0724fa6ff6c4312733e520d59f88ea12740d408d
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "quote.h"
   6#include "diff.h"
   7#include "diffcore.h"
   8#include "delta.h"
   9#include "xdiff-interface.h"
  10#include "color.h"
  11#include "attr.h"
  12#include "run-command.h"
  13#include "utf8.h"
  14#include "userdiff.h"
  15#include "sigchain.h"
  16#include "submodule.h"
  17#include "ll-merge.h"
  18#include "string-list.h"
  19
  20#ifdef NO_FAST_WORKING_DIRECTORY
  21#define FAST_WORKING_DIRECTORY 0
  22#else
  23#define FAST_WORKING_DIRECTORY 1
  24#endif
  25
  26static int diff_detect_rename_default;
  27static int diff_rename_limit_default = 400;
  28static int diff_suppress_blank_empty;
  29static int diff_use_color_default = -1;
  30static int diff_context_default = 3;
  31static const char *diff_word_regex_cfg;
  32static const char *external_diff_cmd_cfg;
  33int diff_auto_refresh_index = 1;
  34static int diff_mnemonic_prefix;
  35static int diff_no_prefix;
  36static int diff_stat_graph_width;
  37static int diff_dirstat_permille_default = 30;
  38static struct diff_options default_diff_options;
  39static long diff_algorithm;
  40
  41static char diff_colors[][COLOR_MAXLEN] = {
  42        GIT_COLOR_RESET,
  43        GIT_COLOR_NORMAL,       /* PLAIN */
  44        GIT_COLOR_BOLD,         /* METAINFO */
  45        GIT_COLOR_CYAN,         /* FRAGINFO */
  46        GIT_COLOR_RED,          /* OLD */
  47        GIT_COLOR_GREEN,        /* NEW */
  48        GIT_COLOR_YELLOW,       /* COMMIT */
  49        GIT_COLOR_BG_RED,       /* WHITESPACE */
  50        GIT_COLOR_NORMAL,       /* FUNCINFO */
  51};
  52
  53static int parse_diff_color_slot(const char *var, int ofs)
  54{
  55        if (!strcasecmp(var+ofs, "plain"))
  56                return DIFF_PLAIN;
  57        if (!strcasecmp(var+ofs, "meta"))
  58                return DIFF_METAINFO;
  59        if (!strcasecmp(var+ofs, "frag"))
  60                return DIFF_FRAGINFO;
  61        if (!strcasecmp(var+ofs, "old"))
  62                return DIFF_FILE_OLD;
  63        if (!strcasecmp(var+ofs, "new"))
  64                return DIFF_FILE_NEW;
  65        if (!strcasecmp(var+ofs, "commit"))
  66                return DIFF_COMMIT;
  67        if (!strcasecmp(var+ofs, "whitespace"))
  68                return DIFF_WHITESPACE;
  69        if (!strcasecmp(var+ofs, "func"))
  70                return DIFF_FUNCINFO;
  71        return -1;
  72}
  73
  74static int parse_dirstat_params(struct diff_options *options, const char *params_string,
  75                                struct strbuf *errmsg)
  76{
  77        char *params_copy = xstrdup(params_string);
  78        struct string_list params = STRING_LIST_INIT_NODUP;
  79        int ret = 0;
  80        int i;
  81
  82        if (*params_copy)
  83                string_list_split_in_place(&params, params_copy, ',', -1);
  84        for (i = 0; i < params.nr; i++) {
  85                const char *p = params.items[i].string;
  86                if (!strcmp(p, "changes")) {
  87                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  88                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  89                } else if (!strcmp(p, "lines")) {
  90                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  91                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  92                } else if (!strcmp(p, "files")) {
  93                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  94                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  95                } else if (!strcmp(p, "noncumulative")) {
  96                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
  97                } else if (!strcmp(p, "cumulative")) {
  98                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
  99                } else if (isdigit(*p)) {
 100                        char *end;
 101                        int permille = strtoul(p, &end, 10) * 10;
 102                        if (*end == '.' && isdigit(*++end)) {
 103                                /* only use first digit */
 104                                permille += *end - '0';
 105                                /* .. and ignore any further digits */
 106                                while (isdigit(*++end))
 107                                        ; /* nothing */
 108                        }
 109                        if (!*end)
 110                                options->dirstat_permille = permille;
 111                        else {
 112                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%s'\n"),
 113                                            p);
 114                                ret++;
 115                        }
 116                } else {
 117                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%s'\n"), p);
 118                        ret++;
 119                }
 120
 121        }
 122        string_list_clear(&params, 0);
 123        free(params_copy);
 124        return ret;
 125}
 126
 127static int parse_submodule_params(struct diff_options *options, const char *value)
 128{
 129        if (!strcmp(value, "log"))
 130                DIFF_OPT_SET(options, SUBMODULE_LOG);
 131        else if (!strcmp(value, "short"))
 132                DIFF_OPT_CLR(options, SUBMODULE_LOG);
 133        else
 134                return -1;
 135        return 0;
 136}
 137
 138static int git_config_rename(const char *var, const char *value)
 139{
 140        if (!value)
 141                return DIFF_DETECT_RENAME;
 142        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 143                return  DIFF_DETECT_COPY;
 144        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 145}
 146
 147long parse_algorithm_value(const char *value)
 148{
 149        if (!value)
 150                return -1;
 151        else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
 152                return 0;
 153        else if (!strcasecmp(value, "minimal"))
 154                return XDF_NEED_MINIMAL;
 155        else if (!strcasecmp(value, "patience"))
 156                return XDF_PATIENCE_DIFF;
 157        else if (!strcasecmp(value, "histogram"))
 158                return XDF_HISTOGRAM_DIFF;
 159        return -1;
 160}
 161
 162/*
 163 * These are to give UI layer defaults.
 164 * The core-level commands such as git-diff-files should
 165 * never be affected by the setting of diff.renames
 166 * the user happens to have in the configuration file.
 167 */
 168int git_diff_ui_config(const char *var, const char *value, void *cb)
 169{
 170        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 171                diff_use_color_default = git_config_colorbool(var, value);
 172                return 0;
 173        }
 174        if (!strcmp(var, "diff.context")) {
 175                diff_context_default = git_config_int(var, value);
 176                if (diff_context_default < 0)
 177                        return -1;
 178                return 0;
 179        }
 180        if (!strcmp(var, "diff.renames")) {
 181                diff_detect_rename_default = git_config_rename(var, value);
 182                return 0;
 183        }
 184        if (!strcmp(var, "diff.autorefreshindex")) {
 185                diff_auto_refresh_index = git_config_bool(var, value);
 186                return 0;
 187        }
 188        if (!strcmp(var, "diff.mnemonicprefix")) {
 189                diff_mnemonic_prefix = git_config_bool(var, value);
 190                return 0;
 191        }
 192        if (!strcmp(var, "diff.noprefix")) {
 193                diff_no_prefix = git_config_bool(var, value);
 194                return 0;
 195        }
 196        if (!strcmp(var, "diff.statgraphwidth")) {
 197                diff_stat_graph_width = git_config_int(var, value);
 198                return 0;
 199        }
 200        if (!strcmp(var, "diff.external"))
 201                return git_config_string(&external_diff_cmd_cfg, var, value);
 202        if (!strcmp(var, "diff.wordregex"))
 203                return git_config_string(&diff_word_regex_cfg, var, value);
 204
 205        if (!strcmp(var, "diff.ignoresubmodules"))
 206                handle_ignore_submodules_arg(&default_diff_options, value);
 207
 208        if (!strcmp(var, "diff.submodule")) {
 209                if (parse_submodule_params(&default_diff_options, value))
 210                        warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
 211                                value);
 212                return 0;
 213        }
 214
 215        if (!strcmp(var, "diff.algorithm")) {
 216                diff_algorithm = parse_algorithm_value(value);
 217                if (diff_algorithm < 0)
 218                        return -1;
 219                return 0;
 220        }
 221
 222        if (git_color_config(var, value, cb) < 0)
 223                return -1;
 224
 225        return git_diff_basic_config(var, value, cb);
 226}
 227
 228int git_diff_basic_config(const char *var, const char *value, void *cb)
 229{
 230        if (!strcmp(var, "diff.renamelimit")) {
 231                diff_rename_limit_default = git_config_int(var, value);
 232                return 0;
 233        }
 234
 235        if (userdiff_config(var, value) < 0)
 236                return -1;
 237
 238        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 239                int slot = parse_diff_color_slot(var, 11);
 240                if (slot < 0)
 241                        return 0;
 242                if (!value)
 243                        return config_error_nonbool(var);
 244                color_parse(value, var, diff_colors[slot]);
 245                return 0;
 246        }
 247
 248        /* like GNU diff's --suppress-blank-empty option  */
 249        if (!strcmp(var, "diff.suppressblankempty") ||
 250                        /* for backwards compatibility */
 251                        !strcmp(var, "diff.suppress-blank-empty")) {
 252                diff_suppress_blank_empty = git_config_bool(var, value);
 253                return 0;
 254        }
 255
 256        if (!strcmp(var, "diff.dirstat")) {
 257                struct strbuf errmsg = STRBUF_INIT;
 258                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 259                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 260                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 261                                errmsg.buf);
 262                strbuf_release(&errmsg);
 263                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 264                return 0;
 265        }
 266
 267        if (!prefixcmp(var, "submodule."))
 268                return parse_submodule_config_option(var, value);
 269
 270        return git_default_config(var, value, cb);
 271}
 272
 273static char *quote_two(const char *one, const char *two)
 274{
 275        int need_one = quote_c_style(one, NULL, NULL, 1);
 276        int need_two = quote_c_style(two, NULL, NULL, 1);
 277        struct strbuf res = STRBUF_INIT;
 278
 279        if (need_one + need_two) {
 280                strbuf_addch(&res, '"');
 281                quote_c_style(one, &res, NULL, 1);
 282                quote_c_style(two, &res, NULL, 1);
 283                strbuf_addch(&res, '"');
 284        } else {
 285                strbuf_addstr(&res, one);
 286                strbuf_addstr(&res, two);
 287        }
 288        return strbuf_detach(&res, NULL);
 289}
 290
 291static const char *external_diff(void)
 292{
 293        static const char *external_diff_cmd = NULL;
 294        static int done_preparing = 0;
 295
 296        if (done_preparing)
 297                return external_diff_cmd;
 298        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 299        if (!external_diff_cmd)
 300                external_diff_cmd = external_diff_cmd_cfg;
 301        done_preparing = 1;
 302        return external_diff_cmd;
 303}
 304
 305static struct diff_tempfile {
 306        const char *name; /* filename external diff should read from */
 307        char hex[41];
 308        char mode[10];
 309        char tmp_path[PATH_MAX];
 310} diff_temp[2];
 311
 312typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 313
 314struct emit_callback {
 315        int color_diff;
 316        unsigned ws_rule;
 317        int blank_at_eof_in_preimage;
 318        int blank_at_eof_in_postimage;
 319        int lno_in_preimage;
 320        int lno_in_postimage;
 321        sane_truncate_fn truncate;
 322        const char **label_path;
 323        struct diff_words_data *diff_words;
 324        struct diff_options *opt;
 325        int *found_changesp;
 326        struct strbuf *header;
 327};
 328
 329static int count_lines(const char *data, int size)
 330{
 331        int count, ch, completely_empty = 1, nl_just_seen = 0;
 332        count = 0;
 333        while (0 < size--) {
 334                ch = *data++;
 335                if (ch == '\n') {
 336                        count++;
 337                        nl_just_seen = 1;
 338                        completely_empty = 0;
 339                }
 340                else {
 341                        nl_just_seen = 0;
 342                        completely_empty = 0;
 343                }
 344        }
 345        if (completely_empty)
 346                return 0;
 347        if (!nl_just_seen)
 348                count++; /* no trailing newline */
 349        return count;
 350}
 351
 352static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 353{
 354        if (!DIFF_FILE_VALID(one)) {
 355                mf->ptr = (char *)""; /* does not matter */
 356                mf->size = 0;
 357                return 0;
 358        }
 359        else if (diff_populate_filespec(one, 0))
 360                return -1;
 361
 362        mf->ptr = one->data;
 363        mf->size = one->size;
 364        return 0;
 365}
 366
 367/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 368static unsigned long diff_filespec_size(struct diff_filespec *one)
 369{
 370        if (!DIFF_FILE_VALID(one))
 371                return 0;
 372        diff_populate_filespec(one, 1);
 373        return one->size;
 374}
 375
 376static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 377{
 378        char *ptr = mf->ptr;
 379        long size = mf->size;
 380        int cnt = 0;
 381
 382        if (!size)
 383                return cnt;
 384        ptr += size - 1; /* pointing at the very end */
 385        if (*ptr != '\n')
 386                ; /* incomplete line */
 387        else
 388                ptr--; /* skip the last LF */
 389        while (mf->ptr < ptr) {
 390                char *prev_eol;
 391                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 392                        if (*prev_eol == '\n')
 393                                break;
 394                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 395                        break;
 396                cnt++;
 397                ptr = prev_eol - 1;
 398        }
 399        return cnt;
 400}
 401
 402static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 403                               struct emit_callback *ecbdata)
 404{
 405        int l1, l2, at;
 406        unsigned ws_rule = ecbdata->ws_rule;
 407        l1 = count_trailing_blank(mf1, ws_rule);
 408        l2 = count_trailing_blank(mf2, ws_rule);
 409        if (l2 <= l1) {
 410                ecbdata->blank_at_eof_in_preimage = 0;
 411                ecbdata->blank_at_eof_in_postimage = 0;
 412                return;
 413        }
 414        at = count_lines(mf1->ptr, mf1->size);
 415        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 416
 417        at = count_lines(mf2->ptr, mf2->size);
 418        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 419}
 420
 421static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 422                        int first, const char *line, int len)
 423{
 424        int has_trailing_newline, has_trailing_carriage_return;
 425        int nofirst;
 426        FILE *file = o->file;
 427
 428        if (o->output_prefix) {
 429                struct strbuf *msg = NULL;
 430                msg = o->output_prefix(o, o->output_prefix_data);
 431                assert(msg);
 432                fwrite(msg->buf, msg->len, 1, file);
 433        }
 434
 435        if (len == 0) {
 436                has_trailing_newline = (first == '\n');
 437                has_trailing_carriage_return = (!has_trailing_newline &&
 438                                                (first == '\r'));
 439                nofirst = has_trailing_newline || has_trailing_carriage_return;
 440        } else {
 441                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 442                if (has_trailing_newline)
 443                        len--;
 444                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 445                if (has_trailing_carriage_return)
 446                        len--;
 447                nofirst = 0;
 448        }
 449
 450        if (len || !nofirst) {
 451                fputs(set, file);
 452                if (!nofirst)
 453                        fputc(first, file);
 454                fwrite(line, len, 1, file);
 455                fputs(reset, file);
 456        }
 457        if (has_trailing_carriage_return)
 458                fputc('\r', file);
 459        if (has_trailing_newline)
 460                fputc('\n', file);
 461}
 462
 463static void emit_line(struct diff_options *o, const char *set, const char *reset,
 464                      const char *line, int len)
 465{
 466        emit_line_0(o, set, reset, line[0], line+1, len-1);
 467}
 468
 469static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 470{
 471        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 472              ecbdata->blank_at_eof_in_preimage &&
 473              ecbdata->blank_at_eof_in_postimage &&
 474              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 475              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 476                return 0;
 477        return ws_blank_line(line, len, ecbdata->ws_rule);
 478}
 479
 480static void emit_add_line(const char *reset,
 481                          struct emit_callback *ecbdata,
 482                          const char *line, int len)
 483{
 484        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 485        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 486
 487        if (!*ws)
 488                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 489        else if (new_blank_line_at_eof(ecbdata, line, len))
 490                /* Blank line at EOF - paint '+' as well */
 491                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 492        else {
 493                /* Emit just the prefix, then the rest. */
 494                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 495                ws_check_emit(line, len, ecbdata->ws_rule,
 496                              ecbdata->opt->file, set, reset, ws);
 497        }
 498}
 499
 500static void emit_hunk_header(struct emit_callback *ecbdata,
 501                             const char *line, int len)
 502{
 503        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 504        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 505        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 506        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 507        static const char atat[2] = { '@', '@' };
 508        const char *cp, *ep;
 509        struct strbuf msgbuf = STRBUF_INIT;
 510        int org_len = len;
 511        int i = 1;
 512
 513        /*
 514         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 515         * it always is at least 10 bytes long.
 516         */
 517        if (len < 10 ||
 518            memcmp(line, atat, 2) ||
 519            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 520                emit_line(ecbdata->opt, plain, reset, line, len);
 521                return;
 522        }
 523        ep += 2; /* skip over @@ */
 524
 525        /* The hunk header in fraginfo color */
 526        strbuf_add(&msgbuf, frag, strlen(frag));
 527        strbuf_add(&msgbuf, line, ep - line);
 528        strbuf_add(&msgbuf, reset, strlen(reset));
 529
 530        /*
 531         * trailing "\r\n"
 532         */
 533        for ( ; i < 3; i++)
 534                if (line[len - i] == '\r' || line[len - i] == '\n')
 535                        len--;
 536
 537        /* blank before the func header */
 538        for (cp = ep; ep - line < len; ep++)
 539                if (*ep != ' ' && *ep != '\t')
 540                        break;
 541        if (ep != cp) {
 542                strbuf_add(&msgbuf, plain, strlen(plain));
 543                strbuf_add(&msgbuf, cp, ep - cp);
 544                strbuf_add(&msgbuf, reset, strlen(reset));
 545        }
 546
 547        if (ep < line + len) {
 548                strbuf_add(&msgbuf, func, strlen(func));
 549                strbuf_add(&msgbuf, ep, line + len - ep);
 550                strbuf_add(&msgbuf, reset, strlen(reset));
 551        }
 552
 553        strbuf_add(&msgbuf, line + len, org_len - len);
 554        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 555        strbuf_release(&msgbuf);
 556}
 557
 558static struct diff_tempfile *claim_diff_tempfile(void) {
 559        int i;
 560        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 561                if (!diff_temp[i].name)
 562                        return diff_temp + i;
 563        die("BUG: diff is failing to clean up its tempfiles");
 564}
 565
 566static int remove_tempfile_installed;
 567
 568static void remove_tempfile(void)
 569{
 570        int i;
 571        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 572                if (diff_temp[i].name == diff_temp[i].tmp_path)
 573                        unlink_or_warn(diff_temp[i].name);
 574                diff_temp[i].name = NULL;
 575        }
 576}
 577
 578static void remove_tempfile_on_signal(int signo)
 579{
 580        remove_tempfile();
 581        sigchain_pop(signo);
 582        raise(signo);
 583}
 584
 585static void print_line_count(FILE *file, int count)
 586{
 587        switch (count) {
 588        case 0:
 589                fprintf(file, "0,0");
 590                break;
 591        case 1:
 592                fprintf(file, "1");
 593                break;
 594        default:
 595                fprintf(file, "1,%d", count);
 596                break;
 597        }
 598}
 599
 600static void emit_rewrite_lines(struct emit_callback *ecb,
 601                               int prefix, const char *data, int size)
 602{
 603        const char *endp = NULL;
 604        static const char *nneof = " No newline at end of file\n";
 605        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 606        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 607
 608        while (0 < size) {
 609                int len;
 610
 611                endp = memchr(data, '\n', size);
 612                len = endp ? (endp - data + 1) : size;
 613                if (prefix != '+') {
 614                        ecb->lno_in_preimage++;
 615                        emit_line_0(ecb->opt, old, reset, '-',
 616                                    data, len);
 617                } else {
 618                        ecb->lno_in_postimage++;
 619                        emit_add_line(reset, ecb, data, len);
 620                }
 621                size -= len;
 622                data += len;
 623        }
 624        if (!endp) {
 625                const char *plain = diff_get_color(ecb->color_diff,
 626                                                   DIFF_PLAIN);
 627                putc('\n', ecb->opt->file);
 628                emit_line_0(ecb->opt, plain, reset, '\\',
 629                            nneof, strlen(nneof));
 630        }
 631}
 632
 633static void emit_rewrite_diff(const char *name_a,
 634                              const char *name_b,
 635                              struct diff_filespec *one,
 636                              struct diff_filespec *two,
 637                              struct userdiff_driver *textconv_one,
 638                              struct userdiff_driver *textconv_two,
 639                              struct diff_options *o)
 640{
 641        int lc_a, lc_b;
 642        const char *name_a_tab, *name_b_tab;
 643        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 644        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 645        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 646        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 647        const char *a_prefix, *b_prefix;
 648        char *data_one, *data_two;
 649        size_t size_one, size_two;
 650        struct emit_callback ecbdata;
 651        char *line_prefix = "";
 652        struct strbuf *msgbuf;
 653
 654        if (o && o->output_prefix) {
 655                msgbuf = o->output_prefix(o, o->output_prefix_data);
 656                line_prefix = msgbuf->buf;
 657        }
 658
 659        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 660                a_prefix = o->b_prefix;
 661                b_prefix = o->a_prefix;
 662        } else {
 663                a_prefix = o->a_prefix;
 664                b_prefix = o->b_prefix;
 665        }
 666
 667        name_a += (*name_a == '/');
 668        name_b += (*name_b == '/');
 669        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 670        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 671
 672        strbuf_reset(&a_name);
 673        strbuf_reset(&b_name);
 674        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 675        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 676
 677        size_one = fill_textconv(textconv_one, one, &data_one);
 678        size_two = fill_textconv(textconv_two, two, &data_two);
 679
 680        memset(&ecbdata, 0, sizeof(ecbdata));
 681        ecbdata.color_diff = want_color(o->use_color);
 682        ecbdata.found_changesp = &o->found_changes;
 683        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 684        ecbdata.opt = o;
 685        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 686                mmfile_t mf1, mf2;
 687                mf1.ptr = (char *)data_one;
 688                mf2.ptr = (char *)data_two;
 689                mf1.size = size_one;
 690                mf2.size = size_two;
 691                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 692        }
 693        ecbdata.lno_in_preimage = 1;
 694        ecbdata.lno_in_postimage = 1;
 695
 696        lc_a = count_lines(data_one, size_one);
 697        lc_b = count_lines(data_two, size_two);
 698        fprintf(o->file,
 699                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 700                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 701                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 702                line_prefix, fraginfo);
 703        if (!o->irreversible_delete)
 704                print_line_count(o->file, lc_a);
 705        else
 706                fprintf(o->file, "?,?");
 707        fprintf(o->file, " +");
 708        print_line_count(o->file, lc_b);
 709        fprintf(o->file, " @@%s\n", reset);
 710        if (lc_a && !o->irreversible_delete)
 711                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 712        if (lc_b)
 713                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 714        if (textconv_one)
 715                free((char *)data_one);
 716        if (textconv_two)
 717                free((char *)data_two);
 718}
 719
 720struct diff_words_buffer {
 721        mmfile_t text;
 722        long alloc;
 723        struct diff_words_orig {
 724                const char *begin, *end;
 725        } *orig;
 726        int orig_nr, orig_alloc;
 727};
 728
 729static void diff_words_append(char *line, unsigned long len,
 730                struct diff_words_buffer *buffer)
 731{
 732        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 733        line++;
 734        len--;
 735        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 736        buffer->text.size += len;
 737        buffer->text.ptr[buffer->text.size] = '\0';
 738}
 739
 740struct diff_words_style_elem {
 741        const char *prefix;
 742        const char *suffix;
 743        const char *color; /* NULL; filled in by the setup code if
 744                            * color is enabled */
 745};
 746
 747struct diff_words_style {
 748        enum diff_words_type type;
 749        struct diff_words_style_elem new, old, ctx;
 750        const char *newline;
 751};
 752
 753static struct diff_words_style diff_words_styles[] = {
 754        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 755        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 756        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 757};
 758
 759struct diff_words_data {
 760        struct diff_words_buffer minus, plus;
 761        const char *current_plus;
 762        int last_minus;
 763        struct diff_options *opt;
 764        regex_t *word_regex;
 765        enum diff_words_type type;
 766        struct diff_words_style *style;
 767};
 768
 769static int fn_out_diff_words_write_helper(FILE *fp,
 770                                          struct diff_words_style_elem *st_el,
 771                                          const char *newline,
 772                                          size_t count, const char *buf,
 773                                          const char *line_prefix)
 774{
 775        int print = 0;
 776
 777        while (count) {
 778                char *p = memchr(buf, '\n', count);
 779                if (print)
 780                        fputs(line_prefix, fp);
 781                if (p != buf) {
 782                        if (st_el->color && fputs(st_el->color, fp) < 0)
 783                                return -1;
 784                        if (fputs(st_el->prefix, fp) < 0 ||
 785                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 786                            fputs(st_el->suffix, fp) < 0)
 787                                return -1;
 788                        if (st_el->color && *st_el->color
 789                            && fputs(GIT_COLOR_RESET, fp) < 0)
 790                                return -1;
 791                }
 792                if (!p)
 793                        return 0;
 794                if (fputs(newline, fp) < 0)
 795                        return -1;
 796                count -= p + 1 - buf;
 797                buf = p + 1;
 798                print = 1;
 799        }
 800        return 0;
 801}
 802
 803/*
 804 * '--color-words' algorithm can be described as:
 805 *
 806 *   1. collect a the minus/plus lines of a diff hunk, divided into
 807 *      minus-lines and plus-lines;
 808 *
 809 *   2. break both minus-lines and plus-lines into words and
 810 *      place them into two mmfile_t with one word for each line;
 811 *
 812 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 813 *
 814 * And for the common parts of the both file, we output the plus side text.
 815 * diff_words->current_plus is used to trace the current position of the plus file
 816 * which printed. diff_words->last_minus is used to trace the last minus word
 817 * printed.
 818 *
 819 * For '--graph' to work with '--color-words', we need to output the graph prefix
 820 * on each line of color words output. Generally, there are two conditions on
 821 * which we should output the prefix.
 822 *
 823 *   1. diff_words->last_minus == 0 &&
 824 *      diff_words->current_plus == diff_words->plus.text.ptr
 825 *
 826 *      that is: the plus text must start as a new line, and if there is no minus
 827 *      word printed, a graph prefix must be printed.
 828 *
 829 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 830 *      *(diff_words->current_plus - 1) == '\n'
 831 *
 832 *      that is: a graph prefix must be printed following a '\n'
 833 */
 834static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 835{
 836        if ((diff_words->last_minus == 0 &&
 837                diff_words->current_plus == diff_words->plus.text.ptr) ||
 838                (diff_words->current_plus > diff_words->plus.text.ptr &&
 839                *(diff_words->current_plus - 1) == '\n')) {
 840                return 1;
 841        } else {
 842                return 0;
 843        }
 844}
 845
 846static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 847{
 848        struct diff_words_data *diff_words = priv;
 849        struct diff_words_style *style = diff_words->style;
 850        int minus_first, minus_len, plus_first, plus_len;
 851        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 852        struct diff_options *opt = diff_words->opt;
 853        struct strbuf *msgbuf;
 854        char *line_prefix = "";
 855
 856        if (line[0] != '@' || parse_hunk_header(line, len,
 857                        &minus_first, &minus_len, &plus_first, &plus_len))
 858                return;
 859
 860        assert(opt);
 861        if (opt->output_prefix) {
 862                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 863                line_prefix = msgbuf->buf;
 864        }
 865
 866        /* POSIX requires that first be decremented by one if len == 0... */
 867        if (minus_len) {
 868                minus_begin = diff_words->minus.orig[minus_first].begin;
 869                minus_end =
 870                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 871        } else
 872                minus_begin = minus_end =
 873                        diff_words->minus.orig[minus_first].end;
 874
 875        if (plus_len) {
 876                plus_begin = diff_words->plus.orig[plus_first].begin;
 877                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 878        } else
 879                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 880
 881        if (color_words_output_graph_prefix(diff_words)) {
 882                fputs(line_prefix, diff_words->opt->file);
 883        }
 884        if (diff_words->current_plus != plus_begin) {
 885                fn_out_diff_words_write_helper(diff_words->opt->file,
 886                                &style->ctx, style->newline,
 887                                plus_begin - diff_words->current_plus,
 888                                diff_words->current_plus, line_prefix);
 889                if (*(plus_begin - 1) == '\n')
 890                        fputs(line_prefix, diff_words->opt->file);
 891        }
 892        if (minus_begin != minus_end) {
 893                fn_out_diff_words_write_helper(diff_words->opt->file,
 894                                &style->old, style->newline,
 895                                minus_end - minus_begin, minus_begin,
 896                                line_prefix);
 897        }
 898        if (plus_begin != plus_end) {
 899                fn_out_diff_words_write_helper(diff_words->opt->file,
 900                                &style->new, style->newline,
 901                                plus_end - plus_begin, plus_begin,
 902                                line_prefix);
 903        }
 904
 905        diff_words->current_plus = plus_end;
 906        diff_words->last_minus = minus_first;
 907}
 908
 909/* This function starts looking at *begin, and returns 0 iff a word was found. */
 910static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 911                int *begin, int *end)
 912{
 913        if (word_regex && *begin < buffer->size) {
 914                regmatch_t match[1];
 915                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 916                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 917                                        '\n', match[0].rm_eo - match[0].rm_so);
 918                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 919                        *begin += match[0].rm_so;
 920                        return *begin >= *end;
 921                }
 922                return -1;
 923        }
 924
 925        /* find the next word */
 926        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 927                (*begin)++;
 928        if (*begin >= buffer->size)
 929                return -1;
 930
 931        /* find the end of the word */
 932        *end = *begin + 1;
 933        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 934                (*end)++;
 935
 936        return 0;
 937}
 938
 939/*
 940 * This function splits the words in buffer->text, stores the list with
 941 * newline separator into out, and saves the offsets of the original words
 942 * in buffer->orig.
 943 */
 944static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 945                regex_t *word_regex)
 946{
 947        int i, j;
 948        long alloc = 0;
 949
 950        out->size = 0;
 951        out->ptr = NULL;
 952
 953        /* fake an empty "0th" word */
 954        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 955        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 956        buffer->orig_nr = 1;
 957
 958        for (i = 0; i < buffer->text.size; i++) {
 959                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 960                        return;
 961
 962                /* store original boundaries */
 963                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 964                                buffer->orig_alloc);
 965                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 966                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 967                buffer->orig_nr++;
 968
 969                /* store one word */
 970                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 971                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 972                out->ptr[out->size + j - i] = '\n';
 973                out->size += j - i + 1;
 974
 975                i = j - 1;
 976        }
 977}
 978
 979/* this executes the word diff on the accumulated buffers */
 980static void diff_words_show(struct diff_words_data *diff_words)
 981{
 982        xpparam_t xpp;
 983        xdemitconf_t xecfg;
 984        mmfile_t minus, plus;
 985        struct diff_words_style *style = diff_words->style;
 986
 987        struct diff_options *opt = diff_words->opt;
 988        struct strbuf *msgbuf;
 989        char *line_prefix = "";
 990
 991        assert(opt);
 992        if (opt->output_prefix) {
 993                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 994                line_prefix = msgbuf->buf;
 995        }
 996
 997        /* special case: only removal */
 998        if (!diff_words->plus.text.size) {
 999                fputs(line_prefix, diff_words->opt->file);
1000                fn_out_diff_words_write_helper(diff_words->opt->file,
1001                        &style->old, style->newline,
1002                        diff_words->minus.text.size,
1003                        diff_words->minus.text.ptr, line_prefix);
1004                diff_words->minus.text.size = 0;
1005                return;
1006        }
1007
1008        diff_words->current_plus = diff_words->plus.text.ptr;
1009        diff_words->last_minus = 0;
1010
1011        memset(&xpp, 0, sizeof(xpp));
1012        memset(&xecfg, 0, sizeof(xecfg));
1013        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1014        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1015        xpp.flags = 0;
1016        /* as only the hunk header will be parsed, we need a 0-context */
1017        xecfg.ctxlen = 0;
1018        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1019                      &xpp, &xecfg);
1020        free(minus.ptr);
1021        free(plus.ptr);
1022        if (diff_words->current_plus != diff_words->plus.text.ptr +
1023                        diff_words->plus.text.size) {
1024                if (color_words_output_graph_prefix(diff_words))
1025                        fputs(line_prefix, diff_words->opt->file);
1026                fn_out_diff_words_write_helper(diff_words->opt->file,
1027                        &style->ctx, style->newline,
1028                        diff_words->plus.text.ptr + diff_words->plus.text.size
1029                        - diff_words->current_plus, diff_words->current_plus,
1030                        line_prefix);
1031        }
1032        diff_words->minus.text.size = diff_words->plus.text.size = 0;
1033}
1034
1035/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1036static void diff_words_flush(struct emit_callback *ecbdata)
1037{
1038        if (ecbdata->diff_words->minus.text.size ||
1039            ecbdata->diff_words->plus.text.size)
1040                diff_words_show(ecbdata->diff_words);
1041}
1042
1043static void diff_filespec_load_driver(struct diff_filespec *one)
1044{
1045        /* Use already-loaded driver */
1046        if (one->driver)
1047                return;
1048
1049        if (S_ISREG(one->mode))
1050                one->driver = userdiff_find_by_path(one->path);
1051
1052        /* Fallback to default settings */
1053        if (!one->driver)
1054                one->driver = userdiff_find_by_name("default");
1055}
1056
1057static const char *userdiff_word_regex(struct diff_filespec *one)
1058{
1059        diff_filespec_load_driver(one);
1060        return one->driver->word_regex;
1061}
1062
1063static void init_diff_words_data(struct emit_callback *ecbdata,
1064                                 struct diff_options *orig_opts,
1065                                 struct diff_filespec *one,
1066                                 struct diff_filespec *two)
1067{
1068        int i;
1069        struct diff_options *o = xmalloc(sizeof(struct diff_options));
1070        memcpy(o, orig_opts, sizeof(struct diff_options));
1071
1072        ecbdata->diff_words =
1073                xcalloc(1, sizeof(struct diff_words_data));
1074        ecbdata->diff_words->type = o->word_diff;
1075        ecbdata->diff_words->opt = o;
1076        if (!o->word_regex)
1077                o->word_regex = userdiff_word_regex(one);
1078        if (!o->word_regex)
1079                o->word_regex = userdiff_word_regex(two);
1080        if (!o->word_regex)
1081                o->word_regex = diff_word_regex_cfg;
1082        if (o->word_regex) {
1083                ecbdata->diff_words->word_regex = (regex_t *)
1084                        xmalloc(sizeof(regex_t));
1085                if (regcomp(ecbdata->diff_words->word_regex,
1086                            o->word_regex,
1087                            REG_EXTENDED | REG_NEWLINE))
1088                        die ("Invalid regular expression: %s",
1089                             o->word_regex);
1090        }
1091        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1092                if (o->word_diff == diff_words_styles[i].type) {
1093                        ecbdata->diff_words->style =
1094                                &diff_words_styles[i];
1095                        break;
1096                }
1097        }
1098        if (want_color(o->use_color)) {
1099                struct diff_words_style *st = ecbdata->diff_words->style;
1100                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1101                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1102                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1103        }
1104}
1105
1106static void free_diff_words_data(struct emit_callback *ecbdata)
1107{
1108        if (ecbdata->diff_words) {
1109                diff_words_flush(ecbdata);
1110                free (ecbdata->diff_words->opt);
1111                free (ecbdata->diff_words->minus.text.ptr);
1112                free (ecbdata->diff_words->minus.orig);
1113                free (ecbdata->diff_words->plus.text.ptr);
1114                free (ecbdata->diff_words->plus.orig);
1115                if (ecbdata->diff_words->word_regex) {
1116                        regfree(ecbdata->diff_words->word_regex);
1117                        free(ecbdata->diff_words->word_regex);
1118                }
1119                free(ecbdata->diff_words);
1120                ecbdata->diff_words = NULL;
1121        }
1122}
1123
1124const char *diff_get_color(int diff_use_color, enum color_diff ix)
1125{
1126        if (want_color(diff_use_color))
1127                return diff_colors[ix];
1128        return "";
1129}
1130
1131static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1132{
1133        const char *cp;
1134        unsigned long allot;
1135        size_t l = len;
1136
1137        if (ecb->truncate)
1138                return ecb->truncate(line, len);
1139        cp = line;
1140        allot = l;
1141        while (0 < l) {
1142                (void) utf8_width(&cp, &l);
1143                if (!cp)
1144                        break; /* truncated in the middle? */
1145        }
1146        return allot - l;
1147}
1148
1149static void find_lno(const char *line, struct emit_callback *ecbdata)
1150{
1151        const char *p;
1152        ecbdata->lno_in_preimage = 0;
1153        ecbdata->lno_in_postimage = 0;
1154        p = strchr(line, '-');
1155        if (!p)
1156                return; /* cannot happen */
1157        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1158        p = strchr(p, '+');
1159        if (!p)
1160                return; /* cannot happen */
1161        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1162}
1163
1164static void fn_out_consume(void *priv, char *line, unsigned long len)
1165{
1166        struct emit_callback *ecbdata = priv;
1167        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1168        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1169        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1170        struct diff_options *o = ecbdata->opt;
1171        char *line_prefix = "";
1172        struct strbuf *msgbuf;
1173
1174        if (o && o->output_prefix) {
1175                msgbuf = o->output_prefix(o, o->output_prefix_data);
1176                line_prefix = msgbuf->buf;
1177        }
1178
1179        if (ecbdata->header) {
1180                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1181                strbuf_reset(ecbdata->header);
1182                ecbdata->header = NULL;
1183        }
1184        *(ecbdata->found_changesp) = 1;
1185
1186        if (ecbdata->label_path[0]) {
1187                const char *name_a_tab, *name_b_tab;
1188
1189                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1190                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1191
1192                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1193                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1194                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1195                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1196                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1197        }
1198
1199        if (diff_suppress_blank_empty
1200            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1201                line[0] = '\n';
1202                len = 1;
1203        }
1204
1205        if (line[0] == '@') {
1206                if (ecbdata->diff_words)
1207                        diff_words_flush(ecbdata);
1208                len = sane_truncate_line(ecbdata, line, len);
1209                find_lno(line, ecbdata);
1210                emit_hunk_header(ecbdata, line, len);
1211                if (line[len-1] != '\n')
1212                        putc('\n', ecbdata->opt->file);
1213                return;
1214        }
1215
1216        if (len < 1) {
1217                emit_line(ecbdata->opt, reset, reset, line, len);
1218                if (ecbdata->diff_words
1219                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1220                        fputs("~\n", ecbdata->opt->file);
1221                return;
1222        }
1223
1224        if (ecbdata->diff_words) {
1225                if (line[0] == '-') {
1226                        diff_words_append(line, len,
1227                                          &ecbdata->diff_words->minus);
1228                        return;
1229                } else if (line[0] == '+') {
1230                        diff_words_append(line, len,
1231                                          &ecbdata->diff_words->plus);
1232                        return;
1233                } else if (!prefixcmp(line, "\\ ")) {
1234                        /*
1235                         * Eat the "no newline at eof" marker as if we
1236                         * saw a "+" or "-" line with nothing on it,
1237                         * and return without diff_words_flush() to
1238                         * defer processing. If this is the end of
1239                         * preimage, more "+" lines may come after it.
1240                         */
1241                        return;
1242                }
1243                diff_words_flush(ecbdata);
1244                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1245                        emit_line(ecbdata->opt, plain, reset, line, len);
1246                        fputs("~\n", ecbdata->opt->file);
1247                } else {
1248                        /*
1249                         * Skip the prefix character, if any.  With
1250                         * diff_suppress_blank_empty, there may be
1251                         * none.
1252                         */
1253                        if (line[0] != '\n') {
1254                              line++;
1255                              len--;
1256                        }
1257                        emit_line(ecbdata->opt, plain, reset, line, len);
1258                }
1259                return;
1260        }
1261
1262        if (line[0] != '+') {
1263                const char *color =
1264                        diff_get_color(ecbdata->color_diff,
1265                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1266                ecbdata->lno_in_preimage++;
1267                if (line[0] == ' ')
1268                        ecbdata->lno_in_postimage++;
1269                emit_line(ecbdata->opt, color, reset, line, len);
1270        } else {
1271                ecbdata->lno_in_postimage++;
1272                emit_add_line(reset, ecbdata, line + 1, len - 1);
1273        }
1274}
1275
1276static char *pprint_rename(const char *a, const char *b)
1277{
1278        const char *old = a;
1279        const char *new = b;
1280        struct strbuf name = STRBUF_INIT;
1281        int pfx_length, sfx_length;
1282        int len_a = strlen(a);
1283        int len_b = strlen(b);
1284        int a_midlen, b_midlen;
1285        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1286        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1287
1288        if (qlen_a || qlen_b) {
1289                quote_c_style(a, &name, NULL, 0);
1290                strbuf_addstr(&name, " => ");
1291                quote_c_style(b, &name, NULL, 0);
1292                return strbuf_detach(&name, NULL);
1293        }
1294
1295        /* Find common prefix */
1296        pfx_length = 0;
1297        while (*old && *new && *old == *new) {
1298                if (*old == '/')
1299                        pfx_length = old - a + 1;
1300                old++;
1301                new++;
1302        }
1303
1304        /* Find common suffix */
1305        old = a + len_a;
1306        new = b + len_b;
1307        sfx_length = 0;
1308        while (a <= old && b <= new && *old == *new) {
1309                if (*old == '/')
1310                        sfx_length = len_a - (old - a);
1311                old--;
1312                new--;
1313        }
1314
1315        /*
1316         * pfx{mid-a => mid-b}sfx
1317         * {pfx-a => pfx-b}sfx
1318         * pfx{sfx-a => sfx-b}
1319         * name-a => name-b
1320         */
1321        a_midlen = len_a - pfx_length - sfx_length;
1322        b_midlen = len_b - pfx_length - sfx_length;
1323        if (a_midlen < 0)
1324                a_midlen = 0;
1325        if (b_midlen < 0)
1326                b_midlen = 0;
1327
1328        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1329        if (pfx_length + sfx_length) {
1330                strbuf_add(&name, a, pfx_length);
1331                strbuf_addch(&name, '{');
1332        }
1333        strbuf_add(&name, a + pfx_length, a_midlen);
1334        strbuf_addstr(&name, " => ");
1335        strbuf_add(&name, b + pfx_length, b_midlen);
1336        if (pfx_length + sfx_length) {
1337                strbuf_addch(&name, '}');
1338                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1339        }
1340        return strbuf_detach(&name, NULL);
1341}
1342
1343struct diffstat_t {
1344        int nr;
1345        int alloc;
1346        struct diffstat_file {
1347                char *from_name;
1348                char *name;
1349                char *print_name;
1350                unsigned is_unmerged:1;
1351                unsigned is_binary:1;
1352                unsigned is_renamed:1;
1353                unsigned is_interesting:1;
1354                uintmax_t added, deleted;
1355        } **files;
1356};
1357
1358static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1359                                          const char *name_a,
1360                                          const char *name_b)
1361{
1362        struct diffstat_file *x;
1363        x = xcalloc(sizeof (*x), 1);
1364        if (diffstat->nr == diffstat->alloc) {
1365                diffstat->alloc = alloc_nr(diffstat->alloc);
1366                diffstat->files = xrealloc(diffstat->files,
1367                                diffstat->alloc * sizeof(x));
1368        }
1369        diffstat->files[diffstat->nr++] = x;
1370        if (name_b) {
1371                x->from_name = xstrdup(name_a);
1372                x->name = xstrdup(name_b);
1373                x->is_renamed = 1;
1374        }
1375        else {
1376                x->from_name = NULL;
1377                x->name = xstrdup(name_a);
1378        }
1379        return x;
1380}
1381
1382static void diffstat_consume(void *priv, char *line, unsigned long len)
1383{
1384        struct diffstat_t *diffstat = priv;
1385        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1386
1387        if (line[0] == '+')
1388                x->added++;
1389        else if (line[0] == '-')
1390                x->deleted++;
1391}
1392
1393const char mime_boundary_leader[] = "------------";
1394
1395static int scale_linear(int it, int width, int max_change)
1396{
1397        if (!it)
1398                return 0;
1399        /*
1400         * make sure that at least one '-' or '+' is printed if
1401         * there is any change to this path. The easiest way is to
1402         * scale linearly as if the alloted width is one column shorter
1403         * than it is, and then add 1 to the result.
1404         */
1405        return 1 + (it * (width - 1) / max_change);
1406}
1407
1408static void show_name(FILE *file,
1409                      const char *prefix, const char *name, int len)
1410{
1411        fprintf(file, " %s%-*s |", prefix, len, name);
1412}
1413
1414static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1415{
1416        if (cnt <= 0)
1417                return;
1418        fprintf(file, "%s", set);
1419        while (cnt--)
1420                putc(ch, file);
1421        fprintf(file, "%s", reset);
1422}
1423
1424static void fill_print_name(struct diffstat_file *file)
1425{
1426        char *pname;
1427
1428        if (file->print_name)
1429                return;
1430
1431        if (!file->is_renamed) {
1432                struct strbuf buf = STRBUF_INIT;
1433                if (quote_c_style(file->name, &buf, NULL, 0)) {
1434                        pname = strbuf_detach(&buf, NULL);
1435                } else {
1436                        pname = file->name;
1437                        strbuf_release(&buf);
1438                }
1439        } else {
1440                pname = pprint_rename(file->from_name, file->name);
1441        }
1442        file->print_name = pname;
1443}
1444
1445int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1446{
1447        struct strbuf sb = STRBUF_INIT;
1448        int ret;
1449
1450        if (!files) {
1451                assert(insertions == 0 && deletions == 0);
1452                return fprintf(fp, "%s\n", " 0 files changed");
1453        }
1454
1455        strbuf_addf(&sb,
1456                    (files == 1) ? " %d file changed" : " %d files changed",
1457                    files);
1458
1459        /*
1460         * For binary diff, the caller may want to print "x files
1461         * changed" with insertions == 0 && deletions == 0.
1462         *
1463         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1464         * is probably less confusing (i.e skip over "2 files changed
1465         * but nothing about added/removed lines? Is this a bug in Git?").
1466         */
1467        if (insertions || deletions == 0) {
1468                /*
1469                 * TRANSLATORS: "+" in (+) is a line addition marker;
1470                 * do not translate it.
1471                 */
1472                strbuf_addf(&sb,
1473                            (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1474                            insertions);
1475        }
1476
1477        if (deletions || insertions == 0) {
1478                /*
1479                 * TRANSLATORS: "-" in (-) is a line removal marker;
1480                 * do not translate it.
1481                 */
1482                strbuf_addf(&sb,
1483                            (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1484                            deletions);
1485        }
1486        strbuf_addch(&sb, '\n');
1487        ret = fputs(sb.buf, fp);
1488        strbuf_release(&sb);
1489        return ret;
1490}
1491
1492static void show_stats(struct diffstat_t *data, struct diff_options *options)
1493{
1494        int i, len, add, del, adds = 0, dels = 0;
1495        uintmax_t max_change = 0, max_len = 0;
1496        int total_files = data->nr, count;
1497        int width, name_width, graph_width, number_width = 0, bin_width = 0;
1498        const char *reset, *add_c, *del_c;
1499        const char *line_prefix = "";
1500        int extra_shown = 0;
1501        struct strbuf *msg = NULL;
1502
1503        if (data->nr == 0)
1504                return;
1505
1506        if (options->output_prefix) {
1507                msg = options->output_prefix(options, options->output_prefix_data);
1508                line_prefix = msg->buf;
1509        }
1510
1511        count = options->stat_count ? options->stat_count : data->nr;
1512
1513        reset = diff_get_color_opt(options, DIFF_RESET);
1514        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1515        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1516
1517        /*
1518         * Find the longest filename and max number of changes
1519         */
1520        for (i = 0; (i < count) && (i < data->nr); i++) {
1521                struct diffstat_file *file = data->files[i];
1522                uintmax_t change = file->added + file->deleted;
1523
1524                if (!file->is_interesting && (change == 0)) {
1525                        count++; /* not shown == room for one more */
1526                        continue;
1527                }
1528                fill_print_name(file);
1529                len = strlen(file->print_name);
1530                if (max_len < len)
1531                        max_len = len;
1532
1533                if (file->is_unmerged) {
1534                        /* "Unmerged" is 8 characters */
1535                        bin_width = bin_width < 8 ? 8 : bin_width;
1536                        continue;
1537                }
1538                if (file->is_binary) {
1539                        /* "Bin XXX -> YYY bytes" */
1540                        int w = 14 + decimal_width(file->added)
1541                                + decimal_width(file->deleted);
1542                        bin_width = bin_width < w ? w : bin_width;
1543                        /* Display change counts aligned with "Bin" */
1544                        number_width = 3;
1545                        continue;
1546                }
1547
1548                if (max_change < change)
1549                        max_change = change;
1550        }
1551        count = i; /* where we can stop scanning in data->files[] */
1552
1553        /*
1554         * We have width = stat_width or term_columns() columns total.
1555         * We want a maximum of min(max_len, stat_name_width) for the name part.
1556         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1557         * We also need 1 for " " and 4 + decimal_width(max_change)
1558         * for " | NNNN " and one the empty column at the end, altogether
1559         * 6 + decimal_width(max_change).
1560         *
1561         * If there's not enough space, we will use the smaller of
1562         * stat_name_width (if set) and 5/8*width for the filename,
1563         * and the rest for constant elements + graph part, but no more
1564         * than stat_graph_width for the graph part.
1565         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1566         * for the standard terminal size).
1567         *
1568         * In other words: stat_width limits the maximum width, and
1569         * stat_name_width fixes the maximum width of the filename,
1570         * and is also used to divide available columns if there
1571         * aren't enough.
1572         *
1573         * Binary files are displayed with "Bin XXX -> YYY bytes"
1574         * instead of the change count and graph. This part is treated
1575         * similarly to the graph part, except that it is not
1576         * "scaled". If total width is too small to accomodate the
1577         * guaranteed minimum width of the filename part and the
1578         * separators and this message, this message will "overflow"
1579         * making the line longer than the maximum width.
1580         */
1581
1582        if (options->stat_width == -1)
1583                width = term_columns() - options->output_prefix_length;
1584        else
1585                width = options->stat_width ? options->stat_width : 80;
1586        number_width = decimal_width(max_change) > number_width ?
1587                decimal_width(max_change) : number_width;
1588
1589        if (options->stat_graph_width == -1)
1590                options->stat_graph_width = diff_stat_graph_width;
1591
1592        /*
1593         * Guarantee 3/8*16==6 for the graph part
1594         * and 5/8*16==10 for the filename part
1595         */
1596        if (width < 16 + 6 + number_width)
1597                width = 16 + 6 + number_width;
1598
1599        /*
1600         * First assign sizes that are wanted, ignoring available width.
1601         * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1602         * starting from "XXX" should fit in graph_width.
1603         */
1604        graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1605        if (options->stat_graph_width &&
1606            options->stat_graph_width < graph_width)
1607                graph_width = options->stat_graph_width;
1608
1609        name_width = (options->stat_name_width > 0 &&
1610                      options->stat_name_width < max_len) ?
1611                options->stat_name_width : max_len;
1612
1613        /*
1614         * Adjust adjustable widths not to exceed maximum width
1615         */
1616        if (name_width + number_width + 6 + graph_width > width) {
1617                if (graph_width > width * 3/8 - number_width - 6) {
1618                        graph_width = width * 3/8 - number_width - 6;
1619                        if (graph_width < 6)
1620                                graph_width = 6;
1621                }
1622
1623                if (options->stat_graph_width &&
1624                    graph_width > options->stat_graph_width)
1625                        graph_width = options->stat_graph_width;
1626                if (name_width > width - number_width - 6 - graph_width)
1627                        name_width = width - number_width - 6 - graph_width;
1628                else
1629                        graph_width = width - number_width - 6 - name_width;
1630        }
1631
1632        /*
1633         * From here name_width is the width of the name area,
1634         * and graph_width is the width of the graph area.
1635         * max_change is used to scale graph properly.
1636         */
1637        for (i = 0; i < count; i++) {
1638                const char *prefix = "";
1639                struct diffstat_file *file = data->files[i];
1640                char *name = file->print_name;
1641                uintmax_t added = file->added;
1642                uintmax_t deleted = file->deleted;
1643                int name_len;
1644
1645                if (!file->is_interesting && (added + deleted == 0))
1646                        continue;
1647
1648                /*
1649                 * "scale" the filename
1650                 */
1651                len = name_width;
1652                name_len = strlen(name);
1653                if (name_width < name_len) {
1654                        char *slash;
1655                        prefix = "...";
1656                        len -= 3;
1657                        name += name_len - len;
1658                        slash = strchr(name, '/');
1659                        if (slash)
1660                                name = slash;
1661                }
1662
1663                if (file->is_binary) {
1664                        fprintf(options->file, "%s", line_prefix);
1665                        show_name(options->file, prefix, name, len);
1666                        fprintf(options->file, " %*s", number_width, "Bin");
1667                        if (!added && !deleted) {
1668                                putc('\n', options->file);
1669                                continue;
1670                        }
1671                        fprintf(options->file, " %s%"PRIuMAX"%s",
1672                                del_c, deleted, reset);
1673                        fprintf(options->file, " -> ");
1674                        fprintf(options->file, "%s%"PRIuMAX"%s",
1675                                add_c, added, reset);
1676                        fprintf(options->file, " bytes");
1677                        fprintf(options->file, "\n");
1678                        continue;
1679                }
1680                else if (file->is_unmerged) {
1681                        fprintf(options->file, "%s", line_prefix);
1682                        show_name(options->file, prefix, name, len);
1683                        fprintf(options->file, " Unmerged\n");
1684                        continue;
1685                }
1686
1687                /*
1688                 * scale the add/delete
1689                 */
1690                add = added;
1691                del = deleted;
1692
1693                if (graph_width <= max_change) {
1694                        int total = add + del;
1695
1696                        total = scale_linear(add + del, graph_width, max_change);
1697                        if (total < 2 && add && del)
1698                                /* width >= 2 due to the sanity check */
1699                                total = 2;
1700                        if (add < del) {
1701                                add = scale_linear(add, graph_width, max_change);
1702                                del = total - add;
1703                        } else {
1704                                del = scale_linear(del, graph_width, max_change);
1705                                add = total - del;
1706                        }
1707                }
1708                fprintf(options->file, "%s", line_prefix);
1709                show_name(options->file, prefix, name, len);
1710                fprintf(options->file, " %*"PRIuMAX"%s",
1711                        number_width, added + deleted,
1712                        added + deleted ? " " : "");
1713                show_graph(options->file, '+', add, add_c, reset);
1714                show_graph(options->file, '-', del, del_c, reset);
1715                fprintf(options->file, "\n");
1716        }
1717
1718        for (i = 0; i < data->nr; i++) {
1719                struct diffstat_file *file = data->files[i];
1720                uintmax_t added = file->added;
1721                uintmax_t deleted = file->deleted;
1722
1723                if (file->is_unmerged ||
1724                    (!file->is_interesting && (added + deleted == 0))) {
1725                        total_files--;
1726                        continue;
1727                }
1728
1729                if (!file->is_binary) {
1730                        adds += added;
1731                        dels += deleted;
1732                }
1733                if (i < count)
1734                        continue;
1735                if (!extra_shown)
1736                        fprintf(options->file, "%s ...\n", line_prefix);
1737                extra_shown = 1;
1738        }
1739        fprintf(options->file, "%s", line_prefix);
1740        print_stat_summary(options->file, total_files, adds, dels);
1741}
1742
1743static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1744{
1745        int i, adds = 0, dels = 0, total_files = data->nr;
1746
1747        if (data->nr == 0)
1748                return;
1749
1750        for (i = 0; i < data->nr; i++) {
1751                int added = data->files[i]->added;
1752                int deleted= data->files[i]->deleted;
1753
1754                if (data->files[i]->is_unmerged ||
1755                    (!data->files[i]->is_interesting && (added + deleted == 0))) {
1756                        total_files--;
1757                } else if (!data->files[i]->is_binary) { /* don't count bytes */
1758                        adds += added;
1759                        dels += deleted;
1760                }
1761        }
1762        if (options->output_prefix) {
1763                struct strbuf *msg = NULL;
1764                msg = options->output_prefix(options,
1765                                options->output_prefix_data);
1766                fprintf(options->file, "%s", msg->buf);
1767        }
1768        print_stat_summary(options->file, total_files, adds, dels);
1769}
1770
1771static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1772{
1773        int i;
1774
1775        if (data->nr == 0)
1776                return;
1777
1778        for (i = 0; i < data->nr; i++) {
1779                struct diffstat_file *file = data->files[i];
1780
1781                if (options->output_prefix) {
1782                        struct strbuf *msg = NULL;
1783                        msg = options->output_prefix(options,
1784                                        options->output_prefix_data);
1785                        fprintf(options->file, "%s", msg->buf);
1786                }
1787
1788                if (file->is_binary)
1789                        fprintf(options->file, "-\t-\t");
1790                else
1791                        fprintf(options->file,
1792                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1793                                file->added, file->deleted);
1794                if (options->line_termination) {
1795                        fill_print_name(file);
1796                        if (!file->is_renamed)
1797                                write_name_quoted(file->name, options->file,
1798                                                  options->line_termination);
1799                        else {
1800                                fputs(file->print_name, options->file);
1801                                putc(options->line_termination, options->file);
1802                        }
1803                } else {
1804                        if (file->is_renamed) {
1805                                putc('\0', options->file);
1806                                write_name_quoted(file->from_name, options->file, '\0');
1807                        }
1808                        write_name_quoted(file->name, options->file, '\0');
1809                }
1810        }
1811}
1812
1813struct dirstat_file {
1814        const char *name;
1815        unsigned long changed;
1816};
1817
1818struct dirstat_dir {
1819        struct dirstat_file *files;
1820        int alloc, nr, permille, cumulative;
1821};
1822
1823static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1824                unsigned long changed, const char *base, int baselen)
1825{
1826        unsigned long this_dir = 0;
1827        unsigned int sources = 0;
1828        const char *line_prefix = "";
1829        struct strbuf *msg = NULL;
1830
1831        if (opt->output_prefix) {
1832                msg = opt->output_prefix(opt, opt->output_prefix_data);
1833                line_prefix = msg->buf;
1834        }
1835
1836        while (dir->nr) {
1837                struct dirstat_file *f = dir->files;
1838                int namelen = strlen(f->name);
1839                unsigned long this;
1840                char *slash;
1841
1842                if (namelen < baselen)
1843                        break;
1844                if (memcmp(f->name, base, baselen))
1845                        break;
1846                slash = strchr(f->name + baselen, '/');
1847                if (slash) {
1848                        int newbaselen = slash + 1 - f->name;
1849                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1850                        sources++;
1851                } else {
1852                        this = f->changed;
1853                        dir->files++;
1854                        dir->nr--;
1855                        sources += 2;
1856                }
1857                this_dir += this;
1858        }
1859
1860        /*
1861         * We don't report dirstat's for
1862         *  - the top level
1863         *  - or cases where everything came from a single directory
1864         *    under this directory (sources == 1).
1865         */
1866        if (baselen && sources != 1) {
1867                if (this_dir) {
1868                        int permille = this_dir * 1000 / changed;
1869                        if (permille >= dir->permille) {
1870                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1871                                        permille / 10, permille % 10, baselen, base);
1872                                if (!dir->cumulative)
1873                                        return 0;
1874                        }
1875                }
1876        }
1877        return this_dir;
1878}
1879
1880static int dirstat_compare(const void *_a, const void *_b)
1881{
1882        const struct dirstat_file *a = _a;
1883        const struct dirstat_file *b = _b;
1884        return strcmp(a->name, b->name);
1885}
1886
1887static void show_dirstat(struct diff_options *options)
1888{
1889        int i;
1890        unsigned long changed;
1891        struct dirstat_dir dir;
1892        struct diff_queue_struct *q = &diff_queued_diff;
1893
1894        dir.files = NULL;
1895        dir.alloc = 0;
1896        dir.nr = 0;
1897        dir.permille = options->dirstat_permille;
1898        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1899
1900        changed = 0;
1901        for (i = 0; i < q->nr; i++) {
1902                struct diff_filepair *p = q->queue[i];
1903                const char *name;
1904                unsigned long copied, added, damage;
1905                int content_changed;
1906
1907                name = p->two->path ? p->two->path : p->one->path;
1908
1909                if (p->one->sha1_valid && p->two->sha1_valid)
1910                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1911                else
1912                        content_changed = 1;
1913
1914                if (!content_changed) {
1915                        /*
1916                         * The SHA1 has not changed, so pre-/post-content is
1917                         * identical. We can therefore skip looking at the
1918                         * file contents altogether.
1919                         */
1920                        damage = 0;
1921                        goto found_damage;
1922                }
1923
1924                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1925                        /*
1926                         * In --dirstat-by-file mode, we don't really need to
1927                         * look at the actual file contents at all.
1928                         * The fact that the SHA1 changed is enough for us to
1929                         * add this file to the list of results
1930                         * (with each file contributing equal damage).
1931                         */
1932                        damage = 1;
1933                        goto found_damage;
1934                }
1935
1936                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1937                        diff_populate_filespec(p->one, 0);
1938                        diff_populate_filespec(p->two, 0);
1939                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1940                                               &copied, &added);
1941                        diff_free_filespec_data(p->one);
1942                        diff_free_filespec_data(p->two);
1943                } else if (DIFF_FILE_VALID(p->one)) {
1944                        diff_populate_filespec(p->one, 1);
1945                        copied = added = 0;
1946                        diff_free_filespec_data(p->one);
1947                } else if (DIFF_FILE_VALID(p->two)) {
1948                        diff_populate_filespec(p->two, 1);
1949                        copied = 0;
1950                        added = p->two->size;
1951                        diff_free_filespec_data(p->two);
1952                } else
1953                        continue;
1954
1955                /*
1956                 * Original minus copied is the removed material,
1957                 * added is the new material.  They are both damages
1958                 * made to the preimage.
1959                 * If the resulting damage is zero, we know that
1960                 * diffcore_count_changes() considers the two entries to
1961                 * be identical, but since content_changed is true, we
1962                 * know that there must have been _some_ kind of change,
1963                 * so we force all entries to have damage > 0.
1964                 */
1965                damage = (p->one->size - copied) + added;
1966                if (!damage)
1967                        damage = 1;
1968
1969found_damage:
1970                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1971                dir.files[dir.nr].name = name;
1972                dir.files[dir.nr].changed = damage;
1973                changed += damage;
1974                dir.nr++;
1975        }
1976
1977        /* This can happen even with many files, if everything was renames */
1978        if (!changed)
1979                return;
1980
1981        /* Show all directories with more than x% of the changes */
1982        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1983        gather_dirstat(options, &dir, changed, "", 0);
1984}
1985
1986static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1987{
1988        int i;
1989        unsigned long changed;
1990        struct dirstat_dir dir;
1991
1992        if (data->nr == 0)
1993                return;
1994
1995        dir.files = NULL;
1996        dir.alloc = 0;
1997        dir.nr = 0;
1998        dir.permille = options->dirstat_permille;
1999        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2000
2001        changed = 0;
2002        for (i = 0; i < data->nr; i++) {
2003                struct diffstat_file *file = data->files[i];
2004                unsigned long damage = file->added + file->deleted;
2005                if (file->is_binary)
2006                        /*
2007                         * binary files counts bytes, not lines. Must find some
2008                         * way to normalize binary bytes vs. textual lines.
2009                         * The following heuristic assumes that there are 64
2010                         * bytes per "line".
2011                         * This is stupid and ugly, but very cheap...
2012                         */
2013                        damage = (damage + 63) / 64;
2014                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2015                dir.files[dir.nr].name = file->name;
2016                dir.files[dir.nr].changed = damage;
2017                changed += damage;
2018                dir.nr++;
2019        }
2020
2021        /* This can happen even with many files, if everything was renames */
2022        if (!changed)
2023                return;
2024
2025        /* Show all directories with more than x% of the changes */
2026        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2027        gather_dirstat(options, &dir, changed, "", 0);
2028}
2029
2030static void free_diffstat_info(struct diffstat_t *diffstat)
2031{
2032        int i;
2033        for (i = 0; i < diffstat->nr; i++) {
2034                struct diffstat_file *f = diffstat->files[i];
2035                if (f->name != f->print_name)
2036                        free(f->print_name);
2037                free(f->name);
2038                free(f->from_name);
2039                free(f);
2040        }
2041        free(diffstat->files);
2042}
2043
2044struct checkdiff_t {
2045        const char *filename;
2046        int lineno;
2047        int conflict_marker_size;
2048        struct diff_options *o;
2049        unsigned ws_rule;
2050        unsigned status;
2051};
2052
2053static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2054{
2055        char firstchar;
2056        int cnt;
2057
2058        if (len < marker_size + 1)
2059                return 0;
2060        firstchar = line[0];
2061        switch (firstchar) {
2062        case '=': case '>': case '<': case '|':
2063                break;
2064        default:
2065                return 0;
2066        }
2067        for (cnt = 1; cnt < marker_size; cnt++)
2068                if (line[cnt] != firstchar)
2069                        return 0;
2070        /* line[1] thru line[marker_size-1] are same as firstchar */
2071        if (len < marker_size + 1 || !isspace(line[marker_size]))
2072                return 0;
2073        return 1;
2074}
2075
2076static void checkdiff_consume(void *priv, char *line, unsigned long len)
2077{
2078        struct checkdiff_t *data = priv;
2079        int marker_size = data->conflict_marker_size;
2080        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2081        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2082        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2083        char *err;
2084        char *line_prefix = "";
2085        struct strbuf *msgbuf;
2086
2087        assert(data->o);
2088        if (data->o->output_prefix) {
2089                msgbuf = data->o->output_prefix(data->o,
2090                        data->o->output_prefix_data);
2091                line_prefix = msgbuf->buf;
2092        }
2093
2094        if (line[0] == '+') {
2095                unsigned bad;
2096                data->lineno++;
2097                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2098                        data->status |= 1;
2099                        fprintf(data->o->file,
2100                                "%s%s:%d: leftover conflict marker\n",
2101                                line_prefix, data->filename, data->lineno);
2102                }
2103                bad = ws_check(line + 1, len - 1, data->ws_rule);
2104                if (!bad)
2105                        return;
2106                data->status |= bad;
2107                err = whitespace_error_string(bad);
2108                fprintf(data->o->file, "%s%s:%d: %s.\n",
2109                        line_prefix, data->filename, data->lineno, err);
2110                free(err);
2111                emit_line(data->o, set, reset, line, 1);
2112                ws_check_emit(line + 1, len - 1, data->ws_rule,
2113                              data->o->file, set, reset, ws);
2114        } else if (line[0] == ' ') {
2115                data->lineno++;
2116        } else if (line[0] == '@') {
2117                char *plus = strchr(line, '+');
2118                if (plus)
2119                        data->lineno = strtol(plus, NULL, 10) - 1;
2120                else
2121                        die("invalid diff");
2122        }
2123}
2124
2125static unsigned char *deflate_it(char *data,
2126                                 unsigned long size,
2127                                 unsigned long *result_size)
2128{
2129        int bound;
2130        unsigned char *deflated;
2131        git_zstream stream;
2132
2133        memset(&stream, 0, sizeof(stream));
2134        git_deflate_init(&stream, zlib_compression_level);
2135        bound = git_deflate_bound(&stream, size);
2136        deflated = xmalloc(bound);
2137        stream.next_out = deflated;
2138        stream.avail_out = bound;
2139
2140        stream.next_in = (unsigned char *)data;
2141        stream.avail_in = size;
2142        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2143                ; /* nothing */
2144        git_deflate_end(&stream);
2145        *result_size = stream.total_out;
2146        return deflated;
2147}
2148
2149static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2150{
2151        void *cp;
2152        void *delta;
2153        void *deflated;
2154        void *data;
2155        unsigned long orig_size;
2156        unsigned long delta_size;
2157        unsigned long deflate_size;
2158        unsigned long data_size;
2159
2160        /* We could do deflated delta, or we could do just deflated two,
2161         * whichever is smaller.
2162         */
2163        delta = NULL;
2164        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2165        if (one->size && two->size) {
2166                delta = diff_delta(one->ptr, one->size,
2167                                   two->ptr, two->size,
2168                                   &delta_size, deflate_size);
2169                if (delta) {
2170                        void *to_free = delta;
2171                        orig_size = delta_size;
2172                        delta = deflate_it(delta, delta_size, &delta_size);
2173                        free(to_free);
2174                }
2175        }
2176
2177        if (delta && delta_size < deflate_size) {
2178                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2179                free(deflated);
2180                data = delta;
2181                data_size = delta_size;
2182        }
2183        else {
2184                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2185                free(delta);
2186                data = deflated;
2187                data_size = deflate_size;
2188        }
2189
2190        /* emit data encoded in base85 */
2191        cp = data;
2192        while (data_size) {
2193                int bytes = (52 < data_size) ? 52 : data_size;
2194                char line[70];
2195                data_size -= bytes;
2196                if (bytes <= 26)
2197                        line[0] = bytes + 'A' - 1;
2198                else
2199                        line[0] = bytes - 26 + 'a' - 1;
2200                encode_85(line + 1, cp, bytes);
2201                cp = (char *) cp + bytes;
2202                fprintf(file, "%s", prefix);
2203                fputs(line, file);
2204                fputc('\n', file);
2205        }
2206        fprintf(file, "%s\n", prefix);
2207        free(data);
2208}
2209
2210static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2211{
2212        fprintf(file, "%sGIT binary patch\n", prefix);
2213        emit_binary_diff_body(file, one, two, prefix);
2214        emit_binary_diff_body(file, two, one, prefix);
2215}
2216
2217int diff_filespec_is_binary(struct diff_filespec *one)
2218{
2219        if (one->is_binary == -1) {
2220                diff_filespec_load_driver(one);
2221                if (one->driver->binary != -1)
2222                        one->is_binary = one->driver->binary;
2223                else {
2224                        if (!one->data && DIFF_FILE_VALID(one))
2225                                diff_populate_filespec(one, 0);
2226                        if (one->data)
2227                                one->is_binary = buffer_is_binary(one->data,
2228                                                one->size);
2229                        if (one->is_binary == -1)
2230                                one->is_binary = 0;
2231                }
2232        }
2233        return one->is_binary;
2234}
2235
2236static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2237{
2238        diff_filespec_load_driver(one);
2239        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2240}
2241
2242void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2243{
2244        if (!options->a_prefix)
2245                options->a_prefix = a;
2246        if (!options->b_prefix)
2247                options->b_prefix = b;
2248}
2249
2250struct userdiff_driver *get_textconv(struct diff_filespec *one)
2251{
2252        if (!DIFF_FILE_VALID(one))
2253                return NULL;
2254
2255        diff_filespec_load_driver(one);
2256        return userdiff_get_textconv(one->driver);
2257}
2258
2259static void builtin_diff(const char *name_a,
2260                         const char *name_b,
2261                         struct diff_filespec *one,
2262                         struct diff_filespec *two,
2263                         const char *xfrm_msg,
2264                         int must_show_header,
2265                         struct diff_options *o,
2266                         int complete_rewrite)
2267{
2268        mmfile_t mf1, mf2;
2269        const char *lbl[2];
2270        char *a_one, *b_two;
2271        const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2272        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2273        const char *a_prefix, *b_prefix;
2274        struct userdiff_driver *textconv_one = NULL;
2275        struct userdiff_driver *textconv_two = NULL;
2276        struct strbuf header = STRBUF_INIT;
2277        struct strbuf *msgbuf;
2278        char *line_prefix = "";
2279
2280        if (o->output_prefix) {
2281                msgbuf = o->output_prefix(o, o->output_prefix_data);
2282                line_prefix = msgbuf->buf;
2283        }
2284
2285        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2286                        (!one->mode || S_ISGITLINK(one->mode)) &&
2287                        (!two->mode || S_ISGITLINK(two->mode))) {
2288                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2289                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2290                show_submodule_summary(o->file, one ? one->path : two->path,
2291                                one->sha1, two->sha1, two->dirty_submodule,
2292                                meta, del, add, reset);
2293                return;
2294        }
2295
2296        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2297                textconv_one = get_textconv(one);
2298                textconv_two = get_textconv(two);
2299        }
2300
2301        diff_set_mnemonic_prefix(o, "a/", "b/");
2302        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2303                a_prefix = o->b_prefix;
2304                b_prefix = o->a_prefix;
2305        } else {
2306                a_prefix = o->a_prefix;
2307                b_prefix = o->b_prefix;
2308        }
2309
2310        /* Never use a non-valid filename anywhere if at all possible */
2311        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2312        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2313
2314        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2315        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2316        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2317        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2318        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2319        if (lbl[0][0] == '/') {
2320                /* /dev/null */
2321                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2322                if (xfrm_msg)
2323                        strbuf_addstr(&header, xfrm_msg);
2324                must_show_header = 1;
2325        }
2326        else if (lbl[1][0] == '/') {
2327                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2328                if (xfrm_msg)
2329                        strbuf_addstr(&header, xfrm_msg);
2330                must_show_header = 1;
2331        }
2332        else {
2333                if (one->mode != two->mode) {
2334                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2335                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2336                        must_show_header = 1;
2337                }
2338                if (xfrm_msg)
2339                        strbuf_addstr(&header, xfrm_msg);
2340
2341                /*
2342                 * we do not run diff between different kind
2343                 * of objects.
2344                 */
2345                if ((one->mode ^ two->mode) & S_IFMT)
2346                        goto free_ab_and_return;
2347                if (complete_rewrite &&
2348                    (textconv_one || !diff_filespec_is_binary(one)) &&
2349                    (textconv_two || !diff_filespec_is_binary(two))) {
2350                        fprintf(o->file, "%s", header.buf);
2351                        strbuf_reset(&header);
2352                        emit_rewrite_diff(name_a, name_b, one, two,
2353                                                textconv_one, textconv_two, o);
2354                        o->found_changes = 1;
2355                        goto free_ab_and_return;
2356                }
2357        }
2358
2359        if (o->irreversible_delete && lbl[1][0] == '/') {
2360                fprintf(o->file, "%s", header.buf);
2361                strbuf_reset(&header);
2362                goto free_ab_and_return;
2363        } else if (!DIFF_OPT_TST(o, TEXT) &&
2364            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2365              (!textconv_two && diff_filespec_is_binary(two)) )) {
2366                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2367                        die("unable to read files to diff");
2368                /* Quite common confusing case */
2369                if (mf1.size == mf2.size &&
2370                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2371                        if (must_show_header)
2372                                fprintf(o->file, "%s", header.buf);
2373                        goto free_ab_and_return;
2374                }
2375                fprintf(o->file, "%s", header.buf);
2376                strbuf_reset(&header);
2377                if (DIFF_OPT_TST(o, BINARY))
2378                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2379                else
2380                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2381                                line_prefix, lbl[0], lbl[1]);
2382                o->found_changes = 1;
2383        } else {
2384                /* Crazy xdl interfaces.. */
2385                const char *diffopts = getenv("GIT_DIFF_OPTS");
2386                xpparam_t xpp;
2387                xdemitconf_t xecfg;
2388                struct emit_callback ecbdata;
2389                const struct userdiff_funcname *pe;
2390
2391                if (must_show_header) {
2392                        fprintf(o->file, "%s", header.buf);
2393                        strbuf_reset(&header);
2394                }
2395
2396                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2397                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2398
2399                pe = diff_funcname_pattern(one);
2400                if (!pe)
2401                        pe = diff_funcname_pattern(two);
2402
2403                memset(&xpp, 0, sizeof(xpp));
2404                memset(&xecfg, 0, sizeof(xecfg));
2405                memset(&ecbdata, 0, sizeof(ecbdata));
2406                ecbdata.label_path = lbl;
2407                ecbdata.color_diff = want_color(o->use_color);
2408                ecbdata.found_changesp = &o->found_changes;
2409                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2410                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2411                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2412                ecbdata.opt = o;
2413                ecbdata.header = header.len ? &header : NULL;
2414                xpp.flags = o->xdl_opts;
2415                xecfg.ctxlen = o->context;
2416                xecfg.interhunkctxlen = o->interhunkcontext;
2417                xecfg.flags = XDL_EMIT_FUNCNAMES;
2418                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2419                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2420                if (pe)
2421                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2422                if (!diffopts)
2423                        ;
2424                else if (!prefixcmp(diffopts, "--unified="))
2425                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2426                else if (!prefixcmp(diffopts, "-u"))
2427                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2428                if (o->word_diff)
2429                        init_diff_words_data(&ecbdata, o, one, two);
2430                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2431                              &xpp, &xecfg);
2432                if (o->word_diff)
2433                        free_diff_words_data(&ecbdata);
2434                if (textconv_one)
2435                        free(mf1.ptr);
2436                if (textconv_two)
2437                        free(mf2.ptr);
2438                xdiff_clear_find_func(&xecfg);
2439        }
2440
2441 free_ab_and_return:
2442        strbuf_release(&header);
2443        diff_free_filespec_data(one);
2444        diff_free_filespec_data(two);
2445        free(a_one);
2446        free(b_two);
2447        return;
2448}
2449
2450static void builtin_diffstat(const char *name_a, const char *name_b,
2451                             struct diff_filespec *one,
2452                             struct diff_filespec *two,
2453                             struct diffstat_t *diffstat,
2454                             struct diff_options *o,
2455                             struct diff_filepair *p)
2456{
2457        mmfile_t mf1, mf2;
2458        struct diffstat_file *data;
2459        int same_contents;
2460        int complete_rewrite = 0;
2461
2462        if (!DIFF_PAIR_UNMERGED(p)) {
2463                if (p->status == DIFF_STATUS_MODIFIED && p->score)
2464                        complete_rewrite = 1;
2465        }
2466
2467        data = diffstat_add(diffstat, name_a, name_b);
2468        data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2469
2470        if (!one || !two) {
2471                data->is_unmerged = 1;
2472                return;
2473        }
2474
2475        same_contents = !hashcmp(one->sha1, two->sha1);
2476
2477        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2478                data->is_binary = 1;
2479                if (same_contents) {
2480                        data->added = 0;
2481                        data->deleted = 0;
2482                } else {
2483                        data->added = diff_filespec_size(two);
2484                        data->deleted = diff_filespec_size(one);
2485                }
2486        }
2487
2488        else if (complete_rewrite) {
2489                diff_populate_filespec(one, 0);
2490                diff_populate_filespec(two, 0);
2491                data->deleted = count_lines(one->data, one->size);
2492                data->added = count_lines(two->data, two->size);
2493        }
2494
2495        else if (!same_contents) {
2496                /* Crazy xdl interfaces.. */
2497                xpparam_t xpp;
2498                xdemitconf_t xecfg;
2499
2500                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2501                        die("unable to read files to diff");
2502
2503                memset(&xpp, 0, sizeof(xpp));
2504                memset(&xecfg, 0, sizeof(xecfg));
2505                xpp.flags = o->xdl_opts;
2506                xecfg.ctxlen = o->context;
2507                xecfg.interhunkctxlen = o->interhunkcontext;
2508                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2509                              &xpp, &xecfg);
2510        }
2511
2512        diff_free_filespec_data(one);
2513        diff_free_filespec_data(two);
2514}
2515
2516static void builtin_checkdiff(const char *name_a, const char *name_b,
2517                              const char *attr_path,
2518                              struct diff_filespec *one,
2519                              struct diff_filespec *two,
2520                              struct diff_options *o)
2521{
2522        mmfile_t mf1, mf2;
2523        struct checkdiff_t data;
2524
2525        if (!two)
2526                return;
2527
2528        memset(&data, 0, sizeof(data));
2529        data.filename = name_b ? name_b : name_a;
2530        data.lineno = 0;
2531        data.o = o;
2532        data.ws_rule = whitespace_rule(attr_path);
2533        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2534
2535        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2536                die("unable to read files to diff");
2537
2538        /*
2539         * All the other codepaths check both sides, but not checking
2540         * the "old" side here is deliberate.  We are checking the newly
2541         * introduced changes, and as long as the "new" side is text, we
2542         * can and should check what it introduces.
2543         */
2544        if (diff_filespec_is_binary(two))
2545                goto free_and_return;
2546        else {
2547                /* Crazy xdl interfaces.. */
2548                xpparam_t xpp;
2549                xdemitconf_t xecfg;
2550
2551                memset(&xpp, 0, sizeof(xpp));
2552                memset(&xecfg, 0, sizeof(xecfg));
2553                xecfg.ctxlen = 1; /* at least one context line */
2554                xpp.flags = 0;
2555                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2556                              &xpp, &xecfg);
2557
2558                if (data.ws_rule & WS_BLANK_AT_EOF) {
2559                        struct emit_callback ecbdata;
2560                        int blank_at_eof;
2561
2562                        ecbdata.ws_rule = data.ws_rule;
2563                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2564                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2565
2566                        if (blank_at_eof) {
2567                                static char *err;
2568                                if (!err)
2569                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2570                                fprintf(o->file, "%s:%d: %s.\n",
2571                                        data.filename, blank_at_eof, err);
2572                                data.status = 1; /* report errors */
2573                        }
2574                }
2575        }
2576 free_and_return:
2577        diff_free_filespec_data(one);
2578        diff_free_filespec_data(two);
2579        if (data.status)
2580                DIFF_OPT_SET(o, CHECK_FAILED);
2581}
2582
2583struct diff_filespec *alloc_filespec(const char *path)
2584{
2585        int namelen = strlen(path);
2586        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2587
2588        memset(spec, 0, sizeof(*spec));
2589        spec->path = (char *)(spec + 1);
2590        memcpy(spec->path, path, namelen+1);
2591        spec->count = 1;
2592        spec->is_binary = -1;
2593        return spec;
2594}
2595
2596void free_filespec(struct diff_filespec *spec)
2597{
2598        if (!--spec->count) {
2599                diff_free_filespec_data(spec);
2600                free(spec);
2601        }
2602}
2603
2604void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2605                   int sha1_valid, unsigned short mode)
2606{
2607        if (mode) {
2608                spec->mode = canon_mode(mode);
2609                hashcpy(spec->sha1, sha1);
2610                spec->sha1_valid = sha1_valid;
2611        }
2612}
2613
2614/*
2615 * Given a name and sha1 pair, if the index tells us the file in
2616 * the work tree has that object contents, return true, so that
2617 * prepare_temp_file() does not have to inflate and extract.
2618 */
2619static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2620{
2621        struct cache_entry *ce;
2622        struct stat st;
2623        int pos, len;
2624
2625        /*
2626         * We do not read the cache ourselves here, because the
2627         * benchmark with my previous version that always reads cache
2628         * shows that it makes things worse for diff-tree comparing
2629         * two linux-2.6 kernel trees in an already checked out work
2630         * tree.  This is because most diff-tree comparisons deal with
2631         * only a small number of files, while reading the cache is
2632         * expensive for a large project, and its cost outweighs the
2633         * savings we get by not inflating the object to a temporary
2634         * file.  Practically, this code only helps when we are used
2635         * by diff-cache --cached, which does read the cache before
2636         * calling us.
2637         */
2638        if (!active_cache)
2639                return 0;
2640
2641        /* We want to avoid the working directory if our caller
2642         * doesn't need the data in a normal file, this system
2643         * is rather slow with its stat/open/mmap/close syscalls,
2644         * and the object is contained in a pack file.  The pack
2645         * is probably already open and will be faster to obtain
2646         * the data through than the working directory.  Loose
2647         * objects however would tend to be slower as they need
2648         * to be individually opened and inflated.
2649         */
2650        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2651                return 0;
2652
2653        len = strlen(name);
2654        pos = cache_name_pos(name, len);
2655        if (pos < 0)
2656                return 0;
2657        ce = active_cache[pos];
2658
2659        /*
2660         * This is not the sha1 we are looking for, or
2661         * unreusable because it is not a regular file.
2662         */
2663        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2664                return 0;
2665
2666        /*
2667         * If ce is marked as "assume unchanged", there is no
2668         * guarantee that work tree matches what we are looking for.
2669         */
2670        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2671                return 0;
2672
2673        /*
2674         * If ce matches the file in the work tree, we can reuse it.
2675         */
2676        if (ce_uptodate(ce) ||
2677            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2678                return 1;
2679
2680        return 0;
2681}
2682
2683static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2684{
2685        int len;
2686        char *data = xmalloc(100), *dirty = "";
2687
2688        /* Are we looking at the work tree? */
2689        if (s->dirty_submodule)
2690                dirty = "-dirty";
2691
2692        len = snprintf(data, 100,
2693                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2694        s->data = data;
2695        s->size = len;
2696        s->should_free = 1;
2697        if (size_only) {
2698                s->data = NULL;
2699                free(data);
2700        }
2701        return 0;
2702}
2703
2704/*
2705 * While doing rename detection and pickaxe operation, we may need to
2706 * grab the data for the blob (or file) for our own in-core comparison.
2707 * diff_filespec has data and size fields for this purpose.
2708 */
2709int diff_populate_filespec(struct diff_filespec *s, int size_only)
2710{
2711        int err = 0;
2712        if (!DIFF_FILE_VALID(s))
2713                die("internal error: asking to populate invalid file.");
2714        if (S_ISDIR(s->mode))
2715                return -1;
2716
2717        if (s->data)
2718                return 0;
2719
2720        if (size_only && 0 < s->size)
2721                return 0;
2722
2723        if (S_ISGITLINK(s->mode))
2724                return diff_populate_gitlink(s, size_only);
2725
2726        if (!s->sha1_valid ||
2727            reuse_worktree_file(s->path, s->sha1, 0)) {
2728                struct strbuf buf = STRBUF_INIT;
2729                struct stat st;
2730                int fd;
2731
2732                if (lstat(s->path, &st) < 0) {
2733                        if (errno == ENOENT) {
2734                        err_empty:
2735                                err = -1;
2736                        empty:
2737                                s->data = (char *)"";
2738                                s->size = 0;
2739                                return err;
2740                        }
2741                }
2742                s->size = xsize_t(st.st_size);
2743                if (!s->size)
2744                        goto empty;
2745                if (S_ISLNK(st.st_mode)) {
2746                        struct strbuf sb = STRBUF_INIT;
2747
2748                        if (strbuf_readlink(&sb, s->path, s->size))
2749                                goto err_empty;
2750                        s->size = sb.len;
2751                        s->data = strbuf_detach(&sb, NULL);
2752                        s->should_free = 1;
2753                        return 0;
2754                }
2755                if (size_only)
2756                        return 0;
2757                fd = open(s->path, O_RDONLY);
2758                if (fd < 0)
2759                        goto err_empty;
2760                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2761                close(fd);
2762                s->should_munmap = 1;
2763
2764                /*
2765                 * Convert from working tree format to canonical git format
2766                 */
2767                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2768                        size_t size = 0;
2769                        munmap(s->data, s->size);
2770                        s->should_munmap = 0;
2771                        s->data = strbuf_detach(&buf, &size);
2772                        s->size = size;
2773                        s->should_free = 1;
2774                }
2775        }
2776        else {
2777                enum object_type type;
2778                if (size_only) {
2779                        type = sha1_object_info(s->sha1, &s->size);
2780                        if (type < 0)
2781                                die("unable to read %s", sha1_to_hex(s->sha1));
2782                } else {
2783                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2784                        if (!s->data)
2785                                die("unable to read %s", sha1_to_hex(s->sha1));
2786                        s->should_free = 1;
2787                }
2788        }
2789        return 0;
2790}
2791
2792void diff_free_filespec_blob(struct diff_filespec *s)
2793{
2794        if (s->should_free)
2795                free(s->data);
2796        else if (s->should_munmap)
2797                munmap(s->data, s->size);
2798
2799        if (s->should_free || s->should_munmap) {
2800                s->should_free = s->should_munmap = 0;
2801                s->data = NULL;
2802        }
2803}
2804
2805void diff_free_filespec_data(struct diff_filespec *s)
2806{
2807        diff_free_filespec_blob(s);
2808        free(s->cnt_data);
2809        s->cnt_data = NULL;
2810}
2811
2812static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2813                           void *blob,
2814                           unsigned long size,
2815                           const unsigned char *sha1,
2816                           int mode)
2817{
2818        int fd;
2819        struct strbuf buf = STRBUF_INIT;
2820        struct strbuf template = STRBUF_INIT;
2821        char *path_dup = xstrdup(path);
2822        const char *base = basename(path_dup);
2823
2824        /* Generate "XXXXXX_basename.ext" */
2825        strbuf_addstr(&template, "XXXXXX_");
2826        strbuf_addstr(&template, base);
2827
2828        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2829                        strlen(base) + 1);
2830        if (fd < 0)
2831                die_errno("unable to create temp-file");
2832        if (convert_to_working_tree(path,
2833                        (const char *)blob, (size_t)size, &buf)) {
2834                blob = buf.buf;
2835                size = buf.len;
2836        }
2837        if (write_in_full(fd, blob, size) != size)
2838                die_errno("unable to write temp-file");
2839        close(fd);
2840        temp->name = temp->tmp_path;
2841        strcpy(temp->hex, sha1_to_hex(sha1));
2842        temp->hex[40] = 0;
2843        sprintf(temp->mode, "%06o", mode);
2844        strbuf_release(&buf);
2845        strbuf_release(&template);
2846        free(path_dup);
2847}
2848
2849static struct diff_tempfile *prepare_temp_file(const char *name,
2850                struct diff_filespec *one)
2851{
2852        struct diff_tempfile *temp = claim_diff_tempfile();
2853
2854        if (!DIFF_FILE_VALID(one)) {
2855        not_a_valid_file:
2856                /* A '-' entry produces this for file-2, and
2857                 * a '+' entry produces this for file-1.
2858                 */
2859                temp->name = "/dev/null";
2860                strcpy(temp->hex, ".");
2861                strcpy(temp->mode, ".");
2862                return temp;
2863        }
2864
2865        if (!remove_tempfile_installed) {
2866                atexit(remove_tempfile);
2867                sigchain_push_common(remove_tempfile_on_signal);
2868                remove_tempfile_installed = 1;
2869        }
2870
2871        if (!one->sha1_valid ||
2872            reuse_worktree_file(name, one->sha1, 1)) {
2873                struct stat st;
2874                if (lstat(name, &st) < 0) {
2875                        if (errno == ENOENT)
2876                                goto not_a_valid_file;
2877                        die_errno("stat(%s)", name);
2878                }
2879                if (S_ISLNK(st.st_mode)) {
2880                        struct strbuf sb = STRBUF_INIT;
2881                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2882                                die_errno("readlink(%s)", name);
2883                        prep_temp_blob(name, temp, sb.buf, sb.len,
2884                                       (one->sha1_valid ?
2885                                        one->sha1 : null_sha1),
2886                                       (one->sha1_valid ?
2887                                        one->mode : S_IFLNK));
2888                        strbuf_release(&sb);
2889                }
2890                else {
2891                        /* we can borrow from the file in the work tree */
2892                        temp->name = name;
2893                        if (!one->sha1_valid)
2894                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2895                        else
2896                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2897                        /* Even though we may sometimes borrow the
2898                         * contents from the work tree, we always want
2899                         * one->mode.  mode is trustworthy even when
2900                         * !(one->sha1_valid), as long as
2901                         * DIFF_FILE_VALID(one).
2902                         */
2903                        sprintf(temp->mode, "%06o", one->mode);
2904                }
2905                return temp;
2906        }
2907        else {
2908                if (diff_populate_filespec(one, 0))
2909                        die("cannot read data blob for %s", one->path);
2910                prep_temp_blob(name, temp, one->data, one->size,
2911                               one->sha1, one->mode);
2912        }
2913        return temp;
2914}
2915
2916/* An external diff command takes:
2917 *
2918 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2919 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2920 *
2921 */
2922static void run_external_diff(const char *pgm,
2923                              const char *name,
2924                              const char *other,
2925                              struct diff_filespec *one,
2926                              struct diff_filespec *two,
2927                              const char *xfrm_msg,
2928                              int complete_rewrite)
2929{
2930        const char *spawn_arg[10];
2931        int retval;
2932        const char **arg = &spawn_arg[0];
2933
2934        if (one && two) {
2935                struct diff_tempfile *temp_one, *temp_two;
2936                const char *othername = (other ? other : name);
2937                temp_one = prepare_temp_file(name, one);
2938                temp_two = prepare_temp_file(othername, two);
2939                *arg++ = pgm;
2940                *arg++ = name;
2941                *arg++ = temp_one->name;
2942                *arg++ = temp_one->hex;
2943                *arg++ = temp_one->mode;
2944                *arg++ = temp_two->name;
2945                *arg++ = temp_two->hex;
2946                *arg++ = temp_two->mode;
2947                if (other) {
2948                        *arg++ = other;
2949                        *arg++ = xfrm_msg;
2950                }
2951        } else {
2952                *arg++ = pgm;
2953                *arg++ = name;
2954        }
2955        *arg = NULL;
2956        fflush(NULL);
2957        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2958        remove_tempfile();
2959        if (retval) {
2960                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2961                exit(1);
2962        }
2963}
2964
2965static int similarity_index(struct diff_filepair *p)
2966{
2967        return p->score * 100 / MAX_SCORE;
2968}
2969
2970static void fill_metainfo(struct strbuf *msg,
2971                          const char *name,
2972                          const char *other,
2973                          struct diff_filespec *one,
2974                          struct diff_filespec *two,
2975                          struct diff_options *o,
2976                          struct diff_filepair *p,
2977                          int *must_show_header,
2978                          int use_color)
2979{
2980        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2981        const char *reset = diff_get_color(use_color, DIFF_RESET);
2982        struct strbuf *msgbuf;
2983        char *line_prefix = "";
2984
2985        *must_show_header = 1;
2986        if (o->output_prefix) {
2987                msgbuf = o->output_prefix(o, o->output_prefix_data);
2988                line_prefix = msgbuf->buf;
2989        }
2990        strbuf_init(msg, PATH_MAX * 2 + 300);
2991        switch (p->status) {
2992        case DIFF_STATUS_COPIED:
2993                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2994                            line_prefix, set, similarity_index(p));
2995                strbuf_addf(msg, "%s\n%s%scopy from ",
2996                            reset,  line_prefix, set);
2997                quote_c_style(name, msg, NULL, 0);
2998                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2999                quote_c_style(other, msg, NULL, 0);
3000                strbuf_addf(msg, "%s\n", reset);
3001                break;
3002        case DIFF_STATUS_RENAMED:
3003                strbuf_addf(msg, "%s%ssimilarity index %d%%",
3004                            line_prefix, set, similarity_index(p));
3005                strbuf_addf(msg, "%s\n%s%srename from ",
3006                            reset, line_prefix, set);
3007                quote_c_style(name, msg, NULL, 0);
3008                strbuf_addf(msg, "%s\n%s%srename to ",
3009                            reset, line_prefix, set);
3010                quote_c_style(other, msg, NULL, 0);
3011                strbuf_addf(msg, "%s\n", reset);
3012                break;
3013        case DIFF_STATUS_MODIFIED:
3014                if (p->score) {
3015                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3016                                    line_prefix,
3017                                    set, similarity_index(p), reset);
3018                        break;
3019                }
3020                /* fallthru */
3021        default:
3022                *must_show_header = 0;
3023        }
3024        if (one && two && hashcmp(one->sha1, two->sha1)) {
3025                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3026
3027                if (DIFF_OPT_TST(o, BINARY)) {
3028                        mmfile_t mf;
3029                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3030                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3031                                abbrev = 40;
3032                }
3033                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3034                            find_unique_abbrev(one->sha1, abbrev));
3035                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3036                if (one->mode == two->mode)
3037                        strbuf_addf(msg, " %06o", one->mode);
3038                strbuf_addf(msg, "%s\n", reset);
3039        }
3040}
3041
3042static void run_diff_cmd(const char *pgm,
3043                         const char *name,
3044                         const char *other,
3045                         const char *attr_path,
3046                         struct diff_filespec *one,
3047                         struct diff_filespec *two,
3048                         struct strbuf *msg,
3049                         struct diff_options *o,
3050                         struct diff_filepair *p)
3051{
3052        const char *xfrm_msg = NULL;
3053        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3054        int must_show_header = 0;
3055
3056
3057        if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3058                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3059                if (drv && drv->external)
3060                        pgm = drv->external;
3061        }
3062
3063        if (msg) {
3064                /*
3065                 * don't use colors when the header is intended for an
3066                 * external diff driver
3067                 */
3068                fill_metainfo(msg, name, other, one, two, o, p,
3069                              &must_show_header,
3070                              want_color(o->use_color) && !pgm);
3071                xfrm_msg = msg->len ? msg->buf : NULL;
3072        }
3073
3074        if (pgm) {
3075                run_external_diff(pgm, name, other, one, two, xfrm_msg,
3076                                  complete_rewrite);
3077                return;
3078        }
3079        if (one && two)
3080                builtin_diff(name, other ? other : name,
3081                             one, two, xfrm_msg, must_show_header,
3082                             o, complete_rewrite);
3083        else
3084                fprintf(o->file, "* Unmerged path %s\n", name);
3085}
3086
3087static void diff_fill_sha1_info(struct diff_filespec *one)
3088{
3089        if (DIFF_FILE_VALID(one)) {
3090                if (!one->sha1_valid) {
3091                        struct stat st;
3092                        if (one->is_stdin) {
3093                                hashcpy(one->sha1, null_sha1);
3094                                return;
3095                        }
3096                        if (lstat(one->path, &st) < 0)
3097                                die_errno("stat '%s'", one->path);
3098                        if (index_path(one->sha1, one->path, &st, 0))
3099                                die("cannot hash %s", one->path);
3100                }
3101        }
3102        else
3103                hashclr(one->sha1);
3104}
3105
3106static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3107{
3108        /* Strip the prefix but do not molest /dev/null and absolute paths */
3109        if (*namep && **namep != '/') {
3110                *namep += prefix_length;
3111                if (**namep == '/')
3112                        ++*namep;
3113        }
3114        if (*otherp && **otherp != '/') {
3115                *otherp += prefix_length;
3116                if (**otherp == '/')
3117                        ++*otherp;
3118        }
3119}
3120
3121static void run_diff(struct diff_filepair *p, struct diff_options *o)
3122{
3123        const char *pgm = external_diff();
3124        struct strbuf msg;
3125        struct diff_filespec *one = p->one;
3126        struct diff_filespec *two = p->two;
3127        const char *name;
3128        const char *other;
3129        const char *attr_path;
3130
3131        name  = p->one->path;
3132        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3133        attr_path = name;
3134        if (o->prefix_length)
3135                strip_prefix(o->prefix_length, &name, &other);
3136
3137        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3138                pgm = NULL;
3139
3140        if (DIFF_PAIR_UNMERGED(p)) {
3141                run_diff_cmd(pgm, name, NULL, attr_path,
3142                             NULL, NULL, NULL, o, p);
3143                return;
3144        }
3145
3146        diff_fill_sha1_info(one);
3147        diff_fill_sha1_info(two);
3148
3149        if (!pgm &&
3150            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3151            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3152                /*
3153                 * a filepair that changes between file and symlink
3154                 * needs to be split into deletion and creation.
3155                 */
3156                struct diff_filespec *null = alloc_filespec(two->path);
3157                run_diff_cmd(NULL, name, other, attr_path,
3158                             one, null, &msg, o, p);
3159                free(null);
3160                strbuf_release(&msg);
3161
3162                null = alloc_filespec(one->path);
3163                run_diff_cmd(NULL, name, other, attr_path,
3164                             null, two, &msg, o, p);
3165                free(null);
3166        }
3167        else
3168                run_diff_cmd(pgm, name, other, attr_path,
3169                             one, two, &msg, o, p);
3170
3171        strbuf_release(&msg);
3172}
3173
3174static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3175                         struct diffstat_t *diffstat)
3176{
3177        const char *name;
3178        const char *other;
3179
3180        if (DIFF_PAIR_UNMERGED(p)) {
3181                /* unmerged */
3182                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3183                return;
3184        }
3185
3186        name = p->one->path;
3187        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3188
3189        if (o->prefix_length)
3190                strip_prefix(o->prefix_length, &name, &other);
3191
3192        diff_fill_sha1_info(p->one);
3193        diff_fill_sha1_info(p->two);
3194
3195        builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3196}
3197
3198static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3199{
3200        const char *name;
3201        const char *other;
3202        const char *attr_path;
3203
3204        if (DIFF_PAIR_UNMERGED(p)) {
3205                /* unmerged */
3206                return;
3207        }
3208
3209        name = p->one->path;
3210        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3211        attr_path = other ? other : name;
3212
3213        if (o->prefix_length)
3214                strip_prefix(o->prefix_length, &name, &other);
3215
3216        diff_fill_sha1_info(p->one);
3217        diff_fill_sha1_info(p->two);
3218
3219        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3220}
3221
3222void diff_setup(struct diff_options *options)
3223{
3224        memcpy(options, &default_diff_options, sizeof(*options));
3225
3226        options->file = stdout;
3227
3228        options->line_termination = '\n';
3229        options->break_opt = -1;
3230        options->rename_limit = -1;
3231        options->dirstat_permille = diff_dirstat_permille_default;
3232        options->context = diff_context_default;
3233        DIFF_OPT_SET(options, RENAME_EMPTY);
3234
3235        options->change = diff_change;
3236        options->add_remove = diff_addremove;
3237        options->use_color = diff_use_color_default;
3238        options->detect_rename = diff_detect_rename_default;
3239        options->xdl_opts |= diff_algorithm;
3240
3241        if (diff_no_prefix) {
3242                options->a_prefix = options->b_prefix = "";
3243        } else if (!diff_mnemonic_prefix) {
3244                options->a_prefix = "a/";
3245                options->b_prefix = "b/";
3246        }
3247}
3248
3249void diff_setup_done(struct diff_options *options)
3250{
3251        int count = 0;
3252
3253        if (options->output_format & DIFF_FORMAT_NAME)
3254                count++;
3255        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3256                count++;
3257        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3258                count++;
3259        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3260                count++;
3261        if (count > 1)
3262                die("--name-only, --name-status, --check and -s are mutually exclusive");
3263
3264        /*
3265         * Most of the time we can say "there are changes"
3266         * only by checking if there are changed paths, but
3267         * --ignore-whitespace* options force us to look
3268         * inside contents.
3269         */
3270
3271        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3272            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3273            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3274                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3275        else
3276                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3277
3278        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3279                options->detect_rename = DIFF_DETECT_COPY;
3280
3281        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3282                options->prefix = NULL;
3283        if (options->prefix)
3284                options->prefix_length = strlen(options->prefix);
3285        else
3286                options->prefix_length = 0;
3287
3288        if (options->output_format & (DIFF_FORMAT_NAME |
3289                                      DIFF_FORMAT_NAME_STATUS |
3290                                      DIFF_FORMAT_CHECKDIFF |
3291                                      DIFF_FORMAT_NO_OUTPUT))
3292                options->output_format &= ~(DIFF_FORMAT_RAW |
3293                                            DIFF_FORMAT_NUMSTAT |
3294                                            DIFF_FORMAT_DIFFSTAT |
3295                                            DIFF_FORMAT_SHORTSTAT |
3296                                            DIFF_FORMAT_DIRSTAT |
3297                                            DIFF_FORMAT_SUMMARY |
3298                                            DIFF_FORMAT_PATCH);
3299
3300        /*
3301         * These cases always need recursive; we do not drop caller-supplied
3302         * recursive bits for other formats here.
3303         */
3304        if (options->output_format & (DIFF_FORMAT_PATCH |
3305                                      DIFF_FORMAT_NUMSTAT |
3306                                      DIFF_FORMAT_DIFFSTAT |
3307                                      DIFF_FORMAT_SHORTSTAT |
3308                                      DIFF_FORMAT_DIRSTAT |
3309                                      DIFF_FORMAT_SUMMARY |
3310                                      DIFF_FORMAT_CHECKDIFF))
3311                DIFF_OPT_SET(options, RECURSIVE);
3312        /*
3313         * Also pickaxe would not work very well if you do not say recursive
3314         */
3315        if (options->pickaxe)
3316                DIFF_OPT_SET(options, RECURSIVE);
3317        /*
3318         * When patches are generated, submodules diffed against the work tree
3319         * must be checked for dirtiness too so it can be shown in the output
3320         */
3321        if (options->output_format & DIFF_FORMAT_PATCH)
3322                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3323
3324        if (options->detect_rename && options->rename_limit < 0)
3325                options->rename_limit = diff_rename_limit_default;
3326        if (options->setup & DIFF_SETUP_USE_CACHE) {
3327                if (!active_cache)
3328                        /* read-cache does not die even when it fails
3329                         * so it is safe for us to do this here.  Also
3330                         * it does not smudge active_cache or active_nr
3331                         * when it fails, so we do not have to worry about
3332                         * cleaning it up ourselves either.
3333                         */
3334                        read_cache();
3335        }
3336        if (options->abbrev <= 0 || 40 < options->abbrev)
3337                options->abbrev = 40; /* full */
3338
3339        /*
3340         * It does not make sense to show the first hit we happened
3341         * to have found.  It does not make sense not to return with
3342         * exit code in such a case either.
3343         */
3344        if (DIFF_OPT_TST(options, QUICK)) {
3345                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3346                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3347        }
3348}
3349
3350static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3351{
3352        char c, *eq;
3353        int len;
3354
3355        if (*arg != '-')
3356                return 0;
3357        c = *++arg;
3358        if (!c)
3359                return 0;
3360        if (c == arg_short) {
3361                c = *++arg;
3362                if (!c)
3363                        return 1;
3364                if (val && isdigit(c)) {
3365                        char *end;
3366                        int n = strtoul(arg, &end, 10);
3367                        if (*end)
3368                                return 0;
3369                        *val = n;
3370                        return 1;
3371                }
3372                return 0;
3373        }
3374        if (c != '-')
3375                return 0;
3376        arg++;
3377        eq = strchr(arg, '=');
3378        if (eq)
3379                len = eq - arg;
3380        else
3381                len = strlen(arg);
3382        if (!len || strncmp(arg, arg_long, len))
3383                return 0;
3384        if (eq) {
3385                int n;
3386                char *end;
3387                if (!isdigit(*++eq))
3388                        return 0;
3389                n = strtoul(eq, &end, 10);
3390                if (*end)
3391                        return 0;
3392                *val = n;
3393        }
3394        return 1;
3395}
3396
3397static int diff_scoreopt_parse(const char *opt);
3398
3399static inline int short_opt(char opt, const char **argv,
3400                            const char **optarg)
3401{
3402        const char *arg = argv[0];
3403        if (arg[0] != '-' || arg[1] != opt)
3404                return 0;
3405        if (arg[2] != '\0') {
3406                *optarg = arg + 2;
3407                return 1;
3408        }
3409        if (!argv[1])
3410                die("Option '%c' requires a value", opt);
3411        *optarg = argv[1];
3412        return 2;
3413}
3414
3415int parse_long_opt(const char *opt, const char **argv,
3416                   const char **optarg)
3417{
3418        const char *arg = argv[0];
3419        if (arg[0] != '-' || arg[1] != '-')
3420                return 0;
3421        arg += strlen("--");
3422        if (prefixcmp(arg, opt))
3423                return 0;
3424        arg += strlen(opt);
3425        if (*arg == '=') { /* sticked form: --option=value */
3426                *optarg = arg + 1;
3427                return 1;
3428        }
3429        if (*arg != '\0')
3430                return 0;
3431        /* separate form: --option value */
3432        if (!argv[1])
3433                die("Option '--%s' requires a value", opt);
3434        *optarg = argv[1];
3435        return 2;
3436}
3437
3438static int stat_opt(struct diff_options *options, const char **av)
3439{
3440        const char *arg = av[0];
3441        char *end;
3442        int width = options->stat_width;
3443        int name_width = options->stat_name_width;
3444        int graph_width = options->stat_graph_width;
3445        int count = options->stat_count;
3446        int argcount = 1;
3447
3448        arg += strlen("--stat");
3449        end = (char *)arg;
3450
3451        switch (*arg) {
3452        case '-':
3453                if (!prefixcmp(arg, "-width")) {
3454                        arg += strlen("-width");
3455                        if (*arg == '=')
3456                                width = strtoul(arg + 1, &end, 10);
3457                        else if (!*arg && !av[1])
3458                                die("Option '--stat-width' requires a value");
3459                        else if (!*arg) {
3460                                width = strtoul(av[1], &end, 10);
3461                                argcount = 2;
3462                        }
3463                } else if (!prefixcmp(arg, "-name-width")) {
3464                        arg += strlen("-name-width");
3465                        if (*arg == '=')
3466                                name_width = strtoul(arg + 1, &end, 10);
3467                        else if (!*arg && !av[1])
3468                                die("Option '--stat-name-width' requires a value");
3469                        else if (!*arg) {
3470                                name_width = strtoul(av[1], &end, 10);
3471                                argcount = 2;
3472                        }
3473                } else if (!prefixcmp(arg, "-graph-width")) {
3474                        arg += strlen("-graph-width");
3475                        if (*arg == '=')
3476                                graph_width = strtoul(arg + 1, &end, 10);
3477                        else if (!*arg && !av[1])
3478                                die("Option '--stat-graph-width' requires a value");
3479                        else if (!*arg) {
3480                                graph_width = strtoul(av[1], &end, 10);
3481                                argcount = 2;
3482                        }
3483                } else if (!prefixcmp(arg, "-count")) {
3484                        arg += strlen("-count");
3485                        if (*arg == '=')
3486                                count = strtoul(arg + 1, &end, 10);
3487                        else if (!*arg && !av[1])
3488                                die("Option '--stat-count' requires a value");
3489                        else if (!*arg) {
3490                                count = strtoul(av[1], &end, 10);
3491                                argcount = 2;
3492                        }
3493                }
3494                break;
3495        case '=':
3496                width = strtoul(arg+1, &end, 10);
3497                if (*end == ',')
3498                        name_width = strtoul(end+1, &end, 10);
3499                if (*end == ',')
3500                        count = strtoul(end+1, &end, 10);
3501        }
3502
3503        /* Important! This checks all the error cases! */
3504        if (*end)
3505                return 0;
3506        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3507        options->stat_name_width = name_width;
3508        options->stat_graph_width = graph_width;
3509        options->stat_width = width;
3510        options->stat_count = count;
3511        return argcount;
3512}
3513
3514static int parse_dirstat_opt(struct diff_options *options, const char *params)
3515{
3516        struct strbuf errmsg = STRBUF_INIT;
3517        if (parse_dirstat_params(options, params, &errmsg))
3518                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3519                    errmsg.buf);
3520        strbuf_release(&errmsg);
3521        /*
3522         * The caller knows a dirstat-related option is given from the command
3523         * line; allow it to say "return this_function();"
3524         */
3525        options->output_format |= DIFF_FORMAT_DIRSTAT;
3526        return 1;
3527}
3528
3529static int parse_submodule_opt(struct diff_options *options, const char *value)
3530{
3531        if (parse_submodule_params(options, value))
3532                die(_("Failed to parse --submodule option parameter: '%s'"),
3533                        value);
3534        return 1;
3535}
3536
3537int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3538{
3539        const char *arg = av[0];
3540        const char *optarg;
3541        int argcount;
3542
3543        /* Output format options */
3544        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3545                options->output_format |= DIFF_FORMAT_PATCH;
3546        else if (opt_arg(arg, 'U', "unified", &options->context))
3547                options->output_format |= DIFF_FORMAT_PATCH;
3548        else if (!strcmp(arg, "--raw"))
3549                options->output_format |= DIFF_FORMAT_RAW;
3550        else if (!strcmp(arg, "--patch-with-raw"))
3551                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3552        else if (!strcmp(arg, "--numstat"))
3553                options->output_format |= DIFF_FORMAT_NUMSTAT;
3554        else if (!strcmp(arg, "--shortstat"))
3555                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3556        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3557                return parse_dirstat_opt(options, "");
3558        else if (!prefixcmp(arg, "-X"))
3559                return parse_dirstat_opt(options, arg + 2);
3560        else if (!prefixcmp(arg, "--dirstat="))
3561                return parse_dirstat_opt(options, arg + 10);
3562        else if (!strcmp(arg, "--cumulative"))
3563                return parse_dirstat_opt(options, "cumulative");
3564        else if (!strcmp(arg, "--dirstat-by-file"))
3565                return parse_dirstat_opt(options, "files");
3566        else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3567                parse_dirstat_opt(options, "files");
3568                return parse_dirstat_opt(options, arg + 18);
3569        }
3570        else if (!strcmp(arg, "--check"))
3571                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3572        else if (!strcmp(arg, "--summary"))
3573                options->output_format |= DIFF_FORMAT_SUMMARY;
3574        else if (!strcmp(arg, "--patch-with-stat"))
3575                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3576        else if (!strcmp(arg, "--name-only"))
3577                options->output_format |= DIFF_FORMAT_NAME;
3578        else if (!strcmp(arg, "--name-status"))
3579                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3580        else if (!strcmp(arg, "-s"))
3581                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3582        else if (!prefixcmp(arg, "--stat"))
3583                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3584                return stat_opt(options, av);
3585
3586        /* renames options */
3587        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3588                 !strcmp(arg, "--break-rewrites")) {
3589                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3590                        return error("invalid argument to -B: %s", arg+2);
3591        }
3592        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3593                 !strcmp(arg, "--find-renames")) {
3594                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3595                        return error("invalid argument to -M: %s", arg+2);
3596                options->detect_rename = DIFF_DETECT_RENAME;
3597        }
3598        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3599                options->irreversible_delete = 1;
3600        }
3601        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3602                 !strcmp(arg, "--find-copies")) {
3603                if (options->detect_rename == DIFF_DETECT_COPY)
3604                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3605                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3606                        return error("invalid argument to -C: %s", arg+2);
3607                options->detect_rename = DIFF_DETECT_COPY;
3608        }
3609        else if (!strcmp(arg, "--no-renames"))
3610                options->detect_rename = 0;
3611        else if (!strcmp(arg, "--rename-empty"))
3612                DIFF_OPT_SET(options, RENAME_EMPTY);
3613        else if (!strcmp(arg, "--no-rename-empty"))
3614                DIFF_OPT_CLR(options, RENAME_EMPTY);
3615        else if (!strcmp(arg, "--relative"))
3616                DIFF_OPT_SET(options, RELATIVE_NAME);
3617        else if (!prefixcmp(arg, "--relative=")) {
3618                DIFF_OPT_SET(options, RELATIVE_NAME);
3619                options->prefix = arg + 11;
3620        }
3621
3622        /* xdiff options */
3623        else if (!strcmp(arg, "--minimal"))
3624                DIFF_XDL_SET(options, NEED_MINIMAL);
3625        else if (!strcmp(arg, "--no-minimal"))
3626                DIFF_XDL_CLR(options, NEED_MINIMAL);
3627        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3628                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3629        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3630                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3631        else if (!strcmp(arg, "--ignore-space-at-eol"))
3632                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3633        else if (!strcmp(arg, "--patience"))
3634                options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3635        else if (!strcmp(arg, "--histogram"))
3636                options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3637        else if (!prefixcmp(arg, "--diff-algorithm=")) {
3638                long value = parse_algorithm_value(arg+17);
3639                if (value < 0)
3640                        return error("option diff-algorithm accepts \"myers\", "
3641                                     "\"minimal\", \"patience\" and \"histogram\"");
3642                /* clear out previous settings */
3643                DIFF_XDL_CLR(options, NEED_MINIMAL);
3644                options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3645                options->xdl_opts |= value;
3646        }
3647
3648        /* flags options */
3649        else if (!strcmp(arg, "--binary")) {
3650                options->output_format |= DIFF_FORMAT_PATCH;
3651                DIFF_OPT_SET(options, BINARY);
3652        }
3653        else if (!strcmp(arg, "--full-index"))
3654                DIFF_OPT_SET(options, FULL_INDEX);
3655        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3656                DIFF_OPT_SET(options, TEXT);
3657        else if (!strcmp(arg, "-R"))
3658                DIFF_OPT_SET(options, REVERSE_DIFF);
3659        else if (!strcmp(arg, "--find-copies-harder"))
3660                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3661        else if (!strcmp(arg, "--follow"))
3662                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3663        else if (!strcmp(arg, "--color"))
3664                options->use_color = 1;
3665        else if (!prefixcmp(arg, "--color=")) {
3666                int value = git_config_colorbool(NULL, arg+8);
3667                if (value < 0)
3668                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3669                options->use_color = value;
3670        }
3671        else if (!strcmp(arg, "--no-color"))
3672                options->use_color = 0;
3673        else if (!strcmp(arg, "--color-words")) {
3674                options->use_color = 1;
3675                options->word_diff = DIFF_WORDS_COLOR;
3676        }
3677        else if (!prefixcmp(arg, "--color-words=")) {
3678                options->use_color = 1;
3679                options->word_diff = DIFF_WORDS_COLOR;
3680                options->word_regex = arg + 14;
3681        }
3682        else if (!strcmp(arg, "--word-diff")) {
3683                if (options->word_diff == DIFF_WORDS_NONE)
3684                        options->word_diff = DIFF_WORDS_PLAIN;
3685        }
3686        else if (!prefixcmp(arg, "--word-diff=")) {
3687                const char *type = arg + 12;
3688                if (!strcmp(type, "plain"))
3689                        options->word_diff = DIFF_WORDS_PLAIN;
3690                else if (!strcmp(type, "color")) {
3691                        options->use_color = 1;
3692                        options->word_diff = DIFF_WORDS_COLOR;
3693                }
3694                else if (!strcmp(type, "porcelain"))
3695                        options->word_diff = DIFF_WORDS_PORCELAIN;
3696                else if (!strcmp(type, "none"))
3697                        options->word_diff = DIFF_WORDS_NONE;
3698                else
3699                        die("bad --word-diff argument: %s", type);
3700        }
3701        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3702                if (options->word_diff == DIFF_WORDS_NONE)
3703                        options->word_diff = DIFF_WORDS_PLAIN;
3704                options->word_regex = optarg;
3705                return argcount;
3706        }
3707        else if (!strcmp(arg, "--exit-code"))
3708                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3709        else if (!strcmp(arg, "--quiet"))
3710                DIFF_OPT_SET(options, QUICK);
3711        else if (!strcmp(arg, "--ext-diff"))
3712                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3713        else if (!strcmp(arg, "--no-ext-diff"))
3714                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3715        else if (!strcmp(arg, "--textconv"))
3716                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3717        else if (!strcmp(arg, "--no-textconv"))
3718                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3719        else if (!strcmp(arg, "--ignore-submodules")) {
3720                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3721                handle_ignore_submodules_arg(options, "all");
3722        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3723                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3724                handle_ignore_submodules_arg(options, arg + 20);
3725        } else if (!strcmp(arg, "--submodule"))
3726                DIFF_OPT_SET(options, SUBMODULE_LOG);
3727        else if (!prefixcmp(arg, "--submodule="))
3728                return parse_submodule_opt(options, arg + 12);
3729
3730        /* misc options */
3731        else if (!strcmp(arg, "-z"))
3732                options->line_termination = 0;
3733        else if ((argcount = short_opt('l', av, &optarg))) {
3734                options->rename_limit = strtoul(optarg, NULL, 10);
3735                return argcount;
3736        }
3737        else if ((argcount = short_opt('S', av, &optarg))) {
3738                options->pickaxe = optarg;
3739                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3740                return argcount;
3741        } else if ((argcount = short_opt('G', av, &optarg))) {
3742                options->pickaxe = optarg;
3743                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3744                return argcount;
3745        }
3746        else if (!strcmp(arg, "--pickaxe-all"))
3747                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3748        else if (!strcmp(arg, "--pickaxe-regex"))
3749                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3750        else if ((argcount = short_opt('O', av, &optarg))) {
3751                options->orderfile = optarg;
3752                return argcount;
3753        }
3754        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3755                options->filter = optarg;
3756                return argcount;
3757        }
3758        else if (!strcmp(arg, "--abbrev"))
3759                options->abbrev = DEFAULT_ABBREV;
3760        else if (!prefixcmp(arg, "--abbrev=")) {
3761                options->abbrev = strtoul(arg + 9, NULL, 10);
3762                if (options->abbrev < MINIMUM_ABBREV)
3763                        options->abbrev = MINIMUM_ABBREV;
3764                else if (40 < options->abbrev)
3765                        options->abbrev = 40;
3766        }
3767        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3768                options->a_prefix = optarg;
3769                return argcount;
3770        }
3771        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3772                options->b_prefix = optarg;
3773                return argcount;
3774        }
3775        else if (!strcmp(arg, "--no-prefix"))
3776                options->a_prefix = options->b_prefix = "";
3777        else if (opt_arg(arg, '\0', "inter-hunk-context",
3778                         &options->interhunkcontext))
3779                ;
3780        else if (!strcmp(arg, "-W"))
3781                DIFF_OPT_SET(options, FUNCCONTEXT);
3782        else if (!strcmp(arg, "--function-context"))
3783                DIFF_OPT_SET(options, FUNCCONTEXT);
3784        else if (!strcmp(arg, "--no-function-context"))
3785                DIFF_OPT_CLR(options, FUNCCONTEXT);
3786        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3787                options->file = fopen(optarg, "w");
3788                if (!options->file)
3789                        die_errno("Could not open '%s'", optarg);
3790                options->close_file = 1;
3791                return argcount;
3792        } else
3793                return 0;
3794        return 1;
3795}
3796
3797int parse_rename_score(const char **cp_p)
3798{
3799        unsigned long num, scale;
3800        int ch, dot;
3801        const char *cp = *cp_p;
3802
3803        num = 0;
3804        scale = 1;
3805        dot = 0;
3806        for (;;) {
3807                ch = *cp;
3808                if ( !dot && ch == '.' ) {
3809                        scale = 1;
3810                        dot = 1;
3811                } else if ( ch == '%' ) {
3812                        scale = dot ? scale*100 : 100;
3813                        cp++;   /* % is always at the end */
3814                        break;
3815                } else if ( ch >= '0' && ch <= '9' ) {
3816                        if ( scale < 100000 ) {
3817                                scale *= 10;
3818                                num = (num*10) + (ch-'0');
3819                        }
3820                } else {
3821                        break;
3822                }
3823                cp++;
3824        }
3825        *cp_p = cp;
3826
3827        /* user says num divided by scale and we say internally that
3828         * is MAX_SCORE * num / scale.
3829         */
3830        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3831}
3832
3833static int diff_scoreopt_parse(const char *opt)
3834{
3835        int opt1, opt2, cmd;
3836
3837        if (*opt++ != '-')
3838                return -1;
3839        cmd = *opt++;
3840        if (cmd == '-') {
3841                /* convert the long-form arguments into short-form versions */
3842                if (!prefixcmp(opt, "break-rewrites")) {
3843                        opt += strlen("break-rewrites");
3844                        if (*opt == 0 || *opt++ == '=')
3845                                cmd = 'B';
3846                } else if (!prefixcmp(opt, "find-copies")) {
3847                        opt += strlen("find-copies");
3848                        if (*opt == 0 || *opt++ == '=')
3849                                cmd = 'C';
3850                } else if (!prefixcmp(opt, "find-renames")) {
3851                        opt += strlen("find-renames");
3852                        if (*opt == 0 || *opt++ == '=')
3853                                cmd = 'M';
3854                }
3855        }
3856        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3857                return -1; /* that is not a -M, -C nor -B option */
3858
3859        opt1 = parse_rename_score(&opt);
3860        if (cmd != 'B')
3861                opt2 = 0;
3862        else {
3863                if (*opt == 0)
3864                        opt2 = 0;
3865                else if (*opt != '/')
3866                        return -1; /* we expect -B80/99 or -B80 */
3867                else {
3868                        opt++;
3869                        opt2 = parse_rename_score(&opt);
3870                }
3871        }
3872        if (*opt != 0)
3873                return -1;
3874        return opt1 | (opt2 << 16);
3875}
3876
3877struct diff_queue_struct diff_queued_diff;
3878
3879void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3880{
3881        if (queue->alloc <= queue->nr) {
3882                queue->alloc = alloc_nr(queue->alloc);
3883                queue->queue = xrealloc(queue->queue,
3884                                        sizeof(dp) * queue->alloc);
3885        }
3886        queue->queue[queue->nr++] = dp;
3887}
3888
3889struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3890                                 struct diff_filespec *one,
3891                                 struct diff_filespec *two)
3892{
3893        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3894        dp->one = one;
3895        dp->two = two;
3896        if (queue)
3897                diff_q(queue, dp);
3898        return dp;
3899}
3900
3901void diff_free_filepair(struct diff_filepair *p)
3902{
3903        free_filespec(p->one);
3904        free_filespec(p->two);
3905        free(p);
3906}
3907
3908/* This is different from find_unique_abbrev() in that
3909 * it stuffs the result with dots for alignment.
3910 */
3911const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3912{
3913        int abblen;
3914        const char *abbrev;
3915        if (len == 40)
3916                return sha1_to_hex(sha1);
3917
3918        abbrev = find_unique_abbrev(sha1, len);
3919        abblen = strlen(abbrev);
3920        if (abblen < 37) {
3921                static char hex[41];
3922                if (len < abblen && abblen <= len + 2)
3923                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3924                else
3925                        sprintf(hex, "%s...", abbrev);
3926                return hex;
3927        }
3928        return sha1_to_hex(sha1);
3929}
3930
3931static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3932{
3933        int line_termination = opt->line_termination;
3934        int inter_name_termination = line_termination ? '\t' : '\0';
3935        if (opt->output_prefix) {
3936                struct strbuf *msg = NULL;
3937                msg = opt->output_prefix(opt, opt->output_prefix_data);
3938                fprintf(opt->file, "%s", msg->buf);
3939        }
3940
3941        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3942                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3943                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3944                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3945        }
3946        if (p->score) {
3947                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3948                        inter_name_termination);
3949        } else {
3950                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3951        }
3952
3953        if (p->status == DIFF_STATUS_COPIED ||
3954            p->status == DIFF_STATUS_RENAMED) {
3955                const char *name_a, *name_b;
3956                name_a = p->one->path;
3957                name_b = p->two->path;
3958                strip_prefix(opt->prefix_length, &name_a, &name_b);
3959                write_name_quoted(name_a, opt->file, inter_name_termination);
3960                write_name_quoted(name_b, opt->file, line_termination);
3961        } else {
3962                const char *name_a, *name_b;
3963                name_a = p->one->mode ? p->one->path : p->two->path;
3964                name_b = NULL;
3965                strip_prefix(opt->prefix_length, &name_a, &name_b);
3966                write_name_quoted(name_a, opt->file, line_termination);
3967        }
3968}
3969
3970int diff_unmodified_pair(struct diff_filepair *p)
3971{
3972        /* This function is written stricter than necessary to support
3973         * the currently implemented transformers, but the idea is to
3974         * let transformers to produce diff_filepairs any way they want,
3975         * and filter and clean them up here before producing the output.
3976         */
3977        struct diff_filespec *one = p->one, *two = p->two;
3978
3979        if (DIFF_PAIR_UNMERGED(p))
3980                return 0; /* unmerged is interesting */
3981
3982        /* deletion, addition, mode or type change
3983         * and rename are all interesting.
3984         */
3985        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3986            DIFF_PAIR_MODE_CHANGED(p) ||
3987            strcmp(one->path, two->path))
3988                return 0;
3989
3990        /* both are valid and point at the same path.  that is, we are
3991         * dealing with a change.
3992         */
3993        if (one->sha1_valid && two->sha1_valid &&
3994            !hashcmp(one->sha1, two->sha1) &&
3995            !one->dirty_submodule && !two->dirty_submodule)
3996                return 1; /* no change */
3997        if (!one->sha1_valid && !two->sha1_valid)
3998                return 1; /* both look at the same file on the filesystem. */
3999        return 0;
4000}
4001
4002static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4003{
4004        if (diff_unmodified_pair(p))
4005                return;
4006
4007        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4008            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4009                return; /* no tree diffs in patch format */
4010
4011        run_diff(p, o);
4012}
4013
4014static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4015                            struct diffstat_t *diffstat)
4016{
4017        if (diff_unmodified_pair(p))
4018                return;
4019
4020        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4021            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4022                return; /* no useful stat for tree diffs */
4023
4024        run_diffstat(p, o, diffstat);
4025}
4026
4027static void diff_flush_checkdiff(struct diff_filepair *p,
4028                struct diff_options *o)
4029{
4030        if (diff_unmodified_pair(p))
4031                return;
4032
4033        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4034            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4035                return; /* nothing to check in tree diffs */
4036
4037        run_checkdiff(p, o);
4038}
4039
4040int diff_queue_is_empty(void)
4041{
4042        struct diff_queue_struct *q = &diff_queued_diff;
4043        int i;
4044        for (i = 0; i < q->nr; i++)
4045                if (!diff_unmodified_pair(q->queue[i]))
4046                        return 0;
4047        return 1;
4048}
4049
4050#if DIFF_DEBUG
4051void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4052{
4053        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4054                x, one ? one : "",
4055                s->path,
4056                DIFF_FILE_VALID(s) ? "valid" : "invalid",
4057                s->mode,
4058                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4059        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
4060                x, one ? one : "",
4061                s->size, s->xfrm_flags);
4062}
4063
4064void diff_debug_filepair(const struct diff_filepair *p, int i)
4065{
4066        diff_debug_filespec(p->one, i, "one");
4067        diff_debug_filespec(p->two, i, "two");
4068        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4069                p->score, p->status ? p->status : '?',
4070                p->one->rename_used, p->broken_pair);
4071}
4072
4073void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4074{
4075        int i;
4076        if (msg)
4077                fprintf(stderr, "%s\n", msg);
4078        fprintf(stderr, "q->nr = %d\n", q->nr);
4079        for (i = 0; i < q->nr; i++) {
4080                struct diff_filepair *p = q->queue[i];
4081                diff_debug_filepair(p, i);
4082        }
4083}
4084#endif
4085
4086static void diff_resolve_rename_copy(void)
4087{
4088        int i;
4089        struct diff_filepair *p;
4090        struct diff_queue_struct *q = &diff_queued_diff;
4091
4092        diff_debug_queue("resolve-rename-copy", q);
4093
4094        for (i = 0; i < q->nr; i++) {
4095                p = q->queue[i];
4096                p->status = 0; /* undecided */
4097                if (DIFF_PAIR_UNMERGED(p))
4098                        p->status = DIFF_STATUS_UNMERGED;
4099                else if (!DIFF_FILE_VALID(p->one))
4100                        p->status = DIFF_STATUS_ADDED;
4101                else if (!DIFF_FILE_VALID(p->two))
4102                        p->status = DIFF_STATUS_DELETED;
4103                else if (DIFF_PAIR_TYPE_CHANGED(p))
4104                        p->status = DIFF_STATUS_TYPE_CHANGED;
4105
4106                /* from this point on, we are dealing with a pair
4107                 * whose both sides are valid and of the same type, i.e.
4108                 * either in-place edit or rename/copy edit.
4109                 */
4110                else if (DIFF_PAIR_RENAME(p)) {
4111                        /*
4112                         * A rename might have re-connected a broken
4113                         * pair up, causing the pathnames to be the
4114                         * same again. If so, that's not a rename at
4115                         * all, just a modification..
4116                         *
4117                         * Otherwise, see if this source was used for
4118                         * multiple renames, in which case we decrement
4119                         * the count, and call it a copy.
4120                         */
4121                        if (!strcmp(p->one->path, p->two->path))
4122                                p->status = DIFF_STATUS_MODIFIED;
4123                        else if (--p->one->rename_used > 0)
4124                                p->status = DIFF_STATUS_COPIED;
4125                        else
4126                                p->status = DIFF_STATUS_RENAMED;
4127                }
4128                else if (hashcmp(p->one->sha1, p->two->sha1) ||
4129                         p->one->mode != p->two->mode ||
4130                         p->one->dirty_submodule ||
4131                         p->two->dirty_submodule ||
4132                         is_null_sha1(p->one->sha1))
4133                        p->status = DIFF_STATUS_MODIFIED;
4134                else {
4135                        /* This is a "no-change" entry and should not
4136                         * happen anymore, but prepare for broken callers.
4137                         */
4138                        error("feeding unmodified %s to diffcore",
4139                              p->one->path);
4140                        p->status = DIFF_STATUS_UNKNOWN;
4141                }
4142        }
4143        diff_debug_queue("resolve-rename-copy done", q);
4144}
4145
4146static int check_pair_status(struct diff_filepair *p)
4147{
4148        switch (p->status) {
4149        case DIFF_STATUS_UNKNOWN:
4150                return 0;
4151        case 0:
4152                die("internal error in diff-resolve-rename-copy");
4153        default:
4154                return 1;
4155        }
4156}
4157
4158static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4159{
4160        int fmt = opt->output_format;
4161
4162        if (fmt & DIFF_FORMAT_CHECKDIFF)
4163                diff_flush_checkdiff(p, opt);
4164        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4165                diff_flush_raw(p, opt);
4166        else if (fmt & DIFF_FORMAT_NAME) {
4167                const char *name_a, *name_b;
4168                name_a = p->two->path;
4169                name_b = NULL;
4170                strip_prefix(opt->prefix_length, &name_a, &name_b);
4171                write_name_quoted(name_a, opt->file, opt->line_termination);
4172        }
4173}
4174
4175static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4176{
4177        if (fs->mode)
4178                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4179        else
4180                fprintf(file, " %s ", newdelete);
4181        write_name_quoted(fs->path, file, '\n');
4182}
4183
4184
4185static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4186                const char *line_prefix)
4187{
4188        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4189                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4190                        p->two->mode, show_name ? ' ' : '\n');
4191                if (show_name) {
4192                        write_name_quoted(p->two->path, file, '\n');
4193                }
4194        }
4195}
4196
4197static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4198                        const char *line_prefix)
4199{
4200        char *names = pprint_rename(p->one->path, p->two->path);
4201
4202        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4203        free(names);
4204        show_mode_change(file, p, 0, line_prefix);
4205}
4206
4207static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4208{
4209        FILE *file = opt->file;
4210        char *line_prefix = "";
4211
4212        if (opt->output_prefix) {
4213                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4214                line_prefix = buf->buf;
4215        }
4216
4217        switch(p->status) {
4218        case DIFF_STATUS_DELETED:
4219                fputs(line_prefix, file);
4220                show_file_mode_name(file, "delete", p->one);
4221                break;
4222        case DIFF_STATUS_ADDED:
4223                fputs(line_prefix, file);
4224                show_file_mode_name(file, "create", p->two);
4225                break;
4226        case DIFF_STATUS_COPIED:
4227                fputs(line_prefix, file);
4228                show_rename_copy(file, "copy", p, line_prefix);
4229                break;
4230        case DIFF_STATUS_RENAMED:
4231                fputs(line_prefix, file);
4232                show_rename_copy(file, "rename", p, line_prefix);
4233                break;
4234        default:
4235                if (p->score) {
4236                        fprintf(file, "%s rewrite ", line_prefix);
4237                        write_name_quoted(p->two->path, file, ' ');
4238                        fprintf(file, "(%d%%)\n", similarity_index(p));
4239                }
4240                show_mode_change(file, p, !p->score, line_prefix);
4241                break;
4242        }
4243}
4244
4245struct patch_id_t {
4246        git_SHA_CTX *ctx;
4247        int patchlen;
4248};
4249
4250static int remove_space(char *line, int len)
4251{
4252        int i;
4253        char *dst = line;
4254        unsigned char c;
4255
4256        for (i = 0; i < len; i++)
4257                if (!isspace((c = line[i])))
4258                        *dst++ = c;
4259
4260        return dst - line;
4261}
4262
4263static void patch_id_consume(void *priv, char *line, unsigned long len)
4264{
4265        struct patch_id_t *data = priv;
4266        int new_len;
4267
4268        /* Ignore line numbers when computing the SHA1 of the patch */
4269        if (!prefixcmp(line, "@@ -"))
4270                return;
4271
4272        new_len = remove_space(line, len);
4273
4274        git_SHA1_Update(data->ctx, line, new_len);
4275        data->patchlen += new_len;
4276}
4277
4278/* returns 0 upon success, and writes result into sha1 */
4279static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4280{
4281        struct diff_queue_struct *q = &diff_queued_diff;
4282        int i;
4283        git_SHA_CTX ctx;
4284        struct patch_id_t data;
4285        char buffer[PATH_MAX * 4 + 20];
4286
4287        git_SHA1_Init(&ctx);
4288        memset(&data, 0, sizeof(struct patch_id_t));
4289        data.ctx = &ctx;
4290
4291        for (i = 0; i < q->nr; i++) {
4292                xpparam_t xpp;
4293                xdemitconf_t xecfg;
4294                mmfile_t mf1, mf2;
4295                struct diff_filepair *p = q->queue[i];
4296                int len1, len2;
4297
4298                memset(&xpp, 0, sizeof(xpp));
4299                memset(&xecfg, 0, sizeof(xecfg));
4300                if (p->status == 0)
4301                        return error("internal diff status error");
4302                if (p->status == DIFF_STATUS_UNKNOWN)
4303                        continue;
4304                if (diff_unmodified_pair(p))
4305                        continue;
4306                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4307                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4308                        continue;
4309                if (DIFF_PAIR_UNMERGED(p))
4310                        continue;
4311
4312                diff_fill_sha1_info(p->one);
4313                diff_fill_sha1_info(p->two);
4314                if (fill_mmfile(&mf1, p->one) < 0 ||
4315                                fill_mmfile(&mf2, p->two) < 0)
4316                        return error("unable to read files to diff");
4317
4318                len1 = remove_space(p->one->path, strlen(p->one->path));
4319                len2 = remove_space(p->two->path, strlen(p->two->path));
4320                if (p->one->mode == 0)
4321                        len1 = snprintf(buffer, sizeof(buffer),
4322                                        "diff--gita/%.*sb/%.*s"
4323                                        "newfilemode%06o"
4324                                        "---/dev/null"
4325                                        "+++b/%.*s",
4326                                        len1, p->one->path,
4327                                        len2, p->two->path,
4328                                        p->two->mode,
4329                                        len2, p->two->path);
4330                else if (p->two->mode == 0)
4331                        len1 = snprintf(buffer, sizeof(buffer),
4332                                        "diff--gita/%.*sb/%.*s"
4333                                        "deletedfilemode%06o"
4334                                        "---a/%.*s"
4335                                        "+++/dev/null",
4336                                        len1, p->one->path,
4337                                        len2, p->two->path,
4338                                        p->one->mode,
4339                                        len1, p->one->path);
4340                else
4341                        len1 = snprintf(buffer, sizeof(buffer),
4342                                        "diff--gita/%.*sb/%.*s"
4343                                        "---a/%.*s"
4344                                        "+++b/%.*s",
4345                                        len1, p->one->path,
4346                                        len2, p->two->path,
4347                                        len1, p->one->path,
4348                                        len2, p->two->path);
4349                git_SHA1_Update(&ctx, buffer, len1);
4350
4351                if (diff_filespec_is_binary(p->one) ||
4352                    diff_filespec_is_binary(p->two)) {
4353                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4354                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4355                        continue;
4356                }
4357
4358                xpp.flags = 0;
4359                xecfg.ctxlen = 3;
4360                xecfg.flags = 0;
4361                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4362                              &xpp, &xecfg);
4363        }
4364
4365        git_SHA1_Final(sha1, &ctx);
4366        return 0;
4367}
4368
4369int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4370{
4371        struct diff_queue_struct *q = &diff_queued_diff;
4372        int i;
4373        int result = diff_get_patch_id(options, sha1);
4374
4375        for (i = 0; i < q->nr; i++)
4376                diff_free_filepair(q->queue[i]);
4377
4378        free(q->queue);
4379        DIFF_QUEUE_CLEAR(q);
4380
4381        return result;
4382}
4383
4384static int is_summary_empty(const struct diff_queue_struct *q)
4385{
4386        int i;
4387
4388        for (i = 0; i < q->nr; i++) {
4389                const struct diff_filepair *p = q->queue[i];
4390
4391                switch (p->status) {
4392                case DIFF_STATUS_DELETED:
4393                case DIFF_STATUS_ADDED:
4394                case DIFF_STATUS_COPIED:
4395                case DIFF_STATUS_RENAMED:
4396                        return 0;
4397                default:
4398                        if (p->score)
4399                                return 0;
4400                        if (p->one->mode && p->two->mode &&
4401                            p->one->mode != p->two->mode)
4402                                return 0;
4403                        break;
4404                }
4405        }
4406        return 1;
4407}
4408
4409static const char rename_limit_warning[] =
4410"inexact rename detection was skipped due to too many files.";
4411
4412static const char degrade_cc_to_c_warning[] =
4413"only found copies from modified paths due to too many files.";
4414
4415static const char rename_limit_advice[] =
4416"you may want to set your %s variable to at least "
4417"%d and retry the command.";
4418
4419void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4420{
4421        if (degraded_cc)
4422                warning(degrade_cc_to_c_warning);
4423        else if (needed)
4424                warning(rename_limit_warning);
4425        else
4426                return;
4427        if (0 < needed && needed < 32767)
4428                warning(rename_limit_advice, varname, needed);
4429}
4430
4431void diff_flush(struct diff_options *options)
4432{
4433        struct diff_queue_struct *q = &diff_queued_diff;
4434        int i, output_format = options->output_format;
4435        int separator = 0;
4436        int dirstat_by_line = 0;
4437
4438        /*
4439         * Order: raw, stat, summary, patch
4440         * or:    name/name-status/checkdiff (other bits clear)
4441         */
4442        if (!q->nr)
4443                goto free_queue;
4444
4445        if (output_format & (DIFF_FORMAT_RAW |
4446                             DIFF_FORMAT_NAME |
4447                             DIFF_FORMAT_NAME_STATUS |
4448                             DIFF_FORMAT_CHECKDIFF)) {
4449                for (i = 0; i < q->nr; i++) {
4450                        struct diff_filepair *p = q->queue[i];
4451                        if (check_pair_status(p))
4452                                flush_one_pair(p, options);
4453                }
4454                separator++;
4455        }
4456
4457        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4458                dirstat_by_line = 1;
4459
4460        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4461            dirstat_by_line) {
4462                struct diffstat_t diffstat;
4463
4464                memset(&diffstat, 0, sizeof(struct diffstat_t));
4465                for (i = 0; i < q->nr; i++) {
4466                        struct diff_filepair *p = q->queue[i];
4467                        if (check_pair_status(p))
4468                                diff_flush_stat(p, options, &diffstat);
4469                }
4470                if (output_format & DIFF_FORMAT_NUMSTAT)
4471                        show_numstat(&diffstat, options);
4472                if (output_format & DIFF_FORMAT_DIFFSTAT)
4473                        show_stats(&diffstat, options);
4474                if (output_format & DIFF_FORMAT_SHORTSTAT)
4475                        show_shortstats(&diffstat, options);
4476                if (output_format & DIFF_FORMAT_DIRSTAT)
4477                        show_dirstat_by_line(&diffstat, options);
4478                free_diffstat_info(&diffstat);
4479                separator++;
4480        }
4481        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4482                show_dirstat(options);
4483
4484        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4485                for (i = 0; i < q->nr; i++) {
4486                        diff_summary(options, q->queue[i]);
4487                }
4488                separator++;
4489        }
4490
4491        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4492            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4493            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4494                /*
4495                 * run diff_flush_patch for the exit status. setting
4496                 * options->file to /dev/null should be safe, becaue we
4497                 * aren't supposed to produce any output anyway.
4498                 */
4499                if (options->close_file)
4500                        fclose(options->file);
4501                options->file = fopen("/dev/null", "w");
4502                if (!options->file)
4503                        die_errno("Could not open /dev/null");
4504                options->close_file = 1;
4505                for (i = 0; i < q->nr; i++) {
4506                        struct diff_filepair *p = q->queue[i];
4507                        if (check_pair_status(p))
4508                                diff_flush_patch(p, options);
4509                        if (options->found_changes)
4510                                break;
4511                }
4512        }
4513
4514        if (output_format & DIFF_FORMAT_PATCH) {
4515                if (separator) {
4516                        if (options->output_prefix) {
4517                                struct strbuf *msg = NULL;
4518                                msg = options->output_prefix(options,
4519                                        options->output_prefix_data);
4520                                fwrite(msg->buf, msg->len, 1, stdout);
4521                        }
4522                        putc(options->line_termination, options->file);
4523                        if (options->stat_sep) {
4524                                /* attach patch instead of inline */
4525                                fputs(options->stat_sep, options->file);
4526                        }
4527                }
4528
4529                for (i = 0; i < q->nr; i++) {
4530                        struct diff_filepair *p = q->queue[i];
4531                        if (check_pair_status(p))
4532                                diff_flush_patch(p, options);
4533                }
4534        }
4535
4536        if (output_format & DIFF_FORMAT_CALLBACK)
4537                options->format_callback(q, options, options->format_callback_data);
4538
4539        for (i = 0; i < q->nr; i++)
4540                diff_free_filepair(q->queue[i]);
4541free_queue:
4542        free(q->queue);
4543        DIFF_QUEUE_CLEAR(q);
4544        if (options->close_file)
4545                fclose(options->file);
4546
4547        /*
4548         * Report the content-level differences with HAS_CHANGES;
4549         * diff_addremove/diff_change does not set the bit when
4550         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4551         */
4552        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4553                if (options->found_changes)
4554                        DIFF_OPT_SET(options, HAS_CHANGES);
4555                else
4556                        DIFF_OPT_CLR(options, HAS_CHANGES);
4557        }
4558}
4559
4560static void diffcore_apply_filter(const char *filter)
4561{
4562        int i;
4563        struct diff_queue_struct *q = &diff_queued_diff;
4564        struct diff_queue_struct outq;
4565        DIFF_QUEUE_CLEAR(&outq);
4566
4567        if (!filter)
4568                return;
4569
4570        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4571                int found;
4572                for (i = found = 0; !found && i < q->nr; i++) {
4573                        struct diff_filepair *p = q->queue[i];
4574                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4575                             ((p->score &&
4576                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4577                              (!p->score &&
4578                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4579                            ((p->status != DIFF_STATUS_MODIFIED) &&
4580                             strchr(filter, p->status)))
4581                                found++;
4582                }
4583                if (found)
4584                        return;
4585
4586                /* otherwise we will clear the whole queue
4587                 * by copying the empty outq at the end of this
4588                 * function, but first clear the current entries
4589                 * in the queue.
4590                 */
4591                for (i = 0; i < q->nr; i++)
4592                        diff_free_filepair(q->queue[i]);
4593        }
4594        else {
4595                /* Only the matching ones */
4596                for (i = 0; i < q->nr; i++) {
4597                        struct diff_filepair *p = q->queue[i];
4598
4599                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4600                             ((p->score &&
4601                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4602                              (!p->score &&
4603                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4604                            ((p->status != DIFF_STATUS_MODIFIED) &&
4605                             strchr(filter, p->status)))
4606                                diff_q(&outq, p);
4607                        else
4608                                diff_free_filepair(p);
4609                }
4610        }
4611        free(q->queue);
4612        *q = outq;
4613}
4614
4615/* Check whether two filespecs with the same mode and size are identical */
4616static int diff_filespec_is_identical(struct diff_filespec *one,
4617                                      struct diff_filespec *two)
4618{
4619        if (S_ISGITLINK(one->mode))
4620                return 0;
4621        if (diff_populate_filespec(one, 0))
4622                return 0;
4623        if (diff_populate_filespec(two, 0))
4624                return 0;
4625        return !memcmp(one->data, two->data, one->size);
4626}
4627
4628static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4629{
4630        int i;
4631        struct diff_queue_struct *q = &diff_queued_diff;
4632        struct diff_queue_struct outq;
4633        DIFF_QUEUE_CLEAR(&outq);
4634
4635        for (i = 0; i < q->nr; i++) {
4636                struct diff_filepair *p = q->queue[i];
4637
4638                /*
4639                 * 1. Entries that come from stat info dirtiness
4640                 *    always have both sides (iow, not create/delete),
4641                 *    one side of the object name is unknown, with
4642                 *    the same mode and size.  Keep the ones that
4643                 *    do not match these criteria.  They have real
4644                 *    differences.
4645                 *
4646                 * 2. At this point, the file is known to be modified,
4647                 *    with the same mode and size, and the object
4648                 *    name of one side is unknown.  Need to inspect
4649                 *    the identical contents.
4650                 */
4651                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4652                    !DIFF_FILE_VALID(p->two) ||
4653                    (p->one->sha1_valid && p->two->sha1_valid) ||
4654                    (p->one->mode != p->two->mode) ||
4655                    diff_populate_filespec(p->one, 1) ||
4656                    diff_populate_filespec(p->two, 1) ||
4657                    (p->one->size != p->two->size) ||
4658                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4659                        diff_q(&outq, p);
4660                else {
4661                        /*
4662                         * The caller can subtract 1 from skip_stat_unmatch
4663                         * to determine how many paths were dirty only
4664                         * due to stat info mismatch.
4665                         */
4666                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4667                                diffopt->skip_stat_unmatch++;
4668                        diff_free_filepair(p);
4669                }
4670        }
4671        free(q->queue);
4672        *q = outq;
4673}
4674
4675static int diffnamecmp(const void *a_, const void *b_)
4676{
4677        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4678        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4679        const char *name_a, *name_b;
4680
4681        name_a = a->one ? a->one->path : a->two->path;
4682        name_b = b->one ? b->one->path : b->two->path;
4683        return strcmp(name_a, name_b);
4684}
4685
4686void diffcore_fix_diff_index(struct diff_options *options)
4687{
4688        struct diff_queue_struct *q = &diff_queued_diff;
4689        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4690}
4691
4692void diffcore_std(struct diff_options *options)
4693{
4694        if (options->skip_stat_unmatch)
4695                diffcore_skip_stat_unmatch(options);
4696        if (!options->found_follow) {
4697                /* See try_to_follow_renames() in tree-diff.c */
4698                if (options->break_opt != -1)
4699                        diffcore_break(options->break_opt);
4700                if (options->detect_rename)
4701                        diffcore_rename(options);
4702                if (options->break_opt != -1)
4703                        diffcore_merge_broken();
4704        }
4705        if (options->pickaxe)
4706                diffcore_pickaxe(options);
4707        if (options->orderfile)
4708                diffcore_order(options->orderfile);
4709        if (!options->found_follow)
4710                /* See try_to_follow_renames() in tree-diff.c */
4711                diff_resolve_rename_copy();
4712        diffcore_apply_filter(options->filter);
4713
4714        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4715                DIFF_OPT_SET(options, HAS_CHANGES);
4716        else
4717                DIFF_OPT_CLR(options, HAS_CHANGES);
4718
4719        options->found_follow = 0;
4720}
4721
4722int diff_result_code(struct diff_options *opt, int status)
4723{
4724        int result = 0;
4725
4726        diff_warn_rename_limit("diff.renamelimit",
4727                               opt->needed_rename_limit,
4728                               opt->degraded_cc_to_c);
4729        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4730            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4731                return status;
4732        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4733            DIFF_OPT_TST(opt, HAS_CHANGES))
4734                result |= 01;
4735        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4736            DIFF_OPT_TST(opt, CHECK_FAILED))
4737                result |= 02;
4738        return result;
4739}
4740
4741int diff_can_quit_early(struct diff_options *opt)
4742{
4743        return (DIFF_OPT_TST(opt, QUICK) &&
4744                !opt->filter &&
4745                DIFF_OPT_TST(opt, HAS_CHANGES));
4746}
4747
4748/*
4749 * Shall changes to this submodule be ignored?
4750 *
4751 * Submodule changes can be configured to be ignored separately for each path,
4752 * but that configuration can be overridden from the command line.
4753 */
4754static int is_submodule_ignored(const char *path, struct diff_options *options)
4755{
4756        int ignored = 0;
4757        unsigned orig_flags = options->flags;
4758        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4759                set_diffopt_flags_from_submodule_config(options, path);
4760        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4761                ignored = 1;
4762        options->flags = orig_flags;
4763        return ignored;
4764}
4765
4766void diff_addremove(struct diff_options *options,
4767                    int addremove, unsigned mode,
4768                    const unsigned char *sha1,
4769                    int sha1_valid,
4770                    const char *concatpath, unsigned dirty_submodule)
4771{
4772        struct diff_filespec *one, *two;
4773
4774        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4775                return;
4776
4777        /* This may look odd, but it is a preparation for
4778         * feeding "there are unchanged files which should
4779         * not produce diffs, but when you are doing copy
4780         * detection you would need them, so here they are"
4781         * entries to the diff-core.  They will be prefixed
4782         * with something like '=' or '*' (I haven't decided
4783         * which but should not make any difference).
4784         * Feeding the same new and old to diff_change()
4785         * also has the same effect.
4786         * Before the final output happens, they are pruned after
4787         * merged into rename/copy pairs as appropriate.
4788         */
4789        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4790                addremove = (addremove == '+' ? '-' :
4791                             addremove == '-' ? '+' : addremove);
4792
4793        if (options->prefix &&
4794            strncmp(concatpath, options->prefix, options->prefix_length))
4795                return;
4796
4797        one = alloc_filespec(concatpath);
4798        two = alloc_filespec(concatpath);
4799
4800        if (addremove != '+')
4801                fill_filespec(one, sha1, sha1_valid, mode);
4802        if (addremove != '-') {
4803                fill_filespec(two, sha1, sha1_valid, mode);
4804                two->dirty_submodule = dirty_submodule;
4805        }
4806
4807        diff_queue(&diff_queued_diff, one, two);
4808        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4809                DIFF_OPT_SET(options, HAS_CHANGES);
4810}
4811
4812void diff_change(struct diff_options *options,
4813                 unsigned old_mode, unsigned new_mode,
4814                 const unsigned char *old_sha1,
4815                 const unsigned char *new_sha1,
4816                 int old_sha1_valid, int new_sha1_valid,
4817                 const char *concatpath,
4818                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4819{
4820        struct diff_filespec *one, *two;
4821
4822        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4823            is_submodule_ignored(concatpath, options))
4824                return;
4825
4826        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4827                unsigned tmp;
4828                const unsigned char *tmp_c;
4829                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4830                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4831                tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4832                        new_sha1_valid = tmp;
4833                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4834                        new_dirty_submodule = tmp;
4835        }
4836
4837        if (options->prefix &&
4838            strncmp(concatpath, options->prefix, options->prefix_length))
4839                return;
4840
4841        one = alloc_filespec(concatpath);
4842        two = alloc_filespec(concatpath);
4843        fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4844        fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4845        one->dirty_submodule = old_dirty_submodule;
4846        two->dirty_submodule = new_dirty_submodule;
4847
4848        diff_queue(&diff_queued_diff, one, two);
4849        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4850                DIFF_OPT_SET(options, HAS_CHANGES);
4851}
4852
4853struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4854{
4855        struct diff_filepair *pair;
4856        struct diff_filespec *one, *two;
4857
4858        if (options->prefix &&
4859            strncmp(path, options->prefix, options->prefix_length))
4860                return NULL;
4861
4862        one = alloc_filespec(path);
4863        two = alloc_filespec(path);
4864        pair = diff_queue(&diff_queued_diff, one, two);
4865        pair->is_unmerged = 1;
4866        return pair;
4867}
4868
4869static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4870                size_t *outsize)
4871{
4872        struct diff_tempfile *temp;
4873        const char *argv[3];
4874        const char **arg = argv;
4875        struct child_process child;
4876        struct strbuf buf = STRBUF_INIT;
4877        int err = 0;
4878
4879        temp = prepare_temp_file(spec->path, spec);
4880        *arg++ = pgm;
4881        *arg++ = temp->name;
4882        *arg = NULL;
4883
4884        memset(&child, 0, sizeof(child));
4885        child.use_shell = 1;
4886        child.argv = argv;
4887        child.out = -1;
4888        if (start_command(&child)) {
4889                remove_tempfile();
4890                return NULL;
4891        }
4892
4893        if (strbuf_read(&buf, child.out, 0) < 0)
4894                err = error("error reading from textconv command '%s'", pgm);
4895        close(child.out);
4896
4897        if (finish_command(&child) || err) {
4898                strbuf_release(&buf);
4899                remove_tempfile();
4900                return NULL;
4901        }
4902        remove_tempfile();
4903
4904        return strbuf_detach(&buf, outsize);
4905}
4906
4907size_t fill_textconv(struct userdiff_driver *driver,
4908                     struct diff_filespec *df,
4909                     char **outbuf)
4910{
4911        size_t size;
4912
4913        if (!driver || !driver->textconv) {
4914                if (!DIFF_FILE_VALID(df)) {
4915                        *outbuf = "";
4916                        return 0;
4917                }
4918                if (diff_populate_filespec(df, 0))
4919                        die("unable to read files to diff");
4920                *outbuf = df->data;
4921                return df->size;
4922        }
4923
4924        if (driver->textconv_cache && df->sha1_valid) {
4925                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4926                                          &size);
4927                if (*outbuf)
4928                        return size;
4929        }
4930
4931        *outbuf = run_textconv(driver->textconv, df, &size);
4932        if (!*outbuf)
4933                die("unable to read files to diff");
4934
4935        if (driver->textconv_cache && df->sha1_valid) {
4936                /* ignore errors, as we might be in a readonly repository */
4937                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4938                                size);
4939                /*
4940                 * we could save up changes and flush them all at the end,
4941                 * but we would need an extra call after all diffing is done.
4942                 * Since generating a cache entry is the slow path anyway,
4943                 * this extra overhead probably isn't a big deal.
4944                 */
4945                notes_cache_write(driver->textconv_cache);
4946        }
4947
4948        return size;
4949}
4950
4951void setup_diff_pager(struct diff_options *opt)
4952{
4953        /*
4954         * If the user asked for our exit code, then either they want --quiet
4955         * or --exit-code. We should definitely not bother with a pager in the
4956         * former case, as we will generate no output. Since we still properly
4957         * report our exit code even when a pager is run, we _could_ run a
4958         * pager with --exit-code. But since we have not done so historically,
4959         * and because it is easy to find people oneline advising "git diff
4960         * --exit-code" in hooks and other scripts, we do not do so.
4961         */
4962        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4963            check_pager_config("diff") != 0)
4964                setup_pager();
4965}