3443                return 0;
3444        arg++;
3445        eq = strchrnul(arg, '=');
3446        len = eq - arg;
3447        if (!len || strncmp(arg, arg_long, len))
3448                return 0;
3449        if (*eq) {
3450                int n;
3451                char *end;
3452                if (!isdigit(*++eq))
3453                        return 0;
3454                n = strtoul(eq, &end, 10);
3455                if (*end)
3456                        return 0;
3457                *val = n;
3458        }
3459        return 1;
3460}
3461
3462static int diff_scoreopt_parse(const char *opt);
3463
3464static inline int short_opt(char opt, const char **argv,
3465                            const char **optarg)
3466{
3467        const char *arg = argv[0];
3468        if (arg[0] != '-' || arg[1] != opt)
3469                return 0;
3470        if (arg[2] != '\0') {
3471                *optarg = arg + 2;
3472                return 1;
3473        }
3474        if (!argv[1])
3475                die("Option '%c' requires a value", opt);
3476        *optarg = argv[1];
3477        return 2;
3478}
3479
3480int parse_long_opt(const char *opt, const char **argv,
3481                   const char **optarg)
3482{
3483        const char *arg = argv[0];
3484        if (!skip_prefix(arg, "--", &arg))
3485                return 0;
3486        if (!skip_prefix(arg, opt, &arg))
3487                return 0;
3488        if (*arg == '=') { /* stuck form: --option=value */
3489                *optarg = arg + 1;
3490                return 1;
3491        }
3492        if (*arg != '\0')
3493                return 0;
3494        /* separate form: --option value */
3495        if (!argv[1])
3496                die("Option '--%s' requires a value", opt);
3497        *optarg = argv[1];
3498        return 2;
3499}
3500
3501static int stat_opt(struct diff_options *options, const char **av)
3502{
3503        const char *arg = av[0];
3504        char *end;
3505        int width = options->stat_width;
3506        int name_width = options->stat_name_width;
3507        int graph_width = options->stat_graph_width;
3508        int count = options->stat_count;
3509        int argcount = 1;
3510
3511        if (!skip_prefix(arg, "--stat", &arg))
3512                die("BUG: stat option does not begin with --stat: %s", arg);
3513        end = (char *)arg;
3514
3515        switch (*arg) {
3516        case '-':
3517                if (skip_prefix(arg, "-width", &arg)) {
3518                        if (*arg == '=')
3519                                width = strtoul(arg + 1, &end, 10);
3520                        else if (!*arg && !av[1])
3521                                die("Option '--stat-width' requires a value");
3522                        else if (!*arg) {
3523                                width = strtoul(av[1], &end, 10);
3524                                argcount = 2;
3525                        }
3526                } else if (skip_prefix(arg, "-name-width", &arg)) {
3527                        if (*arg == '=')
3528                                name_width = strtoul(arg + 1, &end, 10);
3529                        else if (!*arg && !av[1])
3530                                die("Option '--stat-name-width' requires a value");
3531                        else if (!*arg) {
3532                                name_width = strtoul(av[1], &end, 10);
3533                                argcount = 2;
3534                        }
3535                } else if (skip_prefix(arg, "-graph-width", &arg)) {
3536                        if (*arg == '=')
3537                                graph_width = strtoul(arg + 1, &end, 10);
3538                        else if (!*arg && !av[1])
3539                                die("Option '--stat-graph-width' requires a value");
3540                        else if (!*arg) {
3541                                graph_width = strtoul(av[1], &end, 10);
3542                                argcount = 2;
3543                        }
3544                } else if (skip_prefix(arg, "-count", &arg)) {
3545                        if (*arg == '=')
3546                                count = strtoul(arg + 1, &end, 10);
3547                        else if (!*arg && !av[1])
3548                                die("Option '--stat-count' requires a value");
3549                        else if (!*arg) {
3550                                count = strtoul(av[1], &end, 10);
3551                                argcount = 2;
3552                        }
3553                }
3554                break;
3555        case '=':
3556                width = strtoul(arg+1, &end, 10);
3557                if (*end == ',')
3558                        name_width = strtoul(end+1, &end, 10);
3559                if (*end == ',')
3560                        count = strtoul(end+1, &end, 10);
3561        }
3562
3563        /* Important! This checks all the error cases! */
3564        if (*end)
3565                return 0;
3566        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3567        options->stat_name_width = name_width;
3568        options->stat_graph_width = graph_width;
3569        options->stat_width = width;
3570        options->stat_count = count;
3571        return argcount;
3572}
3573
3574static int parse_dirstat_opt(struct diff_options *options, const char *params)
3575{
3576        struct strbuf errmsg = STRBUF_INIT;
3577        if (parse_dirstat_params(options, params, &errmsg))
3578                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3579                    errmsg.buf);
3580        strbuf_release(&errmsg);
3581        /*
3582         * The caller knows a dirstat-related option is given from the command
3583         * line; allow it to say "return this_function();"
3584         */
3585        options->output_format |= DIFF_FORMAT_DIRSTAT;
3586        return 1;
3587}
3588
3589static int parse_submodule_opt(struct diff_options *options, const char *value)
3590{
3591        if (parse_submodule_params(options, value))
3592                die(_("Failed to parse --submodule option parameter: '%s'"),
3593                        value);
3594        return 1;
3595}
3596
3597static const char diff_status_letters[] = {
3598        DIFF_STATUS_ADDED,
3599        DIFF_STATUS_COPIED,
3600        DIFF_STATUS_DELETED,
3601        DIFF_STATUS_MODIFIED,
3602        DIFF_STATUS_RENAMED,
3603        DIFF_STATUS_TYPE_CHANGED,
3604        DIFF_STATUS_UNKNOWN,
3605        DIFF_STATUS_UNMERGED,
3606        DIFF_STATUS_FILTER_AON,
3607        DIFF_STATUS_FILTER_BROKEN,
3608        '\0',
3609};
3610
3611static unsigned int filter_bit['Z' + 1];
3612
3613static void prepare_filter_bits(void)
3614{
3615        int i;
3616
3617        if (!filter_bit[DIFF_STATUS_ADDED]) {
3618                for (i = 0; diff_status_letters[i]; i++)
3619                        filter_bit[(int) diff_status_letters[i]] = (1 << i);
3620        }
3621}
3622
3623static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3624{
3625        return opt->filter & filter_bit[(int) status];
3626}
3627
3628static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3629{
3630        int i, optch;
3631
3632        prepare_filter_bits();
3633
3634        /*
3635         * If there is a negation e.g. 'd' in the input, and we haven't
3636         * initialized the filter field with another --diff-filter, start
3637         * from full set of bits, except for AON.
3638         */
3639        if (!opt->filter) {
3640                for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3641                        if (optch < 'a' || 'z' < optch)
3642                                continue;
3643                        opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3644                        opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3645                        break;
3646                }
3647        }
3648
3649        for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3650                unsigned int bit;
3651                int negate;
3652
3653                if ('a' <= optch && optch <= 'z') {
3654                        negate = 1;
3655                        optch = toupper(optch);
3656                } else {
3657                        negate = 0;
3658                }
3659
3660                bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3661                if (!bit)
3662                        return optarg[i];
3663                if (negate)
3664                        opt->filter &= ~bit;
3665                else
3666                        opt->filter |= bit;
3667        }
3668        return 0;
3669}
3670
3671static void enable_patch_output(int *fmt) {
3672        *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3673        *fmt |= DIFF_FORMAT_PATCH;
3674}
3675
3676static int parse_one_token(const char **arg, const char *token)
3677{
3678        const char *rest;
3679        if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
3680                *arg = rest;
3681                return 1;
3682        }
3683        return 0;
3684}
3685
3686static int parse_ws_error_highlight(struct diff_options *opt, const char *arg)
3687{
3688        const char *orig_arg = arg;
3689        unsigned val = 0;
3690        while (*arg) {
3691                if (parse_one_token(&arg, "none"))
3692                        val = 0;
3693                else if (parse_one_token(&arg, "default"))
3694                        val = WSEH_NEW;
3695                else if (parse_one_token(&arg, "all"))
3696                        val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
3697                else if (parse_one_token(&arg, "new"))
3698                        val |= WSEH_NEW;
3699                else if (parse_one_token(&arg, "old"))
3700                        val |= WSEH_OLD;
3701                else if (parse_one_token(&arg, "context"))
3702                        val |= WSEH_CONTEXT;
3703                else {
3704                        error("unknown value after ws-error-highlight=%.*s",
3705                              (int)(arg - orig_arg), orig_arg);
3706                        return 0;
3707                }
3708                if (*arg)
3709                        arg++;
3710        }
3711        opt->ws_error_highlight = val;
3712        return 1;
3713}
3714
3715int diff_opt_parse(struct diff_options *options,
3716                   const char **av, int ac, const char *prefix)
3717{
3718        const char *arg = av[0];
3719        const char *optarg;
3720        int argcount;
3721
3722        if (!prefix)
3723                prefix = "";
3724
3725        /* Output format options */
3726        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3727            || opt_arg(arg, 'U', "unified", &options->context))
3728                enable_patch_output(&options->output_format);
3729        else if (!strcmp(arg, "--raw"))
3730                options->output_format |= DIFF_FORMAT_RAW;
3731        else if (!strcmp(arg, "--patch-with-raw")) {
3732                enable_patch_output(&options->output_format);
3733                options->output_format |= DIFF_FORMAT_RAW;
3734        } else if (!strcmp(arg, "--numstat"))
3735                options->output_format |= DIFF_FORMAT_NUMSTAT;
3736        else if (!strcmp(arg, "--shortstat"))
3737                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3738        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3739                return parse_dirstat_opt(options, "");
3740        else if (skip_prefix(arg, "-X", &arg))
3741                return parse_dirstat_opt(options, arg);
3742        else if (skip_prefix(arg, "--dirstat=", &arg))
3743                return parse_dirstat_opt(options, arg);
3744        else if (!strcmp(arg, "--cumulative"))
3745                return parse_dirstat_opt(options, "cumulative");
3746        else if (!strcmp(arg, "--dirstat-by-file"))
3747                return parse_dirstat_opt(options, "files");
3748        else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3749                parse_dirstat_opt(options, "files");
3750                return parse_dirstat_opt(options, arg);
3751        }
3752        else if (!strcmp(arg, "--check"))
3753                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3754        else if (!strcmp(arg, "--summary"))
3755                options->output_format |= DIFF_FORMAT_SUMMARY;
3756        else if (!strcmp(arg, "--patch-with-stat")) {
3757                enable_patch_output(&options->output_format);
3758                options->output_format |= DIFF_FORMAT_DIFFSTAT;
3759        } else if (!strcmp(arg, "--name-only"))
3760                options->output_format |= DIFF_FORMAT_NAME;
3761        else if (!strcmp(arg, "--name-status"))
3762                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3763        else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3764                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3765        else if (starts_with(arg, "--stat"))
3766                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3767                return stat_opt(options, av);
3768
3769        /* renames options */
3770        else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3771                 !strcmp(arg, "--break-rewrites")) {
3772                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3773                        return error("invalid argument to -B: %s", arg+2);
3774        }
3775        else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3776                 !strcmp(arg, "--find-renames")) {
3777                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3778                        return error("invalid argument to -M: %s", arg+2);
3779                options->detect_rename = DIFF_DETECT_RENAME;
3780        }
3781        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3782                options->irreversible_delete = 1;
3783        }
3784        else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3785                 !strcmp(arg, "--find-copies")) {
3786                if (options->detect_rename == DIFF_DETECT_COPY)
3787                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3788                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3789                        return error("invalid argument to -C: %s", arg+2);
3790                options->detect_rename = DIFF_DETECT_COPY;
3791        }
3792        else if (!strcmp(arg, "--no-renames"))
3793                options->detect_rename = 0;
3794        else if (!strcmp(arg, "--rename-empty"))
3795                DIFF_OPT_SET(options, RENAME_EMPTY);
3796        else if (!strcmp(arg, "--no-rename-empty"))
3797                DIFF_OPT_CLR(options, RENAME_EMPTY);
3798        else if (!strcmp(arg, "--relative"))
3799                DIFF_OPT_SET(options, RELATIVE_NAME);
3800        else if (skip_prefix(arg, "--relative=", &arg)) {
3801                DIFF_OPT_SET(options, RELATIVE_NAME);
3802                options->prefix = arg;
3803        }
3804
3805        /* xdiff options */
3806        else if (!strcmp(arg, "--minimal"))
3807                DIFF_XDL_SET(options, NEED_MINIMAL);
3808        else if (!strcmp(arg, "--no-minimal"))
3809                DIFF_XDL_CLR(options, NEED_MINIMAL);
3810        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3811                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3812        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3813                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3814        else if (!strcmp(arg, "--ignore-space-at-eol"))
3815                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3816        else if (!strcmp(arg, "--ignore-blank-lines"))
3817                DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3818        else if (!strcmp(arg, "--compaction-heuristic"))
3819                DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3820        else if (!strcmp(arg, "--no-compaction-heuristic"))
3821                DIFF_XDL_CLR(options, COMPACTION_HEURISTIC);
3822        else if (!strcmp(arg, "--patience"))
3823                options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3824        else if (!strcmp(arg, "--histogram"))
3825                options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3826        else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3827                long value = parse_algorithm_value(optarg);
3828                if (value < 0)
3829                        return error("option diff-algorithm accepts \"myers\", "
3830                                     "\"minimal\", \"patience\" and \"histogram\"");
3831                /* clear out previous settings */
3832                DIFF_XDL_CLR(options, NEED_MINIMAL);
3833                options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3834                options->xdl_opts |= value;
3835                return argcount;
3836        }
3837
3838        /* flags options */
3839        else if (!strcmp(arg, "--binary")) {
3840                enable_patch_output(&options->output_format);
3841                DIFF_OPT_SET(options, BINARY);
3842        }
3843        else if (!strcmp(arg, "--full-index"))
3844                DIFF_OPT_SET(options, FULL_INDEX);
3845        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3846                DIFF_OPT_SET(options, TEXT);
3847        else if (!strcmp(arg, "-R"))
3848                DIFF_OPT_SET(options, REVERSE_DIFF);
3849        else if (!strcmp(arg, "--find-copies-harder"))
3850                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3851        else if (!strcmp(arg, "--follow"))
3852                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3853        else if (!strcmp(arg, "--no-follow")) {
3854                DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3855                DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3856        } else if (!strcmp(arg, "--color"))
3857                options->use_color = 1;
3858        else if (skip_prefix(arg, "--color=", &arg)) {
3859                int value = git_config_colorbool(NULL, arg);
3860                if (value < 0)
3861                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3862                options->use_color = value;
3863        }
3864        else if (!strcmp(arg, "--no-color"))
3865                options->use_color = 0;
3866        else if (!strcmp(arg, "--color-words")) {
3867                options->use_color = 1;
3868                options->word_diff = DIFF_WORDS_COLOR;
3869        }
3870        else if (skip_prefix(arg, "--color-words=", &arg)) {
3871                options->use_color = 1;
3872                options->word_diff = DIFF_WORDS_COLOR;
3873                options->word_regex = arg;
3874        }
3875        else if (!strcmp(arg, "--word-diff")) {
3876                if (options->word_diff == DIFF_WORDS_NONE)
3877                        options->word_diff = DIFF_WORDS_PLAIN;
3878        }
3879        else if (skip_prefix(arg, "--word-diff=", &arg)) {
3880                if (!strcmp(arg, "plain"))
3881                        options->word_diff = DIFF_WORDS_PLAIN;
3882                else if (!strcmp(arg, "color")) {
3883                        options->use_color = 1;
3884                        options->word_diff = DIFF_WORDS_COLOR;
3885                }
3886                else if (!strcmp(arg, "porcelain"))
3887                        options->word_diff = DIFF_WORDS_PORCELAIN;
3888                else if (!strcmp(arg, "none"))
3889                        options->word_diff = DIFF_WORDS_NONE;
3890                else
3891                        die("bad --word-diff argument: %s", arg);
3892        }
3893        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3894                if (options->word_diff == DIFF_WORDS_NONE)
3895                        options->word_diff = DIFF_WORDS_PLAIN;
3896                options->word_regex = optarg;
3897                return argcount;
3898        }
3899        else if (!strcmp(arg, "--exit-code"))
3900                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3901        else if (!strcmp(arg, "--quiet"))
3902                DIFF_OPT_SET(options, QUICK);
3903        else if (!strcmp(arg, "--ext-diff"))
3904                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3905        else if (!strcmp(arg, "--no-ext-diff"))
3906                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3907        else if (!strcmp(arg, "--textconv"))
3908                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3909        else if (!strcmp(arg, "--no-textconv"))
3910                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3911        else if (!strcmp(arg, "--ignore-submodules")) {
3912                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3913                handle_ignore_submodules_arg(options, "all");
3914        } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3915                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3916                handle_ignore_submodules_arg(options, arg);
3917        } else if (!strcmp(arg, "--submodule"))
3918                DIFF_OPT_SET(options, SUBMODULE_LOG);
3919        else if (skip_prefix(arg, "--submodule=", &arg))
3920                return parse_submodule_opt(options, arg);
3921        else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
3922                return parse_ws_error_highlight(options, arg);
3923
3924        /* misc options */
3925        else if (!strcmp(arg, "-z"))
3926                options->line_termination = 0;
3927        else if ((argcount = short_opt('l', av, &optarg))) {
3928                options->rename_limit = strtoul(optarg, NULL, 10);
3929                return argcount;
3930        }
3931        else if ((argcount = short_opt('S', av, &optarg))) {
3932                options->pickaxe = optarg;
3933                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3934                return argcount;
3935        } else if ((argcount = short_opt('G', av, &optarg))) {
3936                options->pickaxe = optarg;
3937                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3938                return argcount;
3939        }
3940        else if (!strcmp(arg, "--pickaxe-all"))
3941                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3942        else if (!strcmp(arg, "--pickaxe-regex"))
3943                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3944        else if ((argcount = short_opt('O', av, &optarg))) {
3945                const char *path = prefix_filename(prefix, strlen(prefix), optarg);
3946                options->orderfile = xstrdup(path);
3947                return argcount;
3948        }
3949        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3950                int offending = parse_diff_filter_opt(optarg, options);
3951                if (offending)
3952                        die("unknown change class '%c' in --diff-filter=%s",
3953                            offending, optarg);
3954                return argcount;
3955        }
3956        else if (!strcmp(arg, "--abbrev"))
3957                options->abbrev = DEFAULT_ABBREV;
3958        else if (skip_prefix(arg, "--abbrev=", &arg)) {
3959                options->abbrev = strtoul(arg, NULL, 10);
3960                if (options->abbrev < MINIMUM_ABBREV)
3961                        options->abbrev = MINIMUM_ABBREV;
3962                else if (40 < options->abbrev)
3963                        options->abbrev = 40;
3964        }
3965        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3966                options->a_prefix = optarg;
3967                return argcount;
3968        }
3969        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3970                options->b_prefix = optarg;
3971                return argcount;
3972        }
3973        else if (!strcmp(arg, "--no-prefix"))
3974                options->a_prefix = options->b_prefix = "";
3975        else if (opt_arg(arg, '\0', "inter-hunk-context",
3976                         &options->interhunkcontext))
3977                ;
3978        else if (!strcmp(arg, "-W"))
3979                DIFF_OPT_SET(options, FUNCCONTEXT);
3980        else if (!strcmp(arg, "--function-context"))
3981                DIFF_OPT_SET(options, FUNCCONTEXT);
3982        else if (!strcmp(arg, "--no-function-context"))
3983                DIFF_OPT_CLR(options, FUNCCONTEXT);
3984        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3985                const char *path = prefix_filename(prefix, strlen(prefix), optarg);
3986                options->file = fopen(path, "w");
3987                if (!options->file)
3988                        die_errno("Could not open '%s'", path);
3989                options->close_file = 1;
3990                if (options->use_color != GIT_COLOR_ALWAYS)
3991                        options->use_color = GIT_COLOR_NEVER;
3992                return argcount;
3993        } else
3994                return 0;
3995        return 1;
3996}
3997
3998int parse_rename_score(const char **cp_p)
3999{
4000        unsigned long num, scale;
4001        int ch, dot;
4002        const char *cp = *cp_p;
4003
4004        num = 0;
4005        scale = 1;
4006        dot = 0;
4007        for (;;) {
4008                ch = *cp;
4009                if ( !dot && ch == '.' ) {
4010                        scale = 1;
4011                        dot = 1;
4012                } else if ( ch == '%' ) {
4013                        scale = dot ? scale*100 : 100;
4014                        cp++;   /* % is always at the end */
4015                        break;
4016                } else if ( ch >= '0' && ch <= '9' ) {
4017                        if ( scale < 100000 ) {
4018                                scale *= 10;
4019                                num = (num*10) + (ch-'0');
4020                        }
4021                } else {
4022                        break;
4023                }
4024                cp++;
4025        }
4026        *cp_p = cp;
4027
4028        /* user says num divided by scale and we say internally that
4029         * is MAX_SCORE * num / scale.
4030         */
4031        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4032}
4033
4034static int diff_scoreopt_parse(const char *opt)
4035{
4036        int opt1, opt2, cmd;
4037
4038        if (*opt++ != '-')
4039                return -1;
4040        cmd = *opt++;
4041        if (cmd == '-') {
4042                /* convert the long-form arguments into short-form versions */
4043                if (skip_prefix(opt, "break-rewrites", &opt)) {
4044                        if (*opt == 0 || *opt++ == '=')
4045                                cmd = 'B';
4046                } else if (skip_prefix(opt, "find-copies", &opt)) {
4047                        if (*opt == 0 || *opt++ == '=')
4048                                cmd = 'C';
4049                } else if (skip_prefix(opt, "find-renames", &opt)) {
4050                        if (*opt == 0 || *opt++ == '=')
4051                                cmd = 'M';
4052                }
4053        }
4054        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4055                return -1; /* that is not a -M, -C, or -B option */
4056
4057        opt1 = parse_rename_score(&opt);
4058        if (cmd != 'B')
4059                opt2 = 0;
4060        else {
4061                if (*opt == 0)
4062                        opt2 = 0;
4063                else if (*opt != '/')
4064                        return -1; /* we expect -B80/99 or -B80 */
4065                else {
4066                        opt++;
4067                        opt2 = parse_rename_score(&opt);
4068                }
4069        }
4070        if (*opt != 0)
4071                return -1;
4072        return opt1 | (opt2 << 16);
4073}
4074
4075struct diff_queue_struct diff_queued_diff;
4076
4077void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4078{
4079        ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4080        queue->queue[queue->nr++] = dp;
4081}
4082
4083struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4084                                 struct diff_filespec *one,
4085                                 struct diff_filespec *two)
4086{
4087        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4088        dp->one = one;
4089        dp->two = two;
4090        if (queue)
4091                diff_q(queue, dp);
4092        return dp;
4093}
4094
4095void diff_free_filepair(struct diff_filepair *p)
4096{
4097        free_filespec(p->one);
4098        free_filespec(p->two);
4099        free(p);
4100}
4101
4102/* This is different from find_unique_abbrev() in that
4103 * it stuffs the result with dots for alignment.
4104 */
4105const char *diff_unique_abbrev(const unsigned char *sha1, int len)
4106{
4107        int abblen;
4108        const char *abbrev;
4109        if (len == 40)
4110                return sha1_to_hex(sha1);
4111
4112        abbrev = find_unique_abbrev(sha1, len);
4113        abblen = strlen(abbrev);
4114        if (abblen < 37) {
4115                static char hex[41];
4116                if (len < abblen && abblen <= len + 2)
4117                        xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4118                else
4119                        xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4120                return hex;
4121        }
4122        return sha1_to_hex(sha1);
4123}
4124
4125static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4126{
4127        int line_termination = opt->line_termination;
4128        int inter_name_termination = line_termination ? '\t' : '\0';
4129
4130        fprintf(opt->file, "%s", diff_line_prefix(opt));
4131        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4132                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4133                        diff_unique_abbrev(p->one->oid.hash, opt->abbrev));
4134                fprintf(opt->file, "%s ",
4135                        diff_unique_abbrev(p->two->oid.hash, opt->abbrev));
4136        }
4137        if (p->score) {
4138                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4139                        inter_name_termination);
4140        } else {
4141                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4142        }
4143
4144        if (p->status == DIFF_STATUS_COPIED ||
4145            p->status == DIFF_STATUS_RENAMED) {
4146                const char *name_a, *name_b;
4147                name_a = p->one->path;
4148                name_b = p->two->path;
4149                strip_prefix(opt->prefix_length, &name_a, &name_b);
4150                write_name_quoted(name_a, opt->file, inter_name_termination);
4151                write_name_quoted(name_b, opt->file, line_termination);
4152        } else {
4153                const char *name_a, *name_b;
4154                name_a = p->one->mode ? p->one->path : p->two->path;
4155                name_b = NULL;
4156                strip_prefix(opt->prefix_length, &name_a, &name_b);
4157                write_name_quoted(name_a, opt->file, line_termination);
4158        }
4159}
4160
4161int diff_unmodified_pair(struct diff_filepair *p)
4162{
4163        /* This function is written stricter than necessary to support
4164         * the currently implemented transformers, but the idea is to
4165         * let transformers to produce diff_filepairs any way they want,
4166         * and filter and clean them up here before producing the output.
4167         */
4168        struct diff_filespec *one = p->one, *two = p->two;
4169
4170        if (DIFF_PAIR_UNMERGED(p))
4171                return 0; /* unmerged is interesting */
4172
4173        /* deletion, addition, mode or type change
4174         * and rename are all interesting.
4175         */
4176        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4177            DIFF_PAIR_MODE_CHANGED(p) ||
4178            strcmp(one->path, two->path))
4179                return 0;
4180
4181        /* both are valid and point at the same path.  that is, we are
4182         * dealing with a change.
4183         */
4184        if (one->oid_valid && two->oid_valid &&
4185            !oidcmp(&one->oid, &two->oid) &&
4186            !one->dirty_submodule && !two->dirty_submodule)
4187                return 1; /* no change */
4188        if (!one->oid_valid && !two->oid_valid)
4189                return 1; /* both look at the same file on the filesystem. */
4190        return 0;
4191}
4192
4193static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4194{
4195        if (diff_unmodified_pair(p))
4196                return;
4197
4198        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4199            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4200                return; /* no tree diffs in patch format */
4201
4202        run_diff(p, o);
4203}
4204
4205static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4206                            struct diffstat_t *diffstat)
4207{
4208        if (diff_unmodified_pair(p))
4209                return;
4210
4211        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4212            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4213                return; /* no useful stat for tree diffs */
4214
4215        run_diffstat(p, o, diffstat);
4216}
4217
4218static void diff_flush_checkdiff(struct diff_filepair *p,
4219                struct diff_options *o)
4220{
4221        if (diff_unmodified_pair(p))
4222                return;
4223
4224        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4225            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4226                return; /* nothing to check in tree diffs */
4227
4228        run_checkdiff(p, o);
4229}
4230
4231int diff_queue_is_empty(void)
4232{
4233        struct diff_queue_struct *q = &diff_queued_diff;
4234        int i;
4235        for (i = 0; i < q->nr; i++)
4236                if (!diff_unmodified_pair(q->queue[i]))
4237                        return 0;
4238        return 1;
4239}
4240
4241#if DIFF_DEBUG
4242void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4243{
4244        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4245                x, one ? one : "",
4246                s->path,
4247                DIFF_FILE_VALID(s) ? "valid" : "invalid",
4248                s->mode,
4249                s->oid_valid ? oid_to_hex(&s->oid) : "");
4250        fprintf(stderr, "queue[%d] %s size %lu\n",
4251                x, one ? one : "",
4252                s->size);
4253}
4254
4255void diff_debug_filepair(const struct diff_filepair *p, int i)
4256{
4257        diff_debug_filespec(p->one, i, "one");
4258        diff_debug_filespec(p->two, i, "two");
4259        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4260                p->score, p->status ? p->status : '?',
4261                p->one->rename_used, p->broken_pair);
4262}
4263
4264void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4265{
4266        int i;
4267        if (msg)
4268                fprintf(stderr, "%s\n", msg);
4269        fprintf(stderr, "q->nr = %d\n", q->nr);
4270        for (i = 0; i < q->nr; i++) {
4271                struct diff_filepair *p = q->queue[i];
4272                diff_debug_filepair(p, i);
4273        }
4274}
4275#endif
4276
4277static void diff_resolve_rename_copy(void)
4278{
4279        int i;
4280        struct diff_filepair *p;
4281        struct diff_queue_struct *q = &diff_queued_diff;
4282
4283        diff_debug_queue("resolve-rename-copy", q);
4284
4285        for (i = 0; i < q->nr; i++) {
4286                p = q->queue[i];
4287                p->status = 0; /* undecided */
4288                if (DIFF_PAIR_UNMERGED(p))
4289                        p->status = DIFF_STATUS_UNMERGED;
4290                else if (!DIFF_FILE_VALID(p->one))
4291                        p->status = DIFF_STATUS_ADDED;
4292                else if (!DIFF_FILE_VALID(p->two))
4293                        p->status = DIFF_STATUS_DELETED;
4294                else if (DIFF_PAIR_TYPE_CHANGED(p))
4295                        p->status = DIFF_STATUS_TYPE_CHANGED;
4296
4297                /* from this point on, we are dealing with a pair
4298                 * whose both sides are valid and of the same type, i.e.
4299                 * either in-place edit or rename/copy edit.
4300                 */
4301                else if (DIFF_PAIR_RENAME(p)) {
4302                        /*
4303                         * A rename might have re-connected a broken
4304                         * pair up, causing the pathnames to be the
4305                         * same again. If so, that's not a rename at
4306                         * all, just a modification..
4307                         *
4308                         * Otherwise, see if this source was used for
4309                         * multiple renames, in which case we decrement
4310                         * the count, and call it a copy.
4311                         */
4312                        if (!strcmp(p->one->path, p->two->path))
4313                                p->status = DIFF_STATUS_MODIFIED;
4314                        else if (--p->one->rename_used > 0)
4315                                p->status = DIFF_STATUS_COPIED;
4316                        else
4317                                p->status = DIFF_STATUS_RENAMED;
4318                }
4319                else if (oidcmp(&p->one->oid, &p->two->oid) ||
4320                         p->one->mode != p->two->mode ||
4321                         p->one->dirty_submodule ||
4322                         p->two->dirty_submodule ||
4323                         is_null_oid(&p->one->oid))
4324                        p->status = DIFF_STATUS_MODIFIED;
4325                else {
4326                        /* This is a "no-change" entry and should not
4327                         * happen anymore, but prepare for broken callers.
4328                         */
4329                        error("feeding unmodified %s to diffcore",
4330                              p->one->path);
4331                        p->status = DIFF_STATUS_UNKNOWN;
4332                }
4333        }
4334        diff_debug_queue("resolve-rename-copy done", q);
4335}
4336
4337static int check_pair_status(struct diff_filepair *p)
4338{
4339        switch (p->status) {
4340        case DIFF_STATUS_UNKNOWN:
4341                return 0;
4342        case 0:
4343                die("internal error in diff-resolve-rename-copy");
4344        default:
4345                return 1;
4346        }
4347}
4348
4349static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4350{
4351        int fmt = opt->output_format;
4352
4353        if (fmt & DIFF_FORMAT_CHECKDIFF)
4354                diff_flush_checkdiff(p, opt);
4355        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4356                diff_flush_raw(p, opt);
4357        else if (fmt & DIFF_FORMAT_NAME) {
4358                const char *name_a, *name_b;
4359                name_a = p->two->path;
4360                name_b = NULL;
4361                strip_prefix(opt->prefix_length, &name_a, &name_b);
4362                write_name_quoted(name_a, opt->file, opt->line_termination);
4363        }
4364}
4365
4366static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4367{
4368        if (fs->mode)
4369                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4370        else
4371                fprintf(file, " %s ", newdelete);
4372        write_name_quoted(fs->path, file, '\n');
4373}
4374
4375
4376static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4377                const char *line_prefix)
4378{
4379        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4380                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4381                        p->two->mode, show_name ? ' ' : '\n');
4382                if (show_name) {
4383                        write_name_quoted(p->two->path, file, '\n');
4384                }
4385        }
4386}
4387
4388static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4389                        const char *line_prefix)
4390{
4391        char *names = pprint_rename(p->one->path, p->two->path);
4392
4393        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4394        free(names);
4395        show_mode_change(file, p, 0, line_prefix);
4396}
4397
4398static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4399{
4400        FILE *file = opt->file;
4401        const char *line_prefix = diff_line_prefix(opt);
4402
4403        switch(p->status) {
4404        case DIFF_STATUS_DELETED:
4405                fputs(line_prefix, file);
4406                show_file_mode_name(file, "delete", p->one);
4407                break;
4408        case DIFF_STATUS_ADDED:
4409                fputs(line_prefix, file);
4410                show_file_mode_name(file, "create", p->two);
4411                break;
4412        case DIFF_STATUS_COPIED:
4413                fputs(line_prefix, file);
4414                show_rename_copy(file, "copy", p, line_prefix);
4415                break;
4416        case DIFF_STATUS_RENAMED:
4417                fputs(line_prefix, file);
4418                show_rename_copy(file, "rename", p, line_prefix);
4419                break;
4420        default:
4421                if (p->score) {
4422                        fprintf(file, "%s rewrite ", line_prefix);
4423                        write_name_quoted(p->two->path, file, ' ');
4424                        fprintf(file, "(%d%%)\n", similarity_index(p));
4425                }
4426                show_mode_change(file, p, !p->score, line_prefix);
4427                break;
4428        }
4429}
4430
4431struct patch_id_t {
4432        git_SHA_CTX *ctx;
4433        int patchlen;
4434};
4435
4436static int remove_space(char *line, int len)
4437{
4438        int i;
4439        char *dst = line;
4440        unsigned char c;
4441
4442        for (i = 0; i < len; i++)
4443                if (!isspace((c = line[i])))
4444                        *dst++ = c;
4445
4446        return dst - line;
4447}
4448
4449static void patch_id_consume(void *priv, char *line, unsigned long len)
4450{
4451        struct patch_id_t *data = priv;
4452        int new_len;
4453
4454        /* Ignore line numbers when computing the SHA1 of the patch */
4455        if (starts_with(line, "@@ -"))
4456                return;
4457
4458        new_len = remove_space(line, len);
4459
4460        git_SHA1_Update(data->ctx, line, new_len);
4461        data->patchlen += new_len;
4462}
4463
4464/* returns 0 upon success, and writes result into sha1 */
4465static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1, int diff_header_only)
4466{
4467        struct diff_queue_struct *q = &diff_queued_diff;
4468        int i;
4469        git_SHA_CTX ctx;
4470        struct patch_id_t data;
4471        char buffer[PATH_MAX * 4 + 20];
4472
4473        git_SHA1_Init(&ctx);
4474        memset(&data, 0, sizeof(struct patch_id_t));
4475        data.ctx = &ctx;
4476
4477        for (i = 0; i < q->nr; i++) {
4478                xpparam_t xpp;
4479                xdemitconf_t xecfg;
4480                mmfile_t mf1, mf2;
4481                struct diff_filepair *p = q->queue[i];
4482                int len1, len2;
4483
4484                memset(&xpp, 0, sizeof(xpp));
4485                memset(&xecfg, 0, sizeof(xecfg));
4486                if (p->status == 0)
4487                        return error("internal diff status error");
4488                if (p->status == DIFF_STATUS_UNKNOWN)
4489                        continue;
4490                if (diff_unmodified_pair(p))
4491                        continue;
4492                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4493                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4494                        continue;
4495                if (DIFF_PAIR_UNMERGED(p))
4496                        continue;
4497
4498                diff_fill_sha1_info(p->one);
4499                diff_fill_sha1_info(p->two);
4500
4501                len1 = remove_space(p->one->path, strlen(p->one->path));
4502                len2 = remove_space(p->two->path, strlen(p->two->path));
4503                if (p->one->mode == 0)
4504                        len1 = snprintf(buffer, sizeof(buffer),
4505                                        "diff--gita/%.*sb/%.*s"
4506                                        "newfilemode%06o"
4507                                        "---/dev/null"
4508                                        "+++b/%.*s",
4509                                        len1, p->one->path,
4510                                        len2, p->two->path,
4511                                        p->two->mode,
4512                                        len2, p->two->path);
4513                else if (p->two->mode == 0)
4514                        len1 = snprintf(buffer, sizeof(buffer),
4515                                        "diff--gita/%.*sb/%.*s"
4516                                        "deletedfilemode%06o"
4517                                        "---a/%.*s"
4518                                        "+++/dev/null",
4519                                        len1, p->one->path,
4520                                        len2, p->two->path,
4521                                        p->one->mode,
4522                                        len1, p->one->path);
4523                else
4524                        len1 = snprintf(buffer, sizeof(buffer),
4525                                        "diff--gita/%.*sb/%.*s"
4526                                        "---a/%.*s"
4527                                        "+++b/%.*s",
4528                                        len1, p->one->path,
4529                                        len2, p->two->path,
4530                                        len1, p->one->path,
4531                                        len2, p->two->path);
4532                git_SHA1_Update(&ctx, buffer, len1);
4533
4534                if (diff_header_only)
4535                        continue;
4536
4537                if (fill_mmfile(&mf1, p->one) < 0 ||
4538                    fill_mmfile(&mf2, p->two) < 0)
4539                        return error("unable to read files to diff");
4540
4541                if (diff_filespec_is_binary(p->one) ||
4542                    diff_filespec_is_binary(p->two)) {
4543                        git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4544                                        40);
4545                        git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4546                                        40);
4547                        continue;
4548                }
4549
4550                xpp.flags = 0;
4551                xecfg.ctxlen = 3;
4552                xecfg.flags = 0;
4553                if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4554                                  &xpp, &xecfg))
4555                        return error("unable to generate patch-id diff for %s",
4556                                     p->one->path);
4557        }
4558
4559        git_SHA1_Final(sha1, &ctx);
4560        return 0;
4561}
4562
4563int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1, int diff_header_only)
4564{
4565        struct diff_queue_struct *q = &diff_queued_diff;
4566        int i;
4567        int result = diff_get_patch_id(options, sha1, diff_header_only);
4568
4569        for (i = 0; i < q->nr; i++)
4570                diff_free_filepair(q->queue[i]);
4571
4572        free(q->queue);
4573        DIFF_QUEUE_CLEAR(q);
4574
4575        return result;
4576}
4577
4578static int is_summary_empty(const struct diff_queue_struct *q)
4579{
4580        int i;
4581
4582        for (i = 0; i < q->nr; i++) {
4583                const struct diff_filepair *p = q->queue[i];
4584
4585                switch (p->status) {
4586                case DIFF_STATUS_DELETED:
4587                case DIFF_STATUS_ADDED:
4588                case DIFF_STATUS_COPIED:
4589                case DIFF_STATUS_RENAMED:
4590                        return 0;
4591                default:
4592                        if (p->score)
4593                                return 0;
4594                        if (p->one->mode && p->two->mode &&
4595                            p->one->mode != p->two->mode)
4596                                return 0;
4597                        break;
4598                }
4599        }
4600        return 1;
4601}
4602
4603static const char rename_limit_warning[] =
4604"inexact rename detection was skipped due to too many files.";
4605
4606static const char degrade_cc_to_c_warning[] =
4607"only found copies from modified paths due to too many files.";
4608
4609static const char rename_limit_advice[] =
4610"you may want to set your %s variable to at least "
4611"%d and retry the command.";
4612
4613void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4614{
4615        if (degraded_cc)
4616                warning(degrade_cc_to_c_warning);
4617        else if (needed)
4618                warning(rename_limit_warning);
4619        else
4620                return;
4621        if (0 < needed && needed < 32767)
4622                warning(rename_limit_advice, varname, needed);
4623}
4624
4625void diff_flush(struct diff_options *options)
4626{
4627        struct diff_queue_struct *q = &diff_queued_diff;
4628        int i, output_format = options->output_format;
4629        int separator = 0;
4630        int dirstat_by_line = 0;
4631
4632        /*
4633         * Order: raw, stat, summary, patch
4634         * or:    name/name-status/checkdiff (other bits clear)
4635         */
4636        if (!q->nr)
4637                goto free_queue;
4638
4639        if (output_format & (DIFF_FORMAT_RAW |
4640                             DIFF_FORMAT_NAME |
4641                             DIFF_FORMAT_NAME_STATUS |
4642                             DIFF_FORMAT_CHECKDIFF)) {
4643                for (i = 0; i < q->nr; i++) {
4644                        struct diff_filepair *p = q->queue[i];
4645                        if (check_pair_status(p))
4646                                flush_one_pair(p, options);
4647                }
4648                separator++;
4649        }
4650
4651        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4652                dirstat_by_line = 1;
4653
4654        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4655            dirstat_by_line) {
4656                struct diffstat_t diffstat;
4657
4658                memset(&diffstat, 0, sizeof(struct diffstat_t));
4659                for (i = 0; i < q->nr; i++) {
4660                        struct diff_filepair *p = q->queue[i];
4661                        if (check_pair_status(p))
4662                                diff_flush_stat(p, options, &diffstat);
4663                }
4664                if (output_format & DIFF_FORMAT_NUMSTAT)
4665                        show_numstat(&diffstat, options);
4666                if (output_format & DIFF_FORMAT_DIFFSTAT)
4667                        show_stats(&diffstat, options);
4668                if (output_format & DIFF_FORMAT_SHORTSTAT)
4669                        show_shortstats(&diffstat, options);
4670                if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4671                        show_dirstat_by_line(&diffstat, options);
4672                free_diffstat_info(&diffstat);
4673                separator++;
4674        }
4675        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4676                show_dirstat(options);
4677
4678        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4679                for (i = 0; i < q->nr; i++) {
4680                        diff_summary(options, q->queue[i]);
4681                }
4682                separator++;
4683        }
4684
4685        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4686            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4687            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4688                /*
4689                 * run diff_flush_patch for the exit status. setting
4690                 * options->file to /dev/null should be safe, because we
4691                 * aren't supposed to produce any output anyway.
4692                 */
4693                if (options->close_file)
4694                        fclose(options->file);
4695                options->file = fopen("/dev/null", "w");
4696                if (!options->file)
4697                        die_errno("Could not open /dev/null");
4698                options->close_file = 1;
4699                for (i = 0; i < q->nr; i++) {
4700                        struct diff_filepair *p = q->queue[i];
4701                        if (check_pair_status(p))
4702                                diff_flush_patch(p, options);
4703                        if (options->found_changes)
4704                                break;
4705                }
4706        }
4707
4708        if (output_format & DIFF_FORMAT_PATCH) {
4709                if (separator) {
4710                        fprintf(options->file, "%s%c",
4711                                diff_line_prefix(options),
4712                                options->line_termination);
4713                        if (options->stat_sep) {
4714                                /* attach patch instead of inline */
4715                                fputs(options->stat_sep, options->file);
4716                        }
4717                }
4718
4719                for (i = 0; i < q->nr; i++) {
4720                        struct diff_filepair *p = q->queue[i];
4721                        if (check_pair_status(p))
4722                                diff_flush_patch(p, options);
4723                }
4724        }
4725
4726        if (output_format & DIFF_FORMAT_CALLBACK)
4727                options->format_callback(q, options, options->format_callback_data);
4728
4729        for (i = 0; i < q->nr; i++)
4730                diff_free_filepair(q->queue[i]);
4731free_queue:
4732        free(q->queue);
4733        DIFF_QUEUE_CLEAR(q);
4734        if (options->close_file)
4735                fclose(options->file);
4736
4737        /*
4738         * Report the content-level differences with HAS_CHANGES;
4739         * diff_addremove/diff_change does not set the bit when
4740         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4741         */
4742        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4743                if (options->found_changes)
4744                        DIFF_OPT_SET(options, HAS_CHANGES);
4745                else
4746                        DIFF_OPT_CLR(options, HAS_CHANGES);
4747        }
4748}
4749
4750static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4751{
4752        return (((p->status == DIFF_STATUS_MODIFIED) &&
4753                 ((p->score &&
4754                   filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4755                  (!p->score &&
4756                   filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4757                ((p->status != DIFF_STATUS_MODIFIED) &&
4758                 filter_bit_tst(p->status, options)));
4759}
4760
4761static void diffcore_apply_filter(struct diff_options *options)
4762{
4763        int i;
4764        struct diff_queue_struct *q = &diff_queued_diff;
4765        struct diff_queue_struct outq;
4766
4767        DIFF_QUEUE_CLEAR(&outq);
4768
4769        if (!options->filter)
4770                return;
4771
4772        if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4773                int found;
4774                for (i = found = 0; !found && i < q->nr; i++) {
4775                        if (match_filter(options, q->queue[i]))
4776                                found++;
4777                }
4778                if (found)
4779                        return;
4780
4781                /* otherwise we will clear the whole queue
4782                 * by copying the empty outq at the end of this
4783                 * function, but first clear the current entries
4784                 * in the queue.
4785                 */
4786                for (i = 0; i < q->nr; i++)
4787                        diff_free_filepair(q->queue[i]);
4788        }
4789        else {
4790                /* Only the matching ones */
4791                for (i = 0; i < q->nr; i++) {
4792                        struct diff_filepair *p = q->queue[i];
4793                        if (match_filter(options, p))
4794                                diff_q(&outq, p);
4795                        else
4796                                diff_free_filepair(p);
4797                }
4798        }
4799        free(q->queue);
4800        *q = outq;
4801}
4802
4803/* Check whether two filespecs with the same mode and size are identical */
4804static int diff_filespec_is_identical(struct diff_filespec *one,
4805                                      struct diff_filespec *two)
4806{
4807        if (S_ISGITLINK(one->mode))
4808                return 0;
4809        if (diff_populate_filespec(one, 0))
4810                return 0;
4811        if (diff_populate_filespec(two, 0))
4812                return 0;
4813        return !memcmp(one->data, two->data, one->size);
4814}
4815
4816static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4817{
4818        if (p->done_skip_stat_unmatch)
4819                return p->skip_stat_unmatch_result;
4820
4821        p->done_skip_stat_unmatch = 1;
4822        p->skip_stat_unmatch_result = 0;
4823        /*
4824         * 1. Entries that come from stat info dirtiness
4825         *    always have both sides (iow, not create/delete),
4826         *    one side of the object name is unknown, with
4827         *    the same mode and size.  Keep the ones that
4828         *    do not match these criteria.  They have real
4829         *    differences.
4830         *
4831         * 2. At this point, the file is known to be modified,
4832         *    with the same mode and size, and the object
4833         *    name of one side is unknown.  Need to inspect
4834         *    the identical contents.
4835         */
4836        if (!DIFF_FILE_VALID(p->one) || /* (1) */
4837            !DIFF_FILE_VALID(p->two) ||
4838            (p->one->oid_valid && p->two->oid_valid) ||
4839            (p->one->mode != p->two->mode) ||
4840            diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4841            diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4842            (p->one->size != p->two->size) ||
4843            !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4844                p->skip_stat_unmatch_result = 1;
4845        return p->skip_stat_unmatch_result;
4846}
4847
4848static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4849{
4850        int i;
4851        struct diff_queue_struct *q = &diff_queued_diff;
4852        struct diff_queue_struct outq;
4853        DIFF_QUEUE_CLEAR(&outq);
4854
4855        for (i = 0; i < q->nr; i++) {
4856                struct diff_filepair *p = q->queue[i];
4857
4858                if (diff_filespec_check_stat_unmatch(p))
4859                        diff_q(&outq, p);
4860                else {
4861                        /*
4862                         * The caller can subtract 1 from skip_stat_unmatch
4863                         * to determine how many paths were dirty only
4864                         * due to stat info mismatch.
4865                         */
4866                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4867                                diffopt->skip_stat_unmatch++;
4868                        diff_free_filepair(p);
4869                }
4870        }
4871        free(q->queue);
4872        *q = outq;
4873}
4874
4875static int diffnamecmp(const void *a_, const void *b_)
4876{
4877        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4878        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4879        const char *name_a, *name_b;
4880
4881        name_a = a->one ? a->one->path : a->two->path;
4882        name_b = b->one ? b->one->path : b->two->path;
4883        return strcmp(name_a, name_b);
4884}
4885
4886void diffcore_fix_diff_index(struct diff_options *options)
4887{
4888        struct diff_queue_struct *q = &diff_queued_diff;
4889        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4890}
4891
4892void diffcore_std(struct diff_options *options)
4893{
4894        /* NOTE please keep the following in sync with diff_tree_combined() */
4895        if (options->skip_stat_unmatch)
4896                diffcore_skip_stat_unmatch(options);
4897        if (!options->found_follow) {
4898                /* See try_to_follow_renames() in tree-diff.c */
4899                if (options->break_opt != -1)
4900                        diffcore_break(options->break_opt);
4901                if (options->detect_rename)
4902                        diffcore_rename(options);
4903                if (options->break_opt != -1)
4904                        diffcore_merge_broken();
4905        }
4906        if (options->pickaxe)
4907                diffcore_pickaxe(options);
4908        if (options->orderfile)
4909                diffcore_order(options->orderfile);
4910        if (!options->found_follow)
4911                /* See try_to_follow_renames() in tree-diff.c */
4912                diff_resolve_rename_copy();
4913        diffcore_apply_filter(options);
4914
4915        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4916                DIFF_OPT_SET(options, HAS_CHANGES);
4917        else
4918                DIFF_OPT_CLR(options, HAS_CHANGES);
4919
4920        options->found_follow = 0;
4921}
4922
4923int diff_result_code(struct diff_options *opt, int status)
4924{
4925        int result = 0;
4926
4927        diff_warn_rename_limit("diff.renameLimit",
4928                               opt->needed_rename_limit,
4929                               opt->degraded_cc_to_c);
4930        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4931            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4932                return status;
4933        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4934            DIFF_OPT_TST(opt, HAS_CHANGES))
4935                result |= 01;
4936        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4937            DIFF_OPT_TST(opt, CHECK_FAILED))
4938                result |= 02;
4939        return result;
4940}
4941
4942int diff_can_quit_early(struct diff_options *opt)
4943{
4944        return (DIFF_OPT_TST(opt, QUICK) &&
4945                !opt->filter &&
4946                DIFF_OPT_TST(opt, HAS_CHANGES));
4947}
4948
4949/*
4950 * Shall changes to this submodule be ignored?
4951 *
4952 * Submodule changes can be configured to be ignored separately for each path,
4953 * but that configuration can be overridden from the command line.
4954 */
4955static int is_submodule_ignored(const char *path, struct diff_options *options)
4956{
4957        int ignored = 0;
4958        unsigned orig_flags = options->flags;
4959        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4960                set_diffopt_flags_from_submodule_config(options, path);
4961        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4962                ignored = 1;
4963        options->flags = orig_flags;
4964        return ignored;
4965}
4966
4967void diff_addremove(struct diff_options *options,
4968                    int addremove, unsigned mode,
4969                    const unsigned char *sha1,
4970                    int sha1_valid,
4971                    const char *concatpath, unsigned dirty_submodule)
4972{
4973        struct diff_filespec *one, *two;
4974
4975        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4976                return;
4977
4978        /* This may look odd, but it is a preparation for
4979         * feeding "there are unchanged files which should
4980         * not produce diffs, but when you are doing copy
4981         * detection you would need them, so here they are"
4982         * entries to the diff-core.  They will be prefixed
4983         * with something like '=' or '*' (I haven't decided
4984         * which but should not make any difference).
4985         * Feeding the same new and old to diff_change()
4986         * also has the same effect.
4987         * Before the final output happens, they are pruned after
4988         * merged into rename/copy pairs as appropriate.
4989         */
4990        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4991                addremove = (addremove == '+' ? '-' :
4992                             addremove == '-' ? '+' : addremove);
4993
4994        if (options->prefix &&
4995            strncmp(concatpath, options->prefix, options->prefix_length))
4996                return;
4997
4998        one = alloc_filespec(concatpath);
4999        two = alloc_filespec(concatpath);
5000
5001        if (addremove != '+')
5002                fill_filespec(one, sha1, sha1_valid, mode);
5003        if (addremove != '-') {
5004                fill_filespec(two, sha1, sha1_valid, mode);
5005                two->dirty_submodule = dirty_submodule;
5006        }
5007
5008        diff_queue(&diff_queued_diff, one, two);
5009        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5010                DIFF_OPT_SET(options, HAS_CHANGES);
5011}
5012
5013void diff_change(struct diff_options *options,
5014                 unsigned old_mode, unsigned new_mode,
5015                 const unsigned char *old_sha1,
5016                 const unsigned char *new_sha1,
5017                 int old_sha1_valid, int new_sha1_valid,
5018                 const char *concatpath,
5019                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5020{
5021        struct diff_filespec *one, *two;
5022        struct diff_filepair *p;
5023
5024        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5025            is_submodule_ignored(concatpath, options))
5026                return;
5027
5028        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5029                unsigned tmp;
5030                const unsigned char *tmp_c;
5031                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
5032                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
5033                tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
5034                        new_sha1_valid = tmp;
5035                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
5036                        new_dirty_submodule = tmp;
5037        }
5038
5039        if (options->prefix &&
5040            strncmp(concatpath, options->prefix, options->prefix_length))
5041                return;
5042
5043        one = alloc_filespec(concatpath);
5044        two = alloc_filespec(concatpath);
5045        fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
5046        fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
5047        one->dirty_submodule = old_dirty_submodule;
5048        two->dirty_submodule = new_dirty_submodule;
5049        p = diff_queue(&diff_queued_diff, one, two);
5050
5051        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5052                return;
5053
5054        if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5055            !diff_filespec_check_stat_unmatch(p))
5056                return;
5057
5058        DIFF_OPT_SET(options, HAS_CHANGES);
5059}
5060
5061struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5062{
5063        struct diff_filepair *pair;
5064        struct diff_filespec *one, *two;
5065
5066        if (options->prefix &&
5067            strncmp(path, options->prefix, options->prefix_length))
5068                return NULL;
5069
5070        one = alloc_filespec(path);
5071        two = alloc_filespec(path);
5072        pair = diff_queue(&diff_queued_diff, one, two);
5073        pair->is_unmerged = 1;
5074        return pair;
5075}
5076
5077static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5078                size_t *outsize)
5079{
5080        struct diff_tempfile *temp;
5081        const char *argv[3];
5082        const char **arg = argv;
5083        struct child_process child = CHILD_PROCESS_INIT;
5084        struct strbuf buf = STRBUF_INIT;
5085        int err = 0;
5086
5087        temp = prepare_temp_file(spec->path, spec);
5088        *arg++ = pgm;
5089        *arg++ = temp->name;
5090        *arg = NULL;
5091
5092        child.use_shell = 1;
5093        child.argv = argv;
5094        child.out = -1;
5095        if (start_command(&child)) {
5096                remove_tempfile();
5097                return NULL;
5098        }
5099
5100        if (strbuf_read(&buf, child.out, 0) < 0)
5101                err = error("error reading from textconv command '%s'", pgm);
5102        close(child.out);
5103
5104        if (finish_command(&child) || err) {
5105                strbuf_release(&buf);
5106                remove_tempfile();
5107                return NULL;
5108        }
5109        remove_tempfile();
5110
5111        return strbuf_detach(&buf, outsize);
5112}
5113
5114size_t fill_textconv(struct userdiff_driver *driver,
5115                     struct diff_filespec *df,
5116                     char **outbuf)
5117{
5118        size_t size;
5119
5120        if (!driver) {
5121                if (!DIFF_FILE_VALID(df)) {
5122                        *outbuf = "";
5123                        return 0;
5124                }
5125                if (diff_populate_filespec(df, 0))
5126                        die("unable to read files to diff");
5127                *outbuf = df->data;
5128                return df->size;
5129        }
5130
5131        if (!driver->textconv)
5132                die("BUG: fill_textconv called with non-textconv driver");
5133
5134        if (driver->textconv_cache && df->oid_valid) {
5135                *outbuf = notes_cache_get(driver->textconv_cache,
5136                                          df->oid.hash,
5137                                          &size);
5138                if (*outbuf)
5139                        return size;
5140        }
5141
5142        *outbuf = run_textconv(driver->textconv, df, &size);
5143        if (!*outbuf)
5144                die("unable to read files to diff");
5145
5146        if (driver->textconv_cache && df->oid_valid) {
5147                /* ignore errors, as we might be in a readonly repository */
5148                notes_cache_put(driver->textconv_cache, df->oid.hash, *outbuf,
5149                                size);
5150                /*
5151                 * we could save up changes and flush them all at the end,
5152                 * but we would need an extra call after all diffing is done.
5153                 * Since generating a cache entry is the slow path anyway,
5154                 * this extra overhead probably isn't a big deal.
5155                 */
5156                notes_cache_write(driver->textconv_cache);
5157        }
5158
5159        return size;
5160}
5161
5162void setup_diff_pager(struct diff_options *opt)
5163{
5164        /*
5165         * If the user asked for our exit code, then either they want --quiet
5166         * or --exit-code. We should definitely not bother with a pager in the
5167         * former case, as we will generate no output. Since we still properly
5168         * report our exit code even when a pager is run, we _could_ run a
5169         * pager with --exit-code. But since we have not done so historically,
5170         * and because it is easy to find people oneline advising "git diff
5171         * --exit-code" in hooks and other scripts, we do not do so.
5172         */
5173        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5174            check_pager_config("diff") != 0)
5175                setup_pager();
5176}