diff.con commit submodule: display summary header in bold (4e21513)
   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
  19#ifdef NO_FAST_WORKING_DIRECTORY
  20#define FAST_WORKING_DIRECTORY 0
  21#else
  22#define FAST_WORKING_DIRECTORY 1
  23#endif
  24
  25static int diff_detect_rename_default;
  26static int diff_rename_limit_default = 400;
  27static int diff_suppress_blank_empty;
  28static int diff_use_color_default = -1;
  29static int diff_context_default = 3;
  30static const char *diff_word_regex_cfg;
  31static const char *external_diff_cmd_cfg;
  32int diff_auto_refresh_index = 1;
  33static int diff_mnemonic_prefix;
  34static int diff_no_prefix;
  35static int diff_stat_graph_width;
  36static int diff_dirstat_permille_default = 30;
  37static struct diff_options default_diff_options;
  38
  39static char diff_colors[][COLOR_MAXLEN] = {
  40        GIT_COLOR_RESET,
  41        GIT_COLOR_NORMAL,       /* PLAIN */
  42        GIT_COLOR_BOLD,         /* METAINFO */
  43        GIT_COLOR_CYAN,         /* FRAGINFO */
  44        GIT_COLOR_RED,          /* OLD */
  45        GIT_COLOR_GREEN,        /* NEW */
  46        GIT_COLOR_YELLOW,       /* COMMIT */
  47        GIT_COLOR_BG_RED,       /* WHITESPACE */
  48        GIT_COLOR_NORMAL,       /* FUNCINFO */
  49};
  50
  51static int parse_diff_color_slot(const char *var, int ofs)
  52{
  53        if (!strcasecmp(var+ofs, "plain"))
  54                return DIFF_PLAIN;
  55        if (!strcasecmp(var+ofs, "meta"))
  56                return DIFF_METAINFO;
  57        if (!strcasecmp(var+ofs, "frag"))
  58                return DIFF_FRAGINFO;
  59        if (!strcasecmp(var+ofs, "old"))
  60                return DIFF_FILE_OLD;
  61        if (!strcasecmp(var+ofs, "new"))
  62                return DIFF_FILE_NEW;
  63        if (!strcasecmp(var+ofs, "commit"))
  64                return DIFF_COMMIT;
  65        if (!strcasecmp(var+ofs, "whitespace"))
  66                return DIFF_WHITESPACE;
  67        if (!strcasecmp(var+ofs, "func"))
  68                return DIFF_FUNCINFO;
  69        return -1;
  70}
  71
  72static int parse_dirstat_params(struct diff_options *options, const char *params,
  73                                struct strbuf *errmsg)
  74{
  75        const char *p = params;
  76        int p_len, ret = 0;
  77
  78        while (*p) {
  79                p_len = strchrnul(p, ',') - p;
  80                if (!memcmp(p, "changes", p_len)) {
  81                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  82                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  83                } else if (!memcmp(p, "lines", p_len)) {
  84                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  85                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  86                } else if (!memcmp(p, "files", p_len)) {
  87                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  88                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  89                } else if (!memcmp(p, "noncumulative", p_len)) {
  90                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
  91                } else if (!memcmp(p, "cumulative", p_len)) {
  92                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
  93                } else if (isdigit(*p)) {
  94                        char *end;
  95                        int permille = strtoul(p, &end, 10) * 10;
  96                        if (*end == '.' && isdigit(*++end)) {
  97                                /* only use first digit */
  98                                permille += *end - '0';
  99                                /* .. and ignore any further digits */
 100                                while (isdigit(*++end))
 101                                        ; /* nothing */
 102                        }
 103                        if (end - p == p_len)
 104                                options->dirstat_permille = permille;
 105                        else {
 106                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%.*s'\n"),
 107                                            p_len, p);
 108                                ret++;
 109                        }
 110                } else {
 111                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%.*s'\n"),
 112                                    p_len, p);
 113                        ret++;
 114                }
 115
 116                p += p_len;
 117
 118                if (*p)
 119                        p++; /* more parameters, swallow separator */
 120        }
 121        return ret;
 122}
 123
 124static int parse_submodule_params(struct diff_options *options, const char *value)
 125{
 126        if (!strcmp(value, "log"))
 127                DIFF_OPT_SET(options, SUBMODULE_LOG);
 128        else if (!strcmp(value, "short"))
 129                DIFF_OPT_CLR(options, SUBMODULE_LOG);
 130        else
 131                return -1;
 132        return 0;
 133}
 134
 135static int git_config_rename(const char *var, const char *value)
 136{
 137        if (!value)
 138                return DIFF_DETECT_RENAME;
 139        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 140                return  DIFF_DETECT_COPY;
 141        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 142}
 143
 144/*
 145 * These are to give UI layer defaults.
 146 * The core-level commands such as git-diff-files should
 147 * never be affected by the setting of diff.renames
 148 * the user happens to have in the configuration file.
 149 */
 150int git_diff_ui_config(const char *var, const char *value, void *cb)
 151{
 152        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 153                diff_use_color_default = git_config_colorbool(var, value);
 154                return 0;
 155        }
 156        if (!strcmp(var, "diff.context")) {
 157                diff_context_default = git_config_int(var, value);
 158                if (diff_context_default < 0)
 159                        return -1;
 160                return 0;
 161        }
 162        if (!strcmp(var, "diff.renames")) {
 163                diff_detect_rename_default = git_config_rename(var, value);
 164                return 0;
 165        }
 166        if (!strcmp(var, "diff.autorefreshindex")) {
 167                diff_auto_refresh_index = git_config_bool(var, value);
 168                return 0;
 169        }
 170        if (!strcmp(var, "diff.mnemonicprefix")) {
 171                diff_mnemonic_prefix = git_config_bool(var, value);
 172                return 0;
 173        }
 174        if (!strcmp(var, "diff.noprefix")) {
 175                diff_no_prefix = git_config_bool(var, value);
 176                return 0;
 177        }
 178        if (!strcmp(var, "diff.statgraphwidth")) {
 179                diff_stat_graph_width = git_config_int(var, value);
 180                return 0;
 181        }
 182        if (!strcmp(var, "diff.external"))
 183                return git_config_string(&external_diff_cmd_cfg, var, value);
 184        if (!strcmp(var, "diff.wordregex"))
 185                return git_config_string(&diff_word_regex_cfg, var, value);
 186
 187        if (!strcmp(var, "diff.ignoresubmodules"))
 188                handle_ignore_submodules_arg(&default_diff_options, value);
 189
 190        if (!strcmp(var, "diff.submodule")) {
 191                if (parse_submodule_params(&default_diff_options, value))
 192                        warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
 193                                value);
 194                return 0;
 195        }
 196
 197        if (git_color_config(var, value, cb) < 0)
 198                return -1;
 199
 200        return git_diff_basic_config(var, value, cb);
 201}
 202
 203int git_diff_basic_config(const char *var, const char *value, void *cb)
 204{
 205        if (!strcmp(var, "diff.renamelimit")) {
 206                diff_rename_limit_default = git_config_int(var, value);
 207                return 0;
 208        }
 209
 210        if (userdiff_config(var, value) < 0)
 211                return -1;
 212
 213        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 214                int slot = parse_diff_color_slot(var, 11);
 215                if (slot < 0)
 216                        return 0;
 217                if (!value)
 218                        return config_error_nonbool(var);
 219                color_parse(value, var, diff_colors[slot]);
 220                return 0;
 221        }
 222
 223        /* like GNU diff's --suppress-blank-empty option  */
 224        if (!strcmp(var, "diff.suppressblankempty") ||
 225                        /* for backwards compatibility */
 226                        !strcmp(var, "diff.suppress-blank-empty")) {
 227                diff_suppress_blank_empty = git_config_bool(var, value);
 228                return 0;
 229        }
 230
 231        if (!strcmp(var, "diff.dirstat")) {
 232                struct strbuf errmsg = STRBUF_INIT;
 233                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 234                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 235                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 236                                errmsg.buf);
 237                strbuf_release(&errmsg);
 238                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 239                return 0;
 240        }
 241
 242        if (!prefixcmp(var, "submodule."))
 243                return parse_submodule_config_option(var, value);
 244
 245        return git_default_config(var, value, cb);
 246}
 247
 248static char *quote_two(const char *one, const char *two)
 249{
 250        int need_one = quote_c_style(one, NULL, NULL, 1);
 251        int need_two = quote_c_style(two, NULL, NULL, 1);
 252        struct strbuf res = STRBUF_INIT;
 253
 254        if (need_one + need_two) {
 255                strbuf_addch(&res, '"');
 256                quote_c_style(one, &res, NULL, 1);
 257                quote_c_style(two, &res, NULL, 1);
 258                strbuf_addch(&res, '"');
 259        } else {
 260                strbuf_addstr(&res, one);
 261                strbuf_addstr(&res, two);
 262        }
 263        return strbuf_detach(&res, NULL);
 264}
 265
 266static const char *external_diff(void)
 267{
 268        static const char *external_diff_cmd = NULL;
 269        static int done_preparing = 0;
 270
 271        if (done_preparing)
 272                return external_diff_cmd;
 273        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 274        if (!external_diff_cmd)
 275                external_diff_cmd = external_diff_cmd_cfg;
 276        done_preparing = 1;
 277        return external_diff_cmd;
 278}
 279
 280static struct diff_tempfile {
 281        const char *name; /* filename external diff should read from */
 282        char hex[41];
 283        char mode[10];
 284        char tmp_path[PATH_MAX];
 285} diff_temp[2];
 286
 287typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 288
 289struct emit_callback {
 290        int color_diff;
 291        unsigned ws_rule;
 292        int blank_at_eof_in_preimage;
 293        int blank_at_eof_in_postimage;
 294        int lno_in_preimage;
 295        int lno_in_postimage;
 296        sane_truncate_fn truncate;
 297        const char **label_path;
 298        struct diff_words_data *diff_words;
 299        struct diff_options *opt;
 300        int *found_changesp;
 301        struct strbuf *header;
 302};
 303
 304static int count_lines(const char *data, int size)
 305{
 306        int count, ch, completely_empty = 1, nl_just_seen = 0;
 307        count = 0;
 308        while (0 < size--) {
 309                ch = *data++;
 310                if (ch == '\n') {
 311                        count++;
 312                        nl_just_seen = 1;
 313                        completely_empty = 0;
 314                }
 315                else {
 316                        nl_just_seen = 0;
 317                        completely_empty = 0;
 318                }
 319        }
 320        if (completely_empty)
 321                return 0;
 322        if (!nl_just_seen)
 323                count++; /* no trailing newline */
 324        return count;
 325}
 326
 327static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 328{
 329        if (!DIFF_FILE_VALID(one)) {
 330                mf->ptr = (char *)""; /* does not matter */
 331                mf->size = 0;
 332                return 0;
 333        }
 334        else if (diff_populate_filespec(one, 0))
 335                return -1;
 336
 337        mf->ptr = one->data;
 338        mf->size = one->size;
 339        return 0;
 340}
 341
 342/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 343static unsigned long diff_filespec_size(struct diff_filespec *one)
 344{
 345        if (!DIFF_FILE_VALID(one))
 346                return 0;
 347        diff_populate_filespec(one, 1);
 348        return one->size;
 349}
 350
 351static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 352{
 353        char *ptr = mf->ptr;
 354        long size = mf->size;
 355        int cnt = 0;
 356
 357        if (!size)
 358                return cnt;
 359        ptr += size - 1; /* pointing at the very end */
 360        if (*ptr != '\n')
 361                ; /* incomplete line */
 362        else
 363                ptr--; /* skip the last LF */
 364        while (mf->ptr < ptr) {
 365                char *prev_eol;
 366                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 367                        if (*prev_eol == '\n')
 368                                break;
 369                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 370                        break;
 371                cnt++;
 372                ptr = prev_eol - 1;
 373        }
 374        return cnt;
 375}
 376
 377static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 378                               struct emit_callback *ecbdata)
 379{
 380        int l1, l2, at;
 381        unsigned ws_rule = ecbdata->ws_rule;
 382        l1 = count_trailing_blank(mf1, ws_rule);
 383        l2 = count_trailing_blank(mf2, ws_rule);
 384        if (l2 <= l1) {
 385                ecbdata->blank_at_eof_in_preimage = 0;
 386                ecbdata->blank_at_eof_in_postimage = 0;
 387                return;
 388        }
 389        at = count_lines(mf1->ptr, mf1->size);
 390        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 391
 392        at = count_lines(mf2->ptr, mf2->size);
 393        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 394}
 395
 396static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 397                        int first, const char *line, int len)
 398{
 399        int has_trailing_newline, has_trailing_carriage_return;
 400        int nofirst;
 401        FILE *file = o->file;
 402
 403        if (o->output_prefix) {
 404                struct strbuf *msg = NULL;
 405                msg = o->output_prefix(o, o->output_prefix_data);
 406                assert(msg);
 407                fwrite(msg->buf, msg->len, 1, file);
 408        }
 409
 410        if (len == 0) {
 411                has_trailing_newline = (first == '\n');
 412                has_trailing_carriage_return = (!has_trailing_newline &&
 413                                                (first == '\r'));
 414                nofirst = has_trailing_newline || has_trailing_carriage_return;
 415        } else {
 416                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 417                if (has_trailing_newline)
 418                        len--;
 419                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 420                if (has_trailing_carriage_return)
 421                        len--;
 422                nofirst = 0;
 423        }
 424
 425        if (len || !nofirst) {
 426                fputs(set, file);
 427                if (!nofirst)
 428                        fputc(first, file);
 429                fwrite(line, len, 1, file);
 430                fputs(reset, file);
 431        }
 432        if (has_trailing_carriage_return)
 433                fputc('\r', file);
 434        if (has_trailing_newline)
 435                fputc('\n', file);
 436}
 437
 438static void emit_line(struct diff_options *o, const char *set, const char *reset,
 439                      const char *line, int len)
 440{
 441        emit_line_0(o, set, reset, line[0], line+1, len-1);
 442}
 443
 444static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 445{
 446        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 447              ecbdata->blank_at_eof_in_preimage &&
 448              ecbdata->blank_at_eof_in_postimage &&
 449              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 450              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 451                return 0;
 452        return ws_blank_line(line, len, ecbdata->ws_rule);
 453}
 454
 455static void emit_add_line(const char *reset,
 456                          struct emit_callback *ecbdata,
 457                          const char *line, int len)
 458{
 459        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 460        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 461
 462        if (!*ws)
 463                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 464        else if (new_blank_line_at_eof(ecbdata, line, len))
 465                /* Blank line at EOF - paint '+' as well */
 466                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 467        else {
 468                /* Emit just the prefix, then the rest. */
 469                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 470                ws_check_emit(line, len, ecbdata->ws_rule,
 471                              ecbdata->opt->file, set, reset, ws);
 472        }
 473}
 474
 475static void emit_hunk_header(struct emit_callback *ecbdata,
 476                             const char *line, int len)
 477{
 478        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 479        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 480        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 481        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 482        static const char atat[2] = { '@', '@' };
 483        const char *cp, *ep;
 484        struct strbuf msgbuf = STRBUF_INIT;
 485        int org_len = len;
 486        int i = 1;
 487
 488        /*
 489         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 490         * it always is at least 10 bytes long.
 491         */
 492        if (len < 10 ||
 493            memcmp(line, atat, 2) ||
 494            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 495                emit_line(ecbdata->opt, plain, reset, line, len);
 496                return;
 497        }
 498        ep += 2; /* skip over @@ */
 499
 500        /* The hunk header in fraginfo color */
 501        strbuf_add(&msgbuf, frag, strlen(frag));
 502        strbuf_add(&msgbuf, line, ep - line);
 503        strbuf_add(&msgbuf, reset, strlen(reset));
 504
 505        /*
 506         * trailing "\r\n"
 507         */
 508        for ( ; i < 3; i++)
 509                if (line[len - i] == '\r' || line[len - i] == '\n')
 510                        len--;
 511
 512        /* blank before the func header */
 513        for (cp = ep; ep - line < len; ep++)
 514                if (*ep != ' ' && *ep != '\t')
 515                        break;
 516        if (ep != cp) {
 517                strbuf_add(&msgbuf, plain, strlen(plain));
 518                strbuf_add(&msgbuf, cp, ep - cp);
 519                strbuf_add(&msgbuf, reset, strlen(reset));
 520        }
 521
 522        if (ep < line + len) {
 523                strbuf_add(&msgbuf, func, strlen(func));
 524                strbuf_add(&msgbuf, ep, line + len - ep);
 525                strbuf_add(&msgbuf, reset, strlen(reset));
 526        }
 527
 528        strbuf_add(&msgbuf, line + len, org_len - len);
 529        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 530        strbuf_release(&msgbuf);
 531}
 532
 533static struct diff_tempfile *claim_diff_tempfile(void) {
 534        int i;
 535        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 536                if (!diff_temp[i].name)
 537                        return diff_temp + i;
 538        die("BUG: diff is failing to clean up its tempfiles");
 539}
 540
 541static int remove_tempfile_installed;
 542
 543static void remove_tempfile(void)
 544{
 545        int i;
 546        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 547                if (diff_temp[i].name == diff_temp[i].tmp_path)
 548                        unlink_or_warn(diff_temp[i].name);
 549                diff_temp[i].name = NULL;
 550        }
 551}
 552
 553static void remove_tempfile_on_signal(int signo)
 554{
 555        remove_tempfile();
 556        sigchain_pop(signo);
 557        raise(signo);
 558}
 559
 560static void print_line_count(FILE *file, int count)
 561{
 562        switch (count) {
 563        case 0:
 564                fprintf(file, "0,0");
 565                break;
 566        case 1:
 567                fprintf(file, "1");
 568                break;
 569        default:
 570                fprintf(file, "1,%d", count);
 571                break;
 572        }
 573}
 574
 575static void emit_rewrite_lines(struct emit_callback *ecb,
 576                               int prefix, const char *data, int size)
 577{
 578        const char *endp = NULL;
 579        static const char *nneof = " No newline at end of file\n";
 580        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 581        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 582
 583        while (0 < size) {
 584                int len;
 585
 586                endp = memchr(data, '\n', size);
 587                len = endp ? (endp - data + 1) : size;
 588                if (prefix != '+') {
 589                        ecb->lno_in_preimage++;
 590                        emit_line_0(ecb->opt, old, reset, '-',
 591                                    data, len);
 592                } else {
 593                        ecb->lno_in_postimage++;
 594                        emit_add_line(reset, ecb, data, len);
 595                }
 596                size -= len;
 597                data += len;
 598        }
 599        if (!endp) {
 600                const char *plain = diff_get_color(ecb->color_diff,
 601                                                   DIFF_PLAIN);
 602                putc('\n', ecb->opt->file);
 603                emit_line_0(ecb->opt, plain, reset, '\\',
 604                            nneof, strlen(nneof));
 605        }
 606}
 607
 608static void emit_rewrite_diff(const char *name_a,
 609                              const char *name_b,
 610                              struct diff_filespec *one,
 611                              struct diff_filespec *two,
 612                              struct userdiff_driver *textconv_one,
 613                              struct userdiff_driver *textconv_two,
 614                              struct diff_options *o)
 615{
 616        int lc_a, lc_b;
 617        const char *name_a_tab, *name_b_tab;
 618        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 619        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 620        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 621        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 622        const char *a_prefix, *b_prefix;
 623        char *data_one, *data_two;
 624        size_t size_one, size_two;
 625        struct emit_callback ecbdata;
 626        char *line_prefix = "";
 627        struct strbuf *msgbuf;
 628
 629        if (o && o->output_prefix) {
 630                msgbuf = o->output_prefix(o, o->output_prefix_data);
 631                line_prefix = msgbuf->buf;
 632        }
 633
 634        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 635                a_prefix = o->b_prefix;
 636                b_prefix = o->a_prefix;
 637        } else {
 638                a_prefix = o->a_prefix;
 639                b_prefix = o->b_prefix;
 640        }
 641
 642        name_a += (*name_a == '/');
 643        name_b += (*name_b == '/');
 644        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 645        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 646
 647        strbuf_reset(&a_name);
 648        strbuf_reset(&b_name);
 649        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 650        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 651
 652        size_one = fill_textconv(textconv_one, one, &data_one);
 653        size_two = fill_textconv(textconv_two, two, &data_two);
 654
 655        memset(&ecbdata, 0, sizeof(ecbdata));
 656        ecbdata.color_diff = want_color(o->use_color);
 657        ecbdata.found_changesp = &o->found_changes;
 658        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 659        ecbdata.opt = o;
 660        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 661                mmfile_t mf1, mf2;
 662                mf1.ptr = (char *)data_one;
 663                mf2.ptr = (char *)data_two;
 664                mf1.size = size_one;
 665                mf2.size = size_two;
 666                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 667        }
 668        ecbdata.lno_in_preimage = 1;
 669        ecbdata.lno_in_postimage = 1;
 670
 671        lc_a = count_lines(data_one, size_one);
 672        lc_b = count_lines(data_two, size_two);
 673        fprintf(o->file,
 674                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 675                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 676                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 677                line_prefix, fraginfo);
 678        if (!o->irreversible_delete)
 679                print_line_count(o->file, lc_a);
 680        else
 681                fprintf(o->file, "?,?");
 682        fprintf(o->file, " +");
 683        print_line_count(o->file, lc_b);
 684        fprintf(o->file, " @@%s\n", reset);
 685        if (lc_a && !o->irreversible_delete)
 686                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 687        if (lc_b)
 688                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 689        if (textconv_one)
 690                free((char *)data_one);
 691        if (textconv_two)
 692                free((char *)data_two);
 693}
 694
 695struct diff_words_buffer {
 696        mmfile_t text;
 697        long alloc;
 698        struct diff_words_orig {
 699                const char *begin, *end;
 700        } *orig;
 701        int orig_nr, orig_alloc;
 702};
 703
 704static void diff_words_append(char *line, unsigned long len,
 705                struct diff_words_buffer *buffer)
 706{
 707        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 708        line++;
 709        len--;
 710        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 711        buffer->text.size += len;
 712        buffer->text.ptr[buffer->text.size] = '\0';
 713}
 714
 715struct diff_words_style_elem {
 716        const char *prefix;
 717        const char *suffix;
 718        const char *color; /* NULL; filled in by the setup code if
 719                            * color is enabled */
 720};
 721
 722struct diff_words_style {
 723        enum diff_words_type type;
 724        struct diff_words_style_elem new, old, ctx;
 725        const char *newline;
 726};
 727
 728static struct diff_words_style diff_words_styles[] = {
 729        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 730        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 731        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 732};
 733
 734struct diff_words_data {
 735        struct diff_words_buffer minus, plus;
 736        const char *current_plus;
 737        int last_minus;
 738        struct diff_options *opt;
 739        regex_t *word_regex;
 740        enum diff_words_type type;
 741        struct diff_words_style *style;
 742};
 743
 744static int fn_out_diff_words_write_helper(FILE *fp,
 745                                          struct diff_words_style_elem *st_el,
 746                                          const char *newline,
 747                                          size_t count, const char *buf,
 748                                          const char *line_prefix)
 749{
 750        int print = 0;
 751
 752        while (count) {
 753                char *p = memchr(buf, '\n', count);
 754                if (print)
 755                        fputs(line_prefix, fp);
 756                if (p != buf) {
 757                        if (st_el->color && fputs(st_el->color, fp) < 0)
 758                                return -1;
 759                        if (fputs(st_el->prefix, fp) < 0 ||
 760                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 761                            fputs(st_el->suffix, fp) < 0)
 762                                return -1;
 763                        if (st_el->color && *st_el->color
 764                            && fputs(GIT_COLOR_RESET, fp) < 0)
 765                                return -1;
 766                }
 767                if (!p)
 768                        return 0;
 769                if (fputs(newline, fp) < 0)
 770                        return -1;
 771                count -= p + 1 - buf;
 772                buf = p + 1;
 773                print = 1;
 774        }
 775        return 0;
 776}
 777
 778/*
 779 * '--color-words' algorithm can be described as:
 780 *
 781 *   1. collect a the minus/plus lines of a diff hunk, divided into
 782 *      minus-lines and plus-lines;
 783 *
 784 *   2. break both minus-lines and plus-lines into words and
 785 *      place them into two mmfile_t with one word for each line;
 786 *
 787 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 788 *
 789 * And for the common parts of the both file, we output the plus side text.
 790 * diff_words->current_plus is used to trace the current position of the plus file
 791 * which printed. diff_words->last_minus is used to trace the last minus word
 792 * printed.
 793 *
 794 * For '--graph' to work with '--color-words', we need to output the graph prefix
 795 * on each line of color words output. Generally, there are two conditions on
 796 * which we should output the prefix.
 797 *
 798 *   1. diff_words->last_minus == 0 &&
 799 *      diff_words->current_plus == diff_words->plus.text.ptr
 800 *
 801 *      that is: the plus text must start as a new line, and if there is no minus
 802 *      word printed, a graph prefix must be printed.
 803 *
 804 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 805 *      *(diff_words->current_plus - 1) == '\n'
 806 *
 807 *      that is: a graph prefix must be printed following a '\n'
 808 */
 809static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 810{
 811        if ((diff_words->last_minus == 0 &&
 812                diff_words->current_plus == diff_words->plus.text.ptr) ||
 813                (diff_words->current_plus > diff_words->plus.text.ptr &&
 814                *(diff_words->current_plus - 1) == '\n')) {
 815                return 1;
 816        } else {
 817                return 0;
 818        }
 819}
 820
 821static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 822{
 823        struct diff_words_data *diff_words = priv;
 824        struct diff_words_style *style = diff_words->style;
 825        int minus_first, minus_len, plus_first, plus_len;
 826        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 827        struct diff_options *opt = diff_words->opt;
 828        struct strbuf *msgbuf;
 829        char *line_prefix = "";
 830
 831        if (line[0] != '@' || parse_hunk_header(line, len,
 832                        &minus_first, &minus_len, &plus_first, &plus_len))
 833                return;
 834
 835        assert(opt);
 836        if (opt->output_prefix) {
 837                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 838                line_prefix = msgbuf->buf;
 839        }
 840
 841        /* POSIX requires that first be decremented by one if len == 0... */
 842        if (minus_len) {
 843                minus_begin = diff_words->minus.orig[minus_first].begin;
 844                minus_end =
 845                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 846        } else
 847                minus_begin = minus_end =
 848                        diff_words->minus.orig[minus_first].end;
 849
 850        if (plus_len) {
 851                plus_begin = diff_words->plus.orig[plus_first].begin;
 852                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 853        } else
 854                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 855
 856        if (color_words_output_graph_prefix(diff_words)) {
 857                fputs(line_prefix, diff_words->opt->file);
 858        }
 859        if (diff_words->current_plus != plus_begin) {
 860                fn_out_diff_words_write_helper(diff_words->opt->file,
 861                                &style->ctx, style->newline,
 862                                plus_begin - diff_words->current_plus,
 863                                diff_words->current_plus, line_prefix);
 864                if (*(plus_begin - 1) == '\n')
 865                        fputs(line_prefix, diff_words->opt->file);
 866        }
 867        if (minus_begin != minus_end) {
 868                fn_out_diff_words_write_helper(diff_words->opt->file,
 869                                &style->old, style->newline,
 870                                minus_end - minus_begin, minus_begin,
 871                                line_prefix);
 872        }
 873        if (plus_begin != plus_end) {
 874                fn_out_diff_words_write_helper(diff_words->opt->file,
 875                                &style->new, style->newline,
 876                                plus_end - plus_begin, plus_begin,
 877                                line_prefix);
 878        }
 879
 880        diff_words->current_plus = plus_end;
 881        diff_words->last_minus = minus_first;
 882}
 883
 884/* This function starts looking at *begin, and returns 0 iff a word was found. */
 885static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 886                int *begin, int *end)
 887{
 888        if (word_regex && *begin < buffer->size) {
 889                regmatch_t match[1];
 890                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 891                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 892                                        '\n', match[0].rm_eo - match[0].rm_so);
 893                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 894                        *begin += match[0].rm_so;
 895                        return *begin >= *end;
 896                }
 897                return -1;
 898        }
 899
 900        /* find the next word */
 901        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 902                (*begin)++;
 903        if (*begin >= buffer->size)
 904                return -1;
 905
 906        /* find the end of the word */
 907        *end = *begin + 1;
 908        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 909                (*end)++;
 910
 911        return 0;
 912}
 913
 914/*
 915 * This function splits the words in buffer->text, stores the list with
 916 * newline separator into out, and saves the offsets of the original words
 917 * in buffer->orig.
 918 */
 919static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 920                regex_t *word_regex)
 921{
 922        int i, j;
 923        long alloc = 0;
 924
 925        out->size = 0;
 926        out->ptr = NULL;
 927
 928        /* fake an empty "0th" word */
 929        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 930        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 931        buffer->orig_nr = 1;
 932
 933        for (i = 0; i < buffer->text.size; i++) {
 934                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 935                        return;
 936
 937                /* store original boundaries */
 938                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 939                                buffer->orig_alloc);
 940                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 941                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 942                buffer->orig_nr++;
 943
 944                /* store one word */
 945                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 946                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 947                out->ptr[out->size + j - i] = '\n';
 948                out->size += j - i + 1;
 949
 950                i = j - 1;
 951        }
 952}
 953
 954/* this executes the word diff on the accumulated buffers */
 955static void diff_words_show(struct diff_words_data *diff_words)
 956{
 957        xpparam_t xpp;
 958        xdemitconf_t xecfg;
 959        mmfile_t minus, plus;
 960        struct diff_words_style *style = diff_words->style;
 961
 962        struct diff_options *opt = diff_words->opt;
 963        struct strbuf *msgbuf;
 964        char *line_prefix = "";
 965
 966        assert(opt);
 967        if (opt->output_prefix) {
 968                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 969                line_prefix = msgbuf->buf;
 970        }
 971
 972        /* special case: only removal */
 973        if (!diff_words->plus.text.size) {
 974                fputs(line_prefix, diff_words->opt->file);
 975                fn_out_diff_words_write_helper(diff_words->opt->file,
 976                        &style->old, style->newline,
 977                        diff_words->minus.text.size,
 978                        diff_words->minus.text.ptr, line_prefix);
 979                diff_words->minus.text.size = 0;
 980                return;
 981        }
 982
 983        diff_words->current_plus = diff_words->plus.text.ptr;
 984        diff_words->last_minus = 0;
 985
 986        memset(&xpp, 0, sizeof(xpp));
 987        memset(&xecfg, 0, sizeof(xecfg));
 988        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 989        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 990        xpp.flags = 0;
 991        /* as only the hunk header will be parsed, we need a 0-context */
 992        xecfg.ctxlen = 0;
 993        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 994                      &xpp, &xecfg);
 995        free(minus.ptr);
 996        free(plus.ptr);
 997        if (diff_words->current_plus != diff_words->plus.text.ptr +
 998                        diff_words->plus.text.size) {
 999                if (color_words_output_graph_prefix(diff_words))
1000                        fputs(line_prefix, diff_words->opt->file);
1001                fn_out_diff_words_write_helper(diff_words->opt->file,
1002                        &style->ctx, style->newline,
1003                        diff_words->plus.text.ptr + diff_words->plus.text.size
1004                        - diff_words->current_plus, diff_words->current_plus,
1005                        line_prefix);
1006        }
1007        diff_words->minus.text.size = diff_words->plus.text.size = 0;
1008}
1009
1010/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1011static void diff_words_flush(struct emit_callback *ecbdata)
1012{
1013        if (ecbdata->diff_words->minus.text.size ||
1014            ecbdata->diff_words->plus.text.size)
1015                diff_words_show(ecbdata->diff_words);
1016}
1017
1018static void diff_filespec_load_driver(struct diff_filespec *one)
1019{
1020        /* Use already-loaded driver */
1021        if (one->driver)
1022                return;
1023
1024        if (S_ISREG(one->mode))
1025                one->driver = userdiff_find_by_path(one->path);
1026
1027        /* Fallback to default settings */
1028        if (!one->driver)
1029                one->driver = userdiff_find_by_name("default");
1030}
1031
1032static const char *userdiff_word_regex(struct diff_filespec *one)
1033{
1034        diff_filespec_load_driver(one);
1035        return one->driver->word_regex;
1036}
1037
1038static void init_diff_words_data(struct emit_callback *ecbdata,
1039                                 struct diff_options *orig_opts,
1040                                 struct diff_filespec *one,
1041                                 struct diff_filespec *two)
1042{
1043        int i;
1044        struct diff_options *o = xmalloc(sizeof(struct diff_options));
1045        memcpy(o, orig_opts, sizeof(struct diff_options));
1046
1047        ecbdata->diff_words =
1048                xcalloc(1, sizeof(struct diff_words_data));
1049        ecbdata->diff_words->type = o->word_diff;
1050        ecbdata->diff_words->opt = o;
1051        if (!o->word_regex)
1052                o->word_regex = userdiff_word_regex(one);
1053        if (!o->word_regex)
1054                o->word_regex = userdiff_word_regex(two);
1055        if (!o->word_regex)
1056                o->word_regex = diff_word_regex_cfg;
1057        if (o->word_regex) {
1058                ecbdata->diff_words->word_regex = (regex_t *)
1059                        xmalloc(sizeof(regex_t));
1060                if (regcomp(ecbdata->diff_words->word_regex,
1061                            o->word_regex,
1062                            REG_EXTENDED | REG_NEWLINE))
1063                        die ("Invalid regular expression: %s",
1064                             o->word_regex);
1065        }
1066        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1067                if (o->word_diff == diff_words_styles[i].type) {
1068                        ecbdata->diff_words->style =
1069                                &diff_words_styles[i];
1070                        break;
1071                }
1072        }
1073        if (want_color(o->use_color)) {
1074                struct diff_words_style *st = ecbdata->diff_words->style;
1075                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1076                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1077                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1078        }
1079}
1080
1081static void free_diff_words_data(struct emit_callback *ecbdata)
1082{
1083        if (ecbdata->diff_words) {
1084                diff_words_flush(ecbdata);
1085                free (ecbdata->diff_words->opt);
1086                free (ecbdata->diff_words->minus.text.ptr);
1087                free (ecbdata->diff_words->minus.orig);
1088                free (ecbdata->diff_words->plus.text.ptr);
1089                free (ecbdata->diff_words->plus.orig);
1090                if (ecbdata->diff_words->word_regex) {
1091                        regfree(ecbdata->diff_words->word_regex);
1092                        free(ecbdata->diff_words->word_regex);
1093                }
1094                free(ecbdata->diff_words);
1095                ecbdata->diff_words = NULL;
1096        }
1097}
1098
1099const char *diff_get_color(int diff_use_color, enum color_diff ix)
1100{
1101        if (want_color(diff_use_color))
1102                return diff_colors[ix];
1103        return "";
1104}
1105
1106static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1107{
1108        const char *cp;
1109        unsigned long allot;
1110        size_t l = len;
1111
1112        if (ecb->truncate)
1113                return ecb->truncate(line, len);
1114        cp = line;
1115        allot = l;
1116        while (0 < l) {
1117                (void) utf8_width(&cp, &l);
1118                if (!cp)
1119                        break; /* truncated in the middle? */
1120        }
1121        return allot - l;
1122}
1123
1124static void find_lno(const char *line, struct emit_callback *ecbdata)
1125{
1126        const char *p;
1127        ecbdata->lno_in_preimage = 0;
1128        ecbdata->lno_in_postimage = 0;
1129        p = strchr(line, '-');
1130        if (!p)
1131                return; /* cannot happen */
1132        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1133        p = strchr(p, '+');
1134        if (!p)
1135                return; /* cannot happen */
1136        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1137}
1138
1139static void fn_out_consume(void *priv, char *line, unsigned long len)
1140{
1141        struct emit_callback *ecbdata = priv;
1142        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1143        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1144        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1145        struct diff_options *o = ecbdata->opt;
1146        char *line_prefix = "";
1147        struct strbuf *msgbuf;
1148
1149        if (o && o->output_prefix) {
1150                msgbuf = o->output_prefix(o, o->output_prefix_data);
1151                line_prefix = msgbuf->buf;
1152        }
1153
1154        if (ecbdata->header) {
1155                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1156                strbuf_reset(ecbdata->header);
1157                ecbdata->header = NULL;
1158        }
1159        *(ecbdata->found_changesp) = 1;
1160
1161        if (ecbdata->label_path[0]) {
1162                const char *name_a_tab, *name_b_tab;
1163
1164                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1165                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1166
1167                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1168                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1169                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1170                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1171                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1172        }
1173
1174        if (diff_suppress_blank_empty
1175            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1176                line[0] = '\n';
1177                len = 1;
1178        }
1179
1180        if (line[0] == '@') {
1181                if (ecbdata->diff_words)
1182                        diff_words_flush(ecbdata);
1183                len = sane_truncate_line(ecbdata, line, len);
1184                find_lno(line, ecbdata);
1185                emit_hunk_header(ecbdata, line, len);
1186                if (line[len-1] != '\n')
1187                        putc('\n', ecbdata->opt->file);
1188                return;
1189        }
1190
1191        if (len < 1) {
1192                emit_line(ecbdata->opt, reset, reset, line, len);
1193                if (ecbdata->diff_words
1194                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1195                        fputs("~\n", ecbdata->opt->file);
1196                return;
1197        }
1198
1199        if (ecbdata->diff_words) {
1200                if (line[0] == '-') {
1201                        diff_words_append(line, len,
1202                                          &ecbdata->diff_words->minus);
1203                        return;
1204                } else if (line[0] == '+') {
1205                        diff_words_append(line, len,
1206                                          &ecbdata->diff_words->plus);
1207                        return;
1208                } else if (!prefixcmp(line, "\\ ")) {
1209                        /*
1210                         * Eat the "no newline at eof" marker as if we
1211                         * saw a "+" or "-" line with nothing on it,
1212                         * and return without diff_words_flush() to
1213                         * defer processing. If this is the end of
1214                         * preimage, more "+" lines may come after it.
1215                         */
1216                        return;
1217                }
1218                diff_words_flush(ecbdata);
1219                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1220                        emit_line(ecbdata->opt, plain, reset, line, len);
1221                        fputs("~\n", ecbdata->opt->file);
1222                } else {
1223                        /*
1224                         * Skip the prefix character, if any.  With
1225                         * diff_suppress_blank_empty, there may be
1226                         * none.
1227                         */
1228                        if (line[0] != '\n') {
1229                              line++;
1230                              len--;
1231                        }
1232                        emit_line(ecbdata->opt, plain, reset, line, len);
1233                }
1234                return;
1235        }
1236
1237        if (line[0] != '+') {
1238                const char *color =
1239                        diff_get_color(ecbdata->color_diff,
1240                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1241                ecbdata->lno_in_preimage++;
1242                if (line[0] == ' ')
1243                        ecbdata->lno_in_postimage++;
1244                emit_line(ecbdata->opt, color, reset, line, len);
1245        } else {
1246                ecbdata->lno_in_postimage++;
1247                emit_add_line(reset, ecbdata, line + 1, len - 1);
1248        }
1249}
1250
1251static char *pprint_rename(const char *a, const char *b)
1252{
1253        const char *old = a;
1254        const char *new = b;
1255        struct strbuf name = STRBUF_INIT;
1256        int pfx_length, sfx_length;
1257        int len_a = strlen(a);
1258        int len_b = strlen(b);
1259        int a_midlen, b_midlen;
1260        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1261        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1262
1263        if (qlen_a || qlen_b) {
1264                quote_c_style(a, &name, NULL, 0);
1265                strbuf_addstr(&name, " => ");
1266                quote_c_style(b, &name, NULL, 0);
1267                return strbuf_detach(&name, NULL);
1268        }
1269
1270        /* Find common prefix */
1271        pfx_length = 0;
1272        while (*old && *new && *old == *new) {
1273                if (*old == '/')
1274                        pfx_length = old - a + 1;
1275                old++;
1276                new++;
1277        }
1278
1279        /* Find common suffix */
1280        old = a + len_a;
1281        new = b + len_b;
1282        sfx_length = 0;
1283        while (a <= old && b <= new && *old == *new) {
1284                if (*old == '/')
1285                        sfx_length = len_a - (old - a);
1286                old--;
1287                new--;
1288        }
1289
1290        /*
1291         * pfx{mid-a => mid-b}sfx
1292         * {pfx-a => pfx-b}sfx
1293         * pfx{sfx-a => sfx-b}
1294         * name-a => name-b
1295         */
1296        a_midlen = len_a - pfx_length - sfx_length;
1297        b_midlen = len_b - pfx_length - sfx_length;
1298        if (a_midlen < 0)
1299                a_midlen = 0;
1300        if (b_midlen < 0)
1301                b_midlen = 0;
1302
1303        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1304        if (pfx_length + sfx_length) {
1305                strbuf_add(&name, a, pfx_length);
1306                strbuf_addch(&name, '{');
1307        }
1308        strbuf_add(&name, a + pfx_length, a_midlen);
1309        strbuf_addstr(&name, " => ");
1310        strbuf_add(&name, b + pfx_length, b_midlen);
1311        if (pfx_length + sfx_length) {
1312                strbuf_addch(&name, '}');
1313                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1314        }
1315        return strbuf_detach(&name, NULL);
1316}
1317
1318struct diffstat_t {
1319        int nr;
1320        int alloc;
1321        struct diffstat_file {
1322                char *from_name;
1323                char *name;
1324                char *print_name;
1325                unsigned is_unmerged:1;
1326                unsigned is_binary:1;
1327                unsigned is_renamed:1;
1328                uintmax_t added, deleted;
1329        } **files;
1330};
1331
1332static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1333                                          const char *name_a,
1334                                          const char *name_b)
1335{
1336        struct diffstat_file *x;
1337        x = xcalloc(sizeof (*x), 1);
1338        if (diffstat->nr == diffstat->alloc) {
1339                diffstat->alloc = alloc_nr(diffstat->alloc);
1340                diffstat->files = xrealloc(diffstat->files,
1341                                diffstat->alloc * sizeof(x));
1342        }
1343        diffstat->files[diffstat->nr++] = x;
1344        if (name_b) {
1345                x->from_name = xstrdup(name_a);
1346                x->name = xstrdup(name_b);
1347                x->is_renamed = 1;
1348        }
1349        else {
1350                x->from_name = NULL;
1351                x->name = xstrdup(name_a);
1352        }
1353        return x;
1354}
1355
1356static void diffstat_consume(void *priv, char *line, unsigned long len)
1357{
1358        struct diffstat_t *diffstat = priv;
1359        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1360
1361        if (line[0] == '+')
1362                x->added++;
1363        else if (line[0] == '-')
1364                x->deleted++;
1365}
1366
1367const char mime_boundary_leader[] = "------------";
1368
1369static int scale_linear(int it, int width, int max_change)
1370{
1371        if (!it)
1372                return 0;
1373        /*
1374         * make sure that at least one '-' or '+' is printed if
1375         * there is any change to this path. The easiest way is to
1376         * scale linearly as if the alloted width is one column shorter
1377         * than it is, and then add 1 to the result.
1378         */
1379        return 1 + (it * (width - 1) / max_change);
1380}
1381
1382static void show_name(FILE *file,
1383                      const char *prefix, const char *name, int len)
1384{
1385        fprintf(file, " %s%-*s |", prefix, len, name);
1386}
1387
1388static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1389{
1390        if (cnt <= 0)
1391                return;
1392        fprintf(file, "%s", set);
1393        while (cnt--)
1394                putc(ch, file);
1395        fprintf(file, "%s", reset);
1396}
1397
1398static void fill_print_name(struct diffstat_file *file)
1399{
1400        char *pname;
1401
1402        if (file->print_name)
1403                return;
1404
1405        if (!file->is_renamed) {
1406                struct strbuf buf = STRBUF_INIT;
1407                if (quote_c_style(file->name, &buf, NULL, 0)) {
1408                        pname = strbuf_detach(&buf, NULL);
1409                } else {
1410                        pname = file->name;
1411                        strbuf_release(&buf);
1412                }
1413        } else {
1414                pname = pprint_rename(file->from_name, file->name);
1415        }
1416        file->print_name = pname;
1417}
1418
1419int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1420{
1421        struct strbuf sb = STRBUF_INIT;
1422        int ret;
1423
1424        if (!files) {
1425                assert(insertions == 0 && deletions == 0);
1426                return fprintf(fp, "%s\n", " 0 files changed");
1427        }
1428
1429        strbuf_addf(&sb,
1430                    (files == 1) ? " %d file changed" : " %d files changed",
1431                    files);
1432
1433        /*
1434         * For binary diff, the caller may want to print "x files
1435         * changed" with insertions == 0 && deletions == 0.
1436         *
1437         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1438         * is probably less confusing (i.e skip over "2 files changed
1439         * but nothing about added/removed lines? Is this a bug in Git?").
1440         */
1441        if (insertions || deletions == 0) {
1442                /*
1443                 * TRANSLATORS: "+" in (+) is a line addition marker;
1444                 * do not translate it.
1445                 */
1446                strbuf_addf(&sb,
1447                            (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1448                            insertions);
1449        }
1450
1451        if (deletions || insertions == 0) {
1452                /*
1453                 * TRANSLATORS: "-" in (-) is a line removal marker;
1454                 * do not translate it.
1455                 */
1456                strbuf_addf(&sb,
1457                            (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1458                            deletions);
1459        }
1460        strbuf_addch(&sb, '\n');
1461        ret = fputs(sb.buf, fp);
1462        strbuf_release(&sb);
1463        return ret;
1464}
1465
1466static void show_stats(struct diffstat_t *data, struct diff_options *options)
1467{
1468        int i, len, add, del, adds = 0, dels = 0;
1469        uintmax_t max_change = 0, max_len = 0;
1470        int total_files = data->nr, count;
1471        int width, name_width, graph_width, number_width = 0, bin_width = 0;
1472        const char *reset, *add_c, *del_c;
1473        const char *line_prefix = "";
1474        int extra_shown = 0;
1475        struct strbuf *msg = NULL;
1476
1477        if (data->nr == 0)
1478                return;
1479
1480        if (options->output_prefix) {
1481                msg = options->output_prefix(options, options->output_prefix_data);
1482                line_prefix = msg->buf;
1483        }
1484
1485        count = options->stat_count ? options->stat_count : data->nr;
1486
1487        reset = diff_get_color_opt(options, DIFF_RESET);
1488        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1489        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1490
1491        /*
1492         * Find the longest filename and max number of changes
1493         */
1494        for (i = 0; (i < count) && (i < data->nr); i++) {
1495                struct diffstat_file *file = data->files[i];
1496                uintmax_t change = file->added + file->deleted;
1497                if (!data->files[i]->is_renamed &&
1498                         (change == 0)) {
1499                        count++; /* not shown == room for one more */
1500                        continue;
1501                }
1502                fill_print_name(file);
1503                len = strlen(file->print_name);
1504                if (max_len < len)
1505                        max_len = len;
1506
1507                if (file->is_unmerged) {
1508                        /* "Unmerged" is 8 characters */
1509                        bin_width = bin_width < 8 ? 8 : bin_width;
1510                        continue;
1511                }
1512                if (file->is_binary) {
1513                        /* "Bin XXX -> YYY bytes" */
1514                        int w = 14 + decimal_width(file->added)
1515                                + decimal_width(file->deleted);
1516                        bin_width = bin_width < w ? w : bin_width;
1517                        /* Display change counts aligned with "Bin" */
1518                        number_width = 3;
1519                        continue;
1520                }
1521
1522                if (max_change < change)
1523                        max_change = change;
1524        }
1525        count = i; /* min(count, data->nr) */
1526
1527        /*
1528         * We have width = stat_width or term_columns() columns total.
1529         * We want a maximum of min(max_len, stat_name_width) for the name part.
1530         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1531         * We also need 1 for " " and 4 + decimal_width(max_change)
1532         * for " | NNNN " and one the empty column at the end, altogether
1533         * 6 + decimal_width(max_change).
1534         *
1535         * If there's not enough space, we will use the smaller of
1536         * stat_name_width (if set) and 5/8*width for the filename,
1537         * and the rest for constant elements + graph part, but no more
1538         * than stat_graph_width for the graph part.
1539         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1540         * for the standard terminal size).
1541         *
1542         * In other words: stat_width limits the maximum width, and
1543         * stat_name_width fixes the maximum width of the filename,
1544         * and is also used to divide available columns if there
1545         * aren't enough.
1546         *
1547         * Binary files are displayed with "Bin XXX -> YYY bytes"
1548         * instead of the change count and graph. This part is treated
1549         * similarly to the graph part, except that it is not
1550         * "scaled". If total width is too small to accomodate the
1551         * guaranteed minimum width of the filename part and the
1552         * separators and this message, this message will "overflow"
1553         * making the line longer than the maximum width.
1554         */
1555
1556        if (options->stat_width == -1)
1557                width = term_columns() - options->output_prefix_length;
1558        else
1559                width = options->stat_width ? options->stat_width : 80;
1560        number_width = decimal_width(max_change) > number_width ?
1561                decimal_width(max_change) : number_width;
1562
1563        if (options->stat_graph_width == -1)
1564                options->stat_graph_width = diff_stat_graph_width;
1565
1566        /*
1567         * Guarantee 3/8*16==6 for the graph part
1568         * and 5/8*16==10 for the filename part
1569         */
1570        if (width < 16 + 6 + number_width)
1571                width = 16 + 6 + number_width;
1572
1573        /*
1574         * First assign sizes that are wanted, ignoring available width.
1575         * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1576         * starting from "XXX" should fit in graph_width.
1577         */
1578        graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1579        if (options->stat_graph_width &&
1580            options->stat_graph_width < graph_width)
1581                graph_width = options->stat_graph_width;
1582
1583        name_width = (options->stat_name_width > 0 &&
1584                      options->stat_name_width < max_len) ?
1585                options->stat_name_width : max_len;
1586
1587        /*
1588         * Adjust adjustable widths not to exceed maximum width
1589         */
1590        if (name_width + number_width + 6 + graph_width > width) {
1591                if (graph_width > width * 3/8 - number_width - 6) {
1592                        graph_width = width * 3/8 - number_width - 6;
1593                        if (graph_width < 6)
1594                                graph_width = 6;
1595                }
1596
1597                if (options->stat_graph_width &&
1598                    graph_width > options->stat_graph_width)
1599                        graph_width = options->stat_graph_width;
1600                if (name_width > width - number_width - 6 - graph_width)
1601                        name_width = width - number_width - 6 - graph_width;
1602                else
1603                        graph_width = width - number_width - 6 - name_width;
1604        }
1605
1606        /*
1607         * From here name_width is the width of the name area,
1608         * and graph_width is the width of the graph area.
1609         * max_change is used to scale graph properly.
1610         */
1611        for (i = 0; i < count; i++) {
1612                const char *prefix = "";
1613                char *name = data->files[i]->print_name;
1614                uintmax_t added = data->files[i]->added;
1615                uintmax_t deleted = data->files[i]->deleted;
1616                int name_len;
1617
1618                if (!data->files[i]->is_renamed &&
1619                         (added + deleted == 0)) {
1620                        total_files--;
1621                        continue;
1622                }
1623                /*
1624                 * "scale" the filename
1625                 */
1626                len = name_width;
1627                name_len = strlen(name);
1628                if (name_width < name_len) {
1629                        char *slash;
1630                        prefix = "...";
1631                        len -= 3;
1632                        name += name_len - len;
1633                        slash = strchr(name, '/');
1634                        if (slash)
1635                                name = slash;
1636                }
1637
1638                if (data->files[i]->is_binary) {
1639                        fprintf(options->file, "%s", line_prefix);
1640                        show_name(options->file, prefix, name, len);
1641                        fprintf(options->file, " %*s", number_width, "Bin");
1642                        if (!added && !deleted) {
1643                                putc('\n', options->file);
1644                                continue;
1645                        }
1646                        fprintf(options->file, " %s%"PRIuMAX"%s",
1647                                del_c, deleted, reset);
1648                        fprintf(options->file, " -> ");
1649                        fprintf(options->file, "%s%"PRIuMAX"%s",
1650                                add_c, added, reset);
1651                        fprintf(options->file, " bytes");
1652                        fprintf(options->file, "\n");
1653                        continue;
1654                }
1655                else if (data->files[i]->is_unmerged) {
1656                        fprintf(options->file, "%s", line_prefix);
1657                        show_name(options->file, prefix, name, len);
1658                        fprintf(options->file, " Unmerged\n");
1659                        continue;
1660                }
1661
1662                /*
1663                 * scale the add/delete
1664                 */
1665                add = added;
1666                del = deleted;
1667                adds += add;
1668                dels += del;
1669
1670                if (graph_width <= max_change) {
1671                        int total = add + del;
1672
1673                        total = scale_linear(add + del, graph_width, max_change);
1674                        if (total < 2 && add && del)
1675                                /* width >= 2 due to the sanity check */
1676                                total = 2;
1677                        if (add < del) {
1678                                add = scale_linear(add, graph_width, max_change);
1679                                del = total - add;
1680                        } else {
1681                                del = scale_linear(del, graph_width, max_change);
1682                                add = total - del;
1683                        }
1684                }
1685                fprintf(options->file, "%s", line_prefix);
1686                show_name(options->file, prefix, name, len);
1687                fprintf(options->file, " %*"PRIuMAX"%s",
1688                        number_width, added + deleted,
1689                        added + deleted ? " " : "");
1690                show_graph(options->file, '+', add, add_c, reset);
1691                show_graph(options->file, '-', del, del_c, reset);
1692                fprintf(options->file, "\n");
1693        }
1694        for (i = count; i < data->nr; i++) {
1695                uintmax_t added = data->files[i]->added;
1696                uintmax_t deleted = data->files[i]->deleted;
1697                if (!data->files[i]->is_renamed &&
1698                         (added + deleted == 0)) {
1699                        total_files--;
1700                        continue;
1701                }
1702                adds += added;
1703                dels += deleted;
1704                if (!extra_shown)
1705                        fprintf(options->file, "%s ...\n", line_prefix);
1706                extra_shown = 1;
1707        }
1708        fprintf(options->file, "%s", line_prefix);
1709        print_stat_summary(options->file, total_files, adds, dels);
1710}
1711
1712static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1713{
1714        int i, adds = 0, dels = 0, total_files = data->nr;
1715
1716        if (data->nr == 0)
1717                return;
1718
1719        for (i = 0; i < data->nr; i++) {
1720                int added = data->files[i]->added;
1721                int deleted= data->files[i]->deleted;
1722
1723                if (data->files[i]->is_unmerged)
1724                        continue;
1725                if (!data->files[i]->is_renamed && (added + deleted == 0)) {
1726                        total_files--;
1727                } else if (!data->files[i]->is_binary) { /* don't count bytes */
1728                        adds += added;
1729                        dels += deleted;
1730                }
1731        }
1732        if (options->output_prefix) {
1733                struct strbuf *msg = NULL;
1734                msg = options->output_prefix(options,
1735                                options->output_prefix_data);
1736                fprintf(options->file, "%s", msg->buf);
1737        }
1738        print_stat_summary(options->file, total_files, adds, dels);
1739}
1740
1741static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1742{
1743        int i;
1744
1745        if (data->nr == 0)
1746                return;
1747
1748        for (i = 0; i < data->nr; i++) {
1749                struct diffstat_file *file = data->files[i];
1750
1751                if (options->output_prefix) {
1752                        struct strbuf *msg = NULL;
1753                        msg = options->output_prefix(options,
1754                                        options->output_prefix_data);
1755                        fprintf(options->file, "%s", msg->buf);
1756                }
1757
1758                if (file->is_binary)
1759                        fprintf(options->file, "-\t-\t");
1760                else
1761                        fprintf(options->file,
1762                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1763                                file->added, file->deleted);
1764                if (options->line_termination) {
1765                        fill_print_name(file);
1766                        if (!file->is_renamed)
1767                                write_name_quoted(file->name, options->file,
1768                                                  options->line_termination);
1769                        else {
1770                                fputs(file->print_name, options->file);
1771                                putc(options->line_termination, options->file);
1772                        }
1773                } else {
1774                        if (file->is_renamed) {
1775                                putc('\0', options->file);
1776                                write_name_quoted(file->from_name, options->file, '\0');
1777                        }
1778                        write_name_quoted(file->name, options->file, '\0');
1779                }
1780        }
1781}
1782
1783struct dirstat_file {
1784        const char *name;
1785        unsigned long changed;
1786};
1787
1788struct dirstat_dir {
1789        struct dirstat_file *files;
1790        int alloc, nr, permille, cumulative;
1791};
1792
1793static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1794                unsigned long changed, const char *base, int baselen)
1795{
1796        unsigned long this_dir = 0;
1797        unsigned int sources = 0;
1798        const char *line_prefix = "";
1799        struct strbuf *msg = NULL;
1800
1801        if (opt->output_prefix) {
1802                msg = opt->output_prefix(opt, opt->output_prefix_data);
1803                line_prefix = msg->buf;
1804        }
1805
1806        while (dir->nr) {
1807                struct dirstat_file *f = dir->files;
1808                int namelen = strlen(f->name);
1809                unsigned long this;
1810                char *slash;
1811
1812                if (namelen < baselen)
1813                        break;
1814                if (memcmp(f->name, base, baselen))
1815                        break;
1816                slash = strchr(f->name + baselen, '/');
1817                if (slash) {
1818                        int newbaselen = slash + 1 - f->name;
1819                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1820                        sources++;
1821                } else {
1822                        this = f->changed;
1823                        dir->files++;
1824                        dir->nr--;
1825                        sources += 2;
1826                }
1827                this_dir += this;
1828        }
1829
1830        /*
1831         * We don't report dirstat's for
1832         *  - the top level
1833         *  - or cases where everything came from a single directory
1834         *    under this directory (sources == 1).
1835         */
1836        if (baselen && sources != 1) {
1837                if (this_dir) {
1838                        int permille = this_dir * 1000 / changed;
1839                        if (permille >= dir->permille) {
1840                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1841                                        permille / 10, permille % 10, baselen, base);
1842                                if (!dir->cumulative)
1843                                        return 0;
1844                        }
1845                }
1846        }
1847        return this_dir;
1848}
1849
1850static int dirstat_compare(const void *_a, const void *_b)
1851{
1852        const struct dirstat_file *a = _a;
1853        const struct dirstat_file *b = _b;
1854        return strcmp(a->name, b->name);
1855}
1856
1857static void show_dirstat(struct diff_options *options)
1858{
1859        int i;
1860        unsigned long changed;
1861        struct dirstat_dir dir;
1862        struct diff_queue_struct *q = &diff_queued_diff;
1863
1864        dir.files = NULL;
1865        dir.alloc = 0;
1866        dir.nr = 0;
1867        dir.permille = options->dirstat_permille;
1868        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1869
1870        changed = 0;
1871        for (i = 0; i < q->nr; i++) {
1872                struct diff_filepair *p = q->queue[i];
1873                const char *name;
1874                unsigned long copied, added, damage;
1875                int content_changed;
1876
1877                name = p->two->path ? p->two->path : p->one->path;
1878
1879                if (p->one->sha1_valid && p->two->sha1_valid)
1880                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1881                else
1882                        content_changed = 1;
1883
1884                if (!content_changed) {
1885                        /*
1886                         * The SHA1 has not changed, so pre-/post-content is
1887                         * identical. We can therefore skip looking at the
1888                         * file contents altogether.
1889                         */
1890                        damage = 0;
1891                        goto found_damage;
1892                }
1893
1894                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1895                        /*
1896                         * In --dirstat-by-file mode, we don't really need to
1897                         * look at the actual file contents at all.
1898                         * The fact that the SHA1 changed is enough for us to
1899                         * add this file to the list of results
1900                         * (with each file contributing equal damage).
1901                         */
1902                        damage = 1;
1903                        goto found_damage;
1904                }
1905
1906                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1907                        diff_populate_filespec(p->one, 0);
1908                        diff_populate_filespec(p->two, 0);
1909                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1910                                               &copied, &added);
1911                        diff_free_filespec_data(p->one);
1912                        diff_free_filespec_data(p->two);
1913                } else if (DIFF_FILE_VALID(p->one)) {
1914                        diff_populate_filespec(p->one, 1);
1915                        copied = added = 0;
1916                        diff_free_filespec_data(p->one);
1917                } else if (DIFF_FILE_VALID(p->two)) {
1918                        diff_populate_filespec(p->two, 1);
1919                        copied = 0;
1920                        added = p->two->size;
1921                        diff_free_filespec_data(p->two);
1922                } else
1923                        continue;
1924
1925                /*
1926                 * Original minus copied is the removed material,
1927                 * added is the new material.  They are both damages
1928                 * made to the preimage.
1929                 * If the resulting damage is zero, we know that
1930                 * diffcore_count_changes() considers the two entries to
1931                 * be identical, but since content_changed is true, we
1932                 * know that there must have been _some_ kind of change,
1933                 * so we force all entries to have damage > 0.
1934                 */
1935                damage = (p->one->size - copied) + added;
1936                if (!damage)
1937                        damage = 1;
1938
1939found_damage:
1940                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1941                dir.files[dir.nr].name = name;
1942                dir.files[dir.nr].changed = damage;
1943                changed += damage;
1944                dir.nr++;
1945        }
1946
1947        /* This can happen even with many files, if everything was renames */
1948        if (!changed)
1949                return;
1950
1951        /* Show all directories with more than x% of the changes */
1952        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1953        gather_dirstat(options, &dir, changed, "", 0);
1954}
1955
1956static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1957{
1958        int i;
1959        unsigned long changed;
1960        struct dirstat_dir dir;
1961
1962        if (data->nr == 0)
1963                return;
1964
1965        dir.files = NULL;
1966        dir.alloc = 0;
1967        dir.nr = 0;
1968        dir.permille = options->dirstat_permille;
1969        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1970
1971        changed = 0;
1972        for (i = 0; i < data->nr; i++) {
1973                struct diffstat_file *file = data->files[i];
1974                unsigned long damage = file->added + file->deleted;
1975                if (file->is_binary)
1976                        /*
1977                         * binary files counts bytes, not lines. Must find some
1978                         * way to normalize binary bytes vs. textual lines.
1979                         * The following heuristic assumes that there are 64
1980                         * bytes per "line".
1981                         * This is stupid and ugly, but very cheap...
1982                         */
1983                        damage = (damage + 63) / 64;
1984                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1985                dir.files[dir.nr].name = file->name;
1986                dir.files[dir.nr].changed = damage;
1987                changed += damage;
1988                dir.nr++;
1989        }
1990
1991        /* This can happen even with many files, if everything was renames */
1992        if (!changed)
1993                return;
1994
1995        /* Show all directories with more than x% of the changes */
1996        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1997        gather_dirstat(options, &dir, changed, "", 0);
1998}
1999
2000static void free_diffstat_info(struct diffstat_t *diffstat)
2001{
2002        int i;
2003        for (i = 0; i < diffstat->nr; i++) {
2004                struct diffstat_file *f = diffstat->files[i];
2005                if (f->name != f->print_name)
2006                        free(f->print_name);
2007                free(f->name);
2008                free(f->from_name);
2009                free(f);
2010        }
2011        free(diffstat->files);
2012}
2013
2014struct checkdiff_t {
2015        const char *filename;
2016        int lineno;
2017        int conflict_marker_size;
2018        struct diff_options *o;
2019        unsigned ws_rule;
2020        unsigned status;
2021};
2022
2023static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2024{
2025        char firstchar;
2026        int cnt;
2027
2028        if (len < marker_size + 1)
2029                return 0;
2030        firstchar = line[0];
2031        switch (firstchar) {
2032        case '=': case '>': case '<': case '|':
2033                break;
2034        default:
2035                return 0;
2036        }
2037        for (cnt = 1; cnt < marker_size; cnt++)
2038                if (line[cnt] != firstchar)
2039                        return 0;
2040        /* line[1] thru line[marker_size-1] are same as firstchar */
2041        if (len < marker_size + 1 || !isspace(line[marker_size]))
2042                return 0;
2043        return 1;
2044}
2045
2046static void checkdiff_consume(void *priv, char *line, unsigned long len)
2047{
2048        struct checkdiff_t *data = priv;
2049        int marker_size = data->conflict_marker_size;
2050        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2051        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2052        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2053        char *err;
2054        char *line_prefix = "";
2055        struct strbuf *msgbuf;
2056
2057        assert(data->o);
2058        if (data->o->output_prefix) {
2059                msgbuf = data->o->output_prefix(data->o,
2060                        data->o->output_prefix_data);
2061                line_prefix = msgbuf->buf;
2062        }
2063
2064        if (line[0] == '+') {
2065                unsigned bad;
2066                data->lineno++;
2067                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2068                        data->status |= 1;
2069                        fprintf(data->o->file,
2070                                "%s%s:%d: leftover conflict marker\n",
2071                                line_prefix, data->filename, data->lineno);
2072                }
2073                bad = ws_check(line + 1, len - 1, data->ws_rule);
2074                if (!bad)
2075                        return;
2076                data->status |= bad;
2077                err = whitespace_error_string(bad);
2078                fprintf(data->o->file, "%s%s:%d: %s.\n",
2079                        line_prefix, data->filename, data->lineno, err);
2080                free(err);
2081                emit_line(data->o, set, reset, line, 1);
2082                ws_check_emit(line + 1, len - 1, data->ws_rule,
2083                              data->o->file, set, reset, ws);
2084        } else if (line[0] == ' ') {
2085                data->lineno++;
2086        } else if (line[0] == '@') {
2087                char *plus = strchr(line, '+');
2088                if (plus)
2089                        data->lineno = strtol(plus, NULL, 10) - 1;
2090                else
2091                        die("invalid diff");
2092        }
2093}
2094
2095static unsigned char *deflate_it(char *data,
2096                                 unsigned long size,
2097                                 unsigned long *result_size)
2098{
2099        int bound;
2100        unsigned char *deflated;
2101        git_zstream stream;
2102
2103        memset(&stream, 0, sizeof(stream));
2104        git_deflate_init(&stream, zlib_compression_level);
2105        bound = git_deflate_bound(&stream, size);
2106        deflated = xmalloc(bound);
2107        stream.next_out = deflated;
2108        stream.avail_out = bound;
2109
2110        stream.next_in = (unsigned char *)data;
2111        stream.avail_in = size;
2112        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2113                ; /* nothing */
2114        git_deflate_end(&stream);
2115        *result_size = stream.total_out;
2116        return deflated;
2117}
2118
2119static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2120{
2121        void *cp;
2122        void *delta;
2123        void *deflated;
2124        void *data;
2125        unsigned long orig_size;
2126        unsigned long delta_size;
2127        unsigned long deflate_size;
2128        unsigned long data_size;
2129
2130        /* We could do deflated delta, or we could do just deflated two,
2131         * whichever is smaller.
2132         */
2133        delta = NULL;
2134        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2135        if (one->size && two->size) {
2136                delta = diff_delta(one->ptr, one->size,
2137                                   two->ptr, two->size,
2138                                   &delta_size, deflate_size);
2139                if (delta) {
2140                        void *to_free = delta;
2141                        orig_size = delta_size;
2142                        delta = deflate_it(delta, delta_size, &delta_size);
2143                        free(to_free);
2144                }
2145        }
2146
2147        if (delta && delta_size < deflate_size) {
2148                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2149                free(deflated);
2150                data = delta;
2151                data_size = delta_size;
2152        }
2153        else {
2154                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2155                free(delta);
2156                data = deflated;
2157                data_size = deflate_size;
2158        }
2159
2160        /* emit data encoded in base85 */
2161        cp = data;
2162        while (data_size) {
2163                int bytes = (52 < data_size) ? 52 : data_size;
2164                char line[70];
2165                data_size -= bytes;
2166                if (bytes <= 26)
2167                        line[0] = bytes + 'A' - 1;
2168                else
2169                        line[0] = bytes - 26 + 'a' - 1;
2170                encode_85(line + 1, cp, bytes);
2171                cp = (char *) cp + bytes;
2172                fprintf(file, "%s", prefix);
2173                fputs(line, file);
2174                fputc('\n', file);
2175        }
2176        fprintf(file, "%s\n", prefix);
2177        free(data);
2178}
2179
2180static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2181{
2182        fprintf(file, "%sGIT binary patch\n", prefix);
2183        emit_binary_diff_body(file, one, two, prefix);
2184        emit_binary_diff_body(file, two, one, prefix);
2185}
2186
2187int diff_filespec_is_binary(struct diff_filespec *one)
2188{
2189        if (one->is_binary == -1) {
2190                diff_filespec_load_driver(one);
2191                if (one->driver->binary != -1)
2192                        one->is_binary = one->driver->binary;
2193                else {
2194                        if (!one->data && DIFF_FILE_VALID(one))
2195                                diff_populate_filespec(one, 0);
2196                        if (one->data)
2197                                one->is_binary = buffer_is_binary(one->data,
2198                                                one->size);
2199                        if (one->is_binary == -1)
2200                                one->is_binary = 0;
2201                }
2202        }
2203        return one->is_binary;
2204}
2205
2206static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2207{
2208        diff_filespec_load_driver(one);
2209        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2210}
2211
2212void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2213{
2214        if (!options->a_prefix)
2215                options->a_prefix = a;
2216        if (!options->b_prefix)
2217                options->b_prefix = b;
2218}
2219
2220struct userdiff_driver *get_textconv(struct diff_filespec *one)
2221{
2222        if (!DIFF_FILE_VALID(one))
2223                return NULL;
2224
2225        diff_filespec_load_driver(one);
2226        return userdiff_get_textconv(one->driver);
2227}
2228
2229static void builtin_diff(const char *name_a,
2230                         const char *name_b,
2231                         struct diff_filespec *one,
2232                         struct diff_filespec *two,
2233                         const char *xfrm_msg,
2234                         int must_show_header,
2235                         struct diff_options *o,
2236                         int complete_rewrite)
2237{
2238        mmfile_t mf1, mf2;
2239        const char *lbl[2];
2240        char *a_one, *b_two;
2241        const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2242        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2243        const char *a_prefix, *b_prefix;
2244        struct userdiff_driver *textconv_one = NULL;
2245        struct userdiff_driver *textconv_two = NULL;
2246        struct strbuf header = STRBUF_INIT;
2247        struct strbuf *msgbuf;
2248        char *line_prefix = "";
2249
2250        if (o->output_prefix) {
2251                msgbuf = o->output_prefix(o, o->output_prefix_data);
2252                line_prefix = msgbuf->buf;
2253        }
2254
2255        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2256                        (!one->mode || S_ISGITLINK(one->mode)) &&
2257                        (!two->mode || S_ISGITLINK(two->mode))) {
2258                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2259                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2260                show_submodule_summary(o->file, one ? one->path : two->path,
2261                                one->sha1, two->sha1, two->dirty_submodule,
2262                                meta, del, add, reset);
2263                return;
2264        }
2265
2266        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2267                textconv_one = get_textconv(one);
2268                textconv_two = get_textconv(two);
2269        }
2270
2271        diff_set_mnemonic_prefix(o, "a/", "b/");
2272        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2273                a_prefix = o->b_prefix;
2274                b_prefix = o->a_prefix;
2275        } else {
2276                a_prefix = o->a_prefix;
2277                b_prefix = o->b_prefix;
2278        }
2279
2280        /* Never use a non-valid filename anywhere if at all possible */
2281        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2282        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2283
2284        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2285        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2286        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2287        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2288        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2289        if (lbl[0][0] == '/') {
2290                /* /dev/null */
2291                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2292                if (xfrm_msg)
2293                        strbuf_addstr(&header, xfrm_msg);
2294                must_show_header = 1;
2295        }
2296        else if (lbl[1][0] == '/') {
2297                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2298                if (xfrm_msg)
2299                        strbuf_addstr(&header, xfrm_msg);
2300                must_show_header = 1;
2301        }
2302        else {
2303                if (one->mode != two->mode) {
2304                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2305                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2306                        must_show_header = 1;
2307                }
2308                if (xfrm_msg)
2309                        strbuf_addstr(&header, xfrm_msg);
2310
2311                /*
2312                 * we do not run diff between different kind
2313                 * of objects.
2314                 */
2315                if ((one->mode ^ two->mode) & S_IFMT)
2316                        goto free_ab_and_return;
2317                if (complete_rewrite &&
2318                    (textconv_one || !diff_filespec_is_binary(one)) &&
2319                    (textconv_two || !diff_filespec_is_binary(two))) {
2320                        fprintf(o->file, "%s", header.buf);
2321                        strbuf_reset(&header);
2322                        emit_rewrite_diff(name_a, name_b, one, two,
2323                                                textconv_one, textconv_two, o);
2324                        o->found_changes = 1;
2325                        goto free_ab_and_return;
2326                }
2327        }
2328
2329        if (o->irreversible_delete && lbl[1][0] == '/') {
2330                fprintf(o->file, "%s", header.buf);
2331                strbuf_reset(&header);
2332                goto free_ab_and_return;
2333        } else if (!DIFF_OPT_TST(o, TEXT) &&
2334            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2335              (!textconv_two && diff_filespec_is_binary(two)) )) {
2336                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2337                        die("unable to read files to diff");
2338                /* Quite common confusing case */
2339                if (mf1.size == mf2.size &&
2340                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2341                        if (must_show_header)
2342                                fprintf(o->file, "%s", header.buf);
2343                        goto free_ab_and_return;
2344                }
2345                fprintf(o->file, "%s", header.buf);
2346                strbuf_reset(&header);
2347                if (DIFF_OPT_TST(o, BINARY))
2348                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2349                else
2350                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2351                                line_prefix, lbl[0], lbl[1]);
2352                o->found_changes = 1;
2353        } else {
2354                /* Crazy xdl interfaces.. */
2355                const char *diffopts = getenv("GIT_DIFF_OPTS");
2356                xpparam_t xpp;
2357                xdemitconf_t xecfg;
2358                struct emit_callback ecbdata;
2359                const struct userdiff_funcname *pe;
2360
2361                if (must_show_header) {
2362                        fprintf(o->file, "%s", header.buf);
2363                        strbuf_reset(&header);
2364                }
2365
2366                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2367                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2368
2369                pe = diff_funcname_pattern(one);
2370                if (!pe)
2371                        pe = diff_funcname_pattern(two);
2372
2373                memset(&xpp, 0, sizeof(xpp));
2374                memset(&xecfg, 0, sizeof(xecfg));
2375                memset(&ecbdata, 0, sizeof(ecbdata));
2376                ecbdata.label_path = lbl;
2377                ecbdata.color_diff = want_color(o->use_color);
2378                ecbdata.found_changesp = &o->found_changes;
2379                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2380                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2381                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2382                ecbdata.opt = o;
2383                ecbdata.header = header.len ? &header : NULL;
2384                xpp.flags = o->xdl_opts;
2385                xecfg.ctxlen = o->context;
2386                xecfg.interhunkctxlen = o->interhunkcontext;
2387                xecfg.flags = XDL_EMIT_FUNCNAMES;
2388                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2389                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2390                if (pe)
2391                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2392                if (!diffopts)
2393                        ;
2394                else if (!prefixcmp(diffopts, "--unified="))
2395                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2396                else if (!prefixcmp(diffopts, "-u"))
2397                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2398                if (o->word_diff)
2399                        init_diff_words_data(&ecbdata, o, one, two);
2400                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2401                              &xpp, &xecfg);
2402                if (o->word_diff)
2403                        free_diff_words_data(&ecbdata);
2404                if (textconv_one)
2405                        free(mf1.ptr);
2406                if (textconv_two)
2407                        free(mf2.ptr);
2408                xdiff_clear_find_func(&xecfg);
2409        }
2410
2411 free_ab_and_return:
2412        strbuf_release(&header);
2413        diff_free_filespec_data(one);
2414        diff_free_filespec_data(two);
2415        free(a_one);
2416        free(b_two);
2417        return;
2418}
2419
2420static void builtin_diffstat(const char *name_a, const char *name_b,
2421                             struct diff_filespec *one,
2422                             struct diff_filespec *two,
2423                             struct diffstat_t *diffstat,
2424                             struct diff_options *o,
2425                             int complete_rewrite)
2426{
2427        mmfile_t mf1, mf2;
2428        struct diffstat_file *data;
2429        int same_contents;
2430
2431        data = diffstat_add(diffstat, name_a, name_b);
2432
2433        if (!one || !two) {
2434                data->is_unmerged = 1;
2435                return;
2436        }
2437
2438        same_contents = !hashcmp(one->sha1, two->sha1);
2439
2440        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2441                data->is_binary = 1;
2442                if (same_contents) {
2443                        data->added = 0;
2444                        data->deleted = 0;
2445                } else {
2446                        data->added = diff_filespec_size(two);
2447                        data->deleted = diff_filespec_size(one);
2448                }
2449        }
2450
2451        else if (complete_rewrite) {
2452                diff_populate_filespec(one, 0);
2453                diff_populate_filespec(two, 0);
2454                data->deleted = count_lines(one->data, one->size);
2455                data->added = count_lines(two->data, two->size);
2456        }
2457
2458        else if (!same_contents) {
2459                /* Crazy xdl interfaces.. */
2460                xpparam_t xpp;
2461                xdemitconf_t xecfg;
2462
2463                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2464                        die("unable to read files to diff");
2465
2466                memset(&xpp, 0, sizeof(xpp));
2467                memset(&xecfg, 0, sizeof(xecfg));
2468                xpp.flags = o->xdl_opts;
2469                xecfg.ctxlen = o->context;
2470                xecfg.interhunkctxlen = o->interhunkcontext;
2471                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2472                              &xpp, &xecfg);
2473        }
2474
2475        diff_free_filespec_data(one);
2476        diff_free_filespec_data(two);
2477}
2478
2479static void builtin_checkdiff(const char *name_a, const char *name_b,
2480                              const char *attr_path,
2481                              struct diff_filespec *one,
2482                              struct diff_filespec *two,
2483                              struct diff_options *o)
2484{
2485        mmfile_t mf1, mf2;
2486        struct checkdiff_t data;
2487
2488        if (!two)
2489                return;
2490
2491        memset(&data, 0, sizeof(data));
2492        data.filename = name_b ? name_b : name_a;
2493        data.lineno = 0;
2494        data.o = o;
2495        data.ws_rule = whitespace_rule(attr_path);
2496        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2497
2498        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2499                die("unable to read files to diff");
2500
2501        /*
2502         * All the other codepaths check both sides, but not checking
2503         * the "old" side here is deliberate.  We are checking the newly
2504         * introduced changes, and as long as the "new" side is text, we
2505         * can and should check what it introduces.
2506         */
2507        if (diff_filespec_is_binary(two))
2508                goto free_and_return;
2509        else {
2510                /* Crazy xdl interfaces.. */
2511                xpparam_t xpp;
2512                xdemitconf_t xecfg;
2513
2514                memset(&xpp, 0, sizeof(xpp));
2515                memset(&xecfg, 0, sizeof(xecfg));
2516                xecfg.ctxlen = 1; /* at least one context line */
2517                xpp.flags = 0;
2518                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2519                              &xpp, &xecfg);
2520
2521                if (data.ws_rule & WS_BLANK_AT_EOF) {
2522                        struct emit_callback ecbdata;
2523                        int blank_at_eof;
2524
2525                        ecbdata.ws_rule = data.ws_rule;
2526                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2527                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2528
2529                        if (blank_at_eof) {
2530                                static char *err;
2531                                if (!err)
2532                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2533                                fprintf(o->file, "%s:%d: %s.\n",
2534                                        data.filename, blank_at_eof, err);
2535                                data.status = 1; /* report errors */
2536                        }
2537                }
2538        }
2539 free_and_return:
2540        diff_free_filespec_data(one);
2541        diff_free_filespec_data(two);
2542        if (data.status)
2543                DIFF_OPT_SET(o, CHECK_FAILED);
2544}
2545
2546struct diff_filespec *alloc_filespec(const char *path)
2547{
2548        int namelen = strlen(path);
2549        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2550
2551        memset(spec, 0, sizeof(*spec));
2552        spec->path = (char *)(spec + 1);
2553        memcpy(spec->path, path, namelen+1);
2554        spec->count = 1;
2555        spec->is_binary = -1;
2556        return spec;
2557}
2558
2559void free_filespec(struct diff_filespec *spec)
2560{
2561        if (!--spec->count) {
2562                diff_free_filespec_data(spec);
2563                free(spec);
2564        }
2565}
2566
2567void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2568                   int sha1_valid, unsigned short mode)
2569{
2570        if (mode) {
2571                spec->mode = canon_mode(mode);
2572                hashcpy(spec->sha1, sha1);
2573                spec->sha1_valid = sha1_valid;
2574        }
2575}
2576
2577/*
2578 * Given a name and sha1 pair, if the index tells us the file in
2579 * the work tree has that object contents, return true, so that
2580 * prepare_temp_file() does not have to inflate and extract.
2581 */
2582static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2583{
2584        struct cache_entry *ce;
2585        struct stat st;
2586        int pos, len;
2587
2588        /*
2589         * We do not read the cache ourselves here, because the
2590         * benchmark with my previous version that always reads cache
2591         * shows that it makes things worse for diff-tree comparing
2592         * two linux-2.6 kernel trees in an already checked out work
2593         * tree.  This is because most diff-tree comparisons deal with
2594         * only a small number of files, while reading the cache is
2595         * expensive for a large project, and its cost outweighs the
2596         * savings we get by not inflating the object to a temporary
2597         * file.  Practically, this code only helps when we are used
2598         * by diff-cache --cached, which does read the cache before
2599         * calling us.
2600         */
2601        if (!active_cache)
2602                return 0;
2603
2604        /* We want to avoid the working directory if our caller
2605         * doesn't need the data in a normal file, this system
2606         * is rather slow with its stat/open/mmap/close syscalls,
2607         * and the object is contained in a pack file.  The pack
2608         * is probably already open and will be faster to obtain
2609         * the data through than the working directory.  Loose
2610         * objects however would tend to be slower as they need
2611         * to be individually opened and inflated.
2612         */
2613        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2614                return 0;
2615
2616        len = strlen(name);
2617        pos = cache_name_pos(name, len);
2618        if (pos < 0)
2619                return 0;
2620        ce = active_cache[pos];
2621
2622        /*
2623         * This is not the sha1 we are looking for, or
2624         * unreusable because it is not a regular file.
2625         */
2626        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2627                return 0;
2628
2629        /*
2630         * If ce is marked as "assume unchanged", there is no
2631         * guarantee that work tree matches what we are looking for.
2632         */
2633        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2634                return 0;
2635
2636        /*
2637         * If ce matches the file in the work tree, we can reuse it.
2638         */
2639        if (ce_uptodate(ce) ||
2640            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2641                return 1;
2642
2643        return 0;
2644}
2645
2646static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2647{
2648        int len;
2649        char *data = xmalloc(100), *dirty = "";
2650
2651        /* Are we looking at the work tree? */
2652        if (s->dirty_submodule)
2653                dirty = "-dirty";
2654
2655        len = snprintf(data, 100,
2656                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2657        s->data = data;
2658        s->size = len;
2659        s->should_free = 1;
2660        if (size_only) {
2661                s->data = NULL;
2662                free(data);
2663        }
2664        return 0;
2665}
2666
2667/*
2668 * While doing rename detection and pickaxe operation, we may need to
2669 * grab the data for the blob (or file) for our own in-core comparison.
2670 * diff_filespec has data and size fields for this purpose.
2671 */
2672int diff_populate_filespec(struct diff_filespec *s, int size_only)
2673{
2674        int err = 0;
2675        if (!DIFF_FILE_VALID(s))
2676                die("internal error: asking to populate invalid file.");
2677        if (S_ISDIR(s->mode))
2678                return -1;
2679
2680        if (s->data)
2681                return 0;
2682
2683        if (size_only && 0 < s->size)
2684                return 0;
2685
2686        if (S_ISGITLINK(s->mode))
2687                return diff_populate_gitlink(s, size_only);
2688
2689        if (!s->sha1_valid ||
2690            reuse_worktree_file(s->path, s->sha1, 0)) {
2691                struct strbuf buf = STRBUF_INIT;
2692                struct stat st;
2693                int fd;
2694
2695                if (lstat(s->path, &st) < 0) {
2696                        if (errno == ENOENT) {
2697                        err_empty:
2698                                err = -1;
2699                        empty:
2700                                s->data = (char *)"";
2701                                s->size = 0;
2702                                return err;
2703                        }
2704                }
2705                s->size = xsize_t(st.st_size);
2706                if (!s->size)
2707                        goto empty;
2708                if (S_ISLNK(st.st_mode)) {
2709                        struct strbuf sb = STRBUF_INIT;
2710
2711                        if (strbuf_readlink(&sb, s->path, s->size))
2712                                goto err_empty;
2713                        s->size = sb.len;
2714                        s->data = strbuf_detach(&sb, NULL);
2715                        s->should_free = 1;
2716                        return 0;
2717                }
2718                if (size_only)
2719                        return 0;
2720                fd = open(s->path, O_RDONLY);
2721                if (fd < 0)
2722                        goto err_empty;
2723                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2724                close(fd);
2725                s->should_munmap = 1;
2726
2727                /*
2728                 * Convert from working tree format to canonical git format
2729                 */
2730                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2731                        size_t size = 0;
2732                        munmap(s->data, s->size);
2733                        s->should_munmap = 0;
2734                        s->data = strbuf_detach(&buf, &size);
2735                        s->size = size;
2736                        s->should_free = 1;
2737                }
2738        }
2739        else {
2740                enum object_type type;
2741                if (size_only) {
2742                        type = sha1_object_info(s->sha1, &s->size);
2743                        if (type < 0)
2744                                die("unable to read %s", sha1_to_hex(s->sha1));
2745                } else {
2746                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2747                        if (!s->data)
2748                                die("unable to read %s", sha1_to_hex(s->sha1));
2749                        s->should_free = 1;
2750                }
2751        }
2752        return 0;
2753}
2754
2755void diff_free_filespec_blob(struct diff_filespec *s)
2756{
2757        if (s->should_free)
2758                free(s->data);
2759        else if (s->should_munmap)
2760                munmap(s->data, s->size);
2761
2762        if (s->should_free || s->should_munmap) {
2763                s->should_free = s->should_munmap = 0;
2764                s->data = NULL;
2765        }
2766}
2767
2768void diff_free_filespec_data(struct diff_filespec *s)
2769{
2770        diff_free_filespec_blob(s);
2771        free(s->cnt_data);
2772        s->cnt_data = NULL;
2773}
2774
2775static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2776                           void *blob,
2777                           unsigned long size,
2778                           const unsigned char *sha1,
2779                           int mode)
2780{
2781        int fd;
2782        struct strbuf buf = STRBUF_INIT;
2783        struct strbuf template = STRBUF_INIT;
2784        char *path_dup = xstrdup(path);
2785        const char *base = basename(path_dup);
2786
2787        /* Generate "XXXXXX_basename.ext" */
2788        strbuf_addstr(&template, "XXXXXX_");
2789        strbuf_addstr(&template, base);
2790
2791        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2792                        strlen(base) + 1);
2793        if (fd < 0)
2794                die_errno("unable to create temp-file");
2795        if (convert_to_working_tree(path,
2796                        (const char *)blob, (size_t)size, &buf)) {
2797                blob = buf.buf;
2798                size = buf.len;
2799        }
2800        if (write_in_full(fd, blob, size) != size)
2801                die_errno("unable to write temp-file");
2802        close(fd);
2803        temp->name = temp->tmp_path;
2804        strcpy(temp->hex, sha1_to_hex(sha1));
2805        temp->hex[40] = 0;
2806        sprintf(temp->mode, "%06o", mode);
2807        strbuf_release(&buf);
2808        strbuf_release(&template);
2809        free(path_dup);
2810}
2811
2812static struct diff_tempfile *prepare_temp_file(const char *name,
2813                struct diff_filespec *one)
2814{
2815        struct diff_tempfile *temp = claim_diff_tempfile();
2816
2817        if (!DIFF_FILE_VALID(one)) {
2818        not_a_valid_file:
2819                /* A '-' entry produces this for file-2, and
2820                 * a '+' entry produces this for file-1.
2821                 */
2822                temp->name = "/dev/null";
2823                strcpy(temp->hex, ".");
2824                strcpy(temp->mode, ".");
2825                return temp;
2826        }
2827
2828        if (!remove_tempfile_installed) {
2829                atexit(remove_tempfile);
2830                sigchain_push_common(remove_tempfile_on_signal);
2831                remove_tempfile_installed = 1;
2832        }
2833
2834        if (!one->sha1_valid ||
2835            reuse_worktree_file(name, one->sha1, 1)) {
2836                struct stat st;
2837                if (lstat(name, &st) < 0) {
2838                        if (errno == ENOENT)
2839                                goto not_a_valid_file;
2840                        die_errno("stat(%s)", name);
2841                }
2842                if (S_ISLNK(st.st_mode)) {
2843                        struct strbuf sb = STRBUF_INIT;
2844                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2845                                die_errno("readlink(%s)", name);
2846                        prep_temp_blob(name, temp, sb.buf, sb.len,
2847                                       (one->sha1_valid ?
2848                                        one->sha1 : null_sha1),
2849                                       (one->sha1_valid ?
2850                                        one->mode : S_IFLNK));
2851                        strbuf_release(&sb);
2852                }
2853                else {
2854                        /* we can borrow from the file in the work tree */
2855                        temp->name = name;
2856                        if (!one->sha1_valid)
2857                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2858                        else
2859                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2860                        /* Even though we may sometimes borrow the
2861                         * contents from the work tree, we always want
2862                         * one->mode.  mode is trustworthy even when
2863                         * !(one->sha1_valid), as long as
2864                         * DIFF_FILE_VALID(one).
2865                         */
2866                        sprintf(temp->mode, "%06o", one->mode);
2867                }
2868                return temp;
2869        }
2870        else {
2871                if (diff_populate_filespec(one, 0))
2872                        die("cannot read data blob for %s", one->path);
2873                prep_temp_blob(name, temp, one->data, one->size,
2874                               one->sha1, one->mode);
2875        }
2876        return temp;
2877}
2878
2879/* An external diff command takes:
2880 *
2881 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2882 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2883 *
2884 */
2885static void run_external_diff(const char *pgm,
2886                              const char *name,
2887                              const char *other,
2888                              struct diff_filespec *one,
2889                              struct diff_filespec *two,
2890                              const char *xfrm_msg,
2891                              int complete_rewrite)
2892{
2893        const char *spawn_arg[10];
2894        int retval;
2895        const char **arg = &spawn_arg[0];
2896
2897        if (one && two) {
2898                struct diff_tempfile *temp_one, *temp_two;
2899                const char *othername = (other ? other : name);
2900                temp_one = prepare_temp_file(name, one);
2901                temp_two = prepare_temp_file(othername, two);
2902                *arg++ = pgm;
2903                *arg++ = name;
2904                *arg++ = temp_one->name;
2905                *arg++ = temp_one->hex;
2906                *arg++ = temp_one->mode;
2907                *arg++ = temp_two->name;
2908                *arg++ = temp_two->hex;
2909                *arg++ = temp_two->mode;
2910                if (other) {
2911                        *arg++ = other;
2912                        *arg++ = xfrm_msg;
2913                }
2914        } else {
2915                *arg++ = pgm;
2916                *arg++ = name;
2917        }
2918        *arg = NULL;
2919        fflush(NULL);
2920        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2921        remove_tempfile();
2922        if (retval) {
2923                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2924                exit(1);
2925        }
2926}
2927
2928static int similarity_index(struct diff_filepair *p)
2929{
2930        return p->score * 100 / MAX_SCORE;
2931}
2932
2933static void fill_metainfo(struct strbuf *msg,
2934                          const char *name,
2935                          const char *other,
2936                          struct diff_filespec *one,
2937                          struct diff_filespec *two,
2938                          struct diff_options *o,
2939                          struct diff_filepair *p,
2940                          int *must_show_header,
2941                          int use_color)
2942{
2943        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2944        const char *reset = diff_get_color(use_color, DIFF_RESET);
2945        struct strbuf *msgbuf;
2946        char *line_prefix = "";
2947
2948        *must_show_header = 1;
2949        if (o->output_prefix) {
2950                msgbuf = o->output_prefix(o, o->output_prefix_data);
2951                line_prefix = msgbuf->buf;
2952        }
2953        strbuf_init(msg, PATH_MAX * 2 + 300);
2954        switch (p->status) {
2955        case DIFF_STATUS_COPIED:
2956                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2957                            line_prefix, set, similarity_index(p));
2958                strbuf_addf(msg, "%s\n%s%scopy from ",
2959                            reset,  line_prefix, set);
2960                quote_c_style(name, msg, NULL, 0);
2961                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2962                quote_c_style(other, msg, NULL, 0);
2963                strbuf_addf(msg, "%s\n", reset);
2964                break;
2965        case DIFF_STATUS_RENAMED:
2966                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2967                            line_prefix, set, similarity_index(p));
2968                strbuf_addf(msg, "%s\n%s%srename from ",
2969                            reset, line_prefix, set);
2970                quote_c_style(name, msg, NULL, 0);
2971                strbuf_addf(msg, "%s\n%s%srename to ",
2972                            reset, line_prefix, set);
2973                quote_c_style(other, msg, NULL, 0);
2974                strbuf_addf(msg, "%s\n", reset);
2975                break;
2976        case DIFF_STATUS_MODIFIED:
2977                if (p->score) {
2978                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2979                                    line_prefix,
2980                                    set, similarity_index(p), reset);
2981                        break;
2982                }
2983                /* fallthru */
2984        default:
2985                *must_show_header = 0;
2986        }
2987        if (one && two && hashcmp(one->sha1, two->sha1)) {
2988                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2989
2990                if (DIFF_OPT_TST(o, BINARY)) {
2991                        mmfile_t mf;
2992                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2993                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2994                                abbrev = 40;
2995                }
2996                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2997                            find_unique_abbrev(one->sha1, abbrev));
2998                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2999                if (one->mode == two->mode)
3000                        strbuf_addf(msg, " %06o", one->mode);
3001                strbuf_addf(msg, "%s\n", reset);
3002        }
3003}
3004
3005static void run_diff_cmd(const char *pgm,
3006                         const char *name,
3007                         const char *other,
3008                         const char *attr_path,
3009                         struct diff_filespec *one,
3010                         struct diff_filespec *two,
3011                         struct strbuf *msg,
3012                         struct diff_options *o,
3013                         struct diff_filepair *p)
3014{
3015        const char *xfrm_msg = NULL;
3016        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3017        int must_show_header = 0;
3018
3019
3020        if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3021                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3022                if (drv && drv->external)
3023                        pgm = drv->external;
3024        }
3025
3026        if (msg) {
3027                /*
3028                 * don't use colors when the header is intended for an
3029                 * external diff driver
3030                 */
3031                fill_metainfo(msg, name, other, one, two, o, p,
3032                              &must_show_header,
3033                              want_color(o->use_color) && !pgm);
3034                xfrm_msg = msg->len ? msg->buf : NULL;
3035        }
3036
3037        if (pgm) {
3038                run_external_diff(pgm, name, other, one, two, xfrm_msg,
3039                                  complete_rewrite);
3040                return;
3041        }
3042        if (one && two)
3043                builtin_diff(name, other ? other : name,
3044                             one, two, xfrm_msg, must_show_header,
3045                             o, complete_rewrite);
3046        else
3047                fprintf(o->file, "* Unmerged path %s\n", name);
3048}
3049
3050static void diff_fill_sha1_info(struct diff_filespec *one)
3051{
3052        if (DIFF_FILE_VALID(one)) {
3053                if (!one->sha1_valid) {
3054                        struct stat st;
3055                        if (one->is_stdin) {
3056                                hashcpy(one->sha1, null_sha1);
3057                                return;
3058                        }
3059                        if (lstat(one->path, &st) < 0)
3060                                die_errno("stat '%s'", one->path);
3061                        if (index_path(one->sha1, one->path, &st, 0))
3062                                die("cannot hash %s", one->path);
3063                }
3064        }
3065        else
3066                hashclr(one->sha1);
3067}
3068
3069static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3070{
3071        /* Strip the prefix but do not molest /dev/null and absolute paths */
3072        if (*namep && **namep != '/') {
3073                *namep += prefix_length;
3074                if (**namep == '/')
3075                        ++*namep;
3076        }
3077        if (*otherp && **otherp != '/') {
3078                *otherp += prefix_length;
3079                if (**otherp == '/')
3080                        ++*otherp;
3081        }
3082}
3083
3084static void run_diff(struct diff_filepair *p, struct diff_options *o)
3085{
3086        const char *pgm = external_diff();
3087        struct strbuf msg;
3088        struct diff_filespec *one = p->one;
3089        struct diff_filespec *two = p->two;
3090        const char *name;
3091        const char *other;
3092        const char *attr_path;
3093
3094        name  = p->one->path;
3095        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3096        attr_path = name;
3097        if (o->prefix_length)
3098                strip_prefix(o->prefix_length, &name, &other);
3099
3100        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3101                pgm = NULL;
3102
3103        if (DIFF_PAIR_UNMERGED(p)) {
3104                run_diff_cmd(pgm, name, NULL, attr_path,
3105                             NULL, NULL, NULL, o, p);
3106                return;
3107        }
3108
3109        diff_fill_sha1_info(one);
3110        diff_fill_sha1_info(two);
3111
3112        if (!pgm &&
3113            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3114            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3115                /*
3116                 * a filepair that changes between file and symlink
3117                 * needs to be split into deletion and creation.
3118                 */
3119                struct diff_filespec *null = alloc_filespec(two->path);
3120                run_diff_cmd(NULL, name, other, attr_path,
3121                             one, null, &msg, o, p);
3122                free(null);
3123                strbuf_release(&msg);
3124
3125                null = alloc_filespec(one->path);
3126                run_diff_cmd(NULL, name, other, attr_path,
3127                             null, two, &msg, o, p);
3128                free(null);
3129        }
3130        else
3131                run_diff_cmd(pgm, name, other, attr_path,
3132                             one, two, &msg, o, p);
3133
3134        strbuf_release(&msg);
3135}
3136
3137static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3138                         struct diffstat_t *diffstat)
3139{
3140        const char *name;
3141        const char *other;
3142        int complete_rewrite = 0;
3143
3144        if (DIFF_PAIR_UNMERGED(p)) {
3145                /* unmerged */
3146                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3147                return;
3148        }
3149
3150        name = p->one->path;
3151        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3152
3153        if (o->prefix_length)
3154                strip_prefix(o->prefix_length, &name, &other);
3155
3156        diff_fill_sha1_info(p->one);
3157        diff_fill_sha1_info(p->two);
3158
3159        if (p->status == DIFF_STATUS_MODIFIED && p->score)
3160                complete_rewrite = 1;
3161        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3162}
3163
3164static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3165{
3166        const char *name;
3167        const char *other;
3168        const char *attr_path;
3169
3170        if (DIFF_PAIR_UNMERGED(p)) {
3171                /* unmerged */
3172                return;
3173        }
3174
3175        name = p->one->path;
3176        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3177        attr_path = other ? other : name;
3178
3179        if (o->prefix_length)
3180                strip_prefix(o->prefix_length, &name, &other);
3181
3182        diff_fill_sha1_info(p->one);
3183        diff_fill_sha1_info(p->two);
3184
3185        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3186}
3187
3188void diff_setup(struct diff_options *options)
3189{
3190        memcpy(options, &default_diff_options, sizeof(*options));
3191
3192        options->file = stdout;
3193
3194        options->line_termination = '\n';
3195        options->break_opt = -1;
3196        options->rename_limit = -1;
3197        options->dirstat_permille = diff_dirstat_permille_default;
3198        options->context = diff_context_default;
3199        DIFF_OPT_SET(options, RENAME_EMPTY);
3200
3201        options->change = diff_change;
3202        options->add_remove = diff_addremove;
3203        options->use_color = diff_use_color_default;
3204        options->detect_rename = diff_detect_rename_default;
3205
3206        if (diff_no_prefix) {
3207                options->a_prefix = options->b_prefix = "";
3208        } else if (!diff_mnemonic_prefix) {
3209                options->a_prefix = "a/";
3210                options->b_prefix = "b/";
3211        }
3212}
3213
3214void diff_setup_done(struct diff_options *options)
3215{
3216        int count = 0;
3217
3218        if (options->output_format & DIFF_FORMAT_NAME)
3219                count++;
3220        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3221                count++;
3222        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3223                count++;
3224        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3225                count++;
3226        if (count > 1)
3227                die("--name-only, --name-status, --check and -s are mutually exclusive");
3228
3229        /*
3230         * Most of the time we can say "there are changes"
3231         * only by checking if there are changed paths, but
3232         * --ignore-whitespace* options force us to look
3233         * inside contents.
3234         */
3235
3236        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3237            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3238            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3239                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3240        else
3241                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3242
3243        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3244                options->detect_rename = DIFF_DETECT_COPY;
3245
3246        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3247                options->prefix = NULL;
3248        if (options->prefix)
3249                options->prefix_length = strlen(options->prefix);
3250        else
3251                options->prefix_length = 0;
3252
3253        if (options->output_format & (DIFF_FORMAT_NAME |
3254                                      DIFF_FORMAT_NAME_STATUS |
3255                                      DIFF_FORMAT_CHECKDIFF |
3256                                      DIFF_FORMAT_NO_OUTPUT))
3257                options->output_format &= ~(DIFF_FORMAT_RAW |
3258                                            DIFF_FORMAT_NUMSTAT |
3259                                            DIFF_FORMAT_DIFFSTAT |
3260                                            DIFF_FORMAT_SHORTSTAT |
3261                                            DIFF_FORMAT_DIRSTAT |
3262                                            DIFF_FORMAT_SUMMARY |
3263                                            DIFF_FORMAT_PATCH);
3264
3265        /*
3266         * These cases always need recursive; we do not drop caller-supplied
3267         * recursive bits for other formats here.
3268         */
3269        if (options->output_format & (DIFF_FORMAT_PATCH |
3270                                      DIFF_FORMAT_NUMSTAT |
3271                                      DIFF_FORMAT_DIFFSTAT |
3272                                      DIFF_FORMAT_SHORTSTAT |
3273                                      DIFF_FORMAT_DIRSTAT |
3274                                      DIFF_FORMAT_SUMMARY |
3275                                      DIFF_FORMAT_CHECKDIFF))
3276                DIFF_OPT_SET(options, RECURSIVE);
3277        /*
3278         * Also pickaxe would not work very well if you do not say recursive
3279         */
3280        if (options->pickaxe)
3281                DIFF_OPT_SET(options, RECURSIVE);
3282        /*
3283         * When patches are generated, submodules diffed against the work tree
3284         * must be checked for dirtiness too so it can be shown in the output
3285         */
3286        if (options->output_format & DIFF_FORMAT_PATCH)
3287                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3288
3289        if (options->detect_rename && options->rename_limit < 0)
3290                options->rename_limit = diff_rename_limit_default;
3291        if (options->setup & DIFF_SETUP_USE_CACHE) {
3292                if (!active_cache)
3293                        /* read-cache does not die even when it fails
3294                         * so it is safe for us to do this here.  Also
3295                         * it does not smudge active_cache or active_nr
3296                         * when it fails, so we do not have to worry about
3297                         * cleaning it up ourselves either.
3298                         */
3299                        read_cache();
3300        }
3301        if (options->abbrev <= 0 || 40 < options->abbrev)
3302                options->abbrev = 40; /* full */
3303
3304        /*
3305         * It does not make sense to show the first hit we happened
3306         * to have found.  It does not make sense not to return with
3307         * exit code in such a case either.
3308         */
3309        if (DIFF_OPT_TST(options, QUICK)) {
3310                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3311                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3312        }
3313}
3314
3315static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3316{
3317        char c, *eq;
3318        int len;
3319
3320        if (*arg != '-')
3321                return 0;
3322        c = *++arg;
3323        if (!c)
3324                return 0;
3325        if (c == arg_short) {
3326                c = *++arg;
3327                if (!c)
3328                        return 1;
3329                if (val && isdigit(c)) {
3330                        char *end;
3331                        int n = strtoul(arg, &end, 10);
3332                        if (*end)
3333                                return 0;
3334                        *val = n;
3335                        return 1;
3336                }
3337                return 0;
3338        }
3339        if (c != '-')
3340                return 0;
3341        arg++;
3342        eq = strchr(arg, '=');
3343        if (eq)
3344                len = eq - arg;
3345        else
3346                len = strlen(arg);
3347        if (!len || strncmp(arg, arg_long, len))
3348                return 0;
3349        if (eq) {
3350                int n;
3351                char *end;
3352                if (!isdigit(*++eq))
3353                        return 0;
3354                n = strtoul(eq, &end, 10);
3355                if (*end)
3356                        return 0;
3357                *val = n;
3358        }
3359        return 1;
3360}
3361
3362static int diff_scoreopt_parse(const char *opt);
3363
3364static inline int short_opt(char opt, const char **argv,
3365                            const char **optarg)
3366{
3367        const char *arg = argv[0];
3368        if (arg[0] != '-' || arg[1] != opt)
3369                return 0;
3370        if (arg[2] != '\0') {
3371                *optarg = arg + 2;
3372                return 1;
3373        }
3374        if (!argv[1])
3375                die("Option '%c' requires a value", opt);
3376        *optarg = argv[1];
3377        return 2;
3378}
3379
3380int parse_long_opt(const char *opt, const char **argv,
3381                   const char **optarg)
3382{
3383        const char *arg = argv[0];
3384        if (arg[0] != '-' || arg[1] != '-')
3385                return 0;
3386        arg += strlen("--");
3387        if (prefixcmp(arg, opt))
3388                return 0;
3389        arg += strlen(opt);
3390        if (*arg == '=') { /* sticked form: --option=value */
3391                *optarg = arg + 1;
3392                return 1;
3393        }
3394        if (*arg != '\0')
3395                return 0;
3396        /* separate form: --option value */
3397        if (!argv[1])
3398                die("Option '--%s' requires a value", opt);
3399        *optarg = argv[1];
3400        return 2;
3401}
3402
3403static int stat_opt(struct diff_options *options, const char **av)
3404{
3405        const char *arg = av[0];
3406        char *end;
3407        int width = options->stat_width;
3408        int name_width = options->stat_name_width;
3409        int graph_width = options->stat_graph_width;
3410        int count = options->stat_count;
3411        int argcount = 1;
3412
3413        arg += strlen("--stat");
3414        end = (char *)arg;
3415
3416        switch (*arg) {
3417        case '-':
3418                if (!prefixcmp(arg, "-width")) {
3419                        arg += strlen("-width");
3420                        if (*arg == '=')
3421                                width = strtoul(arg + 1, &end, 10);
3422                        else if (!*arg && !av[1])
3423                                die("Option '--stat-width' requires a value");
3424                        else if (!*arg) {
3425                                width = strtoul(av[1], &end, 10);
3426                                argcount = 2;
3427                        }
3428                } else if (!prefixcmp(arg, "-name-width")) {
3429                        arg += strlen("-name-width");
3430                        if (*arg == '=')
3431                                name_width = strtoul(arg + 1, &end, 10);
3432                        else if (!*arg && !av[1])
3433                                die("Option '--stat-name-width' requires a value");
3434                        else if (!*arg) {
3435                                name_width = strtoul(av[1], &end, 10);
3436                                argcount = 2;
3437                        }
3438                } else if (!prefixcmp(arg, "-graph-width")) {
3439                        arg += strlen("-graph-width");
3440                        if (*arg == '=')
3441                                graph_width = strtoul(arg + 1, &end, 10);
3442                        else if (!*arg && !av[1])
3443                                die("Option '--stat-graph-width' requires a value");
3444                        else if (!*arg) {
3445                                graph_width = strtoul(av[1], &end, 10);
3446                                argcount = 2;
3447                        }
3448                } else if (!prefixcmp(arg, "-count")) {
3449                        arg += strlen("-count");
3450                        if (*arg == '=')
3451                                count = strtoul(arg + 1, &end, 10);
3452                        else if (!*arg && !av[1])
3453                                die("Option '--stat-count' requires a value");
3454                        else if (!*arg) {
3455                                count = strtoul(av[1], &end, 10);
3456                                argcount = 2;
3457                        }
3458                }
3459                break;
3460        case '=':
3461                width = strtoul(arg+1, &end, 10);
3462                if (*end == ',')
3463                        name_width = strtoul(end+1, &end, 10);
3464                if (*end == ',')
3465                        count = strtoul(end+1, &end, 10);
3466        }
3467
3468        /* Important! This checks all the error cases! */
3469        if (*end)
3470                return 0;
3471        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3472        options->stat_name_width = name_width;
3473        options->stat_graph_width = graph_width;
3474        options->stat_width = width;
3475        options->stat_count = count;
3476        return argcount;
3477}
3478
3479static int parse_dirstat_opt(struct diff_options *options, const char *params)
3480{
3481        struct strbuf errmsg = STRBUF_INIT;
3482        if (parse_dirstat_params(options, params, &errmsg))
3483                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3484                    errmsg.buf);
3485        strbuf_release(&errmsg);
3486        /*
3487         * The caller knows a dirstat-related option is given from the command
3488         * line; allow it to say "return this_function();"
3489         */
3490        options->output_format |= DIFF_FORMAT_DIRSTAT;
3491        return 1;
3492}
3493
3494static int parse_submodule_opt(struct diff_options *options, const char *value)
3495{
3496        if (parse_submodule_params(options, value))
3497                die(_("Failed to parse --submodule option parameter: '%s'"),
3498                        value);
3499        return 1;
3500}
3501
3502int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3503{
3504        const char *arg = av[0];
3505        const char *optarg;
3506        int argcount;
3507
3508        /* Output format options */
3509        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3510                options->output_format |= DIFF_FORMAT_PATCH;
3511        else if (opt_arg(arg, 'U', "unified", &options->context))
3512                options->output_format |= DIFF_FORMAT_PATCH;
3513        else if (!strcmp(arg, "--raw"))
3514                options->output_format |= DIFF_FORMAT_RAW;
3515        else if (!strcmp(arg, "--patch-with-raw"))
3516                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3517        else if (!strcmp(arg, "--numstat"))
3518                options->output_format |= DIFF_FORMAT_NUMSTAT;
3519        else if (!strcmp(arg, "--shortstat"))
3520                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3521        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3522                return parse_dirstat_opt(options, "");
3523        else if (!prefixcmp(arg, "-X"))
3524                return parse_dirstat_opt(options, arg + 2);
3525        else if (!prefixcmp(arg, "--dirstat="))
3526                return parse_dirstat_opt(options, arg + 10);
3527        else if (!strcmp(arg, "--cumulative"))
3528                return parse_dirstat_opt(options, "cumulative");
3529        else if (!strcmp(arg, "--dirstat-by-file"))
3530                return parse_dirstat_opt(options, "files");
3531        else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3532                parse_dirstat_opt(options, "files");
3533                return parse_dirstat_opt(options, arg + 18);
3534        }
3535        else if (!strcmp(arg, "--check"))
3536                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3537        else if (!strcmp(arg, "--summary"))
3538                options->output_format |= DIFF_FORMAT_SUMMARY;
3539        else if (!strcmp(arg, "--patch-with-stat"))
3540                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3541        else if (!strcmp(arg, "--name-only"))
3542                options->output_format |= DIFF_FORMAT_NAME;
3543        else if (!strcmp(arg, "--name-status"))
3544                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3545        else if (!strcmp(arg, "-s"))
3546                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3547        else if (!prefixcmp(arg, "--stat"))
3548                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3549                return stat_opt(options, av);
3550
3551        /* renames options */
3552        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3553                 !strcmp(arg, "--break-rewrites")) {
3554                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3555                        return error("invalid argument to -B: %s", arg+2);
3556        }
3557        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3558                 !strcmp(arg, "--find-renames")) {
3559                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3560                        return error("invalid argument to -M: %s", arg+2);
3561                options->detect_rename = DIFF_DETECT_RENAME;
3562        }
3563        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3564                options->irreversible_delete = 1;
3565        }
3566        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3567                 !strcmp(arg, "--find-copies")) {
3568                if (options->detect_rename == DIFF_DETECT_COPY)
3569                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3570                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3571                        return error("invalid argument to -C: %s", arg+2);
3572                options->detect_rename = DIFF_DETECT_COPY;
3573        }
3574        else if (!strcmp(arg, "--no-renames"))
3575                options->detect_rename = 0;
3576        else if (!strcmp(arg, "--rename-empty"))
3577                DIFF_OPT_SET(options, RENAME_EMPTY);
3578        else if (!strcmp(arg, "--no-rename-empty"))
3579                DIFF_OPT_CLR(options, RENAME_EMPTY);
3580        else if (!strcmp(arg, "--relative"))
3581                DIFF_OPT_SET(options, RELATIVE_NAME);
3582        else if (!prefixcmp(arg, "--relative=")) {
3583                DIFF_OPT_SET(options, RELATIVE_NAME);
3584                options->prefix = arg + 11;
3585        }
3586
3587        /* xdiff options */
3588        else if (!strcmp(arg, "--minimal"))
3589                DIFF_XDL_SET(options, NEED_MINIMAL);
3590        else if (!strcmp(arg, "--no-minimal"))
3591                DIFF_XDL_CLR(options, NEED_MINIMAL);
3592        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3593                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3594        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3595                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3596        else if (!strcmp(arg, "--ignore-space-at-eol"))
3597                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3598        else if (!strcmp(arg, "--patience"))
3599                options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3600        else if (!strcmp(arg, "--histogram"))
3601                options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3602
3603        /* flags options */
3604        else if (!strcmp(arg, "--binary")) {
3605                options->output_format |= DIFF_FORMAT_PATCH;
3606                DIFF_OPT_SET(options, BINARY);
3607        }
3608        else if (!strcmp(arg, "--full-index"))
3609                DIFF_OPT_SET(options, FULL_INDEX);
3610        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3611                DIFF_OPT_SET(options, TEXT);
3612        else if (!strcmp(arg, "-R"))
3613                DIFF_OPT_SET(options, REVERSE_DIFF);
3614        else if (!strcmp(arg, "--find-copies-harder"))
3615                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3616        else if (!strcmp(arg, "--follow"))
3617                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3618        else if (!strcmp(arg, "--color"))
3619                options->use_color = 1;
3620        else if (!prefixcmp(arg, "--color=")) {
3621                int value = git_config_colorbool(NULL, arg+8);
3622                if (value < 0)
3623                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3624                options->use_color = value;
3625        }
3626        else if (!strcmp(arg, "--no-color"))
3627                options->use_color = 0;
3628        else if (!strcmp(arg, "--color-words")) {
3629                options->use_color = 1;
3630                options->word_diff = DIFF_WORDS_COLOR;
3631        }
3632        else if (!prefixcmp(arg, "--color-words=")) {
3633                options->use_color = 1;
3634                options->word_diff = DIFF_WORDS_COLOR;
3635                options->word_regex = arg + 14;
3636        }
3637        else if (!strcmp(arg, "--word-diff")) {
3638                if (options->word_diff == DIFF_WORDS_NONE)
3639                        options->word_diff = DIFF_WORDS_PLAIN;
3640        }
3641        else if (!prefixcmp(arg, "--word-diff=")) {
3642                const char *type = arg + 12;
3643                if (!strcmp(type, "plain"))
3644                        options->word_diff = DIFF_WORDS_PLAIN;
3645                else if (!strcmp(type, "color")) {
3646                        options->use_color = 1;
3647                        options->word_diff = DIFF_WORDS_COLOR;
3648                }
3649                else if (!strcmp(type, "porcelain"))
3650                        options->word_diff = DIFF_WORDS_PORCELAIN;
3651                else if (!strcmp(type, "none"))
3652                        options->word_diff = DIFF_WORDS_NONE;
3653                else
3654                        die("bad --word-diff argument: %s", type);
3655        }
3656        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3657                if (options->word_diff == DIFF_WORDS_NONE)
3658                        options->word_diff = DIFF_WORDS_PLAIN;
3659                options->word_regex = optarg;
3660                return argcount;
3661        }
3662        else if (!strcmp(arg, "--exit-code"))
3663                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3664        else if (!strcmp(arg, "--quiet"))
3665                DIFF_OPT_SET(options, QUICK);
3666        else if (!strcmp(arg, "--ext-diff"))
3667                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3668        else if (!strcmp(arg, "--no-ext-diff"))
3669                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3670        else if (!strcmp(arg, "--textconv"))
3671                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3672        else if (!strcmp(arg, "--no-textconv"))
3673                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3674        else if (!strcmp(arg, "--ignore-submodules")) {
3675                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3676                handle_ignore_submodules_arg(options, "all");
3677        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3678                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3679                handle_ignore_submodules_arg(options, arg + 20);
3680        } else if (!strcmp(arg, "--submodule"))
3681                DIFF_OPT_SET(options, SUBMODULE_LOG);
3682        else if (!prefixcmp(arg, "--submodule="))
3683                return parse_submodule_opt(options, arg + 12);
3684
3685        /* misc options */
3686        else if (!strcmp(arg, "-z"))
3687                options->line_termination = 0;
3688        else if ((argcount = short_opt('l', av, &optarg))) {
3689                options->rename_limit = strtoul(optarg, NULL, 10);
3690                return argcount;
3691        }
3692        else if ((argcount = short_opt('S', av, &optarg))) {
3693                options->pickaxe = optarg;
3694                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3695                return argcount;
3696        } else if ((argcount = short_opt('G', av, &optarg))) {
3697                options->pickaxe = optarg;
3698                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3699                return argcount;
3700        }
3701        else if (!strcmp(arg, "--pickaxe-all"))
3702                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3703        else if (!strcmp(arg, "--pickaxe-regex"))
3704                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3705        else if ((argcount = short_opt('O', av, &optarg))) {
3706                options->orderfile = optarg;
3707                return argcount;
3708        }
3709        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3710                options->filter = optarg;
3711                return argcount;
3712        }
3713        else if (!strcmp(arg, "--abbrev"))
3714                options->abbrev = DEFAULT_ABBREV;
3715        else if (!prefixcmp(arg, "--abbrev=")) {
3716                options->abbrev = strtoul(arg + 9, NULL, 10);
3717                if (options->abbrev < MINIMUM_ABBREV)
3718                        options->abbrev = MINIMUM_ABBREV;
3719                else if (40 < options->abbrev)
3720                        options->abbrev = 40;
3721        }
3722        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3723                options->a_prefix = optarg;
3724                return argcount;
3725        }
3726        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3727                options->b_prefix = optarg;
3728                return argcount;
3729        }
3730        else if (!strcmp(arg, "--no-prefix"))
3731                options->a_prefix = options->b_prefix = "";
3732        else if (opt_arg(arg, '\0', "inter-hunk-context",
3733                         &options->interhunkcontext))
3734                ;
3735        else if (!strcmp(arg, "-W"))
3736                DIFF_OPT_SET(options, FUNCCONTEXT);
3737        else if (!strcmp(arg, "--function-context"))
3738                DIFF_OPT_SET(options, FUNCCONTEXT);
3739        else if (!strcmp(arg, "--no-function-context"))
3740                DIFF_OPT_CLR(options, FUNCCONTEXT);
3741        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3742                options->file = fopen(optarg, "w");
3743                if (!options->file)
3744                        die_errno("Could not open '%s'", optarg);
3745                options->close_file = 1;
3746                return argcount;
3747        } else
3748                return 0;
3749        return 1;
3750}
3751
3752int parse_rename_score(const char **cp_p)
3753{
3754        unsigned long num, scale;
3755        int ch, dot;
3756        const char *cp = *cp_p;
3757
3758        num = 0;
3759        scale = 1;
3760        dot = 0;
3761        for (;;) {
3762                ch = *cp;
3763                if ( !dot && ch == '.' ) {
3764                        scale = 1;
3765                        dot = 1;
3766                } else if ( ch == '%' ) {
3767                        scale = dot ? scale*100 : 100;
3768                        cp++;   /* % is always at the end */
3769                        break;
3770                } else if ( ch >= '0' && ch <= '9' ) {
3771                        if ( scale < 100000 ) {
3772                                scale *= 10;
3773                                num = (num*10) + (ch-'0');
3774                        }
3775                } else {
3776                        break;
3777                }
3778                cp++;
3779        }
3780        *cp_p = cp;
3781
3782        /* user says num divided by scale and we say internally that
3783         * is MAX_SCORE * num / scale.
3784         */
3785        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3786}
3787
3788static int diff_scoreopt_parse(const char *opt)
3789{
3790        int opt1, opt2, cmd;
3791
3792        if (*opt++ != '-')
3793                return -1;
3794        cmd = *opt++;
3795        if (cmd == '-') {
3796                /* convert the long-form arguments into short-form versions */
3797                if (!prefixcmp(opt, "break-rewrites")) {
3798                        opt += strlen("break-rewrites");
3799                        if (*opt == 0 || *opt++ == '=')
3800                                cmd = 'B';
3801                } else if (!prefixcmp(opt, "find-copies")) {
3802                        opt += strlen("find-copies");
3803                        if (*opt == 0 || *opt++ == '=')
3804                                cmd = 'C';
3805                } else if (!prefixcmp(opt, "find-renames")) {
3806                        opt += strlen("find-renames");
3807                        if (*opt == 0 || *opt++ == '=')
3808                                cmd = 'M';
3809                }
3810        }
3811        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3812                return -1; /* that is not a -M, -C nor -B option */
3813
3814        opt1 = parse_rename_score(&opt);
3815        if (cmd != 'B')
3816                opt2 = 0;
3817        else {
3818                if (*opt == 0)
3819                        opt2 = 0;
3820                else if (*opt != '/')
3821                        return -1; /* we expect -B80/99 or -B80 */
3822                else {
3823                        opt++;
3824                        opt2 = parse_rename_score(&opt);
3825                }
3826        }
3827        if (*opt != 0)
3828                return -1;
3829        return opt1 | (opt2 << 16);
3830}
3831
3832struct diff_queue_struct diff_queued_diff;
3833
3834void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3835{
3836        if (queue->alloc <= queue->nr) {
3837                queue->alloc = alloc_nr(queue->alloc);
3838                queue->queue = xrealloc(queue->queue,
3839                                        sizeof(dp) * queue->alloc);
3840        }
3841        queue->queue[queue->nr++] = dp;
3842}
3843
3844struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3845                                 struct diff_filespec *one,
3846                                 struct diff_filespec *two)
3847{
3848        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3849        dp->one = one;
3850        dp->two = two;
3851        if (queue)
3852                diff_q(queue, dp);
3853        return dp;
3854}
3855
3856void diff_free_filepair(struct diff_filepair *p)
3857{
3858        free_filespec(p->one);
3859        free_filespec(p->two);
3860        free(p);
3861}
3862
3863/* This is different from find_unique_abbrev() in that
3864 * it stuffs the result with dots for alignment.
3865 */
3866const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3867{
3868        int abblen;
3869        const char *abbrev;
3870        if (len == 40)
3871                return sha1_to_hex(sha1);
3872
3873        abbrev = find_unique_abbrev(sha1, len);
3874        abblen = strlen(abbrev);
3875        if (abblen < 37) {
3876                static char hex[41];
3877                if (len < abblen && abblen <= len + 2)
3878                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3879                else
3880                        sprintf(hex, "%s...", abbrev);
3881                return hex;
3882        }
3883        return sha1_to_hex(sha1);
3884}
3885
3886static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3887{
3888        int line_termination = opt->line_termination;
3889        int inter_name_termination = line_termination ? '\t' : '\0';
3890        if (opt->output_prefix) {
3891                struct strbuf *msg = NULL;
3892                msg = opt->output_prefix(opt, opt->output_prefix_data);
3893                fprintf(opt->file, "%s", msg->buf);
3894        }
3895
3896        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3897                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3898                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3899                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3900        }
3901        if (p->score) {
3902                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3903                        inter_name_termination);
3904        } else {
3905                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3906        }
3907
3908        if (p->status == DIFF_STATUS_COPIED ||
3909            p->status == DIFF_STATUS_RENAMED) {
3910                const char *name_a, *name_b;
3911                name_a = p->one->path;
3912                name_b = p->two->path;
3913                strip_prefix(opt->prefix_length, &name_a, &name_b);
3914                write_name_quoted(name_a, opt->file, inter_name_termination);
3915                write_name_quoted(name_b, opt->file, line_termination);
3916        } else {
3917                const char *name_a, *name_b;
3918                name_a = p->one->mode ? p->one->path : p->two->path;
3919                name_b = NULL;
3920                strip_prefix(opt->prefix_length, &name_a, &name_b);
3921                write_name_quoted(name_a, opt->file, line_termination);
3922        }
3923}
3924
3925int diff_unmodified_pair(struct diff_filepair *p)
3926{
3927        /* This function is written stricter than necessary to support
3928         * the currently implemented transformers, but the idea is to
3929         * let transformers to produce diff_filepairs any way they want,
3930         * and filter and clean them up here before producing the output.
3931         */
3932        struct diff_filespec *one = p->one, *two = p->two;
3933
3934        if (DIFF_PAIR_UNMERGED(p))
3935                return 0; /* unmerged is interesting */
3936
3937        /* deletion, addition, mode or type change
3938         * and rename are all interesting.
3939         */
3940        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3941            DIFF_PAIR_MODE_CHANGED(p) ||
3942            strcmp(one->path, two->path))
3943                return 0;
3944
3945        /* both are valid and point at the same path.  that is, we are
3946         * dealing with a change.
3947         */
3948        if (one->sha1_valid && two->sha1_valid &&
3949            !hashcmp(one->sha1, two->sha1) &&
3950            !one->dirty_submodule && !two->dirty_submodule)
3951                return 1; /* no change */
3952        if (!one->sha1_valid && !two->sha1_valid)
3953                return 1; /* both look at the same file on the filesystem. */
3954        return 0;
3955}
3956
3957static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3958{
3959        if (diff_unmodified_pair(p))
3960                return;
3961
3962        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3963            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3964                return; /* no tree diffs in patch format */
3965
3966        run_diff(p, o);
3967}
3968
3969static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3970                            struct diffstat_t *diffstat)
3971{
3972        if (diff_unmodified_pair(p))
3973                return;
3974
3975        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3976            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3977                return; /* no useful stat for tree diffs */
3978
3979        run_diffstat(p, o, diffstat);
3980}
3981
3982static void diff_flush_checkdiff(struct diff_filepair *p,
3983                struct diff_options *o)
3984{
3985        if (diff_unmodified_pair(p))
3986                return;
3987
3988        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3989            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3990                return; /* nothing to check in tree diffs */
3991
3992        run_checkdiff(p, o);
3993}
3994
3995int diff_queue_is_empty(void)
3996{
3997        struct diff_queue_struct *q = &diff_queued_diff;
3998        int i;
3999        for (i = 0; i < q->nr; i++)
4000                if (!diff_unmodified_pair(q->queue[i]))
4001                        return 0;
4002        return 1;
4003}
4004
4005#if DIFF_DEBUG
4006void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4007{
4008        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4009                x, one ? one : "",
4010                s->path,
4011                DIFF_FILE_VALID(s) ? "valid" : "invalid",
4012                s->mode,
4013                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4014        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
4015                x, one ? one : "",
4016                s->size, s->xfrm_flags);
4017}
4018
4019void diff_debug_filepair(const struct diff_filepair *p, int i)
4020{
4021        diff_debug_filespec(p->one, i, "one");
4022        diff_debug_filespec(p->two, i, "two");
4023        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4024                p->score, p->status ? p->status : '?',
4025                p->one->rename_used, p->broken_pair);
4026}
4027
4028void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4029{
4030        int i;
4031        if (msg)
4032                fprintf(stderr, "%s\n", msg);
4033        fprintf(stderr, "q->nr = %d\n", q->nr);
4034        for (i = 0; i < q->nr; i++) {
4035                struct diff_filepair *p = q->queue[i];
4036                diff_debug_filepair(p, i);
4037        }
4038}
4039#endif
4040
4041static void diff_resolve_rename_copy(void)
4042{
4043        int i;
4044        struct diff_filepair *p;
4045        struct diff_queue_struct *q = &diff_queued_diff;
4046
4047        diff_debug_queue("resolve-rename-copy", q);
4048
4049        for (i = 0; i < q->nr; i++) {
4050                p = q->queue[i];
4051                p->status = 0; /* undecided */
4052                if (DIFF_PAIR_UNMERGED(p))
4053                        p->status = DIFF_STATUS_UNMERGED;
4054                else if (!DIFF_FILE_VALID(p->one))
4055                        p->status = DIFF_STATUS_ADDED;
4056                else if (!DIFF_FILE_VALID(p->two))
4057                        p->status = DIFF_STATUS_DELETED;
4058                else if (DIFF_PAIR_TYPE_CHANGED(p))
4059                        p->status = DIFF_STATUS_TYPE_CHANGED;
4060
4061                /* from this point on, we are dealing with a pair
4062                 * whose both sides are valid and of the same type, i.e.
4063                 * either in-place edit or rename/copy edit.
4064                 */
4065                else if (DIFF_PAIR_RENAME(p)) {
4066                        /*
4067                         * A rename might have re-connected a broken
4068                         * pair up, causing the pathnames to be the
4069                         * same again. If so, that's not a rename at
4070                         * all, just a modification..
4071                         *
4072                         * Otherwise, see if this source was used for
4073                         * multiple renames, in which case we decrement
4074                         * the count, and call it a copy.
4075                         */
4076                        if (!strcmp(p->one->path, p->two->path))
4077                                p->status = DIFF_STATUS_MODIFIED;
4078                        else if (--p->one->rename_used > 0)
4079                                p->status = DIFF_STATUS_COPIED;
4080                        else
4081                                p->status = DIFF_STATUS_RENAMED;
4082                }
4083                else if (hashcmp(p->one->sha1, p->two->sha1) ||
4084                         p->one->mode != p->two->mode ||
4085                         p->one->dirty_submodule ||
4086                         p->two->dirty_submodule ||
4087                         is_null_sha1(p->one->sha1))
4088                        p->status = DIFF_STATUS_MODIFIED;
4089                else {
4090                        /* This is a "no-change" entry and should not
4091                         * happen anymore, but prepare for broken callers.
4092                         */
4093                        error("feeding unmodified %s to diffcore",
4094                              p->one->path);
4095                        p->status = DIFF_STATUS_UNKNOWN;
4096                }
4097        }
4098        diff_debug_queue("resolve-rename-copy done", q);
4099}
4100
4101static int check_pair_status(struct diff_filepair *p)
4102{
4103        switch (p->status) {
4104        case DIFF_STATUS_UNKNOWN:
4105                return 0;
4106        case 0:
4107                die("internal error in diff-resolve-rename-copy");
4108        default:
4109                return 1;
4110        }
4111}
4112
4113static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4114{
4115        int fmt = opt->output_format;
4116
4117        if (fmt & DIFF_FORMAT_CHECKDIFF)
4118                diff_flush_checkdiff(p, opt);
4119        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4120                diff_flush_raw(p, opt);
4121        else if (fmt & DIFF_FORMAT_NAME) {
4122                const char *name_a, *name_b;
4123                name_a = p->two->path;
4124                name_b = NULL;
4125                strip_prefix(opt->prefix_length, &name_a, &name_b);
4126                write_name_quoted(name_a, opt->file, opt->line_termination);
4127        }
4128}
4129
4130static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4131{
4132        if (fs->mode)
4133                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4134        else
4135                fprintf(file, " %s ", newdelete);
4136        write_name_quoted(fs->path, file, '\n');
4137}
4138
4139
4140static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4141                const char *line_prefix)
4142{
4143        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4144                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4145                        p->two->mode, show_name ? ' ' : '\n');
4146                if (show_name) {
4147                        write_name_quoted(p->two->path, file, '\n');
4148                }
4149        }
4150}
4151
4152static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4153                        const char *line_prefix)
4154{
4155        char *names = pprint_rename(p->one->path, p->two->path);
4156
4157        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4158        free(names);
4159        show_mode_change(file, p, 0, line_prefix);
4160}
4161
4162static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4163{
4164        FILE *file = opt->file;
4165        char *line_prefix = "";
4166
4167        if (opt->output_prefix) {
4168                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4169                line_prefix = buf->buf;
4170        }
4171
4172        switch(p->status) {
4173        case DIFF_STATUS_DELETED:
4174                fputs(line_prefix, file);
4175                show_file_mode_name(file, "delete", p->one);
4176                break;
4177        case DIFF_STATUS_ADDED:
4178                fputs(line_prefix, file);
4179                show_file_mode_name(file, "create", p->two);
4180                break;
4181        case DIFF_STATUS_COPIED:
4182                fputs(line_prefix, file);
4183                show_rename_copy(file, "copy", p, line_prefix);
4184                break;
4185        case DIFF_STATUS_RENAMED:
4186                fputs(line_prefix, file);
4187                show_rename_copy(file, "rename", p, line_prefix);
4188                break;
4189        default:
4190                if (p->score) {
4191                        fprintf(file, "%s rewrite ", line_prefix);
4192                        write_name_quoted(p->two->path, file, ' ');
4193                        fprintf(file, "(%d%%)\n", similarity_index(p));
4194                }
4195                show_mode_change(file, p, !p->score, line_prefix);
4196                break;
4197        }
4198}
4199
4200struct patch_id_t {
4201        git_SHA_CTX *ctx;
4202        int patchlen;
4203};
4204
4205static int remove_space(char *line, int len)
4206{
4207        int i;
4208        char *dst = line;
4209        unsigned char c;
4210
4211        for (i = 0; i < len; i++)
4212                if (!isspace((c = line[i])))
4213                        *dst++ = c;
4214
4215        return dst - line;
4216}
4217
4218static void patch_id_consume(void *priv, char *line, unsigned long len)
4219{
4220        struct patch_id_t *data = priv;
4221        int new_len;
4222
4223        /* Ignore line numbers when computing the SHA1 of the patch */
4224        if (!prefixcmp(line, "@@ -"))
4225                return;
4226
4227        new_len = remove_space(line, len);
4228
4229        git_SHA1_Update(data->ctx, line, new_len);
4230        data->patchlen += new_len;
4231}
4232
4233/* returns 0 upon success, and writes result into sha1 */
4234static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4235{
4236        struct diff_queue_struct *q = &diff_queued_diff;
4237        int i;
4238        git_SHA_CTX ctx;
4239        struct patch_id_t data;
4240        char buffer[PATH_MAX * 4 + 20];
4241
4242        git_SHA1_Init(&ctx);
4243        memset(&data, 0, sizeof(struct patch_id_t));
4244        data.ctx = &ctx;
4245
4246        for (i = 0; i < q->nr; i++) {
4247                xpparam_t xpp;
4248                xdemitconf_t xecfg;
4249                mmfile_t mf1, mf2;
4250                struct diff_filepair *p = q->queue[i];
4251                int len1, len2;
4252
4253                memset(&xpp, 0, sizeof(xpp));
4254                memset(&xecfg, 0, sizeof(xecfg));
4255                if (p->status == 0)
4256                        return error("internal diff status error");
4257                if (p->status == DIFF_STATUS_UNKNOWN)
4258                        continue;
4259                if (diff_unmodified_pair(p))
4260                        continue;
4261                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4262                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4263                        continue;
4264                if (DIFF_PAIR_UNMERGED(p))
4265                        continue;
4266
4267                diff_fill_sha1_info(p->one);
4268                diff_fill_sha1_info(p->two);
4269                if (fill_mmfile(&mf1, p->one) < 0 ||
4270                                fill_mmfile(&mf2, p->two) < 0)
4271                        return error("unable to read files to diff");
4272
4273                len1 = remove_space(p->one->path, strlen(p->one->path));
4274                len2 = remove_space(p->two->path, strlen(p->two->path));
4275                if (p->one->mode == 0)
4276                        len1 = snprintf(buffer, sizeof(buffer),
4277                                        "diff--gita/%.*sb/%.*s"
4278                                        "newfilemode%06o"
4279                                        "---/dev/null"
4280                                        "+++b/%.*s",
4281                                        len1, p->one->path,
4282                                        len2, p->two->path,
4283                                        p->two->mode,
4284                                        len2, p->two->path);
4285                else if (p->two->mode == 0)
4286                        len1 = snprintf(buffer, sizeof(buffer),
4287                                        "diff--gita/%.*sb/%.*s"
4288                                        "deletedfilemode%06o"
4289                                        "---a/%.*s"
4290                                        "+++/dev/null",
4291                                        len1, p->one->path,
4292                                        len2, p->two->path,
4293                                        p->one->mode,
4294                                        len1, p->one->path);
4295                else
4296                        len1 = snprintf(buffer, sizeof(buffer),
4297                                        "diff--gita/%.*sb/%.*s"
4298                                        "---a/%.*s"
4299                                        "+++b/%.*s",
4300                                        len1, p->one->path,
4301                                        len2, p->two->path,
4302                                        len1, p->one->path,
4303                                        len2, p->two->path);
4304                git_SHA1_Update(&ctx, buffer, len1);
4305
4306                if (diff_filespec_is_binary(p->one) ||
4307                    diff_filespec_is_binary(p->two)) {
4308                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4309                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4310                        continue;
4311                }
4312
4313                xpp.flags = 0;
4314                xecfg.ctxlen = 3;
4315                xecfg.flags = 0;
4316                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4317                              &xpp, &xecfg);
4318        }
4319
4320        git_SHA1_Final(sha1, &ctx);
4321        return 0;
4322}
4323
4324int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4325{
4326        struct diff_queue_struct *q = &diff_queued_diff;
4327        int i;
4328        int result = diff_get_patch_id(options, sha1);
4329
4330        for (i = 0; i < q->nr; i++)
4331                diff_free_filepair(q->queue[i]);
4332
4333        free(q->queue);
4334        DIFF_QUEUE_CLEAR(q);
4335
4336        return result;
4337}
4338
4339static int is_summary_empty(const struct diff_queue_struct *q)
4340{
4341        int i;
4342
4343        for (i = 0; i < q->nr; i++) {
4344                const struct diff_filepair *p = q->queue[i];
4345
4346                switch (p->status) {
4347                case DIFF_STATUS_DELETED:
4348                case DIFF_STATUS_ADDED:
4349                case DIFF_STATUS_COPIED:
4350                case DIFF_STATUS_RENAMED:
4351                        return 0;
4352                default:
4353                        if (p->score)
4354                                return 0;
4355                        if (p->one->mode && p->two->mode &&
4356                            p->one->mode != p->two->mode)
4357                                return 0;
4358                        break;
4359                }
4360        }
4361        return 1;
4362}
4363
4364static const char rename_limit_warning[] =
4365"inexact rename detection was skipped due to too many files.";
4366
4367static const char degrade_cc_to_c_warning[] =
4368"only found copies from modified paths due to too many files.";
4369
4370static const char rename_limit_advice[] =
4371"you may want to set your %s variable to at least "
4372"%d and retry the command.";
4373
4374void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4375{
4376        if (degraded_cc)
4377                warning(degrade_cc_to_c_warning);
4378        else if (needed)
4379                warning(rename_limit_warning);
4380        else
4381                return;
4382        if (0 < needed && needed < 32767)
4383                warning(rename_limit_advice, varname, needed);
4384}
4385
4386void diff_flush(struct diff_options *options)
4387{
4388        struct diff_queue_struct *q = &diff_queued_diff;
4389        int i, output_format = options->output_format;
4390        int separator = 0;
4391        int dirstat_by_line = 0;
4392
4393        /*
4394         * Order: raw, stat, summary, patch
4395         * or:    name/name-status/checkdiff (other bits clear)
4396         */
4397        if (!q->nr)
4398                goto free_queue;
4399
4400        if (output_format & (DIFF_FORMAT_RAW |
4401                             DIFF_FORMAT_NAME |
4402                             DIFF_FORMAT_NAME_STATUS |
4403                             DIFF_FORMAT_CHECKDIFF)) {
4404                for (i = 0; i < q->nr; i++) {
4405                        struct diff_filepair *p = q->queue[i];
4406                        if (check_pair_status(p))
4407                                flush_one_pair(p, options);
4408                }
4409                separator++;
4410        }
4411
4412        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4413                dirstat_by_line = 1;
4414
4415        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4416            dirstat_by_line) {
4417                struct diffstat_t diffstat;
4418
4419                memset(&diffstat, 0, sizeof(struct diffstat_t));
4420                for (i = 0; i < q->nr; i++) {
4421                        struct diff_filepair *p = q->queue[i];
4422                        if (check_pair_status(p))
4423                                diff_flush_stat(p, options, &diffstat);
4424                }
4425                if (output_format & DIFF_FORMAT_NUMSTAT)
4426                        show_numstat(&diffstat, options);
4427                if (output_format & DIFF_FORMAT_DIFFSTAT)
4428                        show_stats(&diffstat, options);
4429                if (output_format & DIFF_FORMAT_SHORTSTAT)
4430                        show_shortstats(&diffstat, options);
4431                if (output_format & DIFF_FORMAT_DIRSTAT)
4432                        show_dirstat_by_line(&diffstat, options);
4433                free_diffstat_info(&diffstat);
4434                separator++;
4435        }
4436        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4437                show_dirstat(options);
4438
4439        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4440                for (i = 0; i < q->nr; i++) {
4441                        diff_summary(options, q->queue[i]);
4442                }
4443                separator++;
4444        }
4445
4446        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4447            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4448            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4449                /*
4450                 * run diff_flush_patch for the exit status. setting
4451                 * options->file to /dev/null should be safe, becaue we
4452                 * aren't supposed to produce any output anyway.
4453                 */
4454                if (options->close_file)
4455                        fclose(options->file);
4456                options->file = fopen("/dev/null", "w");
4457                if (!options->file)
4458                        die_errno("Could not open /dev/null");
4459                options->close_file = 1;
4460                for (i = 0; i < q->nr; i++) {
4461                        struct diff_filepair *p = q->queue[i];
4462                        if (check_pair_status(p))
4463                                diff_flush_patch(p, options);
4464                        if (options->found_changes)
4465                                break;
4466                }
4467        }
4468
4469        if (output_format & DIFF_FORMAT_PATCH) {
4470                if (separator) {
4471                        if (options->output_prefix) {
4472                                struct strbuf *msg = NULL;
4473                                msg = options->output_prefix(options,
4474                                        options->output_prefix_data);
4475                                fwrite(msg->buf, msg->len, 1, stdout);
4476                        }
4477                        putc(options->line_termination, options->file);
4478                        if (options->stat_sep) {
4479                                /* attach patch instead of inline */
4480                                fputs(options->stat_sep, options->file);
4481                        }
4482                }
4483
4484                for (i = 0; i < q->nr; i++) {
4485                        struct diff_filepair *p = q->queue[i];
4486                        if (check_pair_status(p))
4487                                diff_flush_patch(p, options);
4488                }
4489        }
4490
4491        if (output_format & DIFF_FORMAT_CALLBACK)
4492                options->format_callback(q, options, options->format_callback_data);
4493
4494        for (i = 0; i < q->nr; i++)
4495                diff_free_filepair(q->queue[i]);
4496free_queue:
4497        free(q->queue);
4498        DIFF_QUEUE_CLEAR(q);
4499        if (options->close_file)
4500                fclose(options->file);
4501
4502        /*
4503         * Report the content-level differences with HAS_CHANGES;
4504         * diff_addremove/diff_change does not set the bit when
4505         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4506         */
4507        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4508                if (options->found_changes)
4509                        DIFF_OPT_SET(options, HAS_CHANGES);
4510                else
4511                        DIFF_OPT_CLR(options, HAS_CHANGES);
4512        }
4513}
4514
4515static void diffcore_apply_filter(const char *filter)
4516{
4517        int i;
4518        struct diff_queue_struct *q = &diff_queued_diff;
4519        struct diff_queue_struct outq;
4520        DIFF_QUEUE_CLEAR(&outq);
4521
4522        if (!filter)
4523                return;
4524
4525        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4526                int found;
4527                for (i = found = 0; !found && i < q->nr; i++) {
4528                        struct diff_filepair *p = q->queue[i];
4529                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4530                             ((p->score &&
4531                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4532                              (!p->score &&
4533                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4534                            ((p->status != DIFF_STATUS_MODIFIED) &&
4535                             strchr(filter, p->status)))
4536                                found++;
4537                }
4538                if (found)
4539                        return;
4540
4541                /* otherwise we will clear the whole queue
4542                 * by copying the empty outq at the end of this
4543                 * function, but first clear the current entries
4544                 * in the queue.
4545                 */
4546                for (i = 0; i < q->nr; i++)
4547                        diff_free_filepair(q->queue[i]);
4548        }
4549        else {
4550                /* Only the matching ones */
4551                for (i = 0; i < q->nr; i++) {
4552                        struct diff_filepair *p = q->queue[i];
4553
4554                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4555                             ((p->score &&
4556                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4557                              (!p->score &&
4558                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4559                            ((p->status != DIFF_STATUS_MODIFIED) &&
4560                             strchr(filter, p->status)))
4561                                diff_q(&outq, p);
4562                        else
4563                                diff_free_filepair(p);
4564                }
4565        }
4566        free(q->queue);
4567        *q = outq;
4568}
4569
4570/* Check whether two filespecs with the same mode and size are identical */
4571static int diff_filespec_is_identical(struct diff_filespec *one,
4572                                      struct diff_filespec *two)
4573{
4574        if (S_ISGITLINK(one->mode))
4575                return 0;
4576        if (diff_populate_filespec(one, 0))
4577                return 0;
4578        if (diff_populate_filespec(two, 0))
4579                return 0;
4580        return !memcmp(one->data, two->data, one->size);
4581}
4582
4583static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4584{
4585        int i;
4586        struct diff_queue_struct *q = &diff_queued_diff;
4587        struct diff_queue_struct outq;
4588        DIFF_QUEUE_CLEAR(&outq);
4589
4590        for (i = 0; i < q->nr; i++) {
4591                struct diff_filepair *p = q->queue[i];
4592
4593                /*
4594                 * 1. Entries that come from stat info dirtiness
4595                 *    always have both sides (iow, not create/delete),
4596                 *    one side of the object name is unknown, with
4597                 *    the same mode and size.  Keep the ones that
4598                 *    do not match these criteria.  They have real
4599                 *    differences.
4600                 *
4601                 * 2. At this point, the file is known to be modified,
4602                 *    with the same mode and size, and the object
4603                 *    name of one side is unknown.  Need to inspect
4604                 *    the identical contents.
4605                 */
4606                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4607                    !DIFF_FILE_VALID(p->two) ||
4608                    (p->one->sha1_valid && p->two->sha1_valid) ||
4609                    (p->one->mode != p->two->mode) ||
4610                    diff_populate_filespec(p->one, 1) ||
4611                    diff_populate_filespec(p->two, 1) ||
4612                    (p->one->size != p->two->size) ||
4613                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4614                        diff_q(&outq, p);
4615                else {
4616                        /*
4617                         * The caller can subtract 1 from skip_stat_unmatch
4618                         * to determine how many paths were dirty only
4619                         * due to stat info mismatch.
4620                         */
4621                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4622                                diffopt->skip_stat_unmatch++;
4623                        diff_free_filepair(p);
4624                }
4625        }
4626        free(q->queue);
4627        *q = outq;
4628}
4629
4630static int diffnamecmp(const void *a_, const void *b_)
4631{
4632        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4633        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4634        const char *name_a, *name_b;
4635
4636        name_a = a->one ? a->one->path : a->two->path;
4637        name_b = b->one ? b->one->path : b->two->path;
4638        return strcmp(name_a, name_b);
4639}
4640
4641void diffcore_fix_diff_index(struct diff_options *options)
4642{
4643        struct diff_queue_struct *q = &diff_queued_diff;
4644        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4645}
4646
4647void diffcore_std(struct diff_options *options)
4648{
4649        if (options->skip_stat_unmatch)
4650                diffcore_skip_stat_unmatch(options);
4651        if (!options->found_follow) {
4652                /* See try_to_follow_renames() in tree-diff.c */
4653                if (options->break_opt != -1)
4654                        diffcore_break(options->break_opt);
4655                if (options->detect_rename)
4656                        diffcore_rename(options);
4657                if (options->break_opt != -1)
4658                        diffcore_merge_broken();
4659        }
4660        if (options->pickaxe)
4661                diffcore_pickaxe(options);
4662        if (options->orderfile)
4663                diffcore_order(options->orderfile);
4664        if (!options->found_follow)
4665                /* See try_to_follow_renames() in tree-diff.c */
4666                diff_resolve_rename_copy();
4667        diffcore_apply_filter(options->filter);
4668
4669        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4670                DIFF_OPT_SET(options, HAS_CHANGES);
4671        else
4672                DIFF_OPT_CLR(options, HAS_CHANGES);
4673
4674        options->found_follow = 0;
4675}
4676
4677int diff_result_code(struct diff_options *opt, int status)
4678{
4679        int result = 0;
4680
4681        diff_warn_rename_limit("diff.renamelimit",
4682                               opt->needed_rename_limit,
4683                               opt->degraded_cc_to_c);
4684        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4685            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4686                return status;
4687        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4688            DIFF_OPT_TST(opt, HAS_CHANGES))
4689                result |= 01;
4690        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4691            DIFF_OPT_TST(opt, CHECK_FAILED))
4692                result |= 02;
4693        return result;
4694}
4695
4696int diff_can_quit_early(struct diff_options *opt)
4697{
4698        return (DIFF_OPT_TST(opt, QUICK) &&
4699                !opt->filter &&
4700                DIFF_OPT_TST(opt, HAS_CHANGES));
4701}
4702
4703/*
4704 * Shall changes to this submodule be ignored?
4705 *
4706 * Submodule changes can be configured to be ignored separately for each path,
4707 * but that configuration can be overridden from the command line.
4708 */
4709static int is_submodule_ignored(const char *path, struct diff_options *options)
4710{
4711        int ignored = 0;
4712        unsigned orig_flags = options->flags;
4713        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4714                set_diffopt_flags_from_submodule_config(options, path);
4715        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4716                ignored = 1;
4717        options->flags = orig_flags;
4718        return ignored;
4719}
4720
4721void diff_addremove(struct diff_options *options,
4722                    int addremove, unsigned mode,
4723                    const unsigned char *sha1,
4724                    int sha1_valid,
4725                    const char *concatpath, unsigned dirty_submodule)
4726{
4727        struct diff_filespec *one, *two;
4728
4729        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4730                return;
4731
4732        /* This may look odd, but it is a preparation for
4733         * feeding "there are unchanged files which should
4734         * not produce diffs, but when you are doing copy
4735         * detection you would need them, so here they are"
4736         * entries to the diff-core.  They will be prefixed
4737         * with something like '=' or '*' (I haven't decided
4738         * which but should not make any difference).
4739         * Feeding the same new and old to diff_change()
4740         * also has the same effect.
4741         * Before the final output happens, they are pruned after
4742         * merged into rename/copy pairs as appropriate.
4743         */
4744        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4745                addremove = (addremove == '+' ? '-' :
4746                             addremove == '-' ? '+' : addremove);
4747
4748        if (options->prefix &&
4749            strncmp(concatpath, options->prefix, options->prefix_length))
4750                return;
4751
4752        one = alloc_filespec(concatpath);
4753        two = alloc_filespec(concatpath);
4754
4755        if (addremove != '+')
4756                fill_filespec(one, sha1, sha1_valid, mode);
4757        if (addremove != '-') {
4758                fill_filespec(two, sha1, sha1_valid, mode);
4759                two->dirty_submodule = dirty_submodule;
4760        }
4761
4762        diff_queue(&diff_queued_diff, one, two);
4763        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4764                DIFF_OPT_SET(options, HAS_CHANGES);
4765}
4766
4767void diff_change(struct diff_options *options,
4768                 unsigned old_mode, unsigned new_mode,
4769                 const unsigned char *old_sha1,
4770                 const unsigned char *new_sha1,
4771                 int old_sha1_valid, int new_sha1_valid,
4772                 const char *concatpath,
4773                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4774{
4775        struct diff_filespec *one, *two;
4776
4777        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4778            is_submodule_ignored(concatpath, options))
4779                return;
4780
4781        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4782                unsigned tmp;
4783                const unsigned char *tmp_c;
4784                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4785                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4786                tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4787                        new_sha1_valid = tmp;
4788                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4789                        new_dirty_submodule = tmp;
4790        }
4791
4792        if (options->prefix &&
4793            strncmp(concatpath, options->prefix, options->prefix_length))
4794                return;
4795
4796        one = alloc_filespec(concatpath);
4797        two = alloc_filespec(concatpath);
4798        fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4799        fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4800        one->dirty_submodule = old_dirty_submodule;
4801        two->dirty_submodule = new_dirty_submodule;
4802
4803        diff_queue(&diff_queued_diff, one, two);
4804        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4805                DIFF_OPT_SET(options, HAS_CHANGES);
4806}
4807
4808struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4809{
4810        struct diff_filepair *pair;
4811        struct diff_filespec *one, *two;
4812
4813        if (options->prefix &&
4814            strncmp(path, options->prefix, options->prefix_length))
4815                return NULL;
4816
4817        one = alloc_filespec(path);
4818        two = alloc_filespec(path);
4819        pair = diff_queue(&diff_queued_diff, one, two);
4820        pair->is_unmerged = 1;
4821        return pair;
4822}
4823
4824static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4825                size_t *outsize)
4826{
4827        struct diff_tempfile *temp;
4828        const char *argv[3];
4829        const char **arg = argv;
4830        struct child_process child;
4831        struct strbuf buf = STRBUF_INIT;
4832        int err = 0;
4833
4834        temp = prepare_temp_file(spec->path, spec);
4835        *arg++ = pgm;
4836        *arg++ = temp->name;
4837        *arg = NULL;
4838
4839        memset(&child, 0, sizeof(child));
4840        child.use_shell = 1;
4841        child.argv = argv;
4842        child.out = -1;
4843        if (start_command(&child)) {
4844                remove_tempfile();
4845                return NULL;
4846        }
4847
4848        if (strbuf_read(&buf, child.out, 0) < 0)
4849                err = error("error reading from textconv command '%s'", pgm);
4850        close(child.out);
4851
4852        if (finish_command(&child) || err) {
4853                strbuf_release(&buf);
4854                remove_tempfile();
4855                return NULL;
4856        }
4857        remove_tempfile();
4858
4859        return strbuf_detach(&buf, outsize);
4860}
4861
4862size_t fill_textconv(struct userdiff_driver *driver,
4863                     struct diff_filespec *df,
4864                     char **outbuf)
4865{
4866        size_t size;
4867
4868        if (!driver || !driver->textconv) {
4869                if (!DIFF_FILE_VALID(df)) {
4870                        *outbuf = "";
4871                        return 0;
4872                }
4873                if (diff_populate_filespec(df, 0))
4874                        die("unable to read files to diff");
4875                *outbuf = df->data;
4876                return df->size;
4877        }
4878
4879        if (driver->textconv_cache && df->sha1_valid) {
4880                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4881                                          &size);
4882                if (*outbuf)
4883                        return size;
4884        }
4885
4886        *outbuf = run_textconv(driver->textconv, df, &size);
4887        if (!*outbuf)
4888                die("unable to read files to diff");
4889
4890        if (driver->textconv_cache && df->sha1_valid) {
4891                /* ignore errors, as we might be in a readonly repository */
4892                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4893                                size);
4894                /*
4895                 * we could save up changes and flush them all at the end,
4896                 * but we would need an extra call after all diffing is done.
4897                 * Since generating a cache entry is the slow path anyway,
4898                 * this extra overhead probably isn't a big deal.
4899                 */
4900                notes_cache_write(driver->textconv_cache);
4901        }
4902
4903        return size;
4904}