526f1980596887dff99b1236ce39a47b6bdb7215
   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;
  28int diff_use_color_default = -1;
  29static const char *diff_word_regex_cfg;
  30static const char *external_diff_cmd_cfg;
  31int diff_auto_refresh_index = 1;
  32static int diff_mnemonic_prefix;
  33static int diff_no_prefix;
  34static int diff_dirstat_permille_default = 30;
  35static struct diff_options default_diff_options;
  36
  37static char diff_colors[][COLOR_MAXLEN] = {
  38        GIT_COLOR_RESET,
  39        GIT_COLOR_NORMAL,       /* PLAIN */
  40        GIT_COLOR_BOLD,         /* METAINFO */
  41        GIT_COLOR_CYAN,         /* FRAGINFO */
  42        GIT_COLOR_RED,          /* OLD */
  43        GIT_COLOR_GREEN,        /* NEW */
  44        GIT_COLOR_YELLOW,       /* COMMIT */
  45        GIT_COLOR_BG_RED,       /* WHITESPACE */
  46        GIT_COLOR_NORMAL,       /* FUNCINFO */
  47};
  48
  49static int parse_diff_color_slot(const char *var, int ofs)
  50{
  51        if (!strcasecmp(var+ofs, "plain"))
  52                return DIFF_PLAIN;
  53        if (!strcasecmp(var+ofs, "meta"))
  54                return DIFF_METAINFO;
  55        if (!strcasecmp(var+ofs, "frag"))
  56                return DIFF_FRAGINFO;
  57        if (!strcasecmp(var+ofs, "old"))
  58                return DIFF_FILE_OLD;
  59        if (!strcasecmp(var+ofs, "new"))
  60                return DIFF_FILE_NEW;
  61        if (!strcasecmp(var+ofs, "commit"))
  62                return DIFF_COMMIT;
  63        if (!strcasecmp(var+ofs, "whitespace"))
  64                return DIFF_WHITESPACE;
  65        if (!strcasecmp(var+ofs, "func"))
  66                return DIFF_FUNCINFO;
  67        return -1;
  68}
  69
  70static int parse_dirstat_params(struct diff_options *options, const char *params,
  71                                struct strbuf *errmsg)
  72{
  73        const char *p = params;
  74        int p_len, ret = 0;
  75
  76        while (*p) {
  77                p_len = strchrnul(p, ',') - p;
  78                if (!memcmp(p, "changes", p_len)) {
  79                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  80                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  81                } else if (!memcmp(p, "lines", p_len)) {
  82                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  83                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  84                } else if (!memcmp(p, "files", p_len)) {
  85                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  86                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  87                } else if (!memcmp(p, "noncumulative", p_len)) {
  88                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
  89                } else if (!memcmp(p, "cumulative", p_len)) {
  90                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
  91                } else if (isdigit(*p)) {
  92                        char *end;
  93                        int permille = strtoul(p, &end, 10) * 10;
  94                        if (*end == '.' && isdigit(*++end)) {
  95                                /* only use first digit */
  96                                permille += *end - '0';
  97                                /* .. and ignore any further digits */
  98                                while (isdigit(*++end))
  99                                        ; /* nothing */
 100                        }
 101                        if (end - p == p_len)
 102                                options->dirstat_permille = permille;
 103                        else {
 104                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%.*s'\n"),
 105                                            p_len, p);
 106                                ret++;
 107                        }
 108                } else {
 109                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%.*s'\n"),
 110                                    p_len, p);
 111                        ret++;
 112                }
 113
 114                p += p_len;
 115
 116                if (*p)
 117                        p++; /* more parameters, swallow separator */
 118        }
 119        return ret;
 120}
 121
 122static int git_config_rename(const char *var, const char *value)
 123{
 124        if (!value)
 125                return DIFF_DETECT_RENAME;
 126        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 127                return  DIFF_DETECT_COPY;
 128        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 129}
 130
 131/*
 132 * These are to give UI layer defaults.
 133 * The core-level commands such as git-diff-files should
 134 * never be affected by the setting of diff.renames
 135 * the user happens to have in the configuration file.
 136 */
 137int git_diff_ui_config(const char *var, const char *value, void *cb)
 138{
 139        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 140                diff_use_color_default = git_config_colorbool(var, value);
 141                return 0;
 142        }
 143        if (!strcmp(var, "diff.renames")) {
 144                diff_detect_rename_default = git_config_rename(var, value);
 145                return 0;
 146        }
 147        if (!strcmp(var, "diff.autorefreshindex")) {
 148                diff_auto_refresh_index = git_config_bool(var, value);
 149                return 0;
 150        }
 151        if (!strcmp(var, "diff.mnemonicprefix")) {
 152                diff_mnemonic_prefix = git_config_bool(var, value);
 153                return 0;
 154        }
 155        if (!strcmp(var, "diff.noprefix")) {
 156                diff_no_prefix = git_config_bool(var, value);
 157                return 0;
 158        }
 159        if (!strcmp(var, "diff.external"))
 160                return git_config_string(&external_diff_cmd_cfg, var, value);
 161        if (!strcmp(var, "diff.wordregex"))
 162                return git_config_string(&diff_word_regex_cfg, var, value);
 163
 164        if (!strcmp(var, "diff.ignoresubmodules"))
 165                handle_ignore_submodules_arg(&default_diff_options, value);
 166
 167        if (git_color_config(var, value, cb) < 0)
 168                return -1;
 169
 170        return git_diff_basic_config(var, value, cb);
 171}
 172
 173int git_diff_basic_config(const char *var, const char *value, void *cb)
 174{
 175        if (!strcmp(var, "diff.renamelimit")) {
 176                diff_rename_limit_default = git_config_int(var, value);
 177                return 0;
 178        }
 179
 180        switch (userdiff_config(var, value)) {
 181                case 0: break;
 182                case -1: return -1;
 183                default: return 0;
 184        }
 185
 186        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 187                int slot = parse_diff_color_slot(var, 11);
 188                if (slot < 0)
 189                        return 0;
 190                if (!value)
 191                        return config_error_nonbool(var);
 192                color_parse(value, var, diff_colors[slot]);
 193                return 0;
 194        }
 195
 196        /* like GNU diff's --suppress-blank-empty option  */
 197        if (!strcmp(var, "diff.suppressblankempty") ||
 198                        /* for backwards compatibility */
 199                        !strcmp(var, "diff.suppress-blank-empty")) {
 200                diff_suppress_blank_empty = git_config_bool(var, value);
 201                return 0;
 202        }
 203
 204        if (!strcmp(var, "diff.dirstat")) {
 205                struct strbuf errmsg = STRBUF_INIT;
 206                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 207                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 208                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 209                                errmsg.buf);
 210                strbuf_release(&errmsg);
 211                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 212                return 0;
 213        }
 214
 215        if (!prefixcmp(var, "submodule."))
 216                return parse_submodule_config_option(var, value);
 217
 218        return git_default_config(var, value, cb);
 219}
 220
 221static char *quote_two(const char *one, const char *two)
 222{
 223        int need_one = quote_c_style(one, NULL, NULL, 1);
 224        int need_two = quote_c_style(two, NULL, NULL, 1);
 225        struct strbuf res = STRBUF_INIT;
 226
 227        if (need_one + need_two) {
 228                strbuf_addch(&res, '"');
 229                quote_c_style(one, &res, NULL, 1);
 230                quote_c_style(two, &res, NULL, 1);
 231                strbuf_addch(&res, '"');
 232        } else {
 233                strbuf_addstr(&res, one);
 234                strbuf_addstr(&res, two);
 235        }
 236        return strbuf_detach(&res, NULL);
 237}
 238
 239static const char *external_diff(void)
 240{
 241        static const char *external_diff_cmd = NULL;
 242        static int done_preparing = 0;
 243
 244        if (done_preparing)
 245                return external_diff_cmd;
 246        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 247        if (!external_diff_cmd)
 248                external_diff_cmd = external_diff_cmd_cfg;
 249        done_preparing = 1;
 250        return external_diff_cmd;
 251}
 252
 253static struct diff_tempfile {
 254        const char *name; /* filename external diff should read from */
 255        char hex[41];
 256        char mode[10];
 257        char tmp_path[PATH_MAX];
 258} diff_temp[2];
 259
 260typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 261
 262struct emit_callback {
 263        int color_diff;
 264        unsigned ws_rule;
 265        int blank_at_eof_in_preimage;
 266        int blank_at_eof_in_postimage;
 267        int lno_in_preimage;
 268        int lno_in_postimage;
 269        sane_truncate_fn truncate;
 270        const char **label_path;
 271        struct diff_words_data *diff_words;
 272        struct diff_options *opt;
 273        int *found_changesp;
 274        struct strbuf *header;
 275};
 276
 277static int count_lines(const char *data, int size)
 278{
 279        int count, ch, completely_empty = 1, nl_just_seen = 0;
 280        count = 0;
 281        while (0 < size--) {
 282                ch = *data++;
 283                if (ch == '\n') {
 284                        count++;
 285                        nl_just_seen = 1;
 286                        completely_empty = 0;
 287                }
 288                else {
 289                        nl_just_seen = 0;
 290                        completely_empty = 0;
 291                }
 292        }
 293        if (completely_empty)
 294                return 0;
 295        if (!nl_just_seen)
 296                count++; /* no trailing newline */
 297        return count;
 298}
 299
 300static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 301{
 302        if (!DIFF_FILE_VALID(one)) {
 303                mf->ptr = (char *)""; /* does not matter */
 304                mf->size = 0;
 305                return 0;
 306        }
 307        else if (diff_populate_filespec(one, 0))
 308                return -1;
 309
 310        mf->ptr = one->data;
 311        mf->size = one->size;
 312        return 0;
 313}
 314
 315/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 316static unsigned long diff_filespec_size(struct diff_filespec *one)
 317{
 318        if (!DIFF_FILE_VALID(one))
 319                return 0;
 320        diff_populate_filespec(one, 1);
 321        return one->size;
 322}
 323
 324static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 325{
 326        char *ptr = mf->ptr;
 327        long size = mf->size;
 328        int cnt = 0;
 329
 330        if (!size)
 331                return cnt;
 332        ptr += size - 1; /* pointing at the very end */
 333        if (*ptr != '\n')
 334                ; /* incomplete line */
 335        else
 336                ptr--; /* skip the last LF */
 337        while (mf->ptr < ptr) {
 338                char *prev_eol;
 339                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 340                        if (*prev_eol == '\n')
 341                                break;
 342                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 343                        break;
 344                cnt++;
 345                ptr = prev_eol - 1;
 346        }
 347        return cnt;
 348}
 349
 350static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 351                               struct emit_callback *ecbdata)
 352{
 353        int l1, l2, at;
 354        unsigned ws_rule = ecbdata->ws_rule;
 355        l1 = count_trailing_blank(mf1, ws_rule);
 356        l2 = count_trailing_blank(mf2, ws_rule);
 357        if (l2 <= l1) {
 358                ecbdata->blank_at_eof_in_preimage = 0;
 359                ecbdata->blank_at_eof_in_postimage = 0;
 360                return;
 361        }
 362        at = count_lines(mf1->ptr, mf1->size);
 363        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 364
 365        at = count_lines(mf2->ptr, mf2->size);
 366        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 367}
 368
 369static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 370                        int first, const char *line, int len)
 371{
 372        int has_trailing_newline, has_trailing_carriage_return;
 373        int nofirst;
 374        FILE *file = o->file;
 375
 376        if (o->output_prefix) {
 377                struct strbuf *msg = NULL;
 378                msg = o->output_prefix(o, o->output_prefix_data);
 379                assert(msg);
 380                fwrite(msg->buf, msg->len, 1, file);
 381        }
 382
 383        if (len == 0) {
 384                has_trailing_newline = (first == '\n');
 385                has_trailing_carriage_return = (!has_trailing_newline &&
 386                                                (first == '\r'));
 387                nofirst = has_trailing_newline || has_trailing_carriage_return;
 388        } else {
 389                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 390                if (has_trailing_newline)
 391                        len--;
 392                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 393                if (has_trailing_carriage_return)
 394                        len--;
 395                nofirst = 0;
 396        }
 397
 398        if (len || !nofirst) {
 399                fputs(set, file);
 400                if (!nofirst)
 401                        fputc(first, file);
 402                fwrite(line, len, 1, file);
 403                fputs(reset, file);
 404        }
 405        if (has_trailing_carriage_return)
 406                fputc('\r', file);
 407        if (has_trailing_newline)
 408                fputc('\n', file);
 409}
 410
 411static void emit_line(struct diff_options *o, const char *set, const char *reset,
 412                      const char *line, int len)
 413{
 414        emit_line_0(o, set, reset, line[0], line+1, len-1);
 415}
 416
 417static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 418{
 419        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 420              ecbdata->blank_at_eof_in_preimage &&
 421              ecbdata->blank_at_eof_in_postimage &&
 422              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 423              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 424                return 0;
 425        return ws_blank_line(line, len, ecbdata->ws_rule);
 426}
 427
 428static void emit_add_line(const char *reset,
 429                          struct emit_callback *ecbdata,
 430                          const char *line, int len)
 431{
 432        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 433        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 434
 435        if (!*ws)
 436                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 437        else if (new_blank_line_at_eof(ecbdata, line, len))
 438                /* Blank line at EOF - paint '+' as well */
 439                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 440        else {
 441                /* Emit just the prefix, then the rest. */
 442                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 443                ws_check_emit(line, len, ecbdata->ws_rule,
 444                              ecbdata->opt->file, set, reset, ws);
 445        }
 446}
 447
 448static void emit_hunk_header(struct emit_callback *ecbdata,
 449                             const char *line, int len)
 450{
 451        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 452        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 453        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 454        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 455        static const char atat[2] = { '@', '@' };
 456        const char *cp, *ep;
 457        struct strbuf msgbuf = STRBUF_INIT;
 458        int org_len = len;
 459        int i = 1;
 460
 461        /*
 462         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 463         * it always is at least 10 bytes long.
 464         */
 465        if (len < 10 ||
 466            memcmp(line, atat, 2) ||
 467            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 468                emit_line(ecbdata->opt, plain, reset, line, len);
 469                return;
 470        }
 471        ep += 2; /* skip over @@ */
 472
 473        /* The hunk header in fraginfo color */
 474        strbuf_add(&msgbuf, frag, strlen(frag));
 475        strbuf_add(&msgbuf, line, ep - line);
 476        strbuf_add(&msgbuf, reset, strlen(reset));
 477
 478        /*
 479         * trailing "\r\n"
 480         */
 481        for ( ; i < 3; i++)
 482                if (line[len - i] == '\r' || line[len - i] == '\n')
 483                        len--;
 484
 485        /* blank before the func header */
 486        for (cp = ep; ep - line < len; ep++)
 487                if (*ep != ' ' && *ep != '\t')
 488                        break;
 489        if (ep != cp) {
 490                strbuf_add(&msgbuf, plain, strlen(plain));
 491                strbuf_add(&msgbuf, cp, ep - cp);
 492                strbuf_add(&msgbuf, reset, strlen(reset));
 493        }
 494
 495        if (ep < line + len) {
 496                strbuf_add(&msgbuf, func, strlen(func));
 497                strbuf_add(&msgbuf, ep, line + len - ep);
 498                strbuf_add(&msgbuf, reset, strlen(reset));
 499        }
 500
 501        strbuf_add(&msgbuf, line + len, org_len - len);
 502        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 503        strbuf_release(&msgbuf);
 504}
 505
 506static struct diff_tempfile *claim_diff_tempfile(void) {
 507        int i;
 508        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 509                if (!diff_temp[i].name)
 510                        return diff_temp + i;
 511        die("BUG: diff is failing to clean up its tempfiles");
 512}
 513
 514static int remove_tempfile_installed;
 515
 516static void remove_tempfile(void)
 517{
 518        int i;
 519        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 520                if (diff_temp[i].name == diff_temp[i].tmp_path)
 521                        unlink_or_warn(diff_temp[i].name);
 522                diff_temp[i].name = NULL;
 523        }
 524}
 525
 526static void remove_tempfile_on_signal(int signo)
 527{
 528        remove_tempfile();
 529        sigchain_pop(signo);
 530        raise(signo);
 531}
 532
 533static void print_line_count(FILE *file, int count)
 534{
 535        switch (count) {
 536        case 0:
 537                fprintf(file, "0,0");
 538                break;
 539        case 1:
 540                fprintf(file, "1");
 541                break;
 542        default:
 543                fprintf(file, "1,%d", count);
 544                break;
 545        }
 546}
 547
 548static void emit_rewrite_lines(struct emit_callback *ecb,
 549                               int prefix, const char *data, int size)
 550{
 551        const char *endp = NULL;
 552        static const char *nneof = " No newline at end of file\n";
 553        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 554        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 555
 556        while (0 < size) {
 557                int len;
 558
 559                endp = memchr(data, '\n', size);
 560                len = endp ? (endp - data + 1) : size;
 561                if (prefix != '+') {
 562                        ecb->lno_in_preimage++;
 563                        emit_line_0(ecb->opt, old, reset, '-',
 564                                    data, len);
 565                } else {
 566                        ecb->lno_in_postimage++;
 567                        emit_add_line(reset, ecb, data, len);
 568                }
 569                size -= len;
 570                data += len;
 571        }
 572        if (!endp) {
 573                const char *plain = diff_get_color(ecb->color_diff,
 574                                                   DIFF_PLAIN);
 575                emit_line_0(ecb->opt, plain, reset, '\\',
 576                            nneof, strlen(nneof));
 577        }
 578}
 579
 580static void emit_rewrite_diff(const char *name_a,
 581                              const char *name_b,
 582                              struct diff_filespec *one,
 583                              struct diff_filespec *two,
 584                              struct userdiff_driver *textconv_one,
 585                              struct userdiff_driver *textconv_two,
 586                              struct diff_options *o)
 587{
 588        int lc_a, lc_b;
 589        const char *name_a_tab, *name_b_tab;
 590        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 591        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 592        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 593        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 594        const char *a_prefix, *b_prefix;
 595        char *data_one, *data_two;
 596        size_t size_one, size_two;
 597        struct emit_callback ecbdata;
 598        char *line_prefix = "";
 599        struct strbuf *msgbuf;
 600
 601        if (o && o->output_prefix) {
 602                msgbuf = o->output_prefix(o, o->output_prefix_data);
 603                line_prefix = msgbuf->buf;
 604        }
 605
 606        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 607                a_prefix = o->b_prefix;
 608                b_prefix = o->a_prefix;
 609        } else {
 610                a_prefix = o->a_prefix;
 611                b_prefix = o->b_prefix;
 612        }
 613
 614        name_a += (*name_a == '/');
 615        name_b += (*name_b == '/');
 616        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 617        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 618
 619        strbuf_reset(&a_name);
 620        strbuf_reset(&b_name);
 621        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 622        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 623
 624        size_one = fill_textconv(textconv_one, one, &data_one);
 625        size_two = fill_textconv(textconv_two, two, &data_two);
 626
 627        memset(&ecbdata, 0, sizeof(ecbdata));
 628        ecbdata.color_diff = want_color(o->use_color);
 629        ecbdata.found_changesp = &o->found_changes;
 630        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 631        ecbdata.opt = o;
 632        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 633                mmfile_t mf1, mf2;
 634                mf1.ptr = (char *)data_one;
 635                mf2.ptr = (char *)data_two;
 636                mf1.size = size_one;
 637                mf2.size = size_two;
 638                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 639        }
 640        ecbdata.lno_in_preimage = 1;
 641        ecbdata.lno_in_postimage = 1;
 642
 643        lc_a = count_lines(data_one, size_one);
 644        lc_b = count_lines(data_two, size_two);
 645        fprintf(o->file,
 646                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 647                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 648                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 649                line_prefix, fraginfo);
 650        if (!o->irreversible_delete)
 651                print_line_count(o->file, lc_a);
 652        else
 653                fprintf(o->file, "?,?");
 654        fprintf(o->file, " +");
 655        print_line_count(o->file, lc_b);
 656        fprintf(o->file, " @@%s\n", reset);
 657        if (lc_a && !o->irreversible_delete)
 658                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 659        if (lc_b)
 660                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 661        if (textconv_one)
 662                free((char *)data_one);
 663        if (textconv_two)
 664                free((char *)data_two);
 665}
 666
 667struct diff_words_buffer {
 668        mmfile_t text;
 669        long alloc;
 670        struct diff_words_orig {
 671                const char *begin, *end;
 672        } *orig;
 673        int orig_nr, orig_alloc;
 674};
 675
 676static void diff_words_append(char *line, unsigned long len,
 677                struct diff_words_buffer *buffer)
 678{
 679        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 680        line++;
 681        len--;
 682        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 683        buffer->text.size += len;
 684        buffer->text.ptr[buffer->text.size] = '\0';
 685}
 686
 687struct diff_words_style_elem {
 688        const char *prefix;
 689        const char *suffix;
 690        const char *color; /* NULL; filled in by the setup code if
 691                            * color is enabled */
 692};
 693
 694struct diff_words_style {
 695        enum diff_words_type type;
 696        struct diff_words_style_elem new, old, ctx;
 697        const char *newline;
 698};
 699
 700static struct diff_words_style diff_words_styles[] = {
 701        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 702        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 703        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 704};
 705
 706struct diff_words_data {
 707        struct diff_words_buffer minus, plus;
 708        const char *current_plus;
 709        int last_minus;
 710        struct diff_options *opt;
 711        regex_t *word_regex;
 712        enum diff_words_type type;
 713        struct diff_words_style *style;
 714};
 715
 716static int fn_out_diff_words_write_helper(FILE *fp,
 717                                          struct diff_words_style_elem *st_el,
 718                                          const char *newline,
 719                                          size_t count, const char *buf,
 720                                          const char *line_prefix)
 721{
 722        int print = 0;
 723
 724        while (count) {
 725                char *p = memchr(buf, '\n', count);
 726                if (print)
 727                        fputs(line_prefix, fp);
 728                if (p != buf) {
 729                        if (st_el->color && fputs(st_el->color, fp) < 0)
 730                                return -1;
 731                        if (fputs(st_el->prefix, fp) < 0 ||
 732                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 733                            fputs(st_el->suffix, fp) < 0)
 734                                return -1;
 735                        if (st_el->color && *st_el->color
 736                            && fputs(GIT_COLOR_RESET, fp) < 0)
 737                                return -1;
 738                }
 739                if (!p)
 740                        return 0;
 741                if (fputs(newline, fp) < 0)
 742                        return -1;
 743                count -= p + 1 - buf;
 744                buf = p + 1;
 745                print = 1;
 746        }
 747        return 0;
 748}
 749
 750/*
 751 * '--color-words' algorithm can be described as:
 752 *
 753 *   1. collect a the minus/plus lines of a diff hunk, divided into
 754 *      minus-lines and plus-lines;
 755 *
 756 *   2. break both minus-lines and plus-lines into words and
 757 *      place them into two mmfile_t with one word for each line;
 758 *
 759 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 760 *
 761 * And for the common parts of the both file, we output the plus side text.
 762 * diff_words->current_plus is used to trace the current position of the plus file
 763 * which printed. diff_words->last_minus is used to trace the last minus word
 764 * printed.
 765 *
 766 * For '--graph' to work with '--color-words', we need to output the graph prefix
 767 * on each line of color words output. Generally, there are two conditions on
 768 * which we should output the prefix.
 769 *
 770 *   1. diff_words->last_minus == 0 &&
 771 *      diff_words->current_plus == diff_words->plus.text.ptr
 772 *
 773 *      that is: the plus text must start as a new line, and if there is no minus
 774 *      word printed, a graph prefix must be printed.
 775 *
 776 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 777 *      *(diff_words->current_plus - 1) == '\n'
 778 *
 779 *      that is: a graph prefix must be printed following a '\n'
 780 */
 781static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 782{
 783        if ((diff_words->last_minus == 0 &&
 784                diff_words->current_plus == diff_words->plus.text.ptr) ||
 785                (diff_words->current_plus > diff_words->plus.text.ptr &&
 786                *(diff_words->current_plus - 1) == '\n')) {
 787                return 1;
 788        } else {
 789                return 0;
 790        }
 791}
 792
 793static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 794{
 795        struct diff_words_data *diff_words = priv;
 796        struct diff_words_style *style = diff_words->style;
 797        int minus_first, minus_len, plus_first, plus_len;
 798        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 799        struct diff_options *opt = diff_words->opt;
 800        struct strbuf *msgbuf;
 801        char *line_prefix = "";
 802
 803        if (line[0] != '@' || parse_hunk_header(line, len,
 804                        &minus_first, &minus_len, &plus_first, &plus_len))
 805                return;
 806
 807        assert(opt);
 808        if (opt->output_prefix) {
 809                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 810                line_prefix = msgbuf->buf;
 811        }
 812
 813        /* POSIX requires that first be decremented by one if len == 0... */
 814        if (minus_len) {
 815                minus_begin = diff_words->minus.orig[minus_first].begin;
 816                minus_end =
 817                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 818        } else
 819                minus_begin = minus_end =
 820                        diff_words->minus.orig[minus_first].end;
 821
 822        if (plus_len) {
 823                plus_begin = diff_words->plus.orig[plus_first].begin;
 824                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 825        } else
 826                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 827
 828        if (color_words_output_graph_prefix(diff_words)) {
 829                fputs(line_prefix, diff_words->opt->file);
 830        }
 831        if (diff_words->current_plus != plus_begin) {
 832                fn_out_diff_words_write_helper(diff_words->opt->file,
 833                                &style->ctx, style->newline,
 834                                plus_begin - diff_words->current_plus,
 835                                diff_words->current_plus, line_prefix);
 836                if (*(plus_begin - 1) == '\n')
 837                        fputs(line_prefix, diff_words->opt->file);
 838        }
 839        if (minus_begin != minus_end) {
 840                fn_out_diff_words_write_helper(diff_words->opt->file,
 841                                &style->old, style->newline,
 842                                minus_end - minus_begin, minus_begin,
 843                                line_prefix);
 844        }
 845        if (plus_begin != plus_end) {
 846                fn_out_diff_words_write_helper(diff_words->opt->file,
 847                                &style->new, style->newline,
 848                                plus_end - plus_begin, plus_begin,
 849                                line_prefix);
 850        }
 851
 852        diff_words->current_plus = plus_end;
 853        diff_words->last_minus = minus_first;
 854}
 855
 856/* This function starts looking at *begin, and returns 0 iff a word was found. */
 857static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 858                int *begin, int *end)
 859{
 860        if (word_regex && *begin < buffer->size) {
 861                regmatch_t match[1];
 862                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 863                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 864                                        '\n', match[0].rm_eo - match[0].rm_so);
 865                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 866                        *begin += match[0].rm_so;
 867                        return *begin >= *end;
 868                }
 869                return -1;
 870        }
 871
 872        /* find the next word */
 873        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 874                (*begin)++;
 875        if (*begin >= buffer->size)
 876                return -1;
 877
 878        /* find the end of the word */
 879        *end = *begin + 1;
 880        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 881                (*end)++;
 882
 883        return 0;
 884}
 885
 886/*
 887 * This function splits the words in buffer->text, stores the list with
 888 * newline separator into out, and saves the offsets of the original words
 889 * in buffer->orig.
 890 */
 891static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 892                regex_t *word_regex)
 893{
 894        int i, j;
 895        long alloc = 0;
 896
 897        out->size = 0;
 898        out->ptr = NULL;
 899
 900        /* fake an empty "0th" word */
 901        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 902        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 903        buffer->orig_nr = 1;
 904
 905        for (i = 0; i < buffer->text.size; i++) {
 906                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 907                        return;
 908
 909                /* store original boundaries */
 910                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 911                                buffer->orig_alloc);
 912                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 913                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 914                buffer->orig_nr++;
 915
 916                /* store one word */
 917                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 918                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 919                out->ptr[out->size + j - i] = '\n';
 920                out->size += j - i + 1;
 921
 922                i = j - 1;
 923        }
 924}
 925
 926/* this executes the word diff on the accumulated buffers */
 927static void diff_words_show(struct diff_words_data *diff_words)
 928{
 929        xpparam_t xpp;
 930        xdemitconf_t xecfg;
 931        mmfile_t minus, plus;
 932        struct diff_words_style *style = diff_words->style;
 933
 934        struct diff_options *opt = diff_words->opt;
 935        struct strbuf *msgbuf;
 936        char *line_prefix = "";
 937
 938        assert(opt);
 939        if (opt->output_prefix) {
 940                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 941                line_prefix = msgbuf->buf;
 942        }
 943
 944        /* special case: only removal */
 945        if (!diff_words->plus.text.size) {
 946                fputs(line_prefix, diff_words->opt->file);
 947                fn_out_diff_words_write_helper(diff_words->opt->file,
 948                        &style->old, style->newline,
 949                        diff_words->minus.text.size,
 950                        diff_words->minus.text.ptr, line_prefix);
 951                diff_words->minus.text.size = 0;
 952                return;
 953        }
 954
 955        diff_words->current_plus = diff_words->plus.text.ptr;
 956        diff_words->last_minus = 0;
 957
 958        memset(&xpp, 0, sizeof(xpp));
 959        memset(&xecfg, 0, sizeof(xecfg));
 960        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 961        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 962        xpp.flags = 0;
 963        /* as only the hunk header will be parsed, we need a 0-context */
 964        xecfg.ctxlen = 0;
 965        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 966                      &xpp, &xecfg);
 967        free(minus.ptr);
 968        free(plus.ptr);
 969        if (diff_words->current_plus != diff_words->plus.text.ptr +
 970                        diff_words->plus.text.size) {
 971                if (color_words_output_graph_prefix(diff_words))
 972                        fputs(line_prefix, diff_words->opt->file);
 973                fn_out_diff_words_write_helper(diff_words->opt->file,
 974                        &style->ctx, style->newline,
 975                        diff_words->plus.text.ptr + diff_words->plus.text.size
 976                        - diff_words->current_plus, diff_words->current_plus,
 977                        line_prefix);
 978        }
 979        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 980}
 981
 982/* In "color-words" mode, show word-diff of words accumulated in the buffer */
 983static void diff_words_flush(struct emit_callback *ecbdata)
 984{
 985        if (ecbdata->diff_words->minus.text.size ||
 986            ecbdata->diff_words->plus.text.size)
 987                diff_words_show(ecbdata->diff_words);
 988}
 989
 990static void diff_filespec_load_driver(struct diff_filespec *one)
 991{
 992        /* Use already-loaded driver */
 993        if (one->driver)
 994                return;
 995
 996        if (S_ISREG(one->mode))
 997                one->driver = userdiff_find_by_path(one->path);
 998
 999        /* Fallback to default settings */
1000        if (!one->driver)
1001                one->driver = userdiff_find_by_name("default");
1002}
1003
1004static const char *userdiff_word_regex(struct diff_filespec *one)
1005{
1006        diff_filespec_load_driver(one);
1007        return one->driver->word_regex;
1008}
1009
1010static void init_diff_words_data(struct emit_callback *ecbdata,
1011                                 struct diff_options *o,
1012                                 struct diff_filespec *one,
1013                                 struct diff_filespec *two)
1014{
1015        int i;
1016
1017        ecbdata->diff_words =
1018                xcalloc(1, sizeof(struct diff_words_data));
1019        ecbdata->diff_words->type = o->word_diff;
1020        ecbdata->diff_words->opt = o;
1021        if (!o->word_regex)
1022                o->word_regex = userdiff_word_regex(one);
1023        if (!o->word_regex)
1024                o->word_regex = userdiff_word_regex(two);
1025        if (!o->word_regex)
1026                o->word_regex = diff_word_regex_cfg;
1027        if (o->word_regex) {
1028                ecbdata->diff_words->word_regex = (regex_t *)
1029                        xmalloc(sizeof(regex_t));
1030                if (regcomp(ecbdata->diff_words->word_regex,
1031                            o->word_regex,
1032                            REG_EXTENDED | REG_NEWLINE))
1033                        die ("Invalid regular expression: %s",
1034                             o->word_regex);
1035        }
1036        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1037                if (o->word_diff == diff_words_styles[i].type) {
1038                        ecbdata->diff_words->style =
1039                                &diff_words_styles[i];
1040                        break;
1041                }
1042        }
1043        if (want_color(o->use_color)) {
1044                struct diff_words_style *st = ecbdata->diff_words->style;
1045                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1046                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1047                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1048        }
1049}
1050
1051static void free_diff_words_data(struct emit_callback *ecbdata)
1052{
1053        if (ecbdata->diff_words) {
1054                diff_words_flush(ecbdata);
1055                free (ecbdata->diff_words->minus.text.ptr);
1056                free (ecbdata->diff_words->minus.orig);
1057                free (ecbdata->diff_words->plus.text.ptr);
1058                free (ecbdata->diff_words->plus.orig);
1059                if (ecbdata->diff_words->word_regex) {
1060                        regfree(ecbdata->diff_words->word_regex);
1061                        free(ecbdata->diff_words->word_regex);
1062                }
1063                free(ecbdata->diff_words);
1064                ecbdata->diff_words = NULL;
1065        }
1066}
1067
1068const char *diff_get_color(int diff_use_color, enum color_diff ix)
1069{
1070        if (want_color(diff_use_color))
1071                return diff_colors[ix];
1072        return "";
1073}
1074
1075static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1076{
1077        const char *cp;
1078        unsigned long allot;
1079        size_t l = len;
1080
1081        if (ecb->truncate)
1082                return ecb->truncate(line, len);
1083        cp = line;
1084        allot = l;
1085        while (0 < l) {
1086                (void) utf8_width(&cp, &l);
1087                if (!cp)
1088                        break; /* truncated in the middle? */
1089        }
1090        return allot - l;
1091}
1092
1093static void find_lno(const char *line, struct emit_callback *ecbdata)
1094{
1095        const char *p;
1096        ecbdata->lno_in_preimage = 0;
1097        ecbdata->lno_in_postimage = 0;
1098        p = strchr(line, '-');
1099        if (!p)
1100                return; /* cannot happen */
1101        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1102        p = strchr(p, '+');
1103        if (!p)
1104                return; /* cannot happen */
1105        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1106}
1107
1108static void fn_out_consume(void *priv, char *line, unsigned long len)
1109{
1110        struct emit_callback *ecbdata = priv;
1111        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1112        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1113        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1114        struct diff_options *o = ecbdata->opt;
1115        char *line_prefix = "";
1116        struct strbuf *msgbuf;
1117
1118        if (o && o->output_prefix) {
1119                msgbuf = o->output_prefix(o, o->output_prefix_data);
1120                line_prefix = msgbuf->buf;
1121        }
1122
1123        if (ecbdata->header) {
1124                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1125                strbuf_reset(ecbdata->header);
1126                ecbdata->header = NULL;
1127        }
1128        *(ecbdata->found_changesp) = 1;
1129
1130        if (ecbdata->label_path[0]) {
1131                const char *name_a_tab, *name_b_tab;
1132
1133                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1134                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1135
1136                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1137                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1138                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1139                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1140                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1141        }
1142
1143        if (diff_suppress_blank_empty
1144            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1145                line[0] = '\n';
1146                len = 1;
1147        }
1148
1149        if (line[0] == '@') {
1150                if (ecbdata->diff_words)
1151                        diff_words_flush(ecbdata);
1152                len = sane_truncate_line(ecbdata, line, len);
1153                find_lno(line, ecbdata);
1154                emit_hunk_header(ecbdata, line, len);
1155                if (line[len-1] != '\n')
1156                        putc('\n', ecbdata->opt->file);
1157                return;
1158        }
1159
1160        if (len < 1) {
1161                emit_line(ecbdata->opt, reset, reset, line, len);
1162                if (ecbdata->diff_words
1163                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1164                        fputs("~\n", ecbdata->opt->file);
1165                return;
1166        }
1167
1168        if (ecbdata->diff_words) {
1169                if (line[0] == '-') {
1170                        diff_words_append(line, len,
1171                                          &ecbdata->diff_words->minus);
1172                        return;
1173                } else if (line[0] == '+') {
1174                        diff_words_append(line, len,
1175                                          &ecbdata->diff_words->plus);
1176                        return;
1177                } else if (!prefixcmp(line, "\\ ")) {
1178                        /*
1179                         * Eat the "no newline at eof" marker as if we
1180                         * saw a "+" or "-" line with nothing on it,
1181                         * and return without diff_words_flush() to
1182                         * defer processing. If this is the end of
1183                         * preimage, more "+" lines may come after it.
1184                         */
1185                        return;
1186                }
1187                diff_words_flush(ecbdata);
1188                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1189                        emit_line(ecbdata->opt, plain, reset, line, len);
1190                        fputs("~\n", ecbdata->opt->file);
1191                } else {
1192                        /*
1193                         * Skip the prefix character, if any.  With
1194                         * diff_suppress_blank_empty, there may be
1195                         * none.
1196                         */
1197                        if (line[0] != '\n') {
1198                              line++;
1199                              len--;
1200                        }
1201                        emit_line(ecbdata->opt, plain, reset, line, len);
1202                }
1203                return;
1204        }
1205
1206        if (line[0] != '+') {
1207                const char *color =
1208                        diff_get_color(ecbdata->color_diff,
1209                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1210                ecbdata->lno_in_preimage++;
1211                if (line[0] == ' ')
1212                        ecbdata->lno_in_postimage++;
1213                emit_line(ecbdata->opt, color, reset, line, len);
1214        } else {
1215                ecbdata->lno_in_postimage++;
1216                emit_add_line(reset, ecbdata, line + 1, len - 1);
1217        }
1218}
1219
1220static char *pprint_rename(const char *a, const char *b)
1221{
1222        const char *old = a;
1223        const char *new = b;
1224        struct strbuf name = STRBUF_INIT;
1225        int pfx_length, sfx_length;
1226        int len_a = strlen(a);
1227        int len_b = strlen(b);
1228        int a_midlen, b_midlen;
1229        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1230        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1231
1232        if (qlen_a || qlen_b) {
1233                quote_c_style(a, &name, NULL, 0);
1234                strbuf_addstr(&name, " => ");
1235                quote_c_style(b, &name, NULL, 0);
1236                return strbuf_detach(&name, NULL);
1237        }
1238
1239        /* Find common prefix */
1240        pfx_length = 0;
1241        while (*old && *new && *old == *new) {
1242                if (*old == '/')
1243                        pfx_length = old - a + 1;
1244                old++;
1245                new++;
1246        }
1247
1248        /* Find common suffix */
1249        old = a + len_a;
1250        new = b + len_b;
1251        sfx_length = 0;
1252        while (a <= old && b <= new && *old == *new) {
1253                if (*old == '/')
1254                        sfx_length = len_a - (old - a);
1255                old--;
1256                new--;
1257        }
1258
1259        /*
1260         * pfx{mid-a => mid-b}sfx
1261         * {pfx-a => pfx-b}sfx
1262         * pfx{sfx-a => sfx-b}
1263         * name-a => name-b
1264         */
1265        a_midlen = len_a - pfx_length - sfx_length;
1266        b_midlen = len_b - pfx_length - sfx_length;
1267        if (a_midlen < 0)
1268                a_midlen = 0;
1269        if (b_midlen < 0)
1270                b_midlen = 0;
1271
1272        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1273        if (pfx_length + sfx_length) {
1274                strbuf_add(&name, a, pfx_length);
1275                strbuf_addch(&name, '{');
1276        }
1277        strbuf_add(&name, a + pfx_length, a_midlen);
1278        strbuf_addstr(&name, " => ");
1279        strbuf_add(&name, b + pfx_length, b_midlen);
1280        if (pfx_length + sfx_length) {
1281                strbuf_addch(&name, '}');
1282                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1283        }
1284        return strbuf_detach(&name, NULL);
1285}
1286
1287struct diffstat_t {
1288        int nr;
1289        int alloc;
1290        struct diffstat_file {
1291                char *from_name;
1292                char *name;
1293                char *print_name;
1294                unsigned is_unmerged:1;
1295                unsigned is_binary:1;
1296                unsigned is_renamed:1;
1297                uintmax_t added, deleted;
1298        } **files;
1299};
1300
1301static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1302                                          const char *name_a,
1303                                          const char *name_b)
1304{
1305        struct diffstat_file *x;
1306        x = xcalloc(sizeof (*x), 1);
1307        if (diffstat->nr == diffstat->alloc) {
1308                diffstat->alloc = alloc_nr(diffstat->alloc);
1309                diffstat->files = xrealloc(diffstat->files,
1310                                diffstat->alloc * sizeof(x));
1311        }
1312        diffstat->files[diffstat->nr++] = x;
1313        if (name_b) {
1314                x->from_name = xstrdup(name_a);
1315                x->name = xstrdup(name_b);
1316                x->is_renamed = 1;
1317        }
1318        else {
1319                x->from_name = NULL;
1320                x->name = xstrdup(name_a);
1321        }
1322        return x;
1323}
1324
1325static void diffstat_consume(void *priv, char *line, unsigned long len)
1326{
1327        struct diffstat_t *diffstat = priv;
1328        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1329
1330        if (line[0] == '+')
1331                x->added++;
1332        else if (line[0] == '-')
1333                x->deleted++;
1334}
1335
1336const char mime_boundary_leader[] = "------------";
1337
1338static int scale_linear(int it, int width, int max_change)
1339{
1340        if (!it)
1341                return 0;
1342        /*
1343         * make sure that at least one '-' or '+' is printed if
1344         * there is any change to this path. The easiest way is to
1345         * scale linearly as if the alloted width is one column shorter
1346         * than it is, and then add 1 to the result.
1347         */
1348        return 1 + (it * (width - 1) / max_change);
1349}
1350
1351static void show_name(FILE *file,
1352                      const char *prefix, const char *name, int len)
1353{
1354        fprintf(file, " %s%-*s |", prefix, len, name);
1355}
1356
1357static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1358{
1359        if (cnt <= 0)
1360                return;
1361        fprintf(file, "%s", set);
1362        while (cnt--)
1363                putc(ch, file);
1364        fprintf(file, "%s", reset);
1365}
1366
1367static void fill_print_name(struct diffstat_file *file)
1368{
1369        char *pname;
1370
1371        if (file->print_name)
1372                return;
1373
1374        if (!file->is_renamed) {
1375                struct strbuf buf = STRBUF_INIT;
1376                if (quote_c_style(file->name, &buf, NULL, 0)) {
1377                        pname = strbuf_detach(&buf, NULL);
1378                } else {
1379                        pname = file->name;
1380                        strbuf_release(&buf);
1381                }
1382        } else {
1383                pname = pprint_rename(file->from_name, file->name);
1384        }
1385        file->print_name = pname;
1386}
1387
1388int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1389{
1390        struct strbuf sb = STRBUF_INIT;
1391        int ret;
1392
1393        if (!files) {
1394                assert(insertions == 0 && deletions == 0);
1395                return fputs(_(" 0 files changed\n"), fp);
1396        }
1397
1398        strbuf_addf(&sb,
1399                    Q_(" %d file changed", " %d files changed", files),
1400                    files);
1401
1402        /*
1403         * For binary diff, the caller may want to print "x files
1404         * changed" with insertions == 0 && deletions == 0.
1405         *
1406         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1407         * is probably less confusing (i.e skip over "2 files changed
1408         * but nothing about added/removed lines? Is this a bug in Git?").
1409         */
1410        if (insertions || deletions == 0) {
1411                /*
1412                 * TRANSLATORS: "+" in (+) is a line addition marker;
1413                 * do not translate it.
1414                 */
1415                strbuf_addf(&sb,
1416                            Q_(", %d insertion(+)", ", %d insertions(+)",
1417                               insertions),
1418                            insertions);
1419        }
1420
1421        if (deletions || insertions == 0) {
1422                /*
1423                 * TRANSLATORS: "-" in (-) is a line removal marker;
1424                 * do not translate it.
1425                 */
1426                strbuf_addf(&sb,
1427                            Q_(", %d deletion(-)", ", %d deletions(-)",
1428                               deletions),
1429                            deletions);
1430        }
1431        strbuf_addch(&sb, '\n');
1432        ret = fputs(sb.buf, fp);
1433        strbuf_release(&sb);
1434        return ret;
1435}
1436
1437static void show_stats(struct diffstat_t *data, struct diff_options *options)
1438{
1439        int i, len, add, del, adds = 0, dels = 0;
1440        uintmax_t max_change = 0, max_len = 0;
1441        int total_files = data->nr;
1442        int width, name_width, count;
1443        const char *reset, *add_c, *del_c;
1444        const char *line_prefix = "";
1445        int extra_shown = 0;
1446        struct strbuf *msg = NULL;
1447
1448        if (data->nr == 0)
1449                return;
1450
1451        if (options->output_prefix) {
1452                msg = options->output_prefix(options, options->output_prefix_data);
1453                line_prefix = msg->buf;
1454        }
1455
1456        width = options->stat_width ? options->stat_width : 80;
1457        name_width = options->stat_name_width ? options->stat_name_width : 50;
1458        count = options->stat_count ? options->stat_count : data->nr;
1459
1460        /* Sanity: give at least 5 columns to the graph,
1461         * but leave at least 10 columns for the name.
1462         */
1463        if (width < 25)
1464                width = 25;
1465        if (name_width < 10)
1466                name_width = 10;
1467        else if (width < name_width + 15)
1468                name_width = width - 15;
1469
1470        /* Find the longest filename and max number of changes */
1471        reset = diff_get_color_opt(options, DIFF_RESET);
1472        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1473        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1474
1475        for (i = 0; (i < count) && (i < data->nr); i++) {
1476                struct diffstat_file *file = data->files[i];
1477                uintmax_t change = file->added + file->deleted;
1478                if (!data->files[i]->is_renamed &&
1479                         (change == 0)) {
1480                        count++; /* not shown == room for one more */
1481                        continue;
1482                }
1483                fill_print_name(file);
1484                len = strlen(file->print_name);
1485                if (max_len < len)
1486                        max_len = len;
1487
1488                if (file->is_binary || file->is_unmerged)
1489                        continue;
1490                if (max_change < change)
1491                        max_change = change;
1492        }
1493        count = i; /* min(count, data->nr) */
1494
1495        /* Compute the width of the graph part;
1496         * 10 is for one blank at the beginning of the line plus
1497         * " | count " between the name and the graph.
1498         *
1499         * From here on, name_width is the width of the name area,
1500         * and width is the width of the graph area.
1501         */
1502        name_width = (name_width < max_len) ? name_width : max_len;
1503        if (width < (name_width + 10) + max_change)
1504                width = width - (name_width + 10);
1505        else
1506                width = max_change;
1507
1508        for (i = 0; i < count; i++) {
1509                const char *prefix = "";
1510                char *name = data->files[i]->print_name;
1511                uintmax_t added = data->files[i]->added;
1512                uintmax_t deleted = data->files[i]->deleted;
1513                int name_len;
1514
1515                if (!data->files[i]->is_renamed &&
1516                         (added + deleted == 0)) {
1517                        total_files--;
1518                        continue;
1519                }
1520                /*
1521                 * "scale" the filename
1522                 */
1523                len = name_width;
1524                name_len = strlen(name);
1525                if (name_width < name_len) {
1526                        char *slash;
1527                        prefix = "...";
1528                        len -= 3;
1529                        name += name_len - len;
1530                        slash = strchr(name, '/');
1531                        if (slash)
1532                                name = slash;
1533                }
1534
1535                if (data->files[i]->is_binary) {
1536                        fprintf(options->file, "%s", line_prefix);
1537                        show_name(options->file, prefix, name, len);
1538                        fprintf(options->file, "  Bin ");
1539                        fprintf(options->file, "%s%"PRIuMAX"%s",
1540                                del_c, deleted, reset);
1541                        fprintf(options->file, " -> ");
1542                        fprintf(options->file, "%s%"PRIuMAX"%s",
1543                                add_c, added, reset);
1544                        fprintf(options->file, " bytes");
1545                        fprintf(options->file, "\n");
1546                        continue;
1547                }
1548                else if (data->files[i]->is_unmerged) {
1549                        fprintf(options->file, "%s", line_prefix);
1550                        show_name(options->file, prefix, name, len);
1551                        fprintf(options->file, "  Unmerged\n");
1552                        continue;
1553                }
1554
1555                /*
1556                 * scale the add/delete
1557                 */
1558                add = added;
1559                del = deleted;
1560                adds += add;
1561                dels += del;
1562
1563                if (width <= max_change) {
1564                        int total = add + del;
1565
1566                        total = scale_linear(add + del, width, max_change);
1567                        if (total < 2 && add && del)
1568                                /* width >= 2 due to the sanity check */
1569                                total = 2;
1570                        if (add < del) {
1571                                add = scale_linear(add, width, max_change);
1572                                del = total - add;
1573                        } else {
1574                                del = scale_linear(del, width, max_change);
1575                                add = total - del;
1576                        }
1577                }
1578                fprintf(options->file, "%s", line_prefix);
1579                show_name(options->file, prefix, name, len);
1580                fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1581                                added + deleted ? " " : "");
1582                show_graph(options->file, '+', add, add_c, reset);
1583                show_graph(options->file, '-', del, del_c, reset);
1584                fprintf(options->file, "\n");
1585        }
1586        for (i = count; i < data->nr; i++) {
1587                uintmax_t added = data->files[i]->added;
1588                uintmax_t deleted = data->files[i]->deleted;
1589                if (!data->files[i]->is_renamed &&
1590                         (added + deleted == 0)) {
1591                        total_files--;
1592                        continue;
1593                }
1594                adds += added;
1595                dels += deleted;
1596                if (!extra_shown)
1597                        fprintf(options->file, "%s ...\n", line_prefix);
1598                extra_shown = 1;
1599        }
1600        fprintf(options->file, "%s", line_prefix);
1601        print_stat_summary(options->file, total_files, adds, dels);
1602}
1603
1604static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1605{
1606        int i, adds = 0, dels = 0, total_files = data->nr;
1607
1608        if (data->nr == 0)
1609                return;
1610
1611        for (i = 0; i < data->nr; i++) {
1612                if (!data->files[i]->is_binary &&
1613                    !data->files[i]->is_unmerged) {
1614                        int added = data->files[i]->added;
1615                        int deleted= data->files[i]->deleted;
1616                        if (!data->files[i]->is_renamed &&
1617                            (added + deleted == 0)) {
1618                                total_files--;
1619                        } else {
1620                                adds += added;
1621                                dels += deleted;
1622                        }
1623                }
1624        }
1625        if (options->output_prefix) {
1626                struct strbuf *msg = NULL;
1627                msg = options->output_prefix(options,
1628                                options->output_prefix_data);
1629                fprintf(options->file, "%s", msg->buf);
1630        }
1631        print_stat_summary(options->file, total_files, adds, dels);
1632}
1633
1634static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1635{
1636        int i;
1637
1638        if (data->nr == 0)
1639                return;
1640
1641        for (i = 0; i < data->nr; i++) {
1642                struct diffstat_file *file = data->files[i];
1643
1644                if (options->output_prefix) {
1645                        struct strbuf *msg = NULL;
1646                        msg = options->output_prefix(options,
1647                                        options->output_prefix_data);
1648                        fprintf(options->file, "%s", msg->buf);
1649                }
1650
1651                if (file->is_binary)
1652                        fprintf(options->file, "-\t-\t");
1653                else
1654                        fprintf(options->file,
1655                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1656                                file->added, file->deleted);
1657                if (options->line_termination) {
1658                        fill_print_name(file);
1659                        if (!file->is_renamed)
1660                                write_name_quoted(file->name, options->file,
1661                                                  options->line_termination);
1662                        else {
1663                                fputs(file->print_name, options->file);
1664                                putc(options->line_termination, options->file);
1665                        }
1666                } else {
1667                        if (file->is_renamed) {
1668                                putc('\0', options->file);
1669                                write_name_quoted(file->from_name, options->file, '\0');
1670                        }
1671                        write_name_quoted(file->name, options->file, '\0');
1672                }
1673        }
1674}
1675
1676struct dirstat_file {
1677        const char *name;
1678        unsigned long changed;
1679};
1680
1681struct dirstat_dir {
1682        struct dirstat_file *files;
1683        int alloc, nr, permille, cumulative;
1684};
1685
1686static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1687                unsigned long changed, const char *base, int baselen)
1688{
1689        unsigned long this_dir = 0;
1690        unsigned int sources = 0;
1691        const char *line_prefix = "";
1692        struct strbuf *msg = NULL;
1693
1694        if (opt->output_prefix) {
1695                msg = opt->output_prefix(opt, opt->output_prefix_data);
1696                line_prefix = msg->buf;
1697        }
1698
1699        while (dir->nr) {
1700                struct dirstat_file *f = dir->files;
1701                int namelen = strlen(f->name);
1702                unsigned long this;
1703                char *slash;
1704
1705                if (namelen < baselen)
1706                        break;
1707                if (memcmp(f->name, base, baselen))
1708                        break;
1709                slash = strchr(f->name + baselen, '/');
1710                if (slash) {
1711                        int newbaselen = slash + 1 - f->name;
1712                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1713                        sources++;
1714                } else {
1715                        this = f->changed;
1716                        dir->files++;
1717                        dir->nr--;
1718                        sources += 2;
1719                }
1720                this_dir += this;
1721        }
1722
1723        /*
1724         * We don't report dirstat's for
1725         *  - the top level
1726         *  - or cases where everything came from a single directory
1727         *    under this directory (sources == 1).
1728         */
1729        if (baselen && sources != 1) {
1730                if (this_dir) {
1731                        int permille = this_dir * 1000 / changed;
1732                        if (permille >= dir->permille) {
1733                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1734                                        permille / 10, permille % 10, baselen, base);
1735                                if (!dir->cumulative)
1736                                        return 0;
1737                        }
1738                }
1739        }
1740        return this_dir;
1741}
1742
1743static int dirstat_compare(const void *_a, const void *_b)
1744{
1745        const struct dirstat_file *a = _a;
1746        const struct dirstat_file *b = _b;
1747        return strcmp(a->name, b->name);
1748}
1749
1750static void show_dirstat(struct diff_options *options)
1751{
1752        int i;
1753        unsigned long changed;
1754        struct dirstat_dir dir;
1755        struct diff_queue_struct *q = &diff_queued_diff;
1756
1757        dir.files = NULL;
1758        dir.alloc = 0;
1759        dir.nr = 0;
1760        dir.permille = options->dirstat_permille;
1761        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1762
1763        changed = 0;
1764        for (i = 0; i < q->nr; i++) {
1765                struct diff_filepair *p = q->queue[i];
1766                const char *name;
1767                unsigned long copied, added, damage;
1768                int content_changed;
1769
1770                name = p->two->path ? p->two->path : p->one->path;
1771
1772                if (p->one->sha1_valid && p->two->sha1_valid)
1773                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1774                else
1775                        content_changed = 1;
1776
1777                if (!content_changed) {
1778                        /*
1779                         * The SHA1 has not changed, so pre-/post-content is
1780                         * identical. We can therefore skip looking at the
1781                         * file contents altogether.
1782                         */
1783                        damage = 0;
1784                        goto found_damage;
1785                }
1786
1787                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1788                        /*
1789                         * In --dirstat-by-file mode, we don't really need to
1790                         * look at the actual file contents at all.
1791                         * The fact that the SHA1 changed is enough for us to
1792                         * add this file to the list of results
1793                         * (with each file contributing equal damage).
1794                         */
1795                        damage = 1;
1796                        goto found_damage;
1797                }
1798
1799                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1800                        diff_populate_filespec(p->one, 0);
1801                        diff_populate_filespec(p->two, 0);
1802                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1803                                               &copied, &added);
1804                        diff_free_filespec_data(p->one);
1805                        diff_free_filespec_data(p->two);
1806                } else if (DIFF_FILE_VALID(p->one)) {
1807                        diff_populate_filespec(p->one, 1);
1808                        copied = added = 0;
1809                        diff_free_filespec_data(p->one);
1810                } else if (DIFF_FILE_VALID(p->two)) {
1811                        diff_populate_filespec(p->two, 1);
1812                        copied = 0;
1813                        added = p->two->size;
1814                        diff_free_filespec_data(p->two);
1815                } else
1816                        continue;
1817
1818                /*
1819                 * Original minus copied is the removed material,
1820                 * added is the new material.  They are both damages
1821                 * made to the preimage.
1822                 * If the resulting damage is zero, we know that
1823                 * diffcore_count_changes() considers the two entries to
1824                 * be identical, but since content_changed is true, we
1825                 * know that there must have been _some_ kind of change,
1826                 * so we force all entries to have damage > 0.
1827                 */
1828                damage = (p->one->size - copied) + added;
1829                if (!damage)
1830                        damage = 1;
1831
1832found_damage:
1833                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1834                dir.files[dir.nr].name = name;
1835                dir.files[dir.nr].changed = damage;
1836                changed += damage;
1837                dir.nr++;
1838        }
1839
1840        /* This can happen even with many files, if everything was renames */
1841        if (!changed)
1842                return;
1843
1844        /* Show all directories with more than x% of the changes */
1845        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1846        gather_dirstat(options, &dir, changed, "", 0);
1847}
1848
1849static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1850{
1851        int i;
1852        unsigned long changed;
1853        struct dirstat_dir dir;
1854
1855        if (data->nr == 0)
1856                return;
1857
1858        dir.files = NULL;
1859        dir.alloc = 0;
1860        dir.nr = 0;
1861        dir.permille = options->dirstat_permille;
1862        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1863
1864        changed = 0;
1865        for (i = 0; i < data->nr; i++) {
1866                struct diffstat_file *file = data->files[i];
1867                unsigned long damage = file->added + file->deleted;
1868                if (file->is_binary)
1869                        /*
1870                         * binary files counts bytes, not lines. Must find some
1871                         * way to normalize binary bytes vs. textual lines.
1872                         * The following heuristic assumes that there are 64
1873                         * bytes per "line".
1874                         * This is stupid and ugly, but very cheap...
1875                         */
1876                        damage = (damage + 63) / 64;
1877                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1878                dir.files[dir.nr].name = file->name;
1879                dir.files[dir.nr].changed = damage;
1880                changed += damage;
1881                dir.nr++;
1882        }
1883
1884        /* This can happen even with many files, if everything was renames */
1885        if (!changed)
1886                return;
1887
1888        /* Show all directories with more than x% of the changes */
1889        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1890        gather_dirstat(options, &dir, changed, "", 0);
1891}
1892
1893static void free_diffstat_info(struct diffstat_t *diffstat)
1894{
1895        int i;
1896        for (i = 0; i < diffstat->nr; i++) {
1897                struct diffstat_file *f = diffstat->files[i];
1898                if (f->name != f->print_name)
1899                        free(f->print_name);
1900                free(f->name);
1901                free(f->from_name);
1902                free(f);
1903        }
1904        free(diffstat->files);
1905}
1906
1907struct checkdiff_t {
1908        const char *filename;
1909        int lineno;
1910        int conflict_marker_size;
1911        struct diff_options *o;
1912        unsigned ws_rule;
1913        unsigned status;
1914};
1915
1916static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1917{
1918        char firstchar;
1919        int cnt;
1920
1921        if (len < marker_size + 1)
1922                return 0;
1923        firstchar = line[0];
1924        switch (firstchar) {
1925        case '=': case '>': case '<': case '|':
1926                break;
1927        default:
1928                return 0;
1929        }
1930        for (cnt = 1; cnt < marker_size; cnt++)
1931                if (line[cnt] != firstchar)
1932                        return 0;
1933        /* line[1] thru line[marker_size-1] are same as firstchar */
1934        if (len < marker_size + 1 || !isspace(line[marker_size]))
1935                return 0;
1936        return 1;
1937}
1938
1939static void checkdiff_consume(void *priv, char *line, unsigned long len)
1940{
1941        struct checkdiff_t *data = priv;
1942        int marker_size = data->conflict_marker_size;
1943        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
1944        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
1945        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
1946        char *err;
1947        char *line_prefix = "";
1948        struct strbuf *msgbuf;
1949
1950        assert(data->o);
1951        if (data->o->output_prefix) {
1952                msgbuf = data->o->output_prefix(data->o,
1953                        data->o->output_prefix_data);
1954                line_prefix = msgbuf->buf;
1955        }
1956
1957        if (line[0] == '+') {
1958                unsigned bad;
1959                data->lineno++;
1960                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1961                        data->status |= 1;
1962                        fprintf(data->o->file,
1963                                "%s%s:%d: leftover conflict marker\n",
1964                                line_prefix, data->filename, data->lineno);
1965                }
1966                bad = ws_check(line + 1, len - 1, data->ws_rule);
1967                if (!bad)
1968                        return;
1969                data->status |= bad;
1970                err = whitespace_error_string(bad);
1971                fprintf(data->o->file, "%s%s:%d: %s.\n",
1972                        line_prefix, data->filename, data->lineno, err);
1973                free(err);
1974                emit_line(data->o, set, reset, line, 1);
1975                ws_check_emit(line + 1, len - 1, data->ws_rule,
1976                              data->o->file, set, reset, ws);
1977        } else if (line[0] == ' ') {
1978                data->lineno++;
1979        } else if (line[0] == '@') {
1980                char *plus = strchr(line, '+');
1981                if (plus)
1982                        data->lineno = strtol(plus, NULL, 10) - 1;
1983                else
1984                        die("invalid diff");
1985        }
1986}
1987
1988static unsigned char *deflate_it(char *data,
1989                                 unsigned long size,
1990                                 unsigned long *result_size)
1991{
1992        int bound;
1993        unsigned char *deflated;
1994        git_zstream stream;
1995
1996        memset(&stream, 0, sizeof(stream));
1997        git_deflate_init(&stream, zlib_compression_level);
1998        bound = git_deflate_bound(&stream, size);
1999        deflated = xmalloc(bound);
2000        stream.next_out = deflated;
2001        stream.avail_out = bound;
2002
2003        stream.next_in = (unsigned char *)data;
2004        stream.avail_in = size;
2005        while (git_deflate(&stream, Z_FINISH) == Z_OK)
2006                ; /* nothing */
2007        git_deflate_end(&stream);
2008        *result_size = stream.total_out;
2009        return deflated;
2010}
2011
2012static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2013{
2014        void *cp;
2015        void *delta;
2016        void *deflated;
2017        void *data;
2018        unsigned long orig_size;
2019        unsigned long delta_size;
2020        unsigned long deflate_size;
2021        unsigned long data_size;
2022
2023        /* We could do deflated delta, or we could do just deflated two,
2024         * whichever is smaller.
2025         */
2026        delta = NULL;
2027        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2028        if (one->size && two->size) {
2029                delta = diff_delta(one->ptr, one->size,
2030                                   two->ptr, two->size,
2031                                   &delta_size, deflate_size);
2032                if (delta) {
2033                        void *to_free = delta;
2034                        orig_size = delta_size;
2035                        delta = deflate_it(delta, delta_size, &delta_size);
2036                        free(to_free);
2037                }
2038        }
2039
2040        if (delta && delta_size < deflate_size) {
2041                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2042                free(deflated);
2043                data = delta;
2044                data_size = delta_size;
2045        }
2046        else {
2047                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2048                free(delta);
2049                data = deflated;
2050                data_size = deflate_size;
2051        }
2052
2053        /* emit data encoded in base85 */
2054        cp = data;
2055        while (data_size) {
2056                int bytes = (52 < data_size) ? 52 : data_size;
2057                char line[70];
2058                data_size -= bytes;
2059                if (bytes <= 26)
2060                        line[0] = bytes + 'A' - 1;
2061                else
2062                        line[0] = bytes - 26 + 'a' - 1;
2063                encode_85(line + 1, cp, bytes);
2064                cp = (char *) cp + bytes;
2065                fprintf(file, "%s", prefix);
2066                fputs(line, file);
2067                fputc('\n', file);
2068        }
2069        fprintf(file, "%s\n", prefix);
2070        free(data);
2071}
2072
2073static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2074{
2075        fprintf(file, "%sGIT binary patch\n", prefix);
2076        emit_binary_diff_body(file, one, two, prefix);
2077        emit_binary_diff_body(file, two, one, prefix);
2078}
2079
2080int diff_filespec_is_binary(struct diff_filespec *one)
2081{
2082        if (one->is_binary == -1) {
2083                diff_filespec_load_driver(one);
2084                if (one->driver->binary != -1)
2085                        one->is_binary = one->driver->binary;
2086                else {
2087                        if (!one->data && DIFF_FILE_VALID(one))
2088                                diff_populate_filespec(one, 0);
2089                        if (one->data)
2090                                one->is_binary = buffer_is_binary(one->data,
2091                                                one->size);
2092                        if (one->is_binary == -1)
2093                                one->is_binary = 0;
2094                }
2095        }
2096        return one->is_binary;
2097}
2098
2099static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2100{
2101        diff_filespec_load_driver(one);
2102        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2103}
2104
2105void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2106{
2107        if (!options->a_prefix)
2108                options->a_prefix = a;
2109        if (!options->b_prefix)
2110                options->b_prefix = b;
2111}
2112
2113struct userdiff_driver *get_textconv(struct diff_filespec *one)
2114{
2115        if (!DIFF_FILE_VALID(one))
2116                return NULL;
2117
2118        diff_filespec_load_driver(one);
2119        return userdiff_get_textconv(one->driver);
2120}
2121
2122static void builtin_diff(const char *name_a,
2123                         const char *name_b,
2124                         struct diff_filespec *one,
2125                         struct diff_filespec *two,
2126                         const char *xfrm_msg,
2127                         int must_show_header,
2128                         struct diff_options *o,
2129                         int complete_rewrite)
2130{
2131        mmfile_t mf1, mf2;
2132        const char *lbl[2];
2133        char *a_one, *b_two;
2134        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2135        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2136        const char *a_prefix, *b_prefix;
2137        struct userdiff_driver *textconv_one = NULL;
2138        struct userdiff_driver *textconv_two = NULL;
2139        struct strbuf header = STRBUF_INIT;
2140        struct strbuf *msgbuf;
2141        char *line_prefix = "";
2142
2143        if (o->output_prefix) {
2144                msgbuf = o->output_prefix(o, o->output_prefix_data);
2145                line_prefix = msgbuf->buf;
2146        }
2147
2148        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2149                        (!one->mode || S_ISGITLINK(one->mode)) &&
2150                        (!two->mode || S_ISGITLINK(two->mode))) {
2151                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2152                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2153                show_submodule_summary(o->file, one ? one->path : two->path,
2154                                one->sha1, two->sha1, two->dirty_submodule,
2155                                del, add, reset);
2156                return;
2157        }
2158
2159        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2160                textconv_one = get_textconv(one);
2161                textconv_two = get_textconv(two);
2162        }
2163
2164        diff_set_mnemonic_prefix(o, "a/", "b/");
2165        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2166                a_prefix = o->b_prefix;
2167                b_prefix = o->a_prefix;
2168        } else {
2169                a_prefix = o->a_prefix;
2170                b_prefix = o->b_prefix;
2171        }
2172
2173        /* Never use a non-valid filename anywhere if at all possible */
2174        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2175        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2176
2177        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2178        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2179        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2180        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2181        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2182        if (lbl[0][0] == '/') {
2183                /* /dev/null */
2184                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2185                if (xfrm_msg)
2186                        strbuf_addstr(&header, xfrm_msg);
2187                must_show_header = 1;
2188        }
2189        else if (lbl[1][0] == '/') {
2190                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2191                if (xfrm_msg)
2192                        strbuf_addstr(&header, xfrm_msg);
2193                must_show_header = 1;
2194        }
2195        else {
2196                if (one->mode != two->mode) {
2197                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2198                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2199                        must_show_header = 1;
2200                }
2201                if (xfrm_msg)
2202                        strbuf_addstr(&header, xfrm_msg);
2203
2204                /*
2205                 * we do not run diff between different kind
2206                 * of objects.
2207                 */
2208                if ((one->mode ^ two->mode) & S_IFMT)
2209                        goto free_ab_and_return;
2210                if (complete_rewrite &&
2211                    (textconv_one || !diff_filespec_is_binary(one)) &&
2212                    (textconv_two || !diff_filespec_is_binary(two))) {
2213                        fprintf(o->file, "%s", header.buf);
2214                        strbuf_reset(&header);
2215                        emit_rewrite_diff(name_a, name_b, one, two,
2216                                                textconv_one, textconv_two, o);
2217                        o->found_changes = 1;
2218                        goto free_ab_and_return;
2219                }
2220        }
2221
2222        if (o->irreversible_delete && lbl[1][0] == '/') {
2223                fprintf(o->file, "%s", header.buf);
2224                strbuf_reset(&header);
2225                goto free_ab_and_return;
2226        } else if (!DIFF_OPT_TST(o, TEXT) &&
2227            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2228              (!textconv_two && diff_filespec_is_binary(two)) )) {
2229                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2230                        die("unable to read files to diff");
2231                /* Quite common confusing case */
2232                if (mf1.size == mf2.size &&
2233                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2234                        if (must_show_header)
2235                                fprintf(o->file, "%s", header.buf);
2236                        goto free_ab_and_return;
2237                }
2238                fprintf(o->file, "%s", header.buf);
2239                strbuf_reset(&header);
2240                if (DIFF_OPT_TST(o, BINARY))
2241                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2242                else
2243                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2244                                line_prefix, lbl[0], lbl[1]);
2245                o->found_changes = 1;
2246        } else {
2247                /* Crazy xdl interfaces.. */
2248                const char *diffopts = getenv("GIT_DIFF_OPTS");
2249                xpparam_t xpp;
2250                xdemitconf_t xecfg;
2251                struct emit_callback ecbdata;
2252                const struct userdiff_funcname *pe;
2253
2254                if (must_show_header) {
2255                        fprintf(o->file, "%s", header.buf);
2256                        strbuf_reset(&header);
2257                }
2258
2259                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2260                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2261
2262                pe = diff_funcname_pattern(one);
2263                if (!pe)
2264                        pe = diff_funcname_pattern(two);
2265
2266                memset(&xpp, 0, sizeof(xpp));
2267                memset(&xecfg, 0, sizeof(xecfg));
2268                memset(&ecbdata, 0, sizeof(ecbdata));
2269                ecbdata.label_path = lbl;
2270                ecbdata.color_diff = want_color(o->use_color);
2271                ecbdata.found_changesp = &o->found_changes;
2272                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2273                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2274                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2275                ecbdata.opt = o;
2276                ecbdata.header = header.len ? &header : NULL;
2277                xpp.flags = o->xdl_opts;
2278                xecfg.ctxlen = o->context;
2279                xecfg.interhunkctxlen = o->interhunkcontext;
2280                xecfg.flags = XDL_EMIT_FUNCNAMES;
2281                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2282                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2283                if (pe)
2284                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2285                if (!diffopts)
2286                        ;
2287                else if (!prefixcmp(diffopts, "--unified="))
2288                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2289                else if (!prefixcmp(diffopts, "-u"))
2290                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2291                if (o->word_diff)
2292                        init_diff_words_data(&ecbdata, o, one, two);
2293                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2294                              &xpp, &xecfg);
2295                if (o->word_diff)
2296                        free_diff_words_data(&ecbdata);
2297                if (textconv_one)
2298                        free(mf1.ptr);
2299                if (textconv_two)
2300                        free(mf2.ptr);
2301                xdiff_clear_find_func(&xecfg);
2302        }
2303
2304 free_ab_and_return:
2305        strbuf_release(&header);
2306        diff_free_filespec_data(one);
2307        diff_free_filespec_data(two);
2308        free(a_one);
2309        free(b_two);
2310        return;
2311}
2312
2313static void builtin_diffstat(const char *name_a, const char *name_b,
2314                             struct diff_filespec *one,
2315                             struct diff_filespec *two,
2316                             struct diffstat_t *diffstat,
2317                             struct diff_options *o,
2318                             int complete_rewrite)
2319{
2320        mmfile_t mf1, mf2;
2321        struct diffstat_file *data;
2322
2323        data = diffstat_add(diffstat, name_a, name_b);
2324
2325        if (!one || !two) {
2326                data->is_unmerged = 1;
2327                return;
2328        }
2329
2330        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2331                data->is_binary = 1;
2332                data->added = diff_filespec_size(two);
2333                data->deleted = diff_filespec_size(one);
2334        }
2335
2336        else if (complete_rewrite) {
2337                diff_populate_filespec(one, 0);
2338                diff_populate_filespec(two, 0);
2339                data->deleted = count_lines(one->data, one->size);
2340                data->added = count_lines(two->data, two->size);
2341        }
2342
2343        else {
2344                /* Crazy xdl interfaces.. */
2345                xpparam_t xpp;
2346                xdemitconf_t xecfg;
2347
2348                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2349                        die("unable to read files to diff");
2350
2351                memset(&xpp, 0, sizeof(xpp));
2352                memset(&xecfg, 0, sizeof(xecfg));
2353                xpp.flags = o->xdl_opts;
2354                xecfg.ctxlen = o->context;
2355                xecfg.interhunkctxlen = o->interhunkcontext;
2356                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2357                              &xpp, &xecfg);
2358        }
2359
2360        diff_free_filespec_data(one);
2361        diff_free_filespec_data(two);
2362}
2363
2364static void builtin_checkdiff(const char *name_a, const char *name_b,
2365                              const char *attr_path,
2366                              struct diff_filespec *one,
2367                              struct diff_filespec *two,
2368                              struct diff_options *o)
2369{
2370        mmfile_t mf1, mf2;
2371        struct checkdiff_t data;
2372
2373        if (!two)
2374                return;
2375
2376        memset(&data, 0, sizeof(data));
2377        data.filename = name_b ? name_b : name_a;
2378        data.lineno = 0;
2379        data.o = o;
2380        data.ws_rule = whitespace_rule(attr_path);
2381        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2382
2383        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2384                die("unable to read files to diff");
2385
2386        /*
2387         * All the other codepaths check both sides, but not checking
2388         * the "old" side here is deliberate.  We are checking the newly
2389         * introduced changes, and as long as the "new" side is text, we
2390         * can and should check what it introduces.
2391         */
2392        if (diff_filespec_is_binary(two))
2393                goto free_and_return;
2394        else {
2395                /* Crazy xdl interfaces.. */
2396                xpparam_t xpp;
2397                xdemitconf_t xecfg;
2398
2399                memset(&xpp, 0, sizeof(xpp));
2400                memset(&xecfg, 0, sizeof(xecfg));
2401                xecfg.ctxlen = 1; /* at least one context line */
2402                xpp.flags = 0;
2403                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2404                              &xpp, &xecfg);
2405
2406                if (data.ws_rule & WS_BLANK_AT_EOF) {
2407                        struct emit_callback ecbdata;
2408                        int blank_at_eof;
2409
2410                        ecbdata.ws_rule = data.ws_rule;
2411                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2412                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2413
2414                        if (blank_at_eof) {
2415                                static char *err;
2416                                if (!err)
2417                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2418                                fprintf(o->file, "%s:%d: %s.\n",
2419                                        data.filename, blank_at_eof, err);
2420                                data.status = 1; /* report errors */
2421                        }
2422                }
2423        }
2424 free_and_return:
2425        diff_free_filespec_data(one);
2426        diff_free_filespec_data(two);
2427        if (data.status)
2428                DIFF_OPT_SET(o, CHECK_FAILED);
2429}
2430
2431struct diff_filespec *alloc_filespec(const char *path)
2432{
2433        int namelen = strlen(path);
2434        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2435
2436        memset(spec, 0, sizeof(*spec));
2437        spec->path = (char *)(spec + 1);
2438        memcpy(spec->path, path, namelen+1);
2439        spec->count = 1;
2440        spec->is_binary = -1;
2441        return spec;
2442}
2443
2444void free_filespec(struct diff_filespec *spec)
2445{
2446        if (!--spec->count) {
2447                diff_free_filespec_data(spec);
2448                free(spec);
2449        }
2450}
2451
2452void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2453                   unsigned short mode)
2454{
2455        if (mode) {
2456                spec->mode = canon_mode(mode);
2457                hashcpy(spec->sha1, sha1);
2458                spec->sha1_valid = !is_null_sha1(sha1);
2459        }
2460}
2461
2462/*
2463 * Given a name and sha1 pair, if the index tells us the file in
2464 * the work tree has that object contents, return true, so that
2465 * prepare_temp_file() does not have to inflate and extract.
2466 */
2467static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2468{
2469        struct cache_entry *ce;
2470        struct stat st;
2471        int pos, len;
2472
2473        /*
2474         * We do not read the cache ourselves here, because the
2475         * benchmark with my previous version that always reads cache
2476         * shows that it makes things worse for diff-tree comparing
2477         * two linux-2.6 kernel trees in an already checked out work
2478         * tree.  This is because most diff-tree comparisons deal with
2479         * only a small number of files, while reading the cache is
2480         * expensive for a large project, and its cost outweighs the
2481         * savings we get by not inflating the object to a temporary
2482         * file.  Practically, this code only helps when we are used
2483         * by diff-cache --cached, which does read the cache before
2484         * calling us.
2485         */
2486        if (!active_cache)
2487                return 0;
2488
2489        /* We want to avoid the working directory if our caller
2490         * doesn't need the data in a normal file, this system
2491         * is rather slow with its stat/open/mmap/close syscalls,
2492         * and the object is contained in a pack file.  The pack
2493         * is probably already open and will be faster to obtain
2494         * the data through than the working directory.  Loose
2495         * objects however would tend to be slower as they need
2496         * to be individually opened and inflated.
2497         */
2498        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2499                return 0;
2500
2501        len = strlen(name);
2502        pos = cache_name_pos(name, len);
2503        if (pos < 0)
2504                return 0;
2505        ce = active_cache[pos];
2506
2507        /*
2508         * This is not the sha1 we are looking for, or
2509         * unreusable because it is not a regular file.
2510         */
2511        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2512                return 0;
2513
2514        /*
2515         * If ce is marked as "assume unchanged", there is no
2516         * guarantee that work tree matches what we are looking for.
2517         */
2518        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2519                return 0;
2520
2521        /*
2522         * If ce matches the file in the work tree, we can reuse it.
2523         */
2524        if (ce_uptodate(ce) ||
2525            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2526                return 1;
2527
2528        return 0;
2529}
2530
2531static int populate_from_stdin(struct diff_filespec *s)
2532{
2533        struct strbuf buf = STRBUF_INIT;
2534        size_t size = 0;
2535
2536        if (strbuf_read(&buf, 0, 0) < 0)
2537                return error("error while reading from stdin %s",
2538                                     strerror(errno));
2539
2540        s->should_munmap = 0;
2541        s->data = strbuf_detach(&buf, &size);
2542        s->size = size;
2543        s->should_free = 1;
2544        return 0;
2545}
2546
2547static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2548{
2549        int len;
2550        char *data = xmalloc(100), *dirty = "";
2551
2552        /* Are we looking at the work tree? */
2553        if (s->dirty_submodule)
2554                dirty = "-dirty";
2555
2556        len = snprintf(data, 100,
2557                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2558        s->data = data;
2559        s->size = len;
2560        s->should_free = 1;
2561        if (size_only) {
2562                s->data = NULL;
2563                free(data);
2564        }
2565        return 0;
2566}
2567
2568/*
2569 * While doing rename detection and pickaxe operation, we may need to
2570 * grab the data for the blob (or file) for our own in-core comparison.
2571 * diff_filespec has data and size fields for this purpose.
2572 */
2573int diff_populate_filespec(struct diff_filespec *s, int size_only)
2574{
2575        int err = 0;
2576        if (!DIFF_FILE_VALID(s))
2577                die("internal error: asking to populate invalid file.");
2578        if (S_ISDIR(s->mode))
2579                return -1;
2580
2581        if (s->data)
2582                return 0;
2583
2584        if (size_only && 0 < s->size)
2585                return 0;
2586
2587        if (S_ISGITLINK(s->mode))
2588                return diff_populate_gitlink(s, size_only);
2589
2590        if (!s->sha1_valid ||
2591            reuse_worktree_file(s->path, s->sha1, 0)) {
2592                struct strbuf buf = STRBUF_INIT;
2593                struct stat st;
2594                int fd;
2595
2596                if (!strcmp(s->path, "-"))
2597                        return populate_from_stdin(s);
2598
2599                if (lstat(s->path, &st) < 0) {
2600                        if (errno == ENOENT) {
2601                        err_empty:
2602                                err = -1;
2603                        empty:
2604                                s->data = (char *)"";
2605                                s->size = 0;
2606                                return err;
2607                        }
2608                }
2609                s->size = xsize_t(st.st_size);
2610                if (!s->size)
2611                        goto empty;
2612                if (S_ISLNK(st.st_mode)) {
2613                        struct strbuf sb = STRBUF_INIT;
2614
2615                        if (strbuf_readlink(&sb, s->path, s->size))
2616                                goto err_empty;
2617                        s->size = sb.len;
2618                        s->data = strbuf_detach(&sb, NULL);
2619                        s->should_free = 1;
2620                        return 0;
2621                }
2622                if (size_only)
2623                        return 0;
2624                fd = open(s->path, O_RDONLY);
2625                if (fd < 0)
2626                        goto err_empty;
2627                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2628                close(fd);
2629                s->should_munmap = 1;
2630
2631                /*
2632                 * Convert from working tree format to canonical git format
2633                 */
2634                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2635                        size_t size = 0;
2636                        munmap(s->data, s->size);
2637                        s->should_munmap = 0;
2638                        s->data = strbuf_detach(&buf, &size);
2639                        s->size = size;
2640                        s->should_free = 1;
2641                }
2642        }
2643        else {
2644                enum object_type type;
2645                if (size_only) {
2646                        type = sha1_object_info(s->sha1, &s->size);
2647                        if (type < 0)
2648                                die("unable to read %s", sha1_to_hex(s->sha1));
2649                } else {
2650                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2651                        if (!s->data)
2652                                die("unable to read %s", sha1_to_hex(s->sha1));
2653                        s->should_free = 1;
2654                }
2655        }
2656        return 0;
2657}
2658
2659void diff_free_filespec_blob(struct diff_filespec *s)
2660{
2661        if (s->should_free)
2662                free(s->data);
2663        else if (s->should_munmap)
2664                munmap(s->data, s->size);
2665
2666        if (s->should_free || s->should_munmap) {
2667                s->should_free = s->should_munmap = 0;
2668                s->data = NULL;
2669        }
2670}
2671
2672void diff_free_filespec_data(struct diff_filespec *s)
2673{
2674        diff_free_filespec_blob(s);
2675        free(s->cnt_data);
2676        s->cnt_data = NULL;
2677}
2678
2679static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2680                           void *blob,
2681                           unsigned long size,
2682                           const unsigned char *sha1,
2683                           int mode)
2684{
2685        int fd;
2686        struct strbuf buf = STRBUF_INIT;
2687        struct strbuf template = STRBUF_INIT;
2688        char *path_dup = xstrdup(path);
2689        const char *base = basename(path_dup);
2690
2691        /* Generate "XXXXXX_basename.ext" */
2692        strbuf_addstr(&template, "XXXXXX_");
2693        strbuf_addstr(&template, base);
2694
2695        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2696                        strlen(base) + 1);
2697        if (fd < 0)
2698                die_errno("unable to create temp-file");
2699        if (convert_to_working_tree(path,
2700                        (const char *)blob, (size_t)size, &buf)) {
2701                blob = buf.buf;
2702                size = buf.len;
2703        }
2704        if (write_in_full(fd, blob, size) != size)
2705                die_errno("unable to write temp-file");
2706        close(fd);
2707        temp->name = temp->tmp_path;
2708        strcpy(temp->hex, sha1_to_hex(sha1));
2709        temp->hex[40] = 0;
2710        sprintf(temp->mode, "%06o", mode);
2711        strbuf_release(&buf);
2712        strbuf_release(&template);
2713        free(path_dup);
2714}
2715
2716static struct diff_tempfile *prepare_temp_file(const char *name,
2717                struct diff_filespec *one)
2718{
2719        struct diff_tempfile *temp = claim_diff_tempfile();
2720
2721        if (!DIFF_FILE_VALID(one)) {
2722        not_a_valid_file:
2723                /* A '-' entry produces this for file-2, and
2724                 * a '+' entry produces this for file-1.
2725                 */
2726                temp->name = "/dev/null";
2727                strcpy(temp->hex, ".");
2728                strcpy(temp->mode, ".");
2729                return temp;
2730        }
2731
2732        if (!remove_tempfile_installed) {
2733                atexit(remove_tempfile);
2734                sigchain_push_common(remove_tempfile_on_signal);
2735                remove_tempfile_installed = 1;
2736        }
2737
2738        if (!one->sha1_valid ||
2739            reuse_worktree_file(name, one->sha1, 1)) {
2740                struct stat st;
2741                if (lstat(name, &st) < 0) {
2742                        if (errno == ENOENT)
2743                                goto not_a_valid_file;
2744                        die_errno("stat(%s)", name);
2745                }
2746                if (S_ISLNK(st.st_mode)) {
2747                        struct strbuf sb = STRBUF_INIT;
2748                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2749                                die_errno("readlink(%s)", name);
2750                        prep_temp_blob(name, temp, sb.buf, sb.len,
2751                                       (one->sha1_valid ?
2752                                        one->sha1 : null_sha1),
2753                                       (one->sha1_valid ?
2754                                        one->mode : S_IFLNK));
2755                        strbuf_release(&sb);
2756                }
2757                else {
2758                        /* we can borrow from the file in the work tree */
2759                        temp->name = name;
2760                        if (!one->sha1_valid)
2761                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2762                        else
2763                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2764                        /* Even though we may sometimes borrow the
2765                         * contents from the work tree, we always want
2766                         * one->mode.  mode is trustworthy even when
2767                         * !(one->sha1_valid), as long as
2768                         * DIFF_FILE_VALID(one).
2769                         */
2770                        sprintf(temp->mode, "%06o", one->mode);
2771                }
2772                return temp;
2773        }
2774        else {
2775                if (diff_populate_filespec(one, 0))
2776                        die("cannot read data blob for %s", one->path);
2777                prep_temp_blob(name, temp, one->data, one->size,
2778                               one->sha1, one->mode);
2779        }
2780        return temp;
2781}
2782
2783/* An external diff command takes:
2784 *
2785 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2786 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2787 *
2788 */
2789static void run_external_diff(const char *pgm,
2790                              const char *name,
2791                              const char *other,
2792                              struct diff_filespec *one,
2793                              struct diff_filespec *two,
2794                              const char *xfrm_msg,
2795                              int complete_rewrite)
2796{
2797        const char *spawn_arg[10];
2798        int retval;
2799        const char **arg = &spawn_arg[0];
2800
2801        if (one && two) {
2802                struct diff_tempfile *temp_one, *temp_two;
2803                const char *othername = (other ? other : name);
2804                temp_one = prepare_temp_file(name, one);
2805                temp_two = prepare_temp_file(othername, two);
2806                *arg++ = pgm;
2807                *arg++ = name;
2808                *arg++ = temp_one->name;
2809                *arg++ = temp_one->hex;
2810                *arg++ = temp_one->mode;
2811                *arg++ = temp_two->name;
2812                *arg++ = temp_two->hex;
2813                *arg++ = temp_two->mode;
2814                if (other) {
2815                        *arg++ = other;
2816                        *arg++ = xfrm_msg;
2817                }
2818        } else {
2819                *arg++ = pgm;
2820                *arg++ = name;
2821        }
2822        *arg = NULL;
2823        fflush(NULL);
2824        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2825        remove_tempfile();
2826        if (retval) {
2827                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2828                exit(1);
2829        }
2830}
2831
2832static int similarity_index(struct diff_filepair *p)
2833{
2834        return p->score * 100 / MAX_SCORE;
2835}
2836
2837static void fill_metainfo(struct strbuf *msg,
2838                          const char *name,
2839                          const char *other,
2840                          struct diff_filespec *one,
2841                          struct diff_filespec *two,
2842                          struct diff_options *o,
2843                          struct diff_filepair *p,
2844                          int *must_show_header,
2845                          int use_color)
2846{
2847        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2848        const char *reset = diff_get_color(use_color, DIFF_RESET);
2849        struct strbuf *msgbuf;
2850        char *line_prefix = "";
2851
2852        *must_show_header = 1;
2853        if (o->output_prefix) {
2854                msgbuf = o->output_prefix(o, o->output_prefix_data);
2855                line_prefix = msgbuf->buf;
2856        }
2857        strbuf_init(msg, PATH_MAX * 2 + 300);
2858        switch (p->status) {
2859        case DIFF_STATUS_COPIED:
2860                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2861                            line_prefix, set, similarity_index(p));
2862                strbuf_addf(msg, "%s\n%s%scopy from ",
2863                            reset,  line_prefix, set);
2864                quote_c_style(name, msg, NULL, 0);
2865                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2866                quote_c_style(other, msg, NULL, 0);
2867                strbuf_addf(msg, "%s\n", reset);
2868                break;
2869        case DIFF_STATUS_RENAMED:
2870                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2871                            line_prefix, set, similarity_index(p));
2872                strbuf_addf(msg, "%s\n%s%srename from ",
2873                            reset, line_prefix, set);
2874                quote_c_style(name, msg, NULL, 0);
2875                strbuf_addf(msg, "%s\n%s%srename to ",
2876                            reset, line_prefix, set);
2877                quote_c_style(other, msg, NULL, 0);
2878                strbuf_addf(msg, "%s\n", reset);
2879                break;
2880        case DIFF_STATUS_MODIFIED:
2881                if (p->score) {
2882                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2883                                    line_prefix,
2884                                    set, similarity_index(p), reset);
2885                        break;
2886                }
2887                /* fallthru */
2888        default:
2889                *must_show_header = 0;
2890        }
2891        if (one && two && hashcmp(one->sha1, two->sha1)) {
2892                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2893
2894                if (DIFF_OPT_TST(o, BINARY)) {
2895                        mmfile_t mf;
2896                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2897                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2898                                abbrev = 40;
2899                }
2900                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2901                            find_unique_abbrev(one->sha1, abbrev));
2902                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2903                if (one->mode == two->mode)
2904                        strbuf_addf(msg, " %06o", one->mode);
2905                strbuf_addf(msg, "%s\n", reset);
2906        }
2907}
2908
2909static void run_diff_cmd(const char *pgm,
2910                         const char *name,
2911                         const char *other,
2912                         const char *attr_path,
2913                         struct diff_filespec *one,
2914                         struct diff_filespec *two,
2915                         struct strbuf *msg,
2916                         struct diff_options *o,
2917                         struct diff_filepair *p)
2918{
2919        const char *xfrm_msg = NULL;
2920        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2921        int must_show_header = 0;
2922
2923        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2924                pgm = NULL;
2925        else {
2926                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2927                if (drv && drv->external)
2928                        pgm = drv->external;
2929        }
2930
2931        if (msg) {
2932                /*
2933                 * don't use colors when the header is intended for an
2934                 * external diff driver
2935                 */
2936                fill_metainfo(msg, name, other, one, two, o, p,
2937                              &must_show_header,
2938                              want_color(o->use_color) && !pgm);
2939                xfrm_msg = msg->len ? msg->buf : NULL;
2940        }
2941
2942        if (pgm) {
2943                run_external_diff(pgm, name, other, one, two, xfrm_msg,
2944                                  complete_rewrite);
2945                return;
2946        }
2947        if (one && two)
2948                builtin_diff(name, other ? other : name,
2949                             one, two, xfrm_msg, must_show_header,
2950                             o, complete_rewrite);
2951        else
2952                fprintf(o->file, "* Unmerged path %s\n", name);
2953}
2954
2955static void diff_fill_sha1_info(struct diff_filespec *one)
2956{
2957        if (DIFF_FILE_VALID(one)) {
2958                if (!one->sha1_valid) {
2959                        struct stat st;
2960                        if (!strcmp(one->path, "-")) {
2961                                hashcpy(one->sha1, null_sha1);
2962                                return;
2963                        }
2964                        if (lstat(one->path, &st) < 0)
2965                                die_errno("stat '%s'", one->path);
2966                        if (index_path(one->sha1, one->path, &st, 0))
2967                                die("cannot hash %s", one->path);
2968                }
2969        }
2970        else
2971                hashclr(one->sha1);
2972}
2973
2974static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2975{
2976        /* Strip the prefix but do not molest /dev/null and absolute paths */
2977        if (*namep && **namep != '/') {
2978                *namep += prefix_length;
2979                if (**namep == '/')
2980                        ++*namep;
2981        }
2982        if (*otherp && **otherp != '/') {
2983                *otherp += prefix_length;
2984                if (**otherp == '/')
2985                        ++*otherp;
2986        }
2987}
2988
2989static void run_diff(struct diff_filepair *p, struct diff_options *o)
2990{
2991        const char *pgm = external_diff();
2992        struct strbuf msg;
2993        struct diff_filespec *one = p->one;
2994        struct diff_filespec *two = p->two;
2995        const char *name;
2996        const char *other;
2997        const char *attr_path;
2998
2999        name  = p->one->path;
3000        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3001        attr_path = name;
3002        if (o->prefix_length)
3003                strip_prefix(o->prefix_length, &name, &other);
3004
3005        if (DIFF_PAIR_UNMERGED(p)) {
3006                run_diff_cmd(pgm, name, NULL, attr_path,
3007                             NULL, NULL, NULL, o, p);
3008                return;
3009        }
3010
3011        diff_fill_sha1_info(one);
3012        diff_fill_sha1_info(two);
3013
3014        if (!pgm &&
3015            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3016            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3017                /*
3018                 * a filepair that changes between file and symlink
3019                 * needs to be split into deletion and creation.
3020                 */
3021                struct diff_filespec *null = alloc_filespec(two->path);
3022                run_diff_cmd(NULL, name, other, attr_path,
3023                             one, null, &msg, o, p);
3024                free(null);
3025                strbuf_release(&msg);
3026
3027                null = alloc_filespec(one->path);
3028                run_diff_cmd(NULL, name, other, attr_path,
3029                             null, two, &msg, o, p);
3030                free(null);
3031        }
3032        else
3033                run_diff_cmd(pgm, name, other, attr_path,
3034                             one, two, &msg, o, p);
3035
3036        strbuf_release(&msg);
3037}
3038
3039static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3040                         struct diffstat_t *diffstat)
3041{
3042        const char *name;
3043        const char *other;
3044        int complete_rewrite = 0;
3045
3046        if (DIFF_PAIR_UNMERGED(p)) {
3047                /* unmerged */
3048                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3049                return;
3050        }
3051
3052        name = p->one->path;
3053        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3054
3055        if (o->prefix_length)
3056                strip_prefix(o->prefix_length, &name, &other);
3057
3058        diff_fill_sha1_info(p->one);
3059        diff_fill_sha1_info(p->two);
3060
3061        if (p->status == DIFF_STATUS_MODIFIED && p->score)
3062                complete_rewrite = 1;
3063        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3064}
3065
3066static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3067{
3068        const char *name;
3069        const char *other;
3070        const char *attr_path;
3071
3072        if (DIFF_PAIR_UNMERGED(p)) {
3073                /* unmerged */
3074                return;
3075        }
3076
3077        name = p->one->path;
3078        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3079        attr_path = other ? other : name;
3080
3081        if (o->prefix_length)
3082                strip_prefix(o->prefix_length, &name, &other);
3083
3084        diff_fill_sha1_info(p->one);
3085        diff_fill_sha1_info(p->two);
3086
3087        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3088}
3089
3090void diff_setup(struct diff_options *options)
3091{
3092        memcpy(options, &default_diff_options, sizeof(*options));
3093
3094        options->file = stdout;
3095
3096        options->line_termination = '\n';
3097        options->break_opt = -1;
3098        options->rename_limit = -1;
3099        options->dirstat_permille = diff_dirstat_permille_default;
3100        options->context = 3;
3101
3102        options->change = diff_change;
3103        options->add_remove = diff_addremove;
3104        options->use_color = diff_use_color_default;
3105        options->detect_rename = diff_detect_rename_default;
3106
3107        if (diff_no_prefix) {
3108                options->a_prefix = options->b_prefix = "";
3109        } else if (!diff_mnemonic_prefix) {
3110                options->a_prefix = "a/";
3111                options->b_prefix = "b/";
3112        }
3113}
3114
3115int diff_setup_done(struct diff_options *options)
3116{
3117        int count = 0;
3118
3119        if (options->output_format & DIFF_FORMAT_NAME)
3120                count++;
3121        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3122                count++;
3123        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3124                count++;
3125        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3126                count++;
3127        if (count > 1)
3128                die("--name-only, --name-status, --check and -s are mutually exclusive");
3129
3130        /*
3131         * Most of the time we can say "there are changes"
3132         * only by checking if there are changed paths, but
3133         * --ignore-whitespace* options force us to look
3134         * inside contents.
3135         */
3136
3137        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3138            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3139            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3140                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3141        else
3142                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3143
3144        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3145                options->detect_rename = DIFF_DETECT_COPY;
3146
3147        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3148                options->prefix = NULL;
3149        if (options->prefix)
3150                options->prefix_length = strlen(options->prefix);
3151        else
3152                options->prefix_length = 0;
3153
3154        if (options->output_format & (DIFF_FORMAT_NAME |
3155                                      DIFF_FORMAT_NAME_STATUS |
3156                                      DIFF_FORMAT_CHECKDIFF |
3157                                      DIFF_FORMAT_NO_OUTPUT))
3158                options->output_format &= ~(DIFF_FORMAT_RAW |
3159                                            DIFF_FORMAT_NUMSTAT |
3160                                            DIFF_FORMAT_DIFFSTAT |
3161                                            DIFF_FORMAT_SHORTSTAT |
3162                                            DIFF_FORMAT_DIRSTAT |
3163                                            DIFF_FORMAT_SUMMARY |
3164                                            DIFF_FORMAT_PATCH);
3165
3166        /*
3167         * These cases always need recursive; we do not drop caller-supplied
3168         * recursive bits for other formats here.
3169         */
3170        if (options->output_format & (DIFF_FORMAT_PATCH |
3171                                      DIFF_FORMAT_NUMSTAT |
3172                                      DIFF_FORMAT_DIFFSTAT |
3173                                      DIFF_FORMAT_SHORTSTAT |
3174                                      DIFF_FORMAT_DIRSTAT |
3175                                      DIFF_FORMAT_SUMMARY |
3176                                      DIFF_FORMAT_CHECKDIFF))
3177                DIFF_OPT_SET(options, RECURSIVE);
3178        /*
3179         * Also pickaxe would not work very well if you do not say recursive
3180         */
3181        if (options->pickaxe)
3182                DIFF_OPT_SET(options, RECURSIVE);
3183        /*
3184         * When patches are generated, submodules diffed against the work tree
3185         * must be checked for dirtiness too so it can be shown in the output
3186         */
3187        if (options->output_format & DIFF_FORMAT_PATCH)
3188                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3189
3190        if (options->detect_rename && options->rename_limit < 0)
3191                options->rename_limit = diff_rename_limit_default;
3192        if (options->setup & DIFF_SETUP_USE_CACHE) {
3193                if (!active_cache)
3194                        /* read-cache does not die even when it fails
3195                         * so it is safe for us to do this here.  Also
3196                         * it does not smudge active_cache or active_nr
3197                         * when it fails, so we do not have to worry about
3198                         * cleaning it up ourselves either.
3199                         */
3200                        read_cache();
3201        }
3202        if (options->abbrev <= 0 || 40 < options->abbrev)
3203                options->abbrev = 40; /* full */
3204
3205        /*
3206         * It does not make sense to show the first hit we happened
3207         * to have found.  It does not make sense not to return with
3208         * exit code in such a case either.
3209         */
3210        if (DIFF_OPT_TST(options, QUICK)) {
3211                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3212                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3213        }
3214
3215        return 0;
3216}
3217
3218static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3219{
3220        char c, *eq;
3221        int len;
3222
3223        if (*arg != '-')
3224                return 0;
3225        c = *++arg;
3226        if (!c)
3227                return 0;
3228        if (c == arg_short) {
3229                c = *++arg;
3230                if (!c)
3231                        return 1;
3232                if (val && isdigit(c)) {
3233                        char *end;
3234                        int n = strtoul(arg, &end, 10);
3235                        if (*end)
3236                                return 0;
3237                        *val = n;
3238                        return 1;
3239                }
3240                return 0;
3241        }
3242        if (c != '-')
3243                return 0;
3244        arg++;
3245        eq = strchr(arg, '=');
3246        if (eq)
3247                len = eq - arg;
3248        else
3249                len = strlen(arg);
3250        if (!len || strncmp(arg, arg_long, len))
3251                return 0;
3252        if (eq) {
3253                int n;
3254                char *end;
3255                if (!isdigit(*++eq))
3256                        return 0;
3257                n = strtoul(eq, &end, 10);
3258                if (*end)
3259                        return 0;
3260                *val = n;
3261        }
3262        return 1;
3263}
3264
3265static int diff_scoreopt_parse(const char *opt);
3266
3267static inline int short_opt(char opt, const char **argv,
3268                            const char **optarg)
3269{
3270        const char *arg = argv[0];
3271        if (arg[0] != '-' || arg[1] != opt)
3272                return 0;
3273        if (arg[2] != '\0') {
3274                *optarg = arg + 2;
3275                return 1;
3276        }
3277        if (!argv[1])
3278                die("Option '%c' requires a value", opt);
3279        *optarg = argv[1];
3280        return 2;
3281}
3282
3283int parse_long_opt(const char *opt, const char **argv,
3284                   const char **optarg)
3285{
3286        const char *arg = argv[0];
3287        if (arg[0] != '-' || arg[1] != '-')
3288                return 0;
3289        arg += strlen("--");
3290        if (prefixcmp(arg, opt))
3291                return 0;
3292        arg += strlen(opt);
3293        if (*arg == '=') { /* sticked form: --option=value */
3294                *optarg = arg + 1;
3295                return 1;
3296        }
3297        if (*arg != '\0')
3298                return 0;
3299        /* separate form: --option value */
3300        if (!argv[1])
3301                die("Option '--%s' requires a value", opt);
3302        *optarg = argv[1];
3303        return 2;
3304}
3305
3306static int stat_opt(struct diff_options *options, const char **av)
3307{
3308        const char *arg = av[0];
3309        char *end;
3310        int width = options->stat_width;
3311        int name_width = options->stat_name_width;
3312        int count = options->stat_count;
3313        int argcount = 1;
3314
3315        arg += strlen("--stat");
3316        end = (char *)arg;
3317
3318        switch (*arg) {
3319        case '-':
3320                if (!prefixcmp(arg, "-width")) {
3321                        arg += strlen("-width");
3322                        if (*arg == '=')
3323                                width = strtoul(arg + 1, &end, 10);
3324                        else if (!*arg && !av[1])
3325                                die("Option '--stat-width' requires a value");
3326                        else if (!*arg) {
3327                                width = strtoul(av[1], &end, 10);
3328                                argcount = 2;
3329                        }
3330                } else if (!prefixcmp(arg, "-name-width")) {
3331                        arg += strlen("-name-width");
3332                        if (*arg == '=')
3333                                name_width = strtoul(arg + 1, &end, 10);
3334                        else if (!*arg && !av[1])
3335                                die("Option '--stat-name-width' requires a value");
3336                        else if (!*arg) {
3337                                name_width = strtoul(av[1], &end, 10);
3338                                argcount = 2;
3339                        }
3340                } else if (!prefixcmp(arg, "-count")) {
3341                        arg += strlen("-count");
3342                        if (*arg == '=')
3343                                count = strtoul(arg + 1, &end, 10);
3344                        else if (!*arg && !av[1])
3345                                die("Option '--stat-count' requires a value");
3346                        else if (!*arg) {
3347                                count = strtoul(av[1], &end, 10);
3348                                argcount = 2;
3349                        }
3350                }
3351                break;
3352        case '=':
3353                width = strtoul(arg+1, &end, 10);
3354                if (*end == ',')
3355                        name_width = strtoul(end+1, &end, 10);
3356                if (*end == ',')
3357                        count = strtoul(end+1, &end, 10);
3358        }
3359
3360        /* Important! This checks all the error cases! */
3361        if (*end)
3362                return 0;
3363        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3364        options->stat_name_width = name_width;
3365        options->stat_width = width;
3366        options->stat_count = count;
3367        return argcount;
3368}
3369
3370static int parse_dirstat_opt(struct diff_options *options, const char *params)
3371{
3372        struct strbuf errmsg = STRBUF_INIT;
3373        if (parse_dirstat_params(options, params, &errmsg))
3374                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3375                    errmsg.buf);
3376        strbuf_release(&errmsg);
3377        /*
3378         * The caller knows a dirstat-related option is given from the command
3379         * line; allow it to say "return this_function();"
3380         */
3381        options->output_format |= DIFF_FORMAT_DIRSTAT;
3382        return 1;
3383}
3384
3385int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3386{
3387        const char *arg = av[0];
3388        const char *optarg;
3389        int argcount;
3390
3391        /* Output format options */
3392        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3393                options->output_format |= DIFF_FORMAT_PATCH;
3394        else if (opt_arg(arg, 'U', "unified", &options->context))
3395                options->output_format |= DIFF_FORMAT_PATCH;
3396        else if (!strcmp(arg, "--raw"))
3397                options->output_format |= DIFF_FORMAT_RAW;
3398        else if (!strcmp(arg, "--patch-with-raw"))
3399                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3400        else if (!strcmp(arg, "--numstat"))
3401                options->output_format |= DIFF_FORMAT_NUMSTAT;
3402        else if (!strcmp(arg, "--shortstat"))
3403                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3404        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3405                return parse_dirstat_opt(options, "");
3406        else if (!prefixcmp(arg, "-X"))
3407                return parse_dirstat_opt(options, arg + 2);
3408        else if (!prefixcmp(arg, "--dirstat="))
3409                return parse_dirstat_opt(options, arg + 10);
3410        else if (!strcmp(arg, "--cumulative"))
3411                return parse_dirstat_opt(options, "cumulative");
3412        else if (!strcmp(arg, "--dirstat-by-file"))
3413                return parse_dirstat_opt(options, "files");
3414        else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3415                parse_dirstat_opt(options, "files");
3416                return parse_dirstat_opt(options, arg + 18);
3417        }
3418        else if (!strcmp(arg, "--check"))
3419                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3420        else if (!strcmp(arg, "--summary"))
3421                options->output_format |= DIFF_FORMAT_SUMMARY;
3422        else if (!strcmp(arg, "--patch-with-stat"))
3423                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3424        else if (!strcmp(arg, "--name-only"))
3425                options->output_format |= DIFF_FORMAT_NAME;
3426        else if (!strcmp(arg, "--name-status"))
3427                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3428        else if (!strcmp(arg, "-s"))
3429                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3430        else if (!prefixcmp(arg, "--stat"))
3431                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3432                return stat_opt(options, av);
3433
3434        /* renames options */
3435        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3436                 !strcmp(arg, "--break-rewrites")) {
3437                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3438                        return error("invalid argument to -B: %s", arg+2);
3439        }
3440        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3441                 !strcmp(arg, "--find-renames")) {
3442                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3443                        return error("invalid argument to -M: %s", arg+2);
3444                options->detect_rename = DIFF_DETECT_RENAME;
3445        }
3446        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3447                options->irreversible_delete = 1;
3448        }
3449        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3450                 !strcmp(arg, "--find-copies")) {
3451                if (options->detect_rename == DIFF_DETECT_COPY)
3452                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3453                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3454                        return error("invalid argument to -C: %s", arg+2);
3455                options->detect_rename = DIFF_DETECT_COPY;
3456        }
3457        else if (!strcmp(arg, "--no-renames"))
3458                options->detect_rename = 0;
3459        else if (!strcmp(arg, "--relative"))
3460                DIFF_OPT_SET(options, RELATIVE_NAME);
3461        else if (!prefixcmp(arg, "--relative=")) {
3462                DIFF_OPT_SET(options, RELATIVE_NAME);
3463                options->prefix = arg + 11;
3464        }
3465
3466        /* xdiff options */
3467        else if (!strcmp(arg, "--minimal"))
3468                DIFF_XDL_SET(options, NEED_MINIMAL);
3469        else if (!strcmp(arg, "--no-minimal"))
3470                DIFF_XDL_CLR(options, NEED_MINIMAL);
3471        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3472                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3473        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3474                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3475        else if (!strcmp(arg, "--ignore-space-at-eol"))
3476                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3477        else if (!strcmp(arg, "--patience"))
3478                DIFF_XDL_SET(options, PATIENCE_DIFF);
3479        else if (!strcmp(arg, "--histogram"))
3480                DIFF_XDL_SET(options, HISTOGRAM_DIFF);
3481
3482        /* flags options */
3483        else if (!strcmp(arg, "--binary")) {
3484                options->output_format |= DIFF_FORMAT_PATCH;
3485                DIFF_OPT_SET(options, BINARY);
3486        }
3487        else if (!strcmp(arg, "--full-index"))
3488                DIFF_OPT_SET(options, FULL_INDEX);
3489        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3490                DIFF_OPT_SET(options, TEXT);
3491        else if (!strcmp(arg, "-R"))
3492                DIFF_OPT_SET(options, REVERSE_DIFF);
3493        else if (!strcmp(arg, "--find-copies-harder"))
3494                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3495        else if (!strcmp(arg, "--follow"))
3496                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3497        else if (!strcmp(arg, "--color"))
3498                options->use_color = 1;
3499        else if (!prefixcmp(arg, "--color=")) {
3500                int value = git_config_colorbool(NULL, arg+8);
3501                if (value < 0)
3502                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3503                options->use_color = value;
3504        }
3505        else if (!strcmp(arg, "--no-color"))
3506                options->use_color = 0;
3507        else if (!strcmp(arg, "--color-words")) {
3508                options->use_color = 1;
3509                options->word_diff = DIFF_WORDS_COLOR;
3510        }
3511        else if (!prefixcmp(arg, "--color-words=")) {
3512                options->use_color = 1;
3513                options->word_diff = DIFF_WORDS_COLOR;
3514                options->word_regex = arg + 14;
3515        }
3516        else if (!strcmp(arg, "--word-diff")) {
3517                if (options->word_diff == DIFF_WORDS_NONE)
3518                        options->word_diff = DIFF_WORDS_PLAIN;
3519        }
3520        else if (!prefixcmp(arg, "--word-diff=")) {
3521                const char *type = arg + 12;
3522                if (!strcmp(type, "plain"))
3523                        options->word_diff = DIFF_WORDS_PLAIN;
3524                else if (!strcmp(type, "color")) {
3525                        options->use_color = 1;
3526                        options->word_diff = DIFF_WORDS_COLOR;
3527                }
3528                else if (!strcmp(type, "porcelain"))
3529                        options->word_diff = DIFF_WORDS_PORCELAIN;
3530                else if (!strcmp(type, "none"))
3531                        options->word_diff = DIFF_WORDS_NONE;
3532                else
3533                        die("bad --word-diff argument: %s", type);
3534        }
3535        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3536                if (options->word_diff == DIFF_WORDS_NONE)
3537                        options->word_diff = DIFF_WORDS_PLAIN;
3538                options->word_regex = optarg;
3539                return argcount;
3540        }
3541        else if (!strcmp(arg, "--exit-code"))
3542                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3543        else if (!strcmp(arg, "--quiet"))
3544                DIFF_OPT_SET(options, QUICK);
3545        else if (!strcmp(arg, "--ext-diff"))
3546                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3547        else if (!strcmp(arg, "--no-ext-diff"))
3548                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3549        else if (!strcmp(arg, "--textconv"))
3550                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3551        else if (!strcmp(arg, "--no-textconv"))
3552                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3553        else if (!strcmp(arg, "--ignore-submodules")) {
3554                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3555                handle_ignore_submodules_arg(options, "all");
3556        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3557                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3558                handle_ignore_submodules_arg(options, arg + 20);
3559        } else if (!strcmp(arg, "--submodule"))
3560                DIFF_OPT_SET(options, SUBMODULE_LOG);
3561        else if (!prefixcmp(arg, "--submodule=")) {
3562                if (!strcmp(arg + 12, "log"))
3563                        DIFF_OPT_SET(options, SUBMODULE_LOG);
3564        }
3565
3566        /* misc options */
3567        else if (!strcmp(arg, "-z"))
3568                options->line_termination = 0;
3569        else if ((argcount = short_opt('l', av, &optarg))) {
3570                options->rename_limit = strtoul(optarg, NULL, 10);
3571                return argcount;
3572        }
3573        else if ((argcount = short_opt('S', av, &optarg))) {
3574                options->pickaxe = optarg;
3575                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3576                return argcount;
3577        } else if ((argcount = short_opt('G', av, &optarg))) {
3578                options->pickaxe = optarg;
3579                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3580                return argcount;
3581        }
3582        else if (!strcmp(arg, "--pickaxe-all"))
3583                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3584        else if (!strcmp(arg, "--pickaxe-regex"))
3585                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3586        else if ((argcount = short_opt('O', av, &optarg))) {
3587                options->orderfile = optarg;
3588                return argcount;
3589        }
3590        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3591                options->filter = optarg;
3592                return argcount;
3593        }
3594        else if (!strcmp(arg, "--abbrev"))
3595                options->abbrev = DEFAULT_ABBREV;
3596        else if (!prefixcmp(arg, "--abbrev=")) {
3597                options->abbrev = strtoul(arg + 9, NULL, 10);
3598                if (options->abbrev < MINIMUM_ABBREV)
3599                        options->abbrev = MINIMUM_ABBREV;
3600                else if (40 < options->abbrev)
3601                        options->abbrev = 40;
3602        }
3603        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3604                options->a_prefix = optarg;
3605                return argcount;
3606        }
3607        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3608                options->b_prefix = optarg;
3609                return argcount;
3610        }
3611        else if (!strcmp(arg, "--no-prefix"))
3612                options->a_prefix = options->b_prefix = "";
3613        else if (opt_arg(arg, '\0', "inter-hunk-context",
3614                         &options->interhunkcontext))
3615                ;
3616        else if (!strcmp(arg, "-W"))
3617                DIFF_OPT_SET(options, FUNCCONTEXT);
3618        else if (!strcmp(arg, "--function-context"))
3619                DIFF_OPT_SET(options, FUNCCONTEXT);
3620        else if (!strcmp(arg, "--no-function-context"))
3621                DIFF_OPT_CLR(options, FUNCCONTEXT);
3622        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3623                options->file = fopen(optarg, "w");
3624                if (!options->file)
3625                        die_errno("Could not open '%s'", optarg);
3626                options->close_file = 1;
3627                return argcount;
3628        } else
3629                return 0;
3630        return 1;
3631}
3632
3633int parse_rename_score(const char **cp_p)
3634{
3635        unsigned long num, scale;
3636        int ch, dot;
3637        const char *cp = *cp_p;
3638
3639        num = 0;
3640        scale = 1;
3641        dot = 0;
3642        for (;;) {
3643                ch = *cp;
3644                if ( !dot && ch == '.' ) {
3645                        scale = 1;
3646                        dot = 1;
3647                } else if ( ch == '%' ) {
3648                        scale = dot ? scale*100 : 100;
3649                        cp++;   /* % is always at the end */
3650                        break;
3651                } else if ( ch >= '0' && ch <= '9' ) {
3652                        if ( scale < 100000 ) {
3653                                scale *= 10;
3654                                num = (num*10) + (ch-'0');
3655                        }
3656                } else {
3657                        break;
3658                }
3659                cp++;
3660        }
3661        *cp_p = cp;
3662
3663        /* user says num divided by scale and we say internally that
3664         * is MAX_SCORE * num / scale.
3665         */
3666        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3667}
3668
3669static int diff_scoreopt_parse(const char *opt)
3670{
3671        int opt1, opt2, cmd;
3672
3673        if (*opt++ != '-')
3674                return -1;
3675        cmd = *opt++;
3676        if (cmd == '-') {
3677                /* convert the long-form arguments into short-form versions */
3678                if (!prefixcmp(opt, "break-rewrites")) {
3679                        opt += strlen("break-rewrites");
3680                        if (*opt == 0 || *opt++ == '=')
3681                                cmd = 'B';
3682                } else if (!prefixcmp(opt, "find-copies")) {
3683                        opt += strlen("find-copies");
3684                        if (*opt == 0 || *opt++ == '=')
3685                                cmd = 'C';
3686                } else if (!prefixcmp(opt, "find-renames")) {
3687                        opt += strlen("find-renames");
3688                        if (*opt == 0 || *opt++ == '=')
3689                                cmd = 'M';
3690                }
3691        }
3692        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3693                return -1; /* that is not a -M, -C nor -B option */
3694
3695        opt1 = parse_rename_score(&opt);
3696        if (cmd != 'B')
3697                opt2 = 0;
3698        else {
3699                if (*opt == 0)
3700                        opt2 = 0;
3701                else if (*opt != '/')
3702                        return -1; /* we expect -B80/99 or -B80 */
3703                else {
3704                        opt++;
3705                        opt2 = parse_rename_score(&opt);
3706                }
3707        }
3708        if (*opt != 0)
3709                return -1;
3710        return opt1 | (opt2 << 16);
3711}
3712
3713struct diff_queue_struct diff_queued_diff;
3714
3715void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3716{
3717        if (queue->alloc <= queue->nr) {
3718                queue->alloc = alloc_nr(queue->alloc);
3719                queue->queue = xrealloc(queue->queue,
3720                                        sizeof(dp) * queue->alloc);
3721        }
3722        queue->queue[queue->nr++] = dp;
3723}
3724
3725struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3726                                 struct diff_filespec *one,
3727                                 struct diff_filespec *two)
3728{
3729        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3730        dp->one = one;
3731        dp->two = two;
3732        if (queue)
3733                diff_q(queue, dp);
3734        return dp;
3735}
3736
3737void diff_free_filepair(struct diff_filepair *p)
3738{
3739        free_filespec(p->one);
3740        free_filespec(p->two);
3741        free(p);
3742}
3743
3744/* This is different from find_unique_abbrev() in that
3745 * it stuffs the result with dots for alignment.
3746 */
3747const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3748{
3749        int abblen;
3750        const char *abbrev;
3751        if (len == 40)
3752                return sha1_to_hex(sha1);
3753
3754        abbrev = find_unique_abbrev(sha1, len);
3755        abblen = strlen(abbrev);
3756        if (abblen < 37) {
3757                static char hex[41];
3758                if (len < abblen && abblen <= len + 2)
3759                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3760                else
3761                        sprintf(hex, "%s...", abbrev);
3762                return hex;
3763        }
3764        return sha1_to_hex(sha1);
3765}
3766
3767static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3768{
3769        int line_termination = opt->line_termination;
3770        int inter_name_termination = line_termination ? '\t' : '\0';
3771        if (opt->output_prefix) {
3772                struct strbuf *msg = NULL;
3773                msg = opt->output_prefix(opt, opt->output_prefix_data);
3774                fprintf(opt->file, "%s", msg->buf);
3775        }
3776
3777        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3778                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3779                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3780                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3781        }
3782        if (p->score) {
3783                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3784                        inter_name_termination);
3785        } else {
3786                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3787        }
3788
3789        if (p->status == DIFF_STATUS_COPIED ||
3790            p->status == DIFF_STATUS_RENAMED) {
3791                const char *name_a, *name_b;
3792                name_a = p->one->path;
3793                name_b = p->two->path;
3794                strip_prefix(opt->prefix_length, &name_a, &name_b);
3795                write_name_quoted(name_a, opt->file, inter_name_termination);
3796                write_name_quoted(name_b, opt->file, line_termination);
3797        } else {
3798                const char *name_a, *name_b;
3799                name_a = p->one->mode ? p->one->path : p->two->path;
3800                name_b = NULL;
3801                strip_prefix(opt->prefix_length, &name_a, &name_b);
3802                write_name_quoted(name_a, opt->file, line_termination);
3803        }
3804}
3805
3806int diff_unmodified_pair(struct diff_filepair *p)
3807{
3808        /* This function is written stricter than necessary to support
3809         * the currently implemented transformers, but the idea is to
3810         * let transformers to produce diff_filepairs any way they want,
3811         * and filter and clean them up here before producing the output.
3812         */
3813        struct diff_filespec *one = p->one, *two = p->two;
3814
3815        if (DIFF_PAIR_UNMERGED(p))
3816                return 0; /* unmerged is interesting */
3817
3818        /* deletion, addition, mode or type change
3819         * and rename are all interesting.
3820         */
3821        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3822            DIFF_PAIR_MODE_CHANGED(p) ||
3823            strcmp(one->path, two->path))
3824                return 0;
3825
3826        /* both are valid and point at the same path.  that is, we are
3827         * dealing with a change.
3828         */
3829        if (one->sha1_valid && two->sha1_valid &&
3830            !hashcmp(one->sha1, two->sha1) &&
3831            !one->dirty_submodule && !two->dirty_submodule)
3832                return 1; /* no change */
3833        if (!one->sha1_valid && !two->sha1_valid)
3834                return 1; /* both look at the same file on the filesystem. */
3835        return 0;
3836}
3837
3838static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3839{
3840        if (diff_unmodified_pair(p))
3841                return;
3842
3843        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3844            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3845                return; /* no tree diffs in patch format */
3846
3847        run_diff(p, o);
3848}
3849
3850static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3851                            struct diffstat_t *diffstat)
3852{
3853        if (diff_unmodified_pair(p))
3854                return;
3855
3856        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3857            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3858                return; /* no useful stat for tree diffs */
3859
3860        run_diffstat(p, o, diffstat);
3861}
3862
3863static void diff_flush_checkdiff(struct diff_filepair *p,
3864                struct diff_options *o)
3865{
3866        if (diff_unmodified_pair(p))
3867                return;
3868
3869        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3870            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3871                return; /* nothing to check in tree diffs */
3872
3873        run_checkdiff(p, o);
3874}
3875
3876int diff_queue_is_empty(void)
3877{
3878        struct diff_queue_struct *q = &diff_queued_diff;
3879        int i;
3880        for (i = 0; i < q->nr; i++)
3881                if (!diff_unmodified_pair(q->queue[i]))
3882                        return 0;
3883        return 1;
3884}
3885
3886#if DIFF_DEBUG
3887void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3888{
3889        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3890                x, one ? one : "",
3891                s->path,
3892                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3893                s->mode,
3894                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3895        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3896                x, one ? one : "",
3897                s->size, s->xfrm_flags);
3898}
3899
3900void diff_debug_filepair(const struct diff_filepair *p, int i)
3901{
3902        diff_debug_filespec(p->one, i, "one");
3903        diff_debug_filespec(p->two, i, "two");
3904        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3905                p->score, p->status ? p->status : '?',
3906                p->one->rename_used, p->broken_pair);
3907}
3908
3909void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3910{
3911        int i;
3912        if (msg)
3913                fprintf(stderr, "%s\n", msg);
3914        fprintf(stderr, "q->nr = %d\n", q->nr);
3915        for (i = 0; i < q->nr; i++) {
3916                struct diff_filepair *p = q->queue[i];
3917                diff_debug_filepair(p, i);
3918        }
3919}
3920#endif
3921
3922static void diff_resolve_rename_copy(void)
3923{
3924        int i;
3925        struct diff_filepair *p;
3926        struct diff_queue_struct *q = &diff_queued_diff;
3927
3928        diff_debug_queue("resolve-rename-copy", q);
3929
3930        for (i = 0; i < q->nr; i++) {
3931                p = q->queue[i];
3932                p->status = 0; /* undecided */
3933                if (DIFF_PAIR_UNMERGED(p))
3934                        p->status = DIFF_STATUS_UNMERGED;
3935                else if (!DIFF_FILE_VALID(p->one))
3936                        p->status = DIFF_STATUS_ADDED;
3937                else if (!DIFF_FILE_VALID(p->two))
3938                        p->status = DIFF_STATUS_DELETED;
3939                else if (DIFF_PAIR_TYPE_CHANGED(p))
3940                        p->status = DIFF_STATUS_TYPE_CHANGED;
3941
3942                /* from this point on, we are dealing with a pair
3943                 * whose both sides are valid and of the same type, i.e.
3944                 * either in-place edit or rename/copy edit.
3945                 */
3946                else if (DIFF_PAIR_RENAME(p)) {
3947                        /*
3948                         * A rename might have re-connected a broken
3949                         * pair up, causing the pathnames to be the
3950                         * same again. If so, that's not a rename at
3951                         * all, just a modification..
3952                         *
3953                         * Otherwise, see if this source was used for
3954                         * multiple renames, in which case we decrement
3955                         * the count, and call it a copy.
3956                         */
3957                        if (!strcmp(p->one->path, p->two->path))
3958                                p->status = DIFF_STATUS_MODIFIED;
3959                        else if (--p->one->rename_used > 0)
3960                                p->status = DIFF_STATUS_COPIED;
3961                        else
3962                                p->status = DIFF_STATUS_RENAMED;
3963                }
3964                else if (hashcmp(p->one->sha1, p->two->sha1) ||
3965                         p->one->mode != p->two->mode ||
3966                         p->one->dirty_submodule ||
3967                         p->two->dirty_submodule ||
3968                         is_null_sha1(p->one->sha1))
3969                        p->status = DIFF_STATUS_MODIFIED;
3970                else {
3971                        /* This is a "no-change" entry and should not
3972                         * happen anymore, but prepare for broken callers.
3973                         */
3974                        error("feeding unmodified %s to diffcore",
3975                              p->one->path);
3976                        p->status = DIFF_STATUS_UNKNOWN;
3977                }
3978        }
3979        diff_debug_queue("resolve-rename-copy done", q);
3980}
3981
3982static int check_pair_status(struct diff_filepair *p)
3983{
3984        switch (p->status) {
3985        case DIFF_STATUS_UNKNOWN:
3986                return 0;
3987        case 0:
3988                die("internal error in diff-resolve-rename-copy");
3989        default:
3990                return 1;
3991        }
3992}
3993
3994static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3995{
3996        int fmt = opt->output_format;
3997
3998        if (fmt & DIFF_FORMAT_CHECKDIFF)
3999                diff_flush_checkdiff(p, opt);
4000        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4001                diff_flush_raw(p, opt);
4002        else if (fmt & DIFF_FORMAT_NAME) {
4003                const char *name_a, *name_b;
4004                name_a = p->two->path;
4005                name_b = NULL;
4006                strip_prefix(opt->prefix_length, &name_a, &name_b);
4007                write_name_quoted(name_a, opt->file, opt->line_termination);
4008        }
4009}
4010
4011static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4012{
4013        if (fs->mode)
4014                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4015        else
4016                fprintf(file, " %s ", newdelete);
4017        write_name_quoted(fs->path, file, '\n');
4018}
4019
4020
4021static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4022                const char *line_prefix)
4023{
4024        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4025                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4026                        p->two->mode, show_name ? ' ' : '\n');
4027                if (show_name) {
4028                        write_name_quoted(p->two->path, file, '\n');
4029                }
4030        }
4031}
4032
4033static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4034                        const char *line_prefix)
4035{
4036        char *names = pprint_rename(p->one->path, p->two->path);
4037
4038        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4039        free(names);
4040        show_mode_change(file, p, 0, line_prefix);
4041}
4042
4043static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4044{
4045        FILE *file = opt->file;
4046        char *line_prefix = "";
4047
4048        if (opt->output_prefix) {
4049                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4050                line_prefix = buf->buf;
4051        }
4052
4053        switch(p->status) {
4054        case DIFF_STATUS_DELETED:
4055                fputs(line_prefix, file);
4056                show_file_mode_name(file, "delete", p->one);
4057                break;
4058        case DIFF_STATUS_ADDED:
4059                fputs(line_prefix, file);
4060                show_file_mode_name(file, "create", p->two);
4061                break;
4062        case DIFF_STATUS_COPIED:
4063                fputs(line_prefix, file);
4064                show_rename_copy(file, "copy", p, line_prefix);
4065                break;
4066        case DIFF_STATUS_RENAMED:
4067                fputs(line_prefix, file);
4068                show_rename_copy(file, "rename", p, line_prefix);
4069                break;
4070        default:
4071                if (p->score) {
4072                        fprintf(file, "%s rewrite ", line_prefix);
4073                        write_name_quoted(p->two->path, file, ' ');
4074                        fprintf(file, "(%d%%)\n", similarity_index(p));
4075                }
4076                show_mode_change(file, p, !p->score, line_prefix);
4077                break;
4078        }
4079}
4080
4081struct patch_id_t {
4082        git_SHA_CTX *ctx;
4083        int patchlen;
4084};
4085
4086static int remove_space(char *line, int len)
4087{
4088        int i;
4089        char *dst = line;
4090        unsigned char c;
4091
4092        for (i = 0; i < len; i++)
4093                if (!isspace((c = line[i])))
4094                        *dst++ = c;
4095
4096        return dst - line;
4097}
4098
4099static void patch_id_consume(void *priv, char *line, unsigned long len)
4100{
4101        struct patch_id_t *data = priv;
4102        int new_len;
4103
4104        /* Ignore line numbers when computing the SHA1 of the patch */
4105        if (!prefixcmp(line, "@@ -"))
4106                return;
4107
4108        new_len = remove_space(line, len);
4109
4110        git_SHA1_Update(data->ctx, line, new_len);
4111        data->patchlen += new_len;
4112}
4113
4114/* returns 0 upon success, and writes result into sha1 */
4115static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4116{
4117        struct diff_queue_struct *q = &diff_queued_diff;
4118        int i;
4119        git_SHA_CTX ctx;
4120        struct patch_id_t data;
4121        char buffer[PATH_MAX * 4 + 20];
4122
4123        git_SHA1_Init(&ctx);
4124        memset(&data, 0, sizeof(struct patch_id_t));
4125        data.ctx = &ctx;
4126
4127        for (i = 0; i < q->nr; i++) {
4128                xpparam_t xpp;
4129                xdemitconf_t xecfg;
4130                mmfile_t mf1, mf2;
4131                struct diff_filepair *p = q->queue[i];
4132                int len1, len2;
4133
4134                memset(&xpp, 0, sizeof(xpp));
4135                memset(&xecfg, 0, sizeof(xecfg));
4136                if (p->status == 0)
4137                        return error("internal diff status error");
4138                if (p->status == DIFF_STATUS_UNKNOWN)
4139                        continue;
4140                if (diff_unmodified_pair(p))
4141                        continue;
4142                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4143                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4144                        continue;
4145                if (DIFF_PAIR_UNMERGED(p))
4146                        continue;
4147
4148                diff_fill_sha1_info(p->one);
4149                diff_fill_sha1_info(p->two);
4150                if (fill_mmfile(&mf1, p->one) < 0 ||
4151                                fill_mmfile(&mf2, p->two) < 0)
4152                        return error("unable to read files to diff");
4153
4154                len1 = remove_space(p->one->path, strlen(p->one->path));
4155                len2 = remove_space(p->two->path, strlen(p->two->path));
4156                if (p->one->mode == 0)
4157                        len1 = snprintf(buffer, sizeof(buffer),
4158                                        "diff--gita/%.*sb/%.*s"
4159                                        "newfilemode%06o"
4160                                        "---/dev/null"
4161                                        "+++b/%.*s",
4162                                        len1, p->one->path,
4163                                        len2, p->two->path,
4164                                        p->two->mode,
4165                                        len2, p->two->path);
4166                else if (p->two->mode == 0)
4167                        len1 = snprintf(buffer, sizeof(buffer),
4168                                        "diff--gita/%.*sb/%.*s"
4169                                        "deletedfilemode%06o"
4170                                        "---a/%.*s"
4171                                        "+++/dev/null",
4172                                        len1, p->one->path,
4173                                        len2, p->two->path,
4174                                        p->one->mode,
4175                                        len1, p->one->path);
4176                else
4177                        len1 = snprintf(buffer, sizeof(buffer),
4178                                        "diff--gita/%.*sb/%.*s"
4179                                        "---a/%.*s"
4180                                        "+++b/%.*s",
4181                                        len1, p->one->path,
4182                                        len2, p->two->path,
4183                                        len1, p->one->path,
4184                                        len2, p->two->path);
4185                git_SHA1_Update(&ctx, buffer, len1);
4186
4187                if (diff_filespec_is_binary(p->one) ||
4188                    diff_filespec_is_binary(p->two)) {
4189                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4190                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4191                        continue;
4192                }
4193
4194                xpp.flags = 0;
4195                xecfg.ctxlen = 3;
4196                xecfg.flags = 0;
4197                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4198                              &xpp, &xecfg);
4199        }
4200
4201        git_SHA1_Final(sha1, &ctx);
4202        return 0;
4203}
4204
4205int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4206{
4207        struct diff_queue_struct *q = &diff_queued_diff;
4208        int i;
4209        int result = diff_get_patch_id(options, sha1);
4210
4211        for (i = 0; i < q->nr; i++)
4212                diff_free_filepair(q->queue[i]);
4213
4214        free(q->queue);
4215        DIFF_QUEUE_CLEAR(q);
4216
4217        return result;
4218}
4219
4220static int is_summary_empty(const struct diff_queue_struct *q)
4221{
4222        int i;
4223
4224        for (i = 0; i < q->nr; i++) {
4225                const struct diff_filepair *p = q->queue[i];
4226
4227                switch (p->status) {
4228                case DIFF_STATUS_DELETED:
4229                case DIFF_STATUS_ADDED:
4230                case DIFF_STATUS_COPIED:
4231                case DIFF_STATUS_RENAMED:
4232                        return 0;
4233                default:
4234                        if (p->score)
4235                                return 0;
4236                        if (p->one->mode && p->two->mode &&
4237                            p->one->mode != p->two->mode)
4238                                return 0;
4239                        break;
4240                }
4241        }
4242        return 1;
4243}
4244
4245static const char rename_limit_warning[] =
4246"inexact rename detection was skipped due to too many files.";
4247
4248static const char degrade_cc_to_c_warning[] =
4249"only found copies from modified paths due to too many files.";
4250
4251static const char rename_limit_advice[] =
4252"you may want to set your %s variable to at least "
4253"%d and retry the command.";
4254
4255void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4256{
4257        if (degraded_cc)
4258                warning(degrade_cc_to_c_warning);
4259        else if (needed)
4260                warning(rename_limit_warning);
4261        else
4262                return;
4263        if (0 < needed && needed < 32767)
4264                warning(rename_limit_advice, varname, needed);
4265}
4266
4267void diff_flush(struct diff_options *options)
4268{
4269        struct diff_queue_struct *q = &diff_queued_diff;
4270        int i, output_format = options->output_format;
4271        int separator = 0;
4272        int dirstat_by_line = 0;
4273
4274        /*
4275         * Order: raw, stat, summary, patch
4276         * or:    name/name-status/checkdiff (other bits clear)
4277         */
4278        if (!q->nr)
4279                goto free_queue;
4280
4281        if (output_format & (DIFF_FORMAT_RAW |
4282                             DIFF_FORMAT_NAME |
4283                             DIFF_FORMAT_NAME_STATUS |
4284                             DIFF_FORMAT_CHECKDIFF)) {
4285                for (i = 0; i < q->nr; i++) {
4286                        struct diff_filepair *p = q->queue[i];
4287                        if (check_pair_status(p))
4288                                flush_one_pair(p, options);
4289                }
4290                separator++;
4291        }
4292
4293        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4294                dirstat_by_line = 1;
4295
4296        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4297            dirstat_by_line) {
4298                struct diffstat_t diffstat;
4299
4300                memset(&diffstat, 0, sizeof(struct diffstat_t));
4301                for (i = 0; i < q->nr; i++) {
4302                        struct diff_filepair *p = q->queue[i];
4303                        if (check_pair_status(p))
4304                                diff_flush_stat(p, options, &diffstat);
4305                }
4306                if (output_format & DIFF_FORMAT_NUMSTAT)
4307                        show_numstat(&diffstat, options);
4308                if (output_format & DIFF_FORMAT_DIFFSTAT)
4309                        show_stats(&diffstat, options);
4310                if (output_format & DIFF_FORMAT_SHORTSTAT)
4311                        show_shortstats(&diffstat, options);
4312                if (output_format & DIFF_FORMAT_DIRSTAT)
4313                        show_dirstat_by_line(&diffstat, options);
4314                free_diffstat_info(&diffstat);
4315                separator++;
4316        }
4317        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4318                show_dirstat(options);
4319
4320        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4321                for (i = 0; i < q->nr; i++) {
4322                        diff_summary(options, q->queue[i]);
4323                }
4324                separator++;
4325        }
4326
4327        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4328            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4329            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4330                /*
4331                 * run diff_flush_patch for the exit status. setting
4332                 * options->file to /dev/null should be safe, becaue we
4333                 * aren't supposed to produce any output anyway.
4334                 */
4335                if (options->close_file)
4336                        fclose(options->file);
4337                options->file = fopen("/dev/null", "w");
4338                if (!options->file)
4339                        die_errno("Could not open /dev/null");
4340                options->close_file = 1;
4341                for (i = 0; i < q->nr; i++) {
4342                        struct diff_filepair *p = q->queue[i];
4343                        if (check_pair_status(p))
4344                                diff_flush_patch(p, options);
4345                        if (options->found_changes)
4346                                break;
4347                }
4348        }
4349
4350        if (output_format & DIFF_FORMAT_PATCH) {
4351                if (separator) {
4352                        putc(options->line_termination, options->file);
4353                        if (options->stat_sep) {
4354                                /* attach patch instead of inline */
4355                                fputs(options->stat_sep, options->file);
4356                        }
4357                }
4358
4359                for (i = 0; i < q->nr; i++) {
4360                        struct diff_filepair *p = q->queue[i];
4361                        if (check_pair_status(p))
4362                                diff_flush_patch(p, options);
4363                }
4364        }
4365
4366        if (output_format & DIFF_FORMAT_CALLBACK)
4367                options->format_callback(q, options, options->format_callback_data);
4368
4369        for (i = 0; i < q->nr; i++)
4370                diff_free_filepair(q->queue[i]);
4371free_queue:
4372        free(q->queue);
4373        DIFF_QUEUE_CLEAR(q);
4374        if (options->close_file)
4375                fclose(options->file);
4376
4377        /*
4378         * Report the content-level differences with HAS_CHANGES;
4379         * diff_addremove/diff_change does not set the bit when
4380         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4381         */
4382        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4383                if (options->found_changes)
4384                        DIFF_OPT_SET(options, HAS_CHANGES);
4385                else
4386                        DIFF_OPT_CLR(options, HAS_CHANGES);
4387        }
4388}
4389
4390static void diffcore_apply_filter(const char *filter)
4391{
4392        int i;
4393        struct diff_queue_struct *q = &diff_queued_diff;
4394        struct diff_queue_struct outq;
4395        DIFF_QUEUE_CLEAR(&outq);
4396
4397        if (!filter)
4398                return;
4399
4400        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4401                int found;
4402                for (i = found = 0; !found && i < q->nr; i++) {
4403                        struct diff_filepair *p = q->queue[i];
4404                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4405                             ((p->score &&
4406                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4407                              (!p->score &&
4408                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4409                            ((p->status != DIFF_STATUS_MODIFIED) &&
4410                             strchr(filter, p->status)))
4411                                found++;
4412                }
4413                if (found)
4414                        return;
4415
4416                /* otherwise we will clear the whole queue
4417                 * by copying the empty outq at the end of this
4418                 * function, but first clear the current entries
4419                 * in the queue.
4420                 */
4421                for (i = 0; i < q->nr; i++)
4422                        diff_free_filepair(q->queue[i]);
4423        }
4424        else {
4425                /* Only the matching ones */
4426                for (i = 0; i < q->nr; i++) {
4427                        struct diff_filepair *p = q->queue[i];
4428
4429                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4430                             ((p->score &&
4431                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4432                              (!p->score &&
4433                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4434                            ((p->status != DIFF_STATUS_MODIFIED) &&
4435                             strchr(filter, p->status)))
4436                                diff_q(&outq, p);
4437                        else
4438                                diff_free_filepair(p);
4439                }
4440        }
4441        free(q->queue);
4442        *q = outq;
4443}
4444
4445/* Check whether two filespecs with the same mode and size are identical */
4446static int diff_filespec_is_identical(struct diff_filespec *one,
4447                                      struct diff_filespec *two)
4448{
4449        if (S_ISGITLINK(one->mode))
4450                return 0;
4451        if (diff_populate_filespec(one, 0))
4452                return 0;
4453        if (diff_populate_filespec(two, 0))
4454                return 0;
4455        return !memcmp(one->data, two->data, one->size);
4456}
4457
4458static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4459{
4460        int i;
4461        struct diff_queue_struct *q = &diff_queued_diff;
4462        struct diff_queue_struct outq;
4463        DIFF_QUEUE_CLEAR(&outq);
4464
4465        for (i = 0; i < q->nr; i++) {
4466                struct diff_filepair *p = q->queue[i];
4467
4468                /*
4469                 * 1. Entries that come from stat info dirtiness
4470                 *    always have both sides (iow, not create/delete),
4471                 *    one side of the object name is unknown, with
4472                 *    the same mode and size.  Keep the ones that
4473                 *    do not match these criteria.  They have real
4474                 *    differences.
4475                 *
4476                 * 2. At this point, the file is known to be modified,
4477                 *    with the same mode and size, and the object
4478                 *    name of one side is unknown.  Need to inspect
4479                 *    the identical contents.
4480                 */
4481                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4482                    !DIFF_FILE_VALID(p->two) ||
4483                    (p->one->sha1_valid && p->two->sha1_valid) ||
4484                    (p->one->mode != p->two->mode) ||
4485                    diff_populate_filespec(p->one, 1) ||
4486                    diff_populate_filespec(p->two, 1) ||
4487                    (p->one->size != p->two->size) ||
4488                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4489                        diff_q(&outq, p);
4490                else {
4491                        /*
4492                         * The caller can subtract 1 from skip_stat_unmatch
4493                         * to determine how many paths were dirty only
4494                         * due to stat info mismatch.
4495                         */
4496                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4497                                diffopt->skip_stat_unmatch++;
4498                        diff_free_filepair(p);
4499                }
4500        }
4501        free(q->queue);
4502        *q = outq;
4503}
4504
4505static int diffnamecmp(const void *a_, const void *b_)
4506{
4507        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4508        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4509        const char *name_a, *name_b;
4510
4511        name_a = a->one ? a->one->path : a->two->path;
4512        name_b = b->one ? b->one->path : b->two->path;
4513        return strcmp(name_a, name_b);
4514}
4515
4516void diffcore_fix_diff_index(struct diff_options *options)
4517{
4518        struct diff_queue_struct *q = &diff_queued_diff;
4519        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4520}
4521
4522void diffcore_std(struct diff_options *options)
4523{
4524        if (options->skip_stat_unmatch)
4525                diffcore_skip_stat_unmatch(options);
4526        if (!options->found_follow) {
4527                /* See try_to_follow_renames() in tree-diff.c */
4528                if (options->break_opt != -1)
4529                        diffcore_break(options->break_opt);
4530                if (options->detect_rename)
4531                        diffcore_rename(options);
4532                if (options->break_opt != -1)
4533                        diffcore_merge_broken();
4534        }
4535        if (options->pickaxe)
4536                diffcore_pickaxe(options);
4537        if (options->orderfile)
4538                diffcore_order(options->orderfile);
4539        if (!options->found_follow)
4540                /* See try_to_follow_renames() in tree-diff.c */
4541                diff_resolve_rename_copy();
4542        diffcore_apply_filter(options->filter);
4543
4544        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4545                DIFF_OPT_SET(options, HAS_CHANGES);
4546        else
4547                DIFF_OPT_CLR(options, HAS_CHANGES);
4548
4549        options->found_follow = 0;
4550}
4551
4552int diff_result_code(struct diff_options *opt, int status)
4553{
4554        int result = 0;
4555
4556        diff_warn_rename_limit("diff.renamelimit",
4557                               opt->needed_rename_limit,
4558                               opt->degraded_cc_to_c);
4559        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4560            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4561                return status;
4562        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4563            DIFF_OPT_TST(opt, HAS_CHANGES))
4564                result |= 01;
4565        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4566            DIFF_OPT_TST(opt, CHECK_FAILED))
4567                result |= 02;
4568        return result;
4569}
4570
4571int diff_can_quit_early(struct diff_options *opt)
4572{
4573        return (DIFF_OPT_TST(opt, QUICK) &&
4574                !opt->filter &&
4575                DIFF_OPT_TST(opt, HAS_CHANGES));
4576}
4577
4578/*
4579 * Shall changes to this submodule be ignored?
4580 *
4581 * Submodule changes can be configured to be ignored separately for each path,
4582 * but that configuration can be overridden from the command line.
4583 */
4584static int is_submodule_ignored(const char *path, struct diff_options *options)
4585{
4586        int ignored = 0;
4587        unsigned orig_flags = options->flags;
4588        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4589                set_diffopt_flags_from_submodule_config(options, path);
4590        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4591                ignored = 1;
4592        options->flags = orig_flags;
4593        return ignored;
4594}
4595
4596void diff_addremove(struct diff_options *options,
4597                    int addremove, unsigned mode,
4598                    const unsigned char *sha1,
4599                    const char *concatpath, unsigned dirty_submodule)
4600{
4601        struct diff_filespec *one, *two;
4602
4603        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4604                return;
4605
4606        /* This may look odd, but it is a preparation for
4607         * feeding "there are unchanged files which should
4608         * not produce diffs, but when you are doing copy
4609         * detection you would need them, so here they are"
4610         * entries to the diff-core.  They will be prefixed
4611         * with something like '=' or '*' (I haven't decided
4612         * which but should not make any difference).
4613         * Feeding the same new and old to diff_change()
4614         * also has the same effect.
4615         * Before the final output happens, they are pruned after
4616         * merged into rename/copy pairs as appropriate.
4617         */
4618        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4619                addremove = (addremove == '+' ? '-' :
4620                             addremove == '-' ? '+' : addremove);
4621
4622        if (options->prefix &&
4623            strncmp(concatpath, options->prefix, options->prefix_length))
4624                return;
4625
4626        one = alloc_filespec(concatpath);
4627        two = alloc_filespec(concatpath);
4628
4629        if (addremove != '+')
4630                fill_filespec(one, sha1, mode);
4631        if (addremove != '-') {
4632                fill_filespec(two, sha1, mode);
4633                two->dirty_submodule = dirty_submodule;
4634        }
4635
4636        diff_queue(&diff_queued_diff, one, two);
4637        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4638                DIFF_OPT_SET(options, HAS_CHANGES);
4639}
4640
4641void diff_change(struct diff_options *options,
4642                 unsigned old_mode, unsigned new_mode,
4643                 const unsigned char *old_sha1,
4644                 const unsigned char *new_sha1,
4645                 const char *concatpath,
4646                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4647{
4648        struct diff_filespec *one, *two;
4649
4650        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4651            is_submodule_ignored(concatpath, options))
4652                return;
4653
4654        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4655                unsigned tmp;
4656                const unsigned char *tmp_c;
4657                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4658                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4659                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4660                        new_dirty_submodule = tmp;
4661        }
4662
4663        if (options->prefix &&
4664            strncmp(concatpath, options->prefix, options->prefix_length))
4665                return;
4666
4667        one = alloc_filespec(concatpath);
4668        two = alloc_filespec(concatpath);
4669        fill_filespec(one, old_sha1, old_mode);
4670        fill_filespec(two, new_sha1, new_mode);
4671        one->dirty_submodule = old_dirty_submodule;
4672        two->dirty_submodule = new_dirty_submodule;
4673
4674        diff_queue(&diff_queued_diff, one, two);
4675        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4676                DIFF_OPT_SET(options, HAS_CHANGES);
4677}
4678
4679struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4680{
4681        struct diff_filepair *pair;
4682        struct diff_filespec *one, *two;
4683
4684        if (options->prefix &&
4685            strncmp(path, options->prefix, options->prefix_length))
4686                return NULL;
4687
4688        one = alloc_filespec(path);
4689        two = alloc_filespec(path);
4690        pair = diff_queue(&diff_queued_diff, one, two);
4691        pair->is_unmerged = 1;
4692        return pair;
4693}
4694
4695static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4696                size_t *outsize)
4697{
4698        struct diff_tempfile *temp;
4699        const char *argv[3];
4700        const char **arg = argv;
4701        struct child_process child;
4702        struct strbuf buf = STRBUF_INIT;
4703        int err = 0;
4704
4705        temp = prepare_temp_file(spec->path, spec);
4706        *arg++ = pgm;
4707        *arg++ = temp->name;
4708        *arg = NULL;
4709
4710        memset(&child, 0, sizeof(child));
4711        child.use_shell = 1;
4712        child.argv = argv;
4713        child.out = -1;
4714        if (start_command(&child)) {
4715                remove_tempfile();
4716                return NULL;
4717        }
4718
4719        if (strbuf_read(&buf, child.out, 0) < 0)
4720                err = error("error reading from textconv command '%s'", pgm);
4721        close(child.out);
4722
4723        if (finish_command(&child) || err) {
4724                strbuf_release(&buf);
4725                remove_tempfile();
4726                return NULL;
4727        }
4728        remove_tempfile();
4729
4730        return strbuf_detach(&buf, outsize);
4731}
4732
4733size_t fill_textconv(struct userdiff_driver *driver,
4734                     struct diff_filespec *df,
4735                     char **outbuf)
4736{
4737        size_t size;
4738
4739        if (!driver || !driver->textconv) {
4740                if (!DIFF_FILE_VALID(df)) {
4741                        *outbuf = "";
4742                        return 0;
4743                }
4744                if (diff_populate_filespec(df, 0))
4745                        die("unable to read files to diff");
4746                *outbuf = df->data;
4747                return df->size;
4748        }
4749
4750        if (driver->textconv_cache && df->sha1_valid) {
4751                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4752                                          &size);
4753                if (*outbuf)
4754                        return size;
4755        }
4756
4757        *outbuf = run_textconv(driver->textconv, df, &size);
4758        if (!*outbuf)
4759                die("unable to read files to diff");
4760
4761        if (driver->textconv_cache && df->sha1_valid) {
4762                /* ignore errors, as we might be in a readonly repository */
4763                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4764                                size);
4765                /*
4766                 * we could save up changes and flush them all at the end,
4767                 * but we would need an extra call after all diffing is done.
4768                 * Since generating a cache entry is the slow path anyway,
4769                 * this extra overhead probably isn't a big deal.
4770                 */
4771                notes_cache_write(driver->textconv_cache);
4772        }
4773
4774        return size;
4775}