diff.con commit diff: use large integers for diffstat calculations (0974c11)
   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
  18#ifdef NO_FAST_WORKING_DIRECTORY
  19#define FAST_WORKING_DIRECTORY 0
  20#else
  21#define FAST_WORKING_DIRECTORY 1
  22#endif
  23
  24static int diff_detect_rename_default;
  25static int diff_rename_limit_default = 200;
  26static int diff_suppress_blank_empty;
  27int diff_use_color_default = -1;
  28static const char *diff_word_regex_cfg;
  29static const char *external_diff_cmd_cfg;
  30int diff_auto_refresh_index = 1;
  31static int diff_mnemonic_prefix;
  32
  33static char diff_colors[][COLOR_MAXLEN] = {
  34        GIT_COLOR_RESET,
  35        GIT_COLOR_NORMAL,       /* PLAIN */
  36        GIT_COLOR_BOLD,         /* METAINFO */
  37        GIT_COLOR_CYAN,         /* FRAGINFO */
  38        GIT_COLOR_RED,          /* OLD */
  39        GIT_COLOR_GREEN,        /* NEW */
  40        GIT_COLOR_YELLOW,       /* COMMIT */
  41        GIT_COLOR_BG_RED,       /* WHITESPACE */
  42        GIT_COLOR_NORMAL,       /* FUNCINFO */
  43};
  44
  45static void diff_filespec_load_driver(struct diff_filespec *one);
  46static char *run_textconv(const char *, struct diff_filespec *, size_t *);
  47
  48static int parse_diff_color_slot(const char *var, int ofs)
  49{
  50        if (!strcasecmp(var+ofs, "plain"))
  51                return DIFF_PLAIN;
  52        if (!strcasecmp(var+ofs, "meta"))
  53                return DIFF_METAINFO;
  54        if (!strcasecmp(var+ofs, "frag"))
  55                return DIFF_FRAGINFO;
  56        if (!strcasecmp(var+ofs, "old"))
  57                return DIFF_FILE_OLD;
  58        if (!strcasecmp(var+ofs, "new"))
  59                return DIFF_FILE_NEW;
  60        if (!strcasecmp(var+ofs, "commit"))
  61                return DIFF_COMMIT;
  62        if (!strcasecmp(var+ofs, "whitespace"))
  63                return DIFF_WHITESPACE;
  64        if (!strcasecmp(var+ofs, "func"))
  65                return DIFF_FUNCINFO;
  66        return -1;
  67}
  68
  69static int git_config_rename(const char *var, const char *value)
  70{
  71        if (!value)
  72                return DIFF_DETECT_RENAME;
  73        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
  74                return  DIFF_DETECT_COPY;
  75        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
  76}
  77
  78/*
  79 * These are to give UI layer defaults.
  80 * The core-level commands such as git-diff-files should
  81 * never be affected by the setting of diff.renames
  82 * the user happens to have in the configuration file.
  83 */
  84int git_diff_ui_config(const char *var, const char *value, void *cb)
  85{
  86        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
  87                diff_use_color_default = git_config_colorbool(var, value, -1);
  88                return 0;
  89        }
  90        if (!strcmp(var, "diff.renames")) {
  91                diff_detect_rename_default = git_config_rename(var, value);
  92                return 0;
  93        }
  94        if (!strcmp(var, "diff.autorefreshindex")) {
  95                diff_auto_refresh_index = git_config_bool(var, value);
  96                return 0;
  97        }
  98        if (!strcmp(var, "diff.mnemonicprefix")) {
  99                diff_mnemonic_prefix = git_config_bool(var, value);
 100                return 0;
 101        }
 102        if (!strcmp(var, "diff.external"))
 103                return git_config_string(&external_diff_cmd_cfg, var, value);
 104        if (!strcmp(var, "diff.wordregex"))
 105                return git_config_string(&diff_word_regex_cfg, var, value);
 106
 107        return git_diff_basic_config(var, value, cb);
 108}
 109
 110int git_diff_basic_config(const char *var, const char *value, void *cb)
 111{
 112        if (!strcmp(var, "diff.renamelimit")) {
 113                diff_rename_limit_default = git_config_int(var, value);
 114                return 0;
 115        }
 116
 117        switch (userdiff_config(var, value)) {
 118                case 0: break;
 119                case -1: return -1;
 120                default: return 0;
 121        }
 122
 123        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 124                int slot = parse_diff_color_slot(var, 11);
 125                if (slot < 0)
 126                        return 0;
 127                if (!value)
 128                        return config_error_nonbool(var);
 129                color_parse(value, var, diff_colors[slot]);
 130                return 0;
 131        }
 132
 133        /* like GNU diff's --suppress-blank-empty option  */
 134        if (!strcmp(var, "diff.suppressblankempty") ||
 135                        /* for backwards compatibility */
 136                        !strcmp(var, "diff.suppress-blank-empty")) {
 137                diff_suppress_blank_empty = git_config_bool(var, value);
 138                return 0;
 139        }
 140
 141        return git_color_default_config(var, value, cb);
 142}
 143
 144static char *quote_two(const char *one, const char *two)
 145{
 146        int need_one = quote_c_style(one, NULL, NULL, 1);
 147        int need_two = quote_c_style(two, NULL, NULL, 1);
 148        struct strbuf res = STRBUF_INIT;
 149
 150        if (need_one + need_two) {
 151                strbuf_addch(&res, '"');
 152                quote_c_style(one, &res, NULL, 1);
 153                quote_c_style(two, &res, NULL, 1);
 154                strbuf_addch(&res, '"');
 155        } else {
 156                strbuf_addstr(&res, one);
 157                strbuf_addstr(&res, two);
 158        }
 159        return strbuf_detach(&res, NULL);
 160}
 161
 162static const char *external_diff(void)
 163{
 164        static const char *external_diff_cmd = NULL;
 165        static int done_preparing = 0;
 166
 167        if (done_preparing)
 168                return external_diff_cmd;
 169        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 170        if (!external_diff_cmd)
 171                external_diff_cmd = external_diff_cmd_cfg;
 172        done_preparing = 1;
 173        return external_diff_cmd;
 174}
 175
 176static struct diff_tempfile {
 177        const char *name; /* filename external diff should read from */
 178        char hex[41];
 179        char mode[10];
 180        char tmp_path[PATH_MAX];
 181} diff_temp[2];
 182
 183typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 184
 185struct emit_callback {
 186        int color_diff;
 187        unsigned ws_rule;
 188        int blank_at_eof_in_preimage;
 189        int blank_at_eof_in_postimage;
 190        int lno_in_preimage;
 191        int lno_in_postimage;
 192        sane_truncate_fn truncate;
 193        const char **label_path;
 194        struct diff_words_data *diff_words;
 195        int *found_changesp;
 196        FILE *file;
 197};
 198
 199static int count_lines(const char *data, int size)
 200{
 201        int count, ch, completely_empty = 1, nl_just_seen = 0;
 202        count = 0;
 203        while (0 < size--) {
 204                ch = *data++;
 205                if (ch == '\n') {
 206                        count++;
 207                        nl_just_seen = 1;
 208                        completely_empty = 0;
 209                }
 210                else {
 211                        nl_just_seen = 0;
 212                        completely_empty = 0;
 213                }
 214        }
 215        if (completely_empty)
 216                return 0;
 217        if (!nl_just_seen)
 218                count++; /* no trailing newline */
 219        return count;
 220}
 221
 222static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 223{
 224        if (!DIFF_FILE_VALID(one)) {
 225                mf->ptr = (char *)""; /* does not matter */
 226                mf->size = 0;
 227                return 0;
 228        }
 229        else if (diff_populate_filespec(one, 0))
 230                return -1;
 231
 232        mf->ptr = one->data;
 233        mf->size = one->size;
 234        return 0;
 235}
 236
 237static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 238{
 239        char *ptr = mf->ptr;
 240        long size = mf->size;
 241        int cnt = 0;
 242
 243        if (!size)
 244                return cnt;
 245        ptr += size - 1; /* pointing at the very end */
 246        if (*ptr != '\n')
 247                ; /* incomplete line */
 248        else
 249                ptr--; /* skip the last LF */
 250        while (mf->ptr < ptr) {
 251                char *prev_eol;
 252                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 253                        if (*prev_eol == '\n')
 254                                break;
 255                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 256                        break;
 257                cnt++;
 258                ptr = prev_eol - 1;
 259        }
 260        return cnt;
 261}
 262
 263static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 264                               struct emit_callback *ecbdata)
 265{
 266        int l1, l2, at;
 267        unsigned ws_rule = ecbdata->ws_rule;
 268        l1 = count_trailing_blank(mf1, ws_rule);
 269        l2 = count_trailing_blank(mf2, ws_rule);
 270        if (l2 <= l1) {
 271                ecbdata->blank_at_eof_in_preimage = 0;
 272                ecbdata->blank_at_eof_in_postimage = 0;
 273                return;
 274        }
 275        at = count_lines(mf1->ptr, mf1->size);
 276        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 277
 278        at = count_lines(mf2->ptr, mf2->size);
 279        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 280}
 281
 282static void emit_line_0(FILE *file, const char *set, const char *reset,
 283                        int first, const char *line, int len)
 284{
 285        int has_trailing_newline, has_trailing_carriage_return;
 286        int nofirst;
 287
 288        if (len == 0) {
 289                has_trailing_newline = (first == '\n');
 290                has_trailing_carriage_return = (!has_trailing_newline &&
 291                                                (first == '\r'));
 292                nofirst = has_trailing_newline || has_trailing_carriage_return;
 293        } else {
 294                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 295                if (has_trailing_newline)
 296                        len--;
 297                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 298                if (has_trailing_carriage_return)
 299                        len--;
 300                nofirst = 0;
 301        }
 302
 303        if (len || !nofirst) {
 304                fputs(set, file);
 305                if (!nofirst)
 306                        fputc(first, file);
 307                fwrite(line, len, 1, file);
 308                fputs(reset, file);
 309        }
 310        if (has_trailing_carriage_return)
 311                fputc('\r', file);
 312        if (has_trailing_newline)
 313                fputc('\n', file);
 314}
 315
 316static void emit_line(FILE *file, const char *set, const char *reset,
 317                      const char *line, int len)
 318{
 319        emit_line_0(file, set, reset, line[0], line+1, len-1);
 320}
 321
 322static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 323{
 324        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 325              ecbdata->blank_at_eof_in_preimage &&
 326              ecbdata->blank_at_eof_in_postimage &&
 327              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 328              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 329                return 0;
 330        return ws_blank_line(line, len, ecbdata->ws_rule);
 331}
 332
 333static void emit_add_line(const char *reset,
 334                          struct emit_callback *ecbdata,
 335                          const char *line, int len)
 336{
 337        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 338        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 339
 340        if (!*ws)
 341                emit_line_0(ecbdata->file, set, reset, '+', line, len);
 342        else if (new_blank_line_at_eof(ecbdata, line, len))
 343                /* Blank line at EOF - paint '+' as well */
 344                emit_line_0(ecbdata->file, ws, reset, '+', line, len);
 345        else {
 346                /* Emit just the prefix, then the rest. */
 347                emit_line_0(ecbdata->file, set, reset, '+', "", 0);
 348                ws_check_emit(line, len, ecbdata->ws_rule,
 349                              ecbdata->file, set, reset, ws);
 350        }
 351}
 352
 353static void emit_hunk_header(struct emit_callback *ecbdata,
 354                             const char *line, int len)
 355{
 356        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 357        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 358        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 359        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 360        static const char atat[2] = { '@', '@' };
 361        const char *cp, *ep;
 362
 363        /*
 364         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 365         * it always is at least 10 bytes long.
 366         */
 367        if (len < 10 ||
 368            memcmp(line, atat, 2) ||
 369            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 370                emit_line(ecbdata->file, plain, reset, line, len);
 371                return;
 372        }
 373        ep += 2; /* skip over @@ */
 374
 375        /* The hunk header in fraginfo color */
 376        emit_line(ecbdata->file, frag, reset, line, ep - line);
 377
 378        /* blank before the func header */
 379        for (cp = ep; ep - line < len; ep++)
 380                if (*ep != ' ' && *ep != '\t')
 381                        break;
 382        if (ep != cp)
 383                emit_line(ecbdata->file, plain, reset, cp, ep - cp);
 384
 385        if (ep < line + len)
 386                emit_line(ecbdata->file, func, reset, ep, line + len - ep);
 387}
 388
 389static struct diff_tempfile *claim_diff_tempfile(void) {
 390        int i;
 391        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 392                if (!diff_temp[i].name)
 393                        return diff_temp + i;
 394        die("BUG: diff is failing to clean up its tempfiles");
 395}
 396
 397static int remove_tempfile_installed;
 398
 399static void remove_tempfile(void)
 400{
 401        int i;
 402        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 403                if (diff_temp[i].name == diff_temp[i].tmp_path)
 404                        unlink_or_warn(diff_temp[i].name);
 405                diff_temp[i].name = NULL;
 406        }
 407}
 408
 409static void remove_tempfile_on_signal(int signo)
 410{
 411        remove_tempfile();
 412        sigchain_pop(signo);
 413        raise(signo);
 414}
 415
 416static void print_line_count(FILE *file, int count)
 417{
 418        switch (count) {
 419        case 0:
 420                fprintf(file, "0,0");
 421                break;
 422        case 1:
 423                fprintf(file, "1");
 424                break;
 425        default:
 426                fprintf(file, "1,%d", count);
 427                break;
 428        }
 429}
 430
 431static void emit_rewrite_lines(struct emit_callback *ecb,
 432                               int prefix, const char *data, int size)
 433{
 434        const char *endp = NULL;
 435        static const char *nneof = " No newline at end of file\n";
 436        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 437        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 438
 439        while (0 < size) {
 440                int len;
 441
 442                endp = memchr(data, '\n', size);
 443                len = endp ? (endp - data + 1) : size;
 444                if (prefix != '+') {
 445                        ecb->lno_in_preimage++;
 446                        emit_line_0(ecb->file, old, reset, '-',
 447                                    data, len);
 448                } else {
 449                        ecb->lno_in_postimage++;
 450                        emit_add_line(reset, ecb, data, len);
 451                }
 452                size -= len;
 453                data += len;
 454        }
 455        if (!endp) {
 456                const char *plain = diff_get_color(ecb->color_diff,
 457                                                   DIFF_PLAIN);
 458                emit_line_0(ecb->file, plain, reset, '\\',
 459                            nneof, strlen(nneof));
 460        }
 461}
 462
 463static void emit_rewrite_diff(const char *name_a,
 464                              const char *name_b,
 465                              struct diff_filespec *one,
 466                              struct diff_filespec *two,
 467                              const char *textconv_one,
 468                              const char *textconv_two,
 469                              struct diff_options *o)
 470{
 471        int lc_a, lc_b;
 472        int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
 473        const char *name_a_tab, *name_b_tab;
 474        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 475        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 476        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 477        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 478        const char *a_prefix, *b_prefix;
 479        const char *data_one, *data_two;
 480        size_t size_one, size_two;
 481        struct emit_callback ecbdata;
 482
 483        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 484                a_prefix = o->b_prefix;
 485                b_prefix = o->a_prefix;
 486        } else {
 487                a_prefix = o->a_prefix;
 488                b_prefix = o->b_prefix;
 489        }
 490
 491        name_a += (*name_a == '/');
 492        name_b += (*name_b == '/');
 493        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 494        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 495
 496        strbuf_reset(&a_name);
 497        strbuf_reset(&b_name);
 498        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 499        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 500
 501        diff_populate_filespec(one, 0);
 502        diff_populate_filespec(two, 0);
 503        if (textconv_one) {
 504                data_one = run_textconv(textconv_one, one, &size_one);
 505                if (!data_one)
 506                        die("unable to read files to diff");
 507        }
 508        else {
 509                data_one = one->data;
 510                size_one = one->size;
 511        }
 512        if (textconv_two) {
 513                data_two = run_textconv(textconv_two, two, &size_two);
 514                if (!data_two)
 515                        die("unable to read files to diff");
 516        }
 517        else {
 518                data_two = two->data;
 519                size_two = two->size;
 520        }
 521
 522        memset(&ecbdata, 0, sizeof(ecbdata));
 523        ecbdata.color_diff = color_diff;
 524        ecbdata.found_changesp = &o->found_changes;
 525        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 526        ecbdata.file = o->file;
 527        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 528                mmfile_t mf1, mf2;
 529                mf1.ptr = (char *)data_one;
 530                mf2.ptr = (char *)data_two;
 531                mf1.size = size_one;
 532                mf2.size = size_two;
 533                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 534        }
 535        ecbdata.lno_in_preimage = 1;
 536        ecbdata.lno_in_postimage = 1;
 537
 538        lc_a = count_lines(data_one, size_one);
 539        lc_b = count_lines(data_two, size_two);
 540        fprintf(o->file,
 541                "%s--- %s%s%s\n%s+++ %s%s%s\n%s@@ -",
 542                metainfo, a_name.buf, name_a_tab, reset,
 543                metainfo, b_name.buf, name_b_tab, reset, fraginfo);
 544        print_line_count(o->file, lc_a);
 545        fprintf(o->file, " +");
 546        print_line_count(o->file, lc_b);
 547        fprintf(o->file, " @@%s\n", reset);
 548        if (lc_a)
 549                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 550        if (lc_b)
 551                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 552}
 553
 554struct diff_words_buffer {
 555        mmfile_t text;
 556        long alloc;
 557        struct diff_words_orig {
 558                const char *begin, *end;
 559        } *orig;
 560        int orig_nr, orig_alloc;
 561};
 562
 563static void diff_words_append(char *line, unsigned long len,
 564                struct diff_words_buffer *buffer)
 565{
 566        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 567        line++;
 568        len--;
 569        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 570        buffer->text.size += len;
 571        buffer->text.ptr[buffer->text.size] = '\0';
 572}
 573
 574struct diff_words_data {
 575        struct diff_words_buffer minus, plus;
 576        const char *current_plus;
 577        FILE *file;
 578        regex_t *word_regex;
 579};
 580
 581static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 582{
 583        struct diff_words_data *diff_words = priv;
 584        int minus_first, minus_len, plus_first, plus_len;
 585        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 586
 587        if (line[0] != '@' || parse_hunk_header(line, len,
 588                        &minus_first, &minus_len, &plus_first, &plus_len))
 589                return;
 590
 591        /* POSIX requires that first be decremented by one if len == 0... */
 592        if (minus_len) {
 593                minus_begin = diff_words->minus.orig[minus_first].begin;
 594                minus_end =
 595                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 596        } else
 597                minus_begin = minus_end =
 598                        diff_words->minus.orig[minus_first].end;
 599
 600        if (plus_len) {
 601                plus_begin = diff_words->plus.orig[plus_first].begin;
 602                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 603        } else
 604                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 605
 606        if (diff_words->current_plus != plus_begin)
 607                fwrite(diff_words->current_plus,
 608                                plus_begin - diff_words->current_plus, 1,
 609                                diff_words->file);
 610        if (minus_begin != minus_end)
 611                color_fwrite_lines(diff_words->file,
 612                                diff_get_color(1, DIFF_FILE_OLD),
 613                                minus_end - minus_begin, minus_begin);
 614        if (plus_begin != plus_end)
 615                color_fwrite_lines(diff_words->file,
 616                                diff_get_color(1, DIFF_FILE_NEW),
 617                                plus_end - plus_begin, plus_begin);
 618
 619        diff_words->current_plus = plus_end;
 620}
 621
 622/* This function starts looking at *begin, and returns 0 iff a word was found. */
 623static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 624                int *begin, int *end)
 625{
 626        if (word_regex && *begin < buffer->size) {
 627                regmatch_t match[1];
 628                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 629                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 630                                        '\n', match[0].rm_eo - match[0].rm_so);
 631                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 632                        *begin += match[0].rm_so;
 633                        return *begin >= *end;
 634                }
 635                return -1;
 636        }
 637
 638        /* find the next word */
 639        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 640                (*begin)++;
 641        if (*begin >= buffer->size)
 642                return -1;
 643
 644        /* find the end of the word */
 645        *end = *begin + 1;
 646        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 647                (*end)++;
 648
 649        return 0;
 650}
 651
 652/*
 653 * This function splits the words in buffer->text, stores the list with
 654 * newline separator into out, and saves the offsets of the original words
 655 * in buffer->orig.
 656 */
 657static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 658                regex_t *word_regex)
 659{
 660        int i, j;
 661        long alloc = 0;
 662
 663        out->size = 0;
 664        out->ptr = NULL;
 665
 666        /* fake an empty "0th" word */
 667        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 668        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 669        buffer->orig_nr = 1;
 670
 671        for (i = 0; i < buffer->text.size; i++) {
 672                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 673                        return;
 674
 675                /* store original boundaries */
 676                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 677                                buffer->orig_alloc);
 678                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 679                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 680                buffer->orig_nr++;
 681
 682                /* store one word */
 683                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 684                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 685                out->ptr[out->size + j - i] = '\n';
 686                out->size += j - i + 1;
 687
 688                i = j - 1;
 689        }
 690}
 691
 692/* this executes the word diff on the accumulated buffers */
 693static void diff_words_show(struct diff_words_data *diff_words)
 694{
 695        xpparam_t xpp;
 696        xdemitconf_t xecfg;
 697        xdemitcb_t ecb;
 698        mmfile_t minus, plus;
 699
 700        /* special case: only removal */
 701        if (!diff_words->plus.text.size) {
 702                color_fwrite_lines(diff_words->file,
 703                        diff_get_color(1, DIFF_FILE_OLD),
 704                        diff_words->minus.text.size, diff_words->minus.text.ptr);
 705                diff_words->minus.text.size = 0;
 706                return;
 707        }
 708
 709        diff_words->current_plus = diff_words->plus.text.ptr;
 710
 711        memset(&xpp, 0, sizeof(xpp));
 712        memset(&xecfg, 0, sizeof(xecfg));
 713        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 714        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 715        xpp.flags = XDF_NEED_MINIMAL;
 716        /* as only the hunk header will be parsed, we need a 0-context */
 717        xecfg.ctxlen = 0;
 718        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 719                      &xpp, &xecfg, &ecb);
 720        free(minus.ptr);
 721        free(plus.ptr);
 722        if (diff_words->current_plus != diff_words->plus.text.ptr +
 723                        diff_words->plus.text.size)
 724                fwrite(diff_words->current_plus,
 725                        diff_words->plus.text.ptr + diff_words->plus.text.size
 726                        - diff_words->current_plus, 1,
 727                        diff_words->file);
 728        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 729}
 730
 731/* In "color-words" mode, show word-diff of words accumulated in the buffer */
 732static void diff_words_flush(struct emit_callback *ecbdata)
 733{
 734        if (ecbdata->diff_words->minus.text.size ||
 735            ecbdata->diff_words->plus.text.size)
 736                diff_words_show(ecbdata->diff_words);
 737}
 738
 739static void free_diff_words_data(struct emit_callback *ecbdata)
 740{
 741        if (ecbdata->diff_words) {
 742                diff_words_flush(ecbdata);
 743                free (ecbdata->diff_words->minus.text.ptr);
 744                free (ecbdata->diff_words->minus.orig);
 745                free (ecbdata->diff_words->plus.text.ptr);
 746                free (ecbdata->diff_words->plus.orig);
 747                free(ecbdata->diff_words->word_regex);
 748                free(ecbdata->diff_words);
 749                ecbdata->diff_words = NULL;
 750        }
 751}
 752
 753const char *diff_get_color(int diff_use_color, enum color_diff ix)
 754{
 755        if (diff_use_color)
 756                return diff_colors[ix];
 757        return "";
 758}
 759
 760static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
 761{
 762        const char *cp;
 763        unsigned long allot;
 764        size_t l = len;
 765
 766        if (ecb->truncate)
 767                return ecb->truncate(line, len);
 768        cp = line;
 769        allot = l;
 770        while (0 < l) {
 771                (void) utf8_width(&cp, &l);
 772                if (!cp)
 773                        break; /* truncated in the middle? */
 774        }
 775        return allot - l;
 776}
 777
 778static void find_lno(const char *line, struct emit_callback *ecbdata)
 779{
 780        const char *p;
 781        ecbdata->lno_in_preimage = 0;
 782        ecbdata->lno_in_postimage = 0;
 783        p = strchr(line, '-');
 784        if (!p)
 785                return; /* cannot happen */
 786        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
 787        p = strchr(p, '+');
 788        if (!p)
 789                return; /* cannot happen */
 790        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
 791}
 792
 793static void fn_out_consume(void *priv, char *line, unsigned long len)
 794{
 795        struct emit_callback *ecbdata = priv;
 796        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 797        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 798        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 799
 800        *(ecbdata->found_changesp) = 1;
 801
 802        if (ecbdata->label_path[0]) {
 803                const char *name_a_tab, *name_b_tab;
 804
 805                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 806                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 807
 808                fprintf(ecbdata->file, "%s--- %s%s%s\n",
 809                        meta, ecbdata->label_path[0], reset, name_a_tab);
 810                fprintf(ecbdata->file, "%s+++ %s%s%s\n",
 811                        meta, ecbdata->label_path[1], reset, name_b_tab);
 812                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 813        }
 814
 815        if (diff_suppress_blank_empty
 816            && len == 2 && line[0] == ' ' && line[1] == '\n') {
 817                line[0] = '\n';
 818                len = 1;
 819        }
 820
 821        if (line[0] == '@') {
 822                if (ecbdata->diff_words)
 823                        diff_words_flush(ecbdata);
 824                len = sane_truncate_line(ecbdata, line, len);
 825                find_lno(line, ecbdata);
 826                emit_hunk_header(ecbdata, line, len);
 827                if (line[len-1] != '\n')
 828                        putc('\n', ecbdata->file);
 829                return;
 830        }
 831
 832        if (len < 1) {
 833                emit_line(ecbdata->file, reset, reset, line, len);
 834                return;
 835        }
 836
 837        if (ecbdata->diff_words) {
 838                if (line[0] == '-') {
 839                        diff_words_append(line, len,
 840                                          &ecbdata->diff_words->minus);
 841                        return;
 842                } else if (line[0] == '+') {
 843                        diff_words_append(line, len,
 844                                          &ecbdata->diff_words->plus);
 845                        return;
 846                }
 847                diff_words_flush(ecbdata);
 848                line++;
 849                len--;
 850                emit_line(ecbdata->file, plain, reset, line, len);
 851                return;
 852        }
 853
 854        if (line[0] != '+') {
 855                const char *color =
 856                        diff_get_color(ecbdata->color_diff,
 857                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
 858                ecbdata->lno_in_preimage++;
 859                if (line[0] == ' ')
 860                        ecbdata->lno_in_postimage++;
 861                emit_line(ecbdata->file, color, reset, line, len);
 862        } else {
 863                ecbdata->lno_in_postimage++;
 864                emit_add_line(reset, ecbdata, line + 1, len - 1);
 865        }
 866}
 867
 868static char *pprint_rename(const char *a, const char *b)
 869{
 870        const char *old = a;
 871        const char *new = b;
 872        struct strbuf name = STRBUF_INIT;
 873        int pfx_length, sfx_length;
 874        int len_a = strlen(a);
 875        int len_b = strlen(b);
 876        int a_midlen, b_midlen;
 877        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 878        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 879
 880        if (qlen_a || qlen_b) {
 881                quote_c_style(a, &name, NULL, 0);
 882                strbuf_addstr(&name, " => ");
 883                quote_c_style(b, &name, NULL, 0);
 884                return strbuf_detach(&name, NULL);
 885        }
 886
 887        /* Find common prefix */
 888        pfx_length = 0;
 889        while (*old && *new && *old == *new) {
 890                if (*old == '/')
 891                        pfx_length = old - a + 1;
 892                old++;
 893                new++;
 894        }
 895
 896        /* Find common suffix */
 897        old = a + len_a;
 898        new = b + len_b;
 899        sfx_length = 0;
 900        while (a <= old && b <= new && *old == *new) {
 901                if (*old == '/')
 902                        sfx_length = len_a - (old - a);
 903                old--;
 904                new--;
 905        }
 906
 907        /*
 908         * pfx{mid-a => mid-b}sfx
 909         * {pfx-a => pfx-b}sfx
 910         * pfx{sfx-a => sfx-b}
 911         * name-a => name-b
 912         */
 913        a_midlen = len_a - pfx_length - sfx_length;
 914        b_midlen = len_b - pfx_length - sfx_length;
 915        if (a_midlen < 0)
 916                a_midlen = 0;
 917        if (b_midlen < 0)
 918                b_midlen = 0;
 919
 920        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 921        if (pfx_length + sfx_length) {
 922                strbuf_add(&name, a, pfx_length);
 923                strbuf_addch(&name, '{');
 924        }
 925        strbuf_add(&name, a + pfx_length, a_midlen);
 926        strbuf_addstr(&name, " => ");
 927        strbuf_add(&name, b + pfx_length, b_midlen);
 928        if (pfx_length + sfx_length) {
 929                strbuf_addch(&name, '}');
 930                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 931        }
 932        return strbuf_detach(&name, NULL);
 933}
 934
 935struct diffstat_t {
 936        int nr;
 937        int alloc;
 938        struct diffstat_file {
 939                char *from_name;
 940                char *name;
 941                char *print_name;
 942                unsigned is_unmerged:1;
 943                unsigned is_binary:1;
 944                unsigned is_renamed:1;
 945                uintmax_t added, deleted;
 946        } **files;
 947};
 948
 949static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 950                                          const char *name_a,
 951                                          const char *name_b)
 952{
 953        struct diffstat_file *x;
 954        x = xcalloc(sizeof (*x), 1);
 955        if (diffstat->nr == diffstat->alloc) {
 956                diffstat->alloc = alloc_nr(diffstat->alloc);
 957                diffstat->files = xrealloc(diffstat->files,
 958                                diffstat->alloc * sizeof(x));
 959        }
 960        diffstat->files[diffstat->nr++] = x;
 961        if (name_b) {
 962                x->from_name = xstrdup(name_a);
 963                x->name = xstrdup(name_b);
 964                x->is_renamed = 1;
 965        }
 966        else {
 967                x->from_name = NULL;
 968                x->name = xstrdup(name_a);
 969        }
 970        return x;
 971}
 972
 973static void diffstat_consume(void *priv, char *line, unsigned long len)
 974{
 975        struct diffstat_t *diffstat = priv;
 976        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 977
 978        if (line[0] == '+')
 979                x->added++;
 980        else if (line[0] == '-')
 981                x->deleted++;
 982}
 983
 984const char mime_boundary_leader[] = "------------";
 985
 986static int scale_linear(int it, int width, int max_change)
 987{
 988        /*
 989         * make sure that at least one '-' is printed if there were deletions,
 990         * and likewise for '+'.
 991         */
 992        if (max_change < 2)
 993                return it;
 994        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 995}
 996
 997static void show_name(FILE *file,
 998                      const char *prefix, const char *name, int len)
 999{
1000        fprintf(file, " %s%-*s |", prefix, len, name);
1001}
1002
1003static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1004{
1005        if (cnt <= 0)
1006                return;
1007        fprintf(file, "%s", set);
1008        while (cnt--)
1009                putc(ch, file);
1010        fprintf(file, "%s", reset);
1011}
1012
1013static void fill_print_name(struct diffstat_file *file)
1014{
1015        char *pname;
1016
1017        if (file->print_name)
1018                return;
1019
1020        if (!file->is_renamed) {
1021                struct strbuf buf = STRBUF_INIT;
1022                if (quote_c_style(file->name, &buf, NULL, 0)) {
1023                        pname = strbuf_detach(&buf, NULL);
1024                } else {
1025                        pname = file->name;
1026                        strbuf_release(&buf);
1027                }
1028        } else {
1029                pname = pprint_rename(file->from_name, file->name);
1030        }
1031        file->print_name = pname;
1032}
1033
1034static void show_stats(struct diffstat_t *data, struct diff_options *options)
1035{
1036        int i, len, add, del, adds = 0, dels = 0;
1037        uintmax_t max_change = 0, max_len = 0;
1038        int total_files = data->nr;
1039        int width, name_width;
1040        const char *reset, *set, *add_c, *del_c;
1041
1042        if (data->nr == 0)
1043                return;
1044
1045        width = options->stat_width ? options->stat_width : 80;
1046        name_width = options->stat_name_width ? options->stat_name_width : 50;
1047
1048        /* Sanity: give at least 5 columns to the graph,
1049         * but leave at least 10 columns for the name.
1050         */
1051        if (width < 25)
1052                width = 25;
1053        if (name_width < 10)
1054                name_width = 10;
1055        else if (width < name_width + 15)
1056                name_width = width - 15;
1057
1058        /* Find the longest filename and max number of changes */
1059        reset = diff_get_color_opt(options, DIFF_RESET);
1060        set   = diff_get_color_opt(options, DIFF_PLAIN);
1061        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1062        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1063
1064        for (i = 0; i < data->nr; i++) {
1065                struct diffstat_file *file = data->files[i];
1066                uintmax_t change = file->added + file->deleted;
1067                fill_print_name(file);
1068                len = strlen(file->print_name);
1069                if (max_len < len)
1070                        max_len = len;
1071
1072                if (file->is_binary || file->is_unmerged)
1073                        continue;
1074                if (max_change < change)
1075                        max_change = change;
1076        }
1077
1078        /* Compute the width of the graph part;
1079         * 10 is for one blank at the beginning of the line plus
1080         * " | count " between the name and the graph.
1081         *
1082         * From here on, name_width is the width of the name area,
1083         * and width is the width of the graph area.
1084         */
1085        name_width = (name_width < max_len) ? name_width : max_len;
1086        if (width < (name_width + 10) + max_change)
1087                width = width - (name_width + 10);
1088        else
1089                width = max_change;
1090
1091        for (i = 0; i < data->nr; i++) {
1092                const char *prefix = "";
1093                char *name = data->files[i]->print_name;
1094                uintmax_t added = data->files[i]->added;
1095                uintmax_t deleted = data->files[i]->deleted;
1096                int name_len;
1097
1098                /*
1099                 * "scale" the filename
1100                 */
1101                len = name_width;
1102                name_len = strlen(name);
1103                if (name_width < name_len) {
1104                        char *slash;
1105                        prefix = "...";
1106                        len -= 3;
1107                        name += name_len - len;
1108                        slash = strchr(name, '/');
1109                        if (slash)
1110                                name = slash;
1111                }
1112
1113                if (data->files[i]->is_binary) {
1114                        show_name(options->file, prefix, name, len);
1115                        fprintf(options->file, "  Bin ");
1116                        fprintf(options->file, "%s%"PRIuMAX"%s",
1117                                del_c, deleted, reset);
1118                        fprintf(options->file, " -> ");
1119                        fprintf(options->file, "%s%"PRIuMAX"%s",
1120                                add_c, added, reset);
1121                        fprintf(options->file, " bytes");
1122                        fprintf(options->file, "\n");
1123                        continue;
1124                }
1125                else if (data->files[i]->is_unmerged) {
1126                        show_name(options->file, prefix, name, len);
1127                        fprintf(options->file, "  Unmerged\n");
1128                        continue;
1129                }
1130                else if (!data->files[i]->is_renamed &&
1131                         (added + deleted == 0)) {
1132                        total_files--;
1133                        continue;
1134                }
1135
1136                /*
1137                 * scale the add/delete
1138                 */
1139                add = added;
1140                del = deleted;
1141                adds += add;
1142                dels += del;
1143
1144                if (width <= max_change) {
1145                        add = scale_linear(add, width, max_change);
1146                        del = scale_linear(del, width, max_change);
1147                }
1148                show_name(options->file, prefix, name, len);
1149                fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1150                                added + deleted ? " " : "");
1151                show_graph(options->file, '+', add, add_c, reset);
1152                show_graph(options->file, '-', del, del_c, reset);
1153                fprintf(options->file, "\n");
1154        }
1155        fprintf(options->file,
1156               " %d files changed, %d insertions(+), %d deletions(-)\n",
1157               total_files, adds, dels);
1158}
1159
1160static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1161{
1162        int i, adds = 0, dels = 0, total_files = data->nr;
1163
1164        if (data->nr == 0)
1165                return;
1166
1167        for (i = 0; i < data->nr; i++) {
1168                if (!data->files[i]->is_binary &&
1169                    !data->files[i]->is_unmerged) {
1170                        int added = data->files[i]->added;
1171                        int deleted= data->files[i]->deleted;
1172                        if (!data->files[i]->is_renamed &&
1173                            (added + deleted == 0)) {
1174                                total_files--;
1175                        } else {
1176                                adds += added;
1177                                dels += deleted;
1178                        }
1179                }
1180        }
1181        fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
1182               total_files, adds, dels);
1183}
1184
1185static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1186{
1187        int i;
1188
1189        if (data->nr == 0)
1190                return;
1191
1192        for (i = 0; i < data->nr; i++) {
1193                struct diffstat_file *file = data->files[i];
1194
1195                if (file->is_binary)
1196                        fprintf(options->file, "-\t-\t");
1197                else
1198                        fprintf(options->file,
1199                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1200                                file->added, file->deleted);
1201                if (options->line_termination) {
1202                        fill_print_name(file);
1203                        if (!file->is_renamed)
1204                                write_name_quoted(file->name, options->file,
1205                                                  options->line_termination);
1206                        else {
1207                                fputs(file->print_name, options->file);
1208                                putc(options->line_termination, options->file);
1209                        }
1210                } else {
1211                        if (file->is_renamed) {
1212                                putc('\0', options->file);
1213                                write_name_quoted(file->from_name, options->file, '\0');
1214                        }
1215                        write_name_quoted(file->name, options->file, '\0');
1216                }
1217        }
1218}
1219
1220struct dirstat_file {
1221        const char *name;
1222        unsigned long changed;
1223};
1224
1225struct dirstat_dir {
1226        struct dirstat_file *files;
1227        int alloc, nr, percent, cumulative;
1228};
1229
1230static long gather_dirstat(FILE *file, struct dirstat_dir *dir, unsigned long changed, const char *base, int baselen)
1231{
1232        unsigned long this_dir = 0;
1233        unsigned int sources = 0;
1234
1235        while (dir->nr) {
1236                struct dirstat_file *f = dir->files;
1237                int namelen = strlen(f->name);
1238                unsigned long this;
1239                char *slash;
1240
1241                if (namelen < baselen)
1242                        break;
1243                if (memcmp(f->name, base, baselen))
1244                        break;
1245                slash = strchr(f->name + baselen, '/');
1246                if (slash) {
1247                        int newbaselen = slash + 1 - f->name;
1248                        this = gather_dirstat(file, dir, changed, f->name, newbaselen);
1249                        sources++;
1250                } else {
1251                        this = f->changed;
1252                        dir->files++;
1253                        dir->nr--;
1254                        sources += 2;
1255                }
1256                this_dir += this;
1257        }
1258
1259        /*
1260         * We don't report dirstat's for
1261         *  - the top level
1262         *  - or cases where everything came from a single directory
1263         *    under this directory (sources == 1).
1264         */
1265        if (baselen && sources != 1) {
1266                int permille = this_dir * 1000 / changed;
1267                if (permille) {
1268                        int percent = permille / 10;
1269                        if (percent >= dir->percent) {
1270                                fprintf(file, "%4d.%01d%% %.*s\n", percent, permille % 10, baselen, base);
1271                                if (!dir->cumulative)
1272                                        return 0;
1273                        }
1274                }
1275        }
1276        return this_dir;
1277}
1278
1279static int dirstat_compare(const void *_a, const void *_b)
1280{
1281        const struct dirstat_file *a = _a;
1282        const struct dirstat_file *b = _b;
1283        return strcmp(a->name, b->name);
1284}
1285
1286static void show_dirstat(struct diff_options *options)
1287{
1288        int i;
1289        unsigned long changed;
1290        struct dirstat_dir dir;
1291        struct diff_queue_struct *q = &diff_queued_diff;
1292
1293        dir.files = NULL;
1294        dir.alloc = 0;
1295        dir.nr = 0;
1296        dir.percent = options->dirstat_percent;
1297        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1298
1299        changed = 0;
1300        for (i = 0; i < q->nr; i++) {
1301                struct diff_filepair *p = q->queue[i];
1302                const char *name;
1303                unsigned long copied, added, damage;
1304
1305                name = p->one->path ? p->one->path : p->two->path;
1306
1307                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1308                        diff_populate_filespec(p->one, 0);
1309                        diff_populate_filespec(p->two, 0);
1310                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1311                                               &copied, &added);
1312                        diff_free_filespec_data(p->one);
1313                        diff_free_filespec_data(p->two);
1314                } else if (DIFF_FILE_VALID(p->one)) {
1315                        diff_populate_filespec(p->one, 1);
1316                        copied = added = 0;
1317                        diff_free_filespec_data(p->one);
1318                } else if (DIFF_FILE_VALID(p->two)) {
1319                        diff_populate_filespec(p->two, 1);
1320                        copied = 0;
1321                        added = p->two->size;
1322                        diff_free_filespec_data(p->two);
1323                } else
1324                        continue;
1325
1326                /*
1327                 * Original minus copied is the removed material,
1328                 * added is the new material.  They are both damages
1329                 * made to the preimage. In --dirstat-by-file mode, count
1330                 * damaged files, not damaged lines. This is done by
1331                 * counting only a single damaged line per file.
1332                 */
1333                damage = (p->one->size - copied) + added;
1334                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE) && damage > 0)
1335                        damage = 1;
1336
1337                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1338                dir.files[dir.nr].name = name;
1339                dir.files[dir.nr].changed = damage;
1340                changed += damage;
1341                dir.nr++;
1342        }
1343
1344        /* This can happen even with many files, if everything was renames */
1345        if (!changed)
1346                return;
1347
1348        /* Show all directories with more than x% of the changes */
1349        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1350        gather_dirstat(options->file, &dir, changed, "", 0);
1351}
1352
1353static void free_diffstat_info(struct diffstat_t *diffstat)
1354{
1355        int i;
1356        for (i = 0; i < diffstat->nr; i++) {
1357                struct diffstat_file *f = diffstat->files[i];
1358                if (f->name != f->print_name)
1359                        free(f->print_name);
1360                free(f->name);
1361                free(f->from_name);
1362                free(f);
1363        }
1364        free(diffstat->files);
1365}
1366
1367struct checkdiff_t {
1368        const char *filename;
1369        int lineno;
1370        struct diff_options *o;
1371        unsigned ws_rule;
1372        unsigned status;
1373};
1374
1375static int is_conflict_marker(const char *line, unsigned long len)
1376{
1377        char firstchar;
1378        int cnt;
1379
1380        if (len < 8)
1381                return 0;
1382        firstchar = line[0];
1383        switch (firstchar) {
1384        case '=': case '>': case '<':
1385                break;
1386        default:
1387                return 0;
1388        }
1389        for (cnt = 1; cnt < 7; cnt++)
1390                if (line[cnt] != firstchar)
1391                        return 0;
1392        /* line[0] thru line[6] are same as firstchar */
1393        if (firstchar == '=') {
1394                /* divider between ours and theirs? */
1395                if (len != 8 || line[7] != '\n')
1396                        return 0;
1397        } else if (len < 8 || !isspace(line[7])) {
1398                /* not divider before ours nor after theirs */
1399                return 0;
1400        }
1401        return 1;
1402}
1403
1404static void checkdiff_consume(void *priv, char *line, unsigned long len)
1405{
1406        struct checkdiff_t *data = priv;
1407        int color_diff = DIFF_OPT_TST(data->o, COLOR_DIFF);
1408        const char *ws = diff_get_color(color_diff, DIFF_WHITESPACE);
1409        const char *reset = diff_get_color(color_diff, DIFF_RESET);
1410        const char *set = diff_get_color(color_diff, DIFF_FILE_NEW);
1411        char *err;
1412
1413        if (line[0] == '+') {
1414                unsigned bad;
1415                data->lineno++;
1416                if (is_conflict_marker(line + 1, len - 1)) {
1417                        data->status |= 1;
1418                        fprintf(data->o->file,
1419                                "%s:%d: leftover conflict marker\n",
1420                                data->filename, data->lineno);
1421                }
1422                bad = ws_check(line + 1, len - 1, data->ws_rule);
1423                if (!bad)
1424                        return;
1425                data->status |= bad;
1426                err = whitespace_error_string(bad);
1427                fprintf(data->o->file, "%s:%d: %s.\n",
1428                        data->filename, data->lineno, err);
1429                free(err);
1430                emit_line(data->o->file, set, reset, line, 1);
1431                ws_check_emit(line + 1, len - 1, data->ws_rule,
1432                              data->o->file, set, reset, ws);
1433        } else if (line[0] == ' ') {
1434                data->lineno++;
1435        } else if (line[0] == '@') {
1436                char *plus = strchr(line, '+');
1437                if (plus)
1438                        data->lineno = strtol(plus, NULL, 10) - 1;
1439                else
1440                        die("invalid diff");
1441        }
1442}
1443
1444static unsigned char *deflate_it(char *data,
1445                                 unsigned long size,
1446                                 unsigned long *result_size)
1447{
1448        int bound;
1449        unsigned char *deflated;
1450        z_stream stream;
1451
1452        memset(&stream, 0, sizeof(stream));
1453        deflateInit(&stream, zlib_compression_level);
1454        bound = deflateBound(&stream, size);
1455        deflated = xmalloc(bound);
1456        stream.next_out = deflated;
1457        stream.avail_out = bound;
1458
1459        stream.next_in = (unsigned char *)data;
1460        stream.avail_in = size;
1461        while (deflate(&stream, Z_FINISH) == Z_OK)
1462                ; /* nothing */
1463        deflateEnd(&stream);
1464        *result_size = stream.total_out;
1465        return deflated;
1466}
1467
1468static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two)
1469{
1470        void *cp;
1471        void *delta;
1472        void *deflated;
1473        void *data;
1474        unsigned long orig_size;
1475        unsigned long delta_size;
1476        unsigned long deflate_size;
1477        unsigned long data_size;
1478
1479        /* We could do deflated delta, or we could do just deflated two,
1480         * whichever is smaller.
1481         */
1482        delta = NULL;
1483        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1484        if (one->size && two->size) {
1485                delta = diff_delta(one->ptr, one->size,
1486                                   two->ptr, two->size,
1487                                   &delta_size, deflate_size);
1488                if (delta) {
1489                        void *to_free = delta;
1490                        orig_size = delta_size;
1491                        delta = deflate_it(delta, delta_size, &delta_size);
1492                        free(to_free);
1493                }
1494        }
1495
1496        if (delta && delta_size < deflate_size) {
1497                fprintf(file, "delta %lu\n", orig_size);
1498                free(deflated);
1499                data = delta;
1500                data_size = delta_size;
1501        }
1502        else {
1503                fprintf(file, "literal %lu\n", two->size);
1504                free(delta);
1505                data = deflated;
1506                data_size = deflate_size;
1507        }
1508
1509        /* emit data encoded in base85 */
1510        cp = data;
1511        while (data_size) {
1512                int bytes = (52 < data_size) ? 52 : data_size;
1513                char line[70];
1514                data_size -= bytes;
1515                if (bytes <= 26)
1516                        line[0] = bytes + 'A' - 1;
1517                else
1518                        line[0] = bytes - 26 + 'a' - 1;
1519                encode_85(line + 1, cp, bytes);
1520                cp = (char *) cp + bytes;
1521                fputs(line, file);
1522                fputc('\n', file);
1523        }
1524        fprintf(file, "\n");
1525        free(data);
1526}
1527
1528static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two)
1529{
1530        fprintf(file, "GIT binary patch\n");
1531        emit_binary_diff_body(file, one, two);
1532        emit_binary_diff_body(file, two, one);
1533}
1534
1535static void diff_filespec_load_driver(struct diff_filespec *one)
1536{
1537        if (!one->driver)
1538                one->driver = userdiff_find_by_path(one->path);
1539        if (!one->driver)
1540                one->driver = userdiff_find_by_name("default");
1541}
1542
1543int diff_filespec_is_binary(struct diff_filespec *one)
1544{
1545        if (one->is_binary == -1) {
1546                diff_filespec_load_driver(one);
1547                if (one->driver->binary != -1)
1548                        one->is_binary = one->driver->binary;
1549                else {
1550                        if (!one->data && DIFF_FILE_VALID(one))
1551                                diff_populate_filespec(one, 0);
1552                        if (one->data)
1553                                one->is_binary = buffer_is_binary(one->data,
1554                                                one->size);
1555                        if (one->is_binary == -1)
1556                                one->is_binary = 0;
1557                }
1558        }
1559        return one->is_binary;
1560}
1561
1562static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
1563{
1564        diff_filespec_load_driver(one);
1565        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
1566}
1567
1568static const char *userdiff_word_regex(struct diff_filespec *one)
1569{
1570        diff_filespec_load_driver(one);
1571        return one->driver->word_regex;
1572}
1573
1574void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
1575{
1576        if (!options->a_prefix)
1577                options->a_prefix = a;
1578        if (!options->b_prefix)
1579                options->b_prefix = b;
1580}
1581
1582static const char *get_textconv(struct diff_filespec *one)
1583{
1584        if (!DIFF_FILE_VALID(one))
1585                return NULL;
1586        if (!S_ISREG(one->mode))
1587                return NULL;
1588        diff_filespec_load_driver(one);
1589        return one->driver->textconv;
1590}
1591
1592static void builtin_diff(const char *name_a,
1593                         const char *name_b,
1594                         struct diff_filespec *one,
1595                         struct diff_filespec *two,
1596                         const char *xfrm_msg,
1597                         struct diff_options *o,
1598                         int complete_rewrite)
1599{
1600        mmfile_t mf1, mf2;
1601        const char *lbl[2];
1602        char *a_one, *b_two;
1603        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1604        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1605        const char *a_prefix, *b_prefix;
1606        const char *textconv_one = NULL, *textconv_two = NULL;
1607
1608        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
1609                        (!one->mode || S_ISGITLINK(one->mode)) &&
1610                        (!two->mode || S_ISGITLINK(two->mode))) {
1611                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
1612                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
1613                show_submodule_summary(o->file, one ? one->path : two->path,
1614                                one->sha1, two->sha1,
1615                                del, add, reset);
1616                return;
1617        }
1618
1619        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
1620                textconv_one = get_textconv(one);
1621                textconv_two = get_textconv(two);
1622        }
1623
1624        diff_set_mnemonic_prefix(o, "a/", "b/");
1625        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
1626                a_prefix = o->b_prefix;
1627                b_prefix = o->a_prefix;
1628        } else {
1629                a_prefix = o->a_prefix;
1630                b_prefix = o->b_prefix;
1631        }
1632
1633        /* Never use a non-valid filename anywhere if at all possible */
1634        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
1635        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
1636
1637        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
1638        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
1639        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1640        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1641        fprintf(o->file, "%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1642        if (lbl[0][0] == '/') {
1643                /* /dev/null */
1644                fprintf(o->file, "%snew file mode %06o%s\n", set, two->mode, reset);
1645                if (xfrm_msg && xfrm_msg[0])
1646                        fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1647        }
1648        else if (lbl[1][0] == '/') {
1649                fprintf(o->file, "%sdeleted file mode %06o%s\n", set, one->mode, reset);
1650                if (xfrm_msg && xfrm_msg[0])
1651                        fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1652        }
1653        else {
1654                if (one->mode != two->mode) {
1655                        fprintf(o->file, "%sold mode %06o%s\n", set, one->mode, reset);
1656                        fprintf(o->file, "%snew mode %06o%s\n", set, two->mode, reset);
1657                }
1658                if (xfrm_msg && xfrm_msg[0])
1659                        fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1660                /*
1661                 * we do not run diff between different kind
1662                 * of objects.
1663                 */
1664                if ((one->mode ^ two->mode) & S_IFMT)
1665                        goto free_ab_and_return;
1666                if (complete_rewrite &&
1667                    (textconv_one || !diff_filespec_is_binary(one)) &&
1668                    (textconv_two || !diff_filespec_is_binary(two))) {
1669                        emit_rewrite_diff(name_a, name_b, one, two,
1670                                                textconv_one, textconv_two, o);
1671                        o->found_changes = 1;
1672                        goto free_ab_and_return;
1673                }
1674        }
1675
1676        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1677                die("unable to read files to diff");
1678
1679        if (!DIFF_OPT_TST(o, TEXT) &&
1680            ( (diff_filespec_is_binary(one) && !textconv_one) ||
1681              (diff_filespec_is_binary(two) && !textconv_two) )) {
1682                /* Quite common confusing case */
1683                if (mf1.size == mf2.size &&
1684                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1685                        goto free_ab_and_return;
1686                if (DIFF_OPT_TST(o, BINARY))
1687                        emit_binary_diff(o->file, &mf1, &mf2);
1688                else
1689                        fprintf(o->file, "Binary files %s and %s differ\n",
1690                                lbl[0], lbl[1]);
1691                o->found_changes = 1;
1692        }
1693        else {
1694                /* Crazy xdl interfaces.. */
1695                const char *diffopts = getenv("GIT_DIFF_OPTS");
1696                xpparam_t xpp;
1697                xdemitconf_t xecfg;
1698                xdemitcb_t ecb;
1699                struct emit_callback ecbdata;
1700                const struct userdiff_funcname *pe;
1701
1702                if (textconv_one) {
1703                        size_t size;
1704                        mf1.ptr = run_textconv(textconv_one, one, &size);
1705                        if (!mf1.ptr)
1706                                die("unable to read files to diff");
1707                        mf1.size = size;
1708                }
1709                if (textconv_two) {
1710                        size_t size;
1711                        mf2.ptr = run_textconv(textconv_two, two, &size);
1712                        if (!mf2.ptr)
1713                                die("unable to read files to diff");
1714                        mf2.size = size;
1715                }
1716
1717                pe = diff_funcname_pattern(one);
1718                if (!pe)
1719                        pe = diff_funcname_pattern(two);
1720
1721                memset(&xpp, 0, sizeof(xpp));
1722                memset(&xecfg, 0, sizeof(xecfg));
1723                memset(&ecbdata, 0, sizeof(ecbdata));
1724                ecbdata.label_path = lbl;
1725                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1726                ecbdata.found_changesp = &o->found_changes;
1727                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1728                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
1729                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
1730                ecbdata.file = o->file;
1731                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1732                xecfg.ctxlen = o->context;
1733                xecfg.interhunkctxlen = o->interhunkcontext;
1734                xecfg.flags = XDL_EMIT_FUNCNAMES;
1735                if (pe)
1736                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
1737                if (!diffopts)
1738                        ;
1739                else if (!prefixcmp(diffopts, "--unified="))
1740                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1741                else if (!prefixcmp(diffopts, "-u"))
1742                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1743                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS)) {
1744                        ecbdata.diff_words =
1745                                xcalloc(1, sizeof(struct diff_words_data));
1746                        ecbdata.diff_words->file = o->file;
1747                        if (!o->word_regex)
1748                                o->word_regex = userdiff_word_regex(one);
1749                        if (!o->word_regex)
1750                                o->word_regex = userdiff_word_regex(two);
1751                        if (!o->word_regex)
1752                                o->word_regex = diff_word_regex_cfg;
1753                        if (o->word_regex) {
1754                                ecbdata.diff_words->word_regex = (regex_t *)
1755                                        xmalloc(sizeof(regex_t));
1756                                if (regcomp(ecbdata.diff_words->word_regex,
1757                                                o->word_regex,
1758                                                REG_EXTENDED | REG_NEWLINE))
1759                                        die ("Invalid regular expression: %s",
1760                                                        o->word_regex);
1761                        }
1762                }
1763                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
1764                              &xpp, &xecfg, &ecb);
1765                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1766                        free_diff_words_data(&ecbdata);
1767                if (textconv_one)
1768                        free(mf1.ptr);
1769                if (textconv_two)
1770                        free(mf2.ptr);
1771                xdiff_clear_find_func(&xecfg);
1772        }
1773
1774 free_ab_and_return:
1775        diff_free_filespec_data(one);
1776        diff_free_filespec_data(two);
1777        free(a_one);
1778        free(b_two);
1779        return;
1780}
1781
1782static void builtin_diffstat(const char *name_a, const char *name_b,
1783                             struct diff_filespec *one,
1784                             struct diff_filespec *two,
1785                             struct diffstat_t *diffstat,
1786                             struct diff_options *o,
1787                             int complete_rewrite)
1788{
1789        mmfile_t mf1, mf2;
1790        struct diffstat_file *data;
1791
1792        data = diffstat_add(diffstat, name_a, name_b);
1793
1794        if (!one || !two) {
1795                data->is_unmerged = 1;
1796                return;
1797        }
1798        if (complete_rewrite) {
1799                diff_populate_filespec(one, 0);
1800                diff_populate_filespec(two, 0);
1801                data->deleted = count_lines(one->data, one->size);
1802                data->added = count_lines(two->data, two->size);
1803                goto free_and_return;
1804        }
1805        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1806                die("unable to read files to diff");
1807
1808        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1809                data->is_binary = 1;
1810                data->added = mf2.size;
1811                data->deleted = mf1.size;
1812        } else {
1813                /* Crazy xdl interfaces.. */
1814                xpparam_t xpp;
1815                xdemitconf_t xecfg;
1816                xdemitcb_t ecb;
1817
1818                memset(&xpp, 0, sizeof(xpp));
1819                memset(&xecfg, 0, sizeof(xecfg));
1820                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1821                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
1822                              &xpp, &xecfg, &ecb);
1823        }
1824
1825 free_and_return:
1826        diff_free_filespec_data(one);
1827        diff_free_filespec_data(two);
1828}
1829
1830static void builtin_checkdiff(const char *name_a, const char *name_b,
1831                              const char *attr_path,
1832                              struct diff_filespec *one,
1833                              struct diff_filespec *two,
1834                              struct diff_options *o)
1835{
1836        mmfile_t mf1, mf2;
1837        struct checkdiff_t data;
1838
1839        if (!two)
1840                return;
1841
1842        memset(&data, 0, sizeof(data));
1843        data.filename = name_b ? name_b : name_a;
1844        data.lineno = 0;
1845        data.o = o;
1846        data.ws_rule = whitespace_rule(attr_path);
1847
1848        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1849                die("unable to read files to diff");
1850
1851        /*
1852         * All the other codepaths check both sides, but not checking
1853         * the "old" side here is deliberate.  We are checking the newly
1854         * introduced changes, and as long as the "new" side is text, we
1855         * can and should check what it introduces.
1856         */
1857        if (diff_filespec_is_binary(two))
1858                goto free_and_return;
1859        else {
1860                /* Crazy xdl interfaces.. */
1861                xpparam_t xpp;
1862                xdemitconf_t xecfg;
1863                xdemitcb_t ecb;
1864
1865                memset(&xpp, 0, sizeof(xpp));
1866                memset(&xecfg, 0, sizeof(xecfg));
1867                xecfg.ctxlen = 1; /* at least one context line */
1868                xpp.flags = XDF_NEED_MINIMAL;
1869                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
1870                              &xpp, &xecfg, &ecb);
1871
1872                if (data.ws_rule & WS_BLANK_AT_EOF) {
1873                        struct emit_callback ecbdata;
1874                        int blank_at_eof;
1875
1876                        ecbdata.ws_rule = data.ws_rule;
1877                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
1878                        blank_at_eof = ecbdata.blank_at_eof_in_preimage;
1879
1880                        if (blank_at_eof) {
1881                                static char *err;
1882                                if (!err)
1883                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
1884                                fprintf(o->file, "%s:%d: %s.\n",
1885                                        data.filename, blank_at_eof, err);
1886                                data.status = 1; /* report errors */
1887                        }
1888                }
1889        }
1890 free_and_return:
1891        diff_free_filespec_data(one);
1892        diff_free_filespec_data(two);
1893        if (data.status)
1894                DIFF_OPT_SET(o, CHECK_FAILED);
1895}
1896
1897struct diff_filespec *alloc_filespec(const char *path)
1898{
1899        int namelen = strlen(path);
1900        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1901
1902        memset(spec, 0, sizeof(*spec));
1903        spec->path = (char *)(spec + 1);
1904        memcpy(spec->path, path, namelen+1);
1905        spec->count = 1;
1906        spec->is_binary = -1;
1907        return spec;
1908}
1909
1910void free_filespec(struct diff_filespec *spec)
1911{
1912        if (!--spec->count) {
1913                diff_free_filespec_data(spec);
1914                free(spec);
1915        }
1916}
1917
1918void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1919                   unsigned short mode)
1920{
1921        if (mode) {
1922                spec->mode = canon_mode(mode);
1923                hashcpy(spec->sha1, sha1);
1924                spec->sha1_valid = !is_null_sha1(sha1);
1925        }
1926}
1927
1928/*
1929 * Given a name and sha1 pair, if the index tells us the file in
1930 * the work tree has that object contents, return true, so that
1931 * prepare_temp_file() does not have to inflate and extract.
1932 */
1933static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1934{
1935        struct cache_entry *ce;
1936        struct stat st;
1937        int pos, len;
1938
1939        /*
1940         * We do not read the cache ourselves here, because the
1941         * benchmark with my previous version that always reads cache
1942         * shows that it makes things worse for diff-tree comparing
1943         * two linux-2.6 kernel trees in an already checked out work
1944         * tree.  This is because most diff-tree comparisons deal with
1945         * only a small number of files, while reading the cache is
1946         * expensive for a large project, and its cost outweighs the
1947         * savings we get by not inflating the object to a temporary
1948         * file.  Practically, this code only helps when we are used
1949         * by diff-cache --cached, which does read the cache before
1950         * calling us.
1951         */
1952        if (!active_cache)
1953                return 0;
1954
1955        /* We want to avoid the working directory if our caller
1956         * doesn't need the data in a normal file, this system
1957         * is rather slow with its stat/open/mmap/close syscalls,
1958         * and the object is contained in a pack file.  The pack
1959         * is probably already open and will be faster to obtain
1960         * the data through than the working directory.  Loose
1961         * objects however would tend to be slower as they need
1962         * to be individually opened and inflated.
1963         */
1964        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
1965                return 0;
1966
1967        len = strlen(name);
1968        pos = cache_name_pos(name, len);
1969        if (pos < 0)
1970                return 0;
1971        ce = active_cache[pos];
1972
1973        /*
1974         * This is not the sha1 we are looking for, or
1975         * unreusable because it is not a regular file.
1976         */
1977        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
1978                return 0;
1979
1980        /*
1981         * If ce is marked as "assume unchanged", there is no
1982         * guarantee that work tree matches what we are looking for.
1983         */
1984        if (ce->ce_flags & CE_VALID)
1985                return 0;
1986
1987        /*
1988         * If ce matches the file in the work tree, we can reuse it.
1989         */
1990        if (ce_uptodate(ce) ||
1991            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
1992                return 1;
1993
1994        return 0;
1995}
1996
1997static int populate_from_stdin(struct diff_filespec *s)
1998{
1999        struct strbuf buf = STRBUF_INIT;
2000        size_t size = 0;
2001
2002        if (strbuf_read(&buf, 0, 0) < 0)
2003                return error("error while reading from stdin %s",
2004                                     strerror(errno));
2005
2006        s->should_munmap = 0;
2007        s->data = strbuf_detach(&buf, &size);
2008        s->size = size;
2009        s->should_free = 1;
2010        return 0;
2011}
2012
2013static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2014{
2015        int len;
2016        char *data = xmalloc(100);
2017        len = snprintf(data, 100,
2018                "Subproject commit %s\n", sha1_to_hex(s->sha1));
2019        s->data = data;
2020        s->size = len;
2021        s->should_free = 1;
2022        if (size_only) {
2023                s->data = NULL;
2024                free(data);
2025        }
2026        return 0;
2027}
2028
2029/*
2030 * While doing rename detection and pickaxe operation, we may need to
2031 * grab the data for the blob (or file) for our own in-core comparison.
2032 * diff_filespec has data and size fields for this purpose.
2033 */
2034int diff_populate_filespec(struct diff_filespec *s, int size_only)
2035{
2036        int err = 0;
2037        if (!DIFF_FILE_VALID(s))
2038                die("internal error: asking to populate invalid file.");
2039        if (S_ISDIR(s->mode))
2040                return -1;
2041
2042        if (s->data)
2043                return 0;
2044
2045        if (size_only && 0 < s->size)
2046                return 0;
2047
2048        if (S_ISGITLINK(s->mode))
2049                return diff_populate_gitlink(s, size_only);
2050
2051        if (!s->sha1_valid ||
2052            reuse_worktree_file(s->path, s->sha1, 0)) {
2053                struct strbuf buf = STRBUF_INIT;
2054                struct stat st;
2055                int fd;
2056
2057                if (!strcmp(s->path, "-"))
2058                        return populate_from_stdin(s);
2059
2060                if (lstat(s->path, &st) < 0) {
2061                        if (errno == ENOENT) {
2062                        err_empty:
2063                                err = -1;
2064                        empty:
2065                                s->data = (char *)"";
2066                                s->size = 0;
2067                                return err;
2068                        }
2069                }
2070                s->size = xsize_t(st.st_size);
2071                if (!s->size)
2072                        goto empty;
2073                if (S_ISLNK(st.st_mode)) {
2074                        struct strbuf sb = STRBUF_INIT;
2075
2076                        if (strbuf_readlink(&sb, s->path, s->size))
2077                                goto err_empty;
2078                        s->size = sb.len;
2079                        s->data = strbuf_detach(&sb, NULL);
2080                        s->should_free = 1;
2081                        return 0;
2082                }
2083                if (size_only)
2084                        return 0;
2085                fd = open(s->path, O_RDONLY);
2086                if (fd < 0)
2087                        goto err_empty;
2088                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2089                close(fd);
2090                s->should_munmap = 1;
2091
2092                /*
2093                 * Convert from working tree format to canonical git format
2094                 */
2095                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2096                        size_t size = 0;
2097                        munmap(s->data, s->size);
2098                        s->should_munmap = 0;
2099                        s->data = strbuf_detach(&buf, &size);
2100                        s->size = size;
2101                        s->should_free = 1;
2102                }
2103        }
2104        else {
2105                enum object_type type;
2106                if (size_only)
2107                        type = sha1_object_info(s->sha1, &s->size);
2108                else {
2109                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2110                        s->should_free = 1;
2111                }
2112        }
2113        return 0;
2114}
2115
2116void diff_free_filespec_blob(struct diff_filespec *s)
2117{
2118        if (s->should_free)
2119                free(s->data);
2120        else if (s->should_munmap)
2121                munmap(s->data, s->size);
2122
2123        if (s->should_free || s->should_munmap) {
2124                s->should_free = s->should_munmap = 0;
2125                s->data = NULL;
2126        }
2127}
2128
2129void diff_free_filespec_data(struct diff_filespec *s)
2130{
2131        diff_free_filespec_blob(s);
2132        free(s->cnt_data);
2133        s->cnt_data = NULL;
2134}
2135
2136static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2137                           void *blob,
2138                           unsigned long size,
2139                           const unsigned char *sha1,
2140                           int mode)
2141{
2142        int fd;
2143        struct strbuf buf = STRBUF_INIT;
2144        struct strbuf template = STRBUF_INIT;
2145        char *path_dup = xstrdup(path);
2146        const char *base = basename(path_dup);
2147
2148        /* Generate "XXXXXX_basename.ext" */
2149        strbuf_addstr(&template, "XXXXXX_");
2150        strbuf_addstr(&template, base);
2151
2152        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2153                        strlen(base) + 1);
2154        if (fd < 0)
2155                die_errno("unable to create temp-file");
2156        if (convert_to_working_tree(path,
2157                        (const char *)blob, (size_t)size, &buf)) {
2158                blob = buf.buf;
2159                size = buf.len;
2160        }
2161        if (write_in_full(fd, blob, size) != size)
2162                die_errno("unable to write temp-file");
2163        close(fd);
2164        temp->name = temp->tmp_path;
2165        strcpy(temp->hex, sha1_to_hex(sha1));
2166        temp->hex[40] = 0;
2167        sprintf(temp->mode, "%06o", mode);
2168        strbuf_release(&buf);
2169        strbuf_release(&template);
2170        free(path_dup);
2171}
2172
2173static struct diff_tempfile *prepare_temp_file(const char *name,
2174                struct diff_filespec *one)
2175{
2176        struct diff_tempfile *temp = claim_diff_tempfile();
2177
2178        if (!DIFF_FILE_VALID(one)) {
2179        not_a_valid_file:
2180                /* A '-' entry produces this for file-2, and
2181                 * a '+' entry produces this for file-1.
2182                 */
2183                temp->name = "/dev/null";
2184                strcpy(temp->hex, ".");
2185                strcpy(temp->mode, ".");
2186                return temp;
2187        }
2188
2189        if (!remove_tempfile_installed) {
2190                atexit(remove_tempfile);
2191                sigchain_push_common(remove_tempfile_on_signal);
2192                remove_tempfile_installed = 1;
2193        }
2194
2195        if (!one->sha1_valid ||
2196            reuse_worktree_file(name, one->sha1, 1)) {
2197                struct stat st;
2198                if (lstat(name, &st) < 0) {
2199                        if (errno == ENOENT)
2200                                goto not_a_valid_file;
2201                        die_errno("stat(%s)", name);
2202                }
2203                if (S_ISLNK(st.st_mode)) {
2204                        struct strbuf sb = STRBUF_INIT;
2205                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2206                                die_errno("readlink(%s)", name);
2207                        prep_temp_blob(name, temp, sb.buf, sb.len,
2208                                       (one->sha1_valid ?
2209                                        one->sha1 : null_sha1),
2210                                       (one->sha1_valid ?
2211                                        one->mode : S_IFLNK));
2212                        strbuf_release(&sb);
2213                }
2214                else {
2215                        /* we can borrow from the file in the work tree */
2216                        temp->name = name;
2217                        if (!one->sha1_valid)
2218                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2219                        else
2220                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2221                        /* Even though we may sometimes borrow the
2222                         * contents from the work tree, we always want
2223                         * one->mode.  mode is trustworthy even when
2224                         * !(one->sha1_valid), as long as
2225                         * DIFF_FILE_VALID(one).
2226                         */
2227                        sprintf(temp->mode, "%06o", one->mode);
2228                }
2229                return temp;
2230        }
2231        else {
2232                if (diff_populate_filespec(one, 0))
2233                        die("cannot read data blob for %s", one->path);
2234                prep_temp_blob(name, temp, one->data, one->size,
2235                               one->sha1, one->mode);
2236        }
2237        return temp;
2238}
2239
2240/* An external diff command takes:
2241 *
2242 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2243 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2244 *
2245 */
2246static void run_external_diff(const char *pgm,
2247                              const char *name,
2248                              const char *other,
2249                              struct diff_filespec *one,
2250                              struct diff_filespec *two,
2251                              const char *xfrm_msg,
2252                              int complete_rewrite)
2253{
2254        const char *spawn_arg[10];
2255        int retval;
2256        const char **arg = &spawn_arg[0];
2257
2258        if (one && two) {
2259                struct diff_tempfile *temp_one, *temp_two;
2260                const char *othername = (other ? other : name);
2261                temp_one = prepare_temp_file(name, one);
2262                temp_two = prepare_temp_file(othername, two);
2263                *arg++ = pgm;
2264                *arg++ = name;
2265                *arg++ = temp_one->name;
2266                *arg++ = temp_one->hex;
2267                *arg++ = temp_one->mode;
2268                *arg++ = temp_two->name;
2269                *arg++ = temp_two->hex;
2270                *arg++ = temp_two->mode;
2271                if (other) {
2272                        *arg++ = other;
2273                        *arg++ = xfrm_msg;
2274                }
2275        } else {
2276                *arg++ = pgm;
2277                *arg++ = name;
2278        }
2279        *arg = NULL;
2280        fflush(NULL);
2281        retval = run_command_v_opt(spawn_arg, 0);
2282        remove_tempfile();
2283        if (retval) {
2284                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2285                exit(1);
2286        }
2287}
2288
2289static int similarity_index(struct diff_filepair *p)
2290{
2291        return p->score * 100 / MAX_SCORE;
2292}
2293
2294static void fill_metainfo(struct strbuf *msg,
2295                          const char *name,
2296                          const char *other,
2297                          struct diff_filespec *one,
2298                          struct diff_filespec *two,
2299                          struct diff_options *o,
2300                          struct diff_filepair *p)
2301{
2302        strbuf_init(msg, PATH_MAX * 2 + 300);
2303        switch (p->status) {
2304        case DIFF_STATUS_COPIED:
2305                strbuf_addf(msg, "similarity index %d%%", similarity_index(p));
2306                strbuf_addstr(msg, "\ncopy from ");
2307                quote_c_style(name, msg, NULL, 0);
2308                strbuf_addstr(msg, "\ncopy to ");
2309                quote_c_style(other, msg, NULL, 0);
2310                strbuf_addch(msg, '\n');
2311                break;
2312        case DIFF_STATUS_RENAMED:
2313                strbuf_addf(msg, "similarity index %d%%", similarity_index(p));
2314                strbuf_addstr(msg, "\nrename from ");
2315                quote_c_style(name, msg, NULL, 0);
2316                strbuf_addstr(msg, "\nrename to ");
2317                quote_c_style(other, msg, NULL, 0);
2318                strbuf_addch(msg, '\n');
2319                break;
2320        case DIFF_STATUS_MODIFIED:
2321                if (p->score) {
2322                        strbuf_addf(msg, "dissimilarity index %d%%\n",
2323                                    similarity_index(p));
2324                        break;
2325                }
2326                /* fallthru */
2327        default:
2328                /* nothing */
2329                ;
2330        }
2331        if (one && two && hashcmp(one->sha1, two->sha1)) {
2332                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2333
2334                if (DIFF_OPT_TST(o, BINARY)) {
2335                        mmfile_t mf;
2336                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2337                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2338                                abbrev = 40;
2339                }
2340                strbuf_addf(msg, "index %.*s..%.*s",
2341                            abbrev, sha1_to_hex(one->sha1),
2342                            abbrev, sha1_to_hex(two->sha1));
2343                if (one->mode == two->mode)
2344                        strbuf_addf(msg, " %06o", one->mode);
2345                strbuf_addch(msg, '\n');
2346        }
2347        if (msg->len)
2348                strbuf_setlen(msg, msg->len - 1);
2349}
2350
2351static void run_diff_cmd(const char *pgm,
2352                         const char *name,
2353                         const char *other,
2354                         const char *attr_path,
2355                         struct diff_filespec *one,
2356                         struct diff_filespec *two,
2357                         struct strbuf *msg,
2358                         struct diff_options *o,
2359                         struct diff_filepair *p)
2360{
2361        const char *xfrm_msg = NULL;
2362        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2363
2364        if (msg) {
2365                fill_metainfo(msg, name, other, one, two, o, p);
2366                xfrm_msg = msg->len ? msg->buf : NULL;
2367        }
2368
2369        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2370                pgm = NULL;
2371        else {
2372                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2373                if (drv && drv->external)
2374                        pgm = drv->external;
2375        }
2376
2377        if (pgm) {
2378                run_external_diff(pgm, name, other, one, two, xfrm_msg,
2379                                  complete_rewrite);
2380                return;
2381        }
2382        if (one && two)
2383                builtin_diff(name, other ? other : name,
2384                             one, two, xfrm_msg, o, complete_rewrite);
2385        else
2386                fprintf(o->file, "* Unmerged path %s\n", name);
2387}
2388
2389static void diff_fill_sha1_info(struct diff_filespec *one)
2390{
2391        if (DIFF_FILE_VALID(one)) {
2392                if (!one->sha1_valid) {
2393                        struct stat st;
2394                        if (!strcmp(one->path, "-")) {
2395                                hashcpy(one->sha1, null_sha1);
2396                                return;
2397                        }
2398                        if (lstat(one->path, &st) < 0)
2399                                die_errno("stat '%s'", one->path);
2400                        if (index_path(one->sha1, one->path, &st, 0))
2401                                die("cannot hash %s", one->path);
2402                }
2403        }
2404        else
2405                hashclr(one->sha1);
2406}
2407
2408static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2409{
2410        /* Strip the prefix but do not molest /dev/null and absolute paths */
2411        if (*namep && **namep != '/')
2412                *namep += prefix_length;
2413        if (*otherp && **otherp != '/')
2414                *otherp += prefix_length;
2415}
2416
2417static void run_diff(struct diff_filepair *p, struct diff_options *o)
2418{
2419        const char *pgm = external_diff();
2420        struct strbuf msg;
2421        struct diff_filespec *one = p->one;
2422        struct diff_filespec *two = p->two;
2423        const char *name;
2424        const char *other;
2425        const char *attr_path;
2426
2427        name  = p->one->path;
2428        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2429        attr_path = name;
2430        if (o->prefix_length)
2431                strip_prefix(o->prefix_length, &name, &other);
2432
2433        if (DIFF_PAIR_UNMERGED(p)) {
2434                run_diff_cmd(pgm, name, NULL, attr_path,
2435                             NULL, NULL, NULL, o, p);
2436                return;
2437        }
2438
2439        diff_fill_sha1_info(one);
2440        diff_fill_sha1_info(two);
2441
2442        if (!pgm &&
2443            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2444            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2445                /*
2446                 * a filepair that changes between file and symlink
2447                 * needs to be split into deletion and creation.
2448                 */
2449                struct diff_filespec *null = alloc_filespec(two->path);
2450                run_diff_cmd(NULL, name, other, attr_path,
2451                             one, null, &msg, o, p);
2452                free(null);
2453                strbuf_release(&msg);
2454
2455                null = alloc_filespec(one->path);
2456                run_diff_cmd(NULL, name, other, attr_path,
2457                             null, two, &msg, o, p);
2458                free(null);
2459        }
2460        else
2461                run_diff_cmd(pgm, name, other, attr_path,
2462                             one, two, &msg, o, p);
2463
2464        strbuf_release(&msg);
2465}
2466
2467static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2468                         struct diffstat_t *diffstat)
2469{
2470        const char *name;
2471        const char *other;
2472        int complete_rewrite = 0;
2473
2474        if (DIFF_PAIR_UNMERGED(p)) {
2475                /* unmerged */
2476                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2477                return;
2478        }
2479
2480        name = p->one->path;
2481        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2482
2483        if (o->prefix_length)
2484                strip_prefix(o->prefix_length, &name, &other);
2485
2486        diff_fill_sha1_info(p->one);
2487        diff_fill_sha1_info(p->two);
2488
2489        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2490                complete_rewrite = 1;
2491        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2492}
2493
2494static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2495{
2496        const char *name;
2497        const char *other;
2498        const char *attr_path;
2499
2500        if (DIFF_PAIR_UNMERGED(p)) {
2501                /* unmerged */
2502                return;
2503        }
2504
2505        name = p->one->path;
2506        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2507        attr_path = other ? other : name;
2508
2509        if (o->prefix_length)
2510                strip_prefix(o->prefix_length, &name, &other);
2511
2512        diff_fill_sha1_info(p->one);
2513        diff_fill_sha1_info(p->two);
2514
2515        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
2516}
2517
2518void diff_setup(struct diff_options *options)
2519{
2520        memset(options, 0, sizeof(*options));
2521
2522        options->file = stdout;
2523
2524        options->line_termination = '\n';
2525        options->break_opt = -1;
2526        options->rename_limit = -1;
2527        options->dirstat_percent = 3;
2528        options->context = 3;
2529
2530        options->change = diff_change;
2531        options->add_remove = diff_addremove;
2532        if (diff_use_color_default > 0)
2533                DIFF_OPT_SET(options, COLOR_DIFF);
2534        options->detect_rename = diff_detect_rename_default;
2535
2536        if (!diff_mnemonic_prefix) {
2537                options->a_prefix = "a/";
2538                options->b_prefix = "b/";
2539        }
2540}
2541
2542int diff_setup_done(struct diff_options *options)
2543{
2544        int count = 0;
2545
2546        if (options->output_format & DIFF_FORMAT_NAME)
2547                count++;
2548        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2549                count++;
2550        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2551                count++;
2552        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2553                count++;
2554        if (count > 1)
2555                die("--name-only, --name-status, --check and -s are mutually exclusive");
2556
2557        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2558                options->detect_rename = DIFF_DETECT_COPY;
2559
2560        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
2561                options->prefix = NULL;
2562        if (options->prefix)
2563                options->prefix_length = strlen(options->prefix);
2564        else
2565                options->prefix_length = 0;
2566
2567        if (options->output_format & (DIFF_FORMAT_NAME |
2568                                      DIFF_FORMAT_NAME_STATUS |
2569                                      DIFF_FORMAT_CHECKDIFF |
2570                                      DIFF_FORMAT_NO_OUTPUT))
2571                options->output_format &= ~(DIFF_FORMAT_RAW |
2572                                            DIFF_FORMAT_NUMSTAT |
2573                                            DIFF_FORMAT_DIFFSTAT |
2574                                            DIFF_FORMAT_SHORTSTAT |
2575                                            DIFF_FORMAT_DIRSTAT |
2576                                            DIFF_FORMAT_SUMMARY |
2577                                            DIFF_FORMAT_PATCH);
2578
2579        /*
2580         * These cases always need recursive; we do not drop caller-supplied
2581         * recursive bits for other formats here.
2582         */
2583        if (options->output_format & (DIFF_FORMAT_PATCH |
2584                                      DIFF_FORMAT_NUMSTAT |
2585                                      DIFF_FORMAT_DIFFSTAT |
2586                                      DIFF_FORMAT_SHORTSTAT |
2587                                      DIFF_FORMAT_DIRSTAT |
2588                                      DIFF_FORMAT_SUMMARY |
2589                                      DIFF_FORMAT_CHECKDIFF))
2590                DIFF_OPT_SET(options, RECURSIVE);
2591        /*
2592         * Also pickaxe would not work very well if you do not say recursive
2593         */
2594        if (options->pickaxe)
2595                DIFF_OPT_SET(options, RECURSIVE);
2596
2597        if (options->detect_rename && options->rename_limit < 0)
2598                options->rename_limit = diff_rename_limit_default;
2599        if (options->setup & DIFF_SETUP_USE_CACHE) {
2600                if (!active_cache)
2601                        /* read-cache does not die even when it fails
2602                         * so it is safe for us to do this here.  Also
2603                         * it does not smudge active_cache or active_nr
2604                         * when it fails, so we do not have to worry about
2605                         * cleaning it up ourselves either.
2606                         */
2607                        read_cache();
2608        }
2609        if (options->abbrev <= 0 || 40 < options->abbrev)
2610                options->abbrev = 40; /* full */
2611
2612        /*
2613         * It does not make sense to show the first hit we happened
2614         * to have found.  It does not make sense not to return with
2615         * exit code in such a case either.
2616         */
2617        if (DIFF_OPT_TST(options, QUIET)) {
2618                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2619                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2620        }
2621
2622        return 0;
2623}
2624
2625static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2626{
2627        char c, *eq;
2628        int len;
2629
2630        if (*arg != '-')
2631                return 0;
2632        c = *++arg;
2633        if (!c)
2634                return 0;
2635        if (c == arg_short) {
2636                c = *++arg;
2637                if (!c)
2638                        return 1;
2639                if (val && isdigit(c)) {
2640                        char *end;
2641                        int n = strtoul(arg, &end, 10);
2642                        if (*end)
2643                                return 0;
2644                        *val = n;
2645                        return 1;
2646                }
2647                return 0;
2648        }
2649        if (c != '-')
2650                return 0;
2651        arg++;
2652        eq = strchr(arg, '=');
2653        if (eq)
2654                len = eq - arg;
2655        else
2656                len = strlen(arg);
2657        if (!len || strncmp(arg, arg_long, len))
2658                return 0;
2659        if (eq) {
2660                int n;
2661                char *end;
2662                if (!isdigit(*++eq))
2663                        return 0;
2664                n = strtoul(eq, &end, 10);
2665                if (*end)
2666                        return 0;
2667                *val = n;
2668        }
2669        return 1;
2670}
2671
2672static int diff_scoreopt_parse(const char *opt);
2673
2674int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2675{
2676        const char *arg = av[0];
2677
2678        /* Output format options */
2679        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2680                options->output_format |= DIFF_FORMAT_PATCH;
2681        else if (opt_arg(arg, 'U', "unified", &options->context))
2682                options->output_format |= DIFF_FORMAT_PATCH;
2683        else if (!strcmp(arg, "--raw"))
2684                options->output_format |= DIFF_FORMAT_RAW;
2685        else if (!strcmp(arg, "--patch-with-raw"))
2686                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2687        else if (!strcmp(arg, "--numstat"))
2688                options->output_format |= DIFF_FORMAT_NUMSTAT;
2689        else if (!strcmp(arg, "--shortstat"))
2690                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2691        else if (opt_arg(arg, 'X', "dirstat", &options->dirstat_percent))
2692                options->output_format |= DIFF_FORMAT_DIRSTAT;
2693        else if (!strcmp(arg, "--cumulative")) {
2694                options->output_format |= DIFF_FORMAT_DIRSTAT;
2695                DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
2696        } else if (opt_arg(arg, 0, "dirstat-by-file",
2697                           &options->dirstat_percent)) {
2698                options->output_format |= DIFF_FORMAT_DIRSTAT;
2699                DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
2700        }
2701        else if (!strcmp(arg, "--check"))
2702                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2703        else if (!strcmp(arg, "--summary"))
2704                options->output_format |= DIFF_FORMAT_SUMMARY;
2705        else if (!strcmp(arg, "--patch-with-stat"))
2706                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2707        else if (!strcmp(arg, "--name-only"))
2708                options->output_format |= DIFF_FORMAT_NAME;
2709        else if (!strcmp(arg, "--name-status"))
2710                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2711        else if (!strcmp(arg, "-s"))
2712                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2713        else if (!prefixcmp(arg, "--stat")) {
2714                char *end;
2715                int width = options->stat_width;
2716                int name_width = options->stat_name_width;
2717                arg += 6;
2718                end = (char *)arg;
2719
2720                switch (*arg) {
2721                case '-':
2722                        if (!prefixcmp(arg, "-width="))
2723                                width = strtoul(arg + 7, &end, 10);
2724                        else if (!prefixcmp(arg, "-name-width="))
2725                                name_width = strtoul(arg + 12, &end, 10);
2726                        break;
2727                case '=':
2728                        width = strtoul(arg+1, &end, 10);
2729                        if (*end == ',')
2730                                name_width = strtoul(end+1, &end, 10);
2731                }
2732
2733                /* Important! This checks all the error cases! */
2734                if (*end)
2735                        return 0;
2736                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2737                options->stat_name_width = name_width;
2738                options->stat_width = width;
2739        }
2740
2741        /* renames options */
2742        else if (!prefixcmp(arg, "-B")) {
2743                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2744                        return -1;
2745        }
2746        else if (!prefixcmp(arg, "-M")) {
2747                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2748                        return -1;
2749                options->detect_rename = DIFF_DETECT_RENAME;
2750        }
2751        else if (!prefixcmp(arg, "-C")) {
2752                if (options->detect_rename == DIFF_DETECT_COPY)
2753                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2754                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2755                        return -1;
2756                options->detect_rename = DIFF_DETECT_COPY;
2757        }
2758        else if (!strcmp(arg, "--no-renames"))
2759                options->detect_rename = 0;
2760        else if (!strcmp(arg, "--relative"))
2761                DIFF_OPT_SET(options, RELATIVE_NAME);
2762        else if (!prefixcmp(arg, "--relative=")) {
2763                DIFF_OPT_SET(options, RELATIVE_NAME);
2764                options->prefix = arg + 11;
2765        }
2766
2767        /* xdiff options */
2768        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2769                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
2770        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2771                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
2772        else if (!strcmp(arg, "--ignore-space-at-eol"))
2773                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
2774        else if (!strcmp(arg, "--patience"))
2775                DIFF_XDL_SET(options, PATIENCE_DIFF);
2776
2777        /* flags options */
2778        else if (!strcmp(arg, "--binary")) {
2779                options->output_format |= DIFF_FORMAT_PATCH;
2780                DIFF_OPT_SET(options, BINARY);
2781        }
2782        else if (!strcmp(arg, "--full-index"))
2783                DIFF_OPT_SET(options, FULL_INDEX);
2784        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2785                DIFF_OPT_SET(options, TEXT);
2786        else if (!strcmp(arg, "-R"))
2787                DIFF_OPT_SET(options, REVERSE_DIFF);
2788        else if (!strcmp(arg, "--find-copies-harder"))
2789                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2790        else if (!strcmp(arg, "--follow"))
2791                DIFF_OPT_SET(options, FOLLOW_RENAMES);
2792        else if (!strcmp(arg, "--color"))
2793                DIFF_OPT_SET(options, COLOR_DIFF);
2794        else if (!strcmp(arg, "--no-color"))
2795                DIFF_OPT_CLR(options, COLOR_DIFF);
2796        else if (!strcmp(arg, "--color-words")) {
2797                DIFF_OPT_SET(options, COLOR_DIFF);
2798                DIFF_OPT_SET(options, COLOR_DIFF_WORDS);
2799        }
2800        else if (!prefixcmp(arg, "--color-words=")) {
2801                DIFF_OPT_SET(options, COLOR_DIFF);
2802                DIFF_OPT_SET(options, COLOR_DIFF_WORDS);
2803                options->word_regex = arg + 14;
2804        }
2805        else if (!strcmp(arg, "--exit-code"))
2806                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2807        else if (!strcmp(arg, "--quiet"))
2808                DIFF_OPT_SET(options, QUIET);
2809        else if (!strcmp(arg, "--ext-diff"))
2810                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2811        else if (!strcmp(arg, "--no-ext-diff"))
2812                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2813        else if (!strcmp(arg, "--textconv"))
2814                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
2815        else if (!strcmp(arg, "--no-textconv"))
2816                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
2817        else if (!strcmp(arg, "--ignore-submodules"))
2818                DIFF_OPT_SET(options, IGNORE_SUBMODULES);
2819        else if (!strcmp(arg, "--submodule"))
2820                DIFF_OPT_SET(options, SUBMODULE_LOG);
2821        else if (!prefixcmp(arg, "--submodule=")) {
2822                if (!strcmp(arg + 12, "log"))
2823                        DIFF_OPT_SET(options, SUBMODULE_LOG);
2824        }
2825
2826        /* misc options */
2827        else if (!strcmp(arg, "-z"))
2828                options->line_termination = 0;
2829        else if (!prefixcmp(arg, "-l"))
2830                options->rename_limit = strtoul(arg+2, NULL, 10);
2831        else if (!prefixcmp(arg, "-S"))
2832                options->pickaxe = arg + 2;
2833        else if (!strcmp(arg, "--pickaxe-all"))
2834                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2835        else if (!strcmp(arg, "--pickaxe-regex"))
2836                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2837        else if (!prefixcmp(arg, "-O"))
2838                options->orderfile = arg + 2;
2839        else if (!prefixcmp(arg, "--diff-filter="))
2840                options->filter = arg + 14;
2841        else if (!strcmp(arg, "--abbrev"))
2842                options->abbrev = DEFAULT_ABBREV;
2843        else if (!prefixcmp(arg, "--abbrev=")) {
2844                options->abbrev = strtoul(arg + 9, NULL, 10);
2845                if (options->abbrev < MINIMUM_ABBREV)
2846                        options->abbrev = MINIMUM_ABBREV;
2847                else if (40 < options->abbrev)
2848                        options->abbrev = 40;
2849        }
2850        else if (!prefixcmp(arg, "--src-prefix="))
2851                options->a_prefix = arg + 13;
2852        else if (!prefixcmp(arg, "--dst-prefix="))
2853                options->b_prefix = arg + 13;
2854        else if (!strcmp(arg, "--no-prefix"))
2855                options->a_prefix = options->b_prefix = "";
2856        else if (opt_arg(arg, '\0', "inter-hunk-context",
2857                         &options->interhunkcontext))
2858                ;
2859        else if (!prefixcmp(arg, "--output=")) {
2860                options->file = fopen(arg + strlen("--output="), "w");
2861                options->close_file = 1;
2862        } else
2863                return 0;
2864        return 1;
2865}
2866
2867static int parse_num(const char **cp_p)
2868{
2869        unsigned long num, scale;
2870        int ch, dot;
2871        const char *cp = *cp_p;
2872
2873        num = 0;
2874        scale = 1;
2875        dot = 0;
2876        for (;;) {
2877                ch = *cp;
2878                if ( !dot && ch == '.' ) {
2879                        scale = 1;
2880                        dot = 1;
2881                } else if ( ch == '%' ) {
2882                        scale = dot ? scale*100 : 100;
2883                        cp++;   /* % is always at the end */
2884                        break;
2885                } else if ( ch >= '0' && ch <= '9' ) {
2886                        if ( scale < 100000 ) {
2887                                scale *= 10;
2888                                num = (num*10) + (ch-'0');
2889                        }
2890                } else {
2891                        break;
2892                }
2893                cp++;
2894        }
2895        *cp_p = cp;
2896
2897        /* user says num divided by scale and we say internally that
2898         * is MAX_SCORE * num / scale.
2899         */
2900        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2901}
2902
2903static int diff_scoreopt_parse(const char *opt)
2904{
2905        int opt1, opt2, cmd;
2906
2907        if (*opt++ != '-')
2908                return -1;
2909        cmd = *opt++;
2910        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2911                return -1; /* that is not a -M, -C nor -B option */
2912
2913        opt1 = parse_num(&opt);
2914        if (cmd != 'B')
2915                opt2 = 0;
2916        else {
2917                if (*opt == 0)
2918                        opt2 = 0;
2919                else if (*opt != '/')
2920                        return -1; /* we expect -B80/99 or -B80 */
2921                else {
2922                        opt++;
2923                        opt2 = parse_num(&opt);
2924                }
2925        }
2926        if (*opt != 0)
2927                return -1;
2928        return opt1 | (opt2 << 16);
2929}
2930
2931struct diff_queue_struct diff_queued_diff;
2932
2933void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2934{
2935        if (queue->alloc <= queue->nr) {
2936                queue->alloc = alloc_nr(queue->alloc);
2937                queue->queue = xrealloc(queue->queue,
2938                                        sizeof(dp) * queue->alloc);
2939        }
2940        queue->queue[queue->nr++] = dp;
2941}
2942
2943struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2944                                 struct diff_filespec *one,
2945                                 struct diff_filespec *two)
2946{
2947        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2948        dp->one = one;
2949        dp->two = two;
2950        if (queue)
2951                diff_q(queue, dp);
2952        return dp;
2953}
2954
2955void diff_free_filepair(struct diff_filepair *p)
2956{
2957        free_filespec(p->one);
2958        free_filespec(p->two);
2959        free(p);
2960}
2961
2962/* This is different from find_unique_abbrev() in that
2963 * it stuffs the result with dots for alignment.
2964 */
2965const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2966{
2967        int abblen;
2968        const char *abbrev;
2969        if (len == 40)
2970                return sha1_to_hex(sha1);
2971
2972        abbrev = find_unique_abbrev(sha1, len);
2973        abblen = strlen(abbrev);
2974        if (abblen < 37) {
2975                static char hex[41];
2976                if (len < abblen && abblen <= len + 2)
2977                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2978                else
2979                        sprintf(hex, "%s...", abbrev);
2980                return hex;
2981        }
2982        return sha1_to_hex(sha1);
2983}
2984
2985static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2986{
2987        int line_termination = opt->line_termination;
2988        int inter_name_termination = line_termination ? '\t' : '\0';
2989
2990        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2991                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
2992                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
2993                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2994        }
2995        if (p->score) {
2996                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
2997                        inter_name_termination);
2998        } else {
2999                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3000        }
3001
3002        if (p->status == DIFF_STATUS_COPIED ||
3003            p->status == DIFF_STATUS_RENAMED) {
3004                const char *name_a, *name_b;
3005                name_a = p->one->path;
3006                name_b = p->two->path;
3007                strip_prefix(opt->prefix_length, &name_a, &name_b);
3008                write_name_quoted(name_a, opt->file, inter_name_termination);
3009                write_name_quoted(name_b, opt->file, line_termination);
3010        } else {
3011                const char *name_a, *name_b;
3012                name_a = p->one->mode ? p->one->path : p->two->path;
3013                name_b = NULL;
3014                strip_prefix(opt->prefix_length, &name_a, &name_b);
3015                write_name_quoted(name_a, opt->file, line_termination);
3016        }
3017}
3018
3019int diff_unmodified_pair(struct diff_filepair *p)
3020{
3021        /* This function is written stricter than necessary to support
3022         * the currently implemented transformers, but the idea is to
3023         * let transformers to produce diff_filepairs any way they want,
3024         * and filter and clean them up here before producing the output.
3025         */
3026        struct diff_filespec *one = p->one, *two = p->two;
3027
3028        if (DIFF_PAIR_UNMERGED(p))
3029                return 0; /* unmerged is interesting */
3030
3031        /* deletion, addition, mode or type change
3032         * and rename are all interesting.
3033         */
3034        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3035            DIFF_PAIR_MODE_CHANGED(p) ||
3036            strcmp(one->path, two->path))
3037                return 0;
3038
3039        /* both are valid and point at the same path.  that is, we are
3040         * dealing with a change.
3041         */
3042        if (one->sha1_valid && two->sha1_valid &&
3043            !hashcmp(one->sha1, two->sha1))
3044                return 1; /* no change */
3045        if (!one->sha1_valid && !two->sha1_valid)
3046                return 1; /* both look at the same file on the filesystem. */
3047        return 0;
3048}
3049
3050static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3051{
3052        if (diff_unmodified_pair(p))
3053                return;
3054
3055        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3056            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3057                return; /* no tree diffs in patch format */
3058
3059        run_diff(p, o);
3060}
3061
3062static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3063                            struct diffstat_t *diffstat)
3064{
3065        if (diff_unmodified_pair(p))
3066                return;
3067
3068        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3069            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3070                return; /* no tree diffs in patch format */
3071
3072        run_diffstat(p, o, diffstat);
3073}
3074
3075static void diff_flush_checkdiff(struct diff_filepair *p,
3076                struct diff_options *o)
3077{
3078        if (diff_unmodified_pair(p))
3079                return;
3080
3081        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3082            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3083                return; /* no tree diffs in patch format */
3084
3085        run_checkdiff(p, o);
3086}
3087
3088int diff_queue_is_empty(void)
3089{
3090        struct diff_queue_struct *q = &diff_queued_diff;
3091        int i;
3092        for (i = 0; i < q->nr; i++)
3093                if (!diff_unmodified_pair(q->queue[i]))
3094                        return 0;
3095        return 1;
3096}
3097
3098#if DIFF_DEBUG
3099void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3100{
3101        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3102                x, one ? one : "",
3103                s->path,
3104                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3105                s->mode,
3106                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3107        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3108                x, one ? one : "",
3109                s->size, s->xfrm_flags);
3110}
3111
3112void diff_debug_filepair(const struct diff_filepair *p, int i)
3113{
3114        diff_debug_filespec(p->one, i, "one");
3115        diff_debug_filespec(p->two, i, "two");
3116        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3117                p->score, p->status ? p->status : '?',
3118                p->one->rename_used, p->broken_pair);
3119}
3120
3121void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3122{
3123        int i;
3124        if (msg)
3125                fprintf(stderr, "%s\n", msg);
3126        fprintf(stderr, "q->nr = %d\n", q->nr);
3127        for (i = 0; i < q->nr; i++) {
3128                struct diff_filepair *p = q->queue[i];
3129                diff_debug_filepair(p, i);
3130        }
3131}
3132#endif
3133
3134static void diff_resolve_rename_copy(void)
3135{
3136        int i;
3137        struct diff_filepair *p;
3138        struct diff_queue_struct *q = &diff_queued_diff;
3139
3140        diff_debug_queue("resolve-rename-copy", q);
3141
3142        for (i = 0; i < q->nr; i++) {
3143                p = q->queue[i];
3144                p->status = 0; /* undecided */
3145                if (DIFF_PAIR_UNMERGED(p))
3146                        p->status = DIFF_STATUS_UNMERGED;
3147                else if (!DIFF_FILE_VALID(p->one))
3148                        p->status = DIFF_STATUS_ADDED;
3149                else if (!DIFF_FILE_VALID(p->two))
3150                        p->status = DIFF_STATUS_DELETED;
3151                else if (DIFF_PAIR_TYPE_CHANGED(p))
3152                        p->status = DIFF_STATUS_TYPE_CHANGED;
3153
3154                /* from this point on, we are dealing with a pair
3155                 * whose both sides are valid and of the same type, i.e.
3156                 * either in-place edit or rename/copy edit.
3157                 */
3158                else if (DIFF_PAIR_RENAME(p)) {
3159                        /*
3160                         * A rename might have re-connected a broken
3161                         * pair up, causing the pathnames to be the
3162                         * same again. If so, that's not a rename at
3163                         * all, just a modification..
3164                         *
3165                         * Otherwise, see if this source was used for
3166                         * multiple renames, in which case we decrement
3167                         * the count, and call it a copy.
3168                         */
3169                        if (!strcmp(p->one->path, p->two->path))
3170                                p->status = DIFF_STATUS_MODIFIED;
3171                        else if (--p->one->rename_used > 0)
3172                                p->status = DIFF_STATUS_COPIED;
3173                        else
3174                                p->status = DIFF_STATUS_RENAMED;
3175                }
3176                else if (hashcmp(p->one->sha1, p->two->sha1) ||
3177                         p->one->mode != p->two->mode ||
3178                         is_null_sha1(p->one->sha1))
3179                        p->status = DIFF_STATUS_MODIFIED;
3180                else {
3181                        /* This is a "no-change" entry and should not
3182                         * happen anymore, but prepare for broken callers.
3183                         */
3184                        error("feeding unmodified %s to diffcore",
3185                              p->one->path);
3186                        p->status = DIFF_STATUS_UNKNOWN;
3187                }
3188        }
3189        diff_debug_queue("resolve-rename-copy done", q);
3190}
3191
3192static int check_pair_status(struct diff_filepair *p)
3193{
3194        switch (p->status) {
3195        case DIFF_STATUS_UNKNOWN:
3196                return 0;
3197        case 0:
3198                die("internal error in diff-resolve-rename-copy");
3199        default:
3200                return 1;
3201        }
3202}
3203
3204static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3205{
3206        int fmt = opt->output_format;
3207
3208        if (fmt & DIFF_FORMAT_CHECKDIFF)
3209                diff_flush_checkdiff(p, opt);
3210        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
3211                diff_flush_raw(p, opt);
3212        else if (fmt & DIFF_FORMAT_NAME) {
3213                const char *name_a, *name_b;
3214                name_a = p->two->path;
3215                name_b = NULL;
3216                strip_prefix(opt->prefix_length, &name_a, &name_b);
3217                write_name_quoted(name_a, opt->file, opt->line_termination);
3218        }
3219}
3220
3221static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
3222{
3223        if (fs->mode)
3224                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
3225        else
3226                fprintf(file, " %s ", newdelete);
3227        write_name_quoted(fs->path, file, '\n');
3228}
3229
3230
3231static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name)
3232{
3233        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
3234                fprintf(file, " mode change %06o => %06o%c", p->one->mode, p->two->mode,
3235                        show_name ? ' ' : '\n');
3236                if (show_name) {
3237                        write_name_quoted(p->two->path, file, '\n');
3238                }
3239        }
3240}
3241
3242static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p)
3243{
3244        char *names = pprint_rename(p->one->path, p->two->path);
3245
3246        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
3247        free(names);
3248        show_mode_change(file, p, 0);
3249}
3250
3251static void diff_summary(FILE *file, struct diff_filepair *p)
3252{
3253        switch(p->status) {
3254        case DIFF_STATUS_DELETED:
3255                show_file_mode_name(file, "delete", p->one);
3256                break;
3257        case DIFF_STATUS_ADDED:
3258                show_file_mode_name(file, "create", p->two);
3259                break;
3260        case DIFF_STATUS_COPIED:
3261                show_rename_copy(file, "copy", p);
3262                break;
3263        case DIFF_STATUS_RENAMED:
3264                show_rename_copy(file, "rename", p);
3265                break;
3266        default:
3267                if (p->score) {
3268                        fputs(" rewrite ", file);
3269                        write_name_quoted(p->two->path, file, ' ');
3270                        fprintf(file, "(%d%%)\n", similarity_index(p));
3271                }
3272                show_mode_change(file, p, !p->score);
3273                break;
3274        }
3275}
3276
3277struct patch_id_t {
3278        git_SHA_CTX *ctx;
3279        int patchlen;
3280};
3281
3282static int remove_space(char *line, int len)
3283{
3284        int i;
3285        char *dst = line;
3286        unsigned char c;
3287
3288        for (i = 0; i < len; i++)
3289                if (!isspace((c = line[i])))
3290                        *dst++ = c;
3291
3292        return dst - line;
3293}
3294
3295static void patch_id_consume(void *priv, char *line, unsigned long len)
3296{
3297        struct patch_id_t *data = priv;
3298        int new_len;
3299
3300        /* Ignore line numbers when computing the SHA1 of the patch */
3301        if (!prefixcmp(line, "@@ -"))
3302                return;
3303
3304        new_len = remove_space(line, len);
3305
3306        git_SHA1_Update(data->ctx, line, new_len);
3307        data->patchlen += new_len;
3308}
3309
3310/* returns 0 upon success, and writes result into sha1 */
3311static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
3312{
3313        struct diff_queue_struct *q = &diff_queued_diff;
3314        int i;
3315        git_SHA_CTX ctx;
3316        struct patch_id_t data;
3317        char buffer[PATH_MAX * 4 + 20];
3318
3319        git_SHA1_Init(&ctx);
3320        memset(&data, 0, sizeof(struct patch_id_t));
3321        data.ctx = &ctx;
3322
3323        for (i = 0; i < q->nr; i++) {
3324                xpparam_t xpp;
3325                xdemitconf_t xecfg;
3326                xdemitcb_t ecb;
3327                mmfile_t mf1, mf2;
3328                struct diff_filepair *p = q->queue[i];
3329                int len1, len2;
3330
3331                memset(&xpp, 0, sizeof(xpp));
3332                memset(&xecfg, 0, sizeof(xecfg));
3333                if (p->status == 0)
3334                        return error("internal diff status error");
3335                if (p->status == DIFF_STATUS_UNKNOWN)
3336                        continue;
3337                if (diff_unmodified_pair(p))
3338                        continue;
3339                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3340                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3341                        continue;
3342                if (DIFF_PAIR_UNMERGED(p))
3343                        continue;
3344
3345                diff_fill_sha1_info(p->one);
3346                diff_fill_sha1_info(p->two);
3347                if (fill_mmfile(&mf1, p->one) < 0 ||
3348                                fill_mmfile(&mf2, p->two) < 0)
3349                        return error("unable to read files to diff");
3350
3351                len1 = remove_space(p->one->path, strlen(p->one->path));
3352                len2 = remove_space(p->two->path, strlen(p->two->path));
3353                if (p->one->mode == 0)
3354                        len1 = snprintf(buffer, sizeof(buffer),
3355                                        "diff--gita/%.*sb/%.*s"
3356                                        "newfilemode%06o"
3357                                        "---/dev/null"
3358                                        "+++b/%.*s",
3359                                        len1, p->one->path,
3360                                        len2, p->two->path,
3361                                        p->two->mode,
3362                                        len2, p->two->path);
3363                else if (p->two->mode == 0)
3364                        len1 = snprintf(buffer, sizeof(buffer),
3365                                        "diff--gita/%.*sb/%.*s"
3366                                        "deletedfilemode%06o"
3367                                        "---a/%.*s"
3368                                        "+++/dev/null",
3369                                        len1, p->one->path,
3370                                        len2, p->two->path,
3371                                        p->one->mode,
3372                                        len1, p->one->path);
3373                else
3374                        len1 = snprintf(buffer, sizeof(buffer),
3375                                        "diff--gita/%.*sb/%.*s"
3376                                        "---a/%.*s"
3377                                        "+++b/%.*s",
3378                                        len1, p->one->path,
3379                                        len2, p->two->path,
3380                                        len1, p->one->path,
3381                                        len2, p->two->path);
3382                git_SHA1_Update(&ctx, buffer, len1);
3383
3384                xpp.flags = XDF_NEED_MINIMAL;
3385                xecfg.ctxlen = 3;
3386                xecfg.flags = XDL_EMIT_FUNCNAMES;
3387                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
3388                              &xpp, &xecfg, &ecb);
3389        }
3390
3391        git_SHA1_Final(sha1, &ctx);
3392        return 0;
3393}
3394
3395int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
3396{
3397        struct diff_queue_struct *q = &diff_queued_diff;
3398        int i;
3399        int result = diff_get_patch_id(options, sha1);
3400
3401        for (i = 0; i < q->nr; i++)
3402                diff_free_filepair(q->queue[i]);
3403
3404        free(q->queue);
3405        q->queue = NULL;
3406        q->nr = q->alloc = 0;
3407
3408        return result;
3409}
3410
3411static int is_summary_empty(const struct diff_queue_struct *q)
3412{
3413        int i;
3414
3415        for (i = 0; i < q->nr; i++) {
3416                const struct diff_filepair *p = q->queue[i];
3417
3418                switch (p->status) {
3419                case DIFF_STATUS_DELETED:
3420                case DIFF_STATUS_ADDED:
3421                case DIFF_STATUS_COPIED:
3422                case DIFF_STATUS_RENAMED:
3423                        return 0;
3424                default:
3425                        if (p->score)
3426                                return 0;
3427                        if (p->one->mode && p->two->mode &&
3428                            p->one->mode != p->two->mode)
3429                                return 0;
3430                        break;
3431                }
3432        }
3433        return 1;
3434}
3435
3436void diff_flush(struct diff_options *options)
3437{
3438        struct diff_queue_struct *q = &diff_queued_diff;
3439        int i, output_format = options->output_format;
3440        int separator = 0;
3441
3442        /*
3443         * Order: raw, stat, summary, patch
3444         * or:    name/name-status/checkdiff (other bits clear)
3445         */
3446        if (!q->nr)
3447                goto free_queue;
3448
3449        if (output_format & (DIFF_FORMAT_RAW |
3450                             DIFF_FORMAT_NAME |
3451                             DIFF_FORMAT_NAME_STATUS |
3452                             DIFF_FORMAT_CHECKDIFF)) {
3453                for (i = 0; i < q->nr; i++) {
3454                        struct diff_filepair *p = q->queue[i];
3455                        if (check_pair_status(p))
3456                                flush_one_pair(p, options);
3457                }
3458                separator++;
3459        }
3460
3461        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
3462                struct diffstat_t diffstat;
3463
3464                memset(&diffstat, 0, sizeof(struct diffstat_t));
3465                for (i = 0; i < q->nr; i++) {
3466                        struct diff_filepair *p = q->queue[i];
3467                        if (check_pair_status(p))
3468                                diff_flush_stat(p, options, &diffstat);
3469                }
3470                if (output_format & DIFF_FORMAT_NUMSTAT)
3471                        show_numstat(&diffstat, options);
3472                if (output_format & DIFF_FORMAT_DIFFSTAT)
3473                        show_stats(&diffstat, options);
3474                if (output_format & DIFF_FORMAT_SHORTSTAT)
3475                        show_shortstats(&diffstat, options);
3476                free_diffstat_info(&diffstat);
3477                separator++;
3478        }
3479        if (output_format & DIFF_FORMAT_DIRSTAT)
3480                show_dirstat(options);
3481
3482        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
3483                for (i = 0; i < q->nr; i++)
3484                        diff_summary(options->file, q->queue[i]);
3485                separator++;
3486        }
3487
3488        if (output_format & DIFF_FORMAT_PATCH) {
3489                if (separator) {
3490                        putc(options->line_termination, options->file);
3491                        if (options->stat_sep) {
3492                                /* attach patch instead of inline */
3493                                fputs(options->stat_sep, options->file);
3494                        }
3495                }
3496
3497                for (i = 0; i < q->nr; i++) {
3498                        struct diff_filepair *p = q->queue[i];
3499                        if (check_pair_status(p))
3500                                diff_flush_patch(p, options);
3501                }
3502        }
3503
3504        if (output_format & DIFF_FORMAT_CALLBACK)
3505                options->format_callback(q, options, options->format_callback_data);
3506
3507        for (i = 0; i < q->nr; i++)
3508                diff_free_filepair(q->queue[i]);
3509free_queue:
3510        free(q->queue);
3511        q->queue = NULL;
3512        q->nr = q->alloc = 0;
3513        if (options->close_file)
3514                fclose(options->file);
3515}
3516
3517static void diffcore_apply_filter(const char *filter)
3518{
3519        int i;
3520        struct diff_queue_struct *q = &diff_queued_diff;
3521        struct diff_queue_struct outq;
3522        outq.queue = NULL;
3523        outq.nr = outq.alloc = 0;
3524
3525        if (!filter)
3526                return;
3527
3528        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3529                int found;
3530                for (i = found = 0; !found && i < q->nr; i++) {
3531                        struct diff_filepair *p = q->queue[i];
3532                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3533                             ((p->score &&
3534                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3535                              (!p->score &&
3536                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3537                            ((p->status != DIFF_STATUS_MODIFIED) &&
3538                             strchr(filter, p->status)))
3539                                found++;
3540                }
3541                if (found)
3542                        return;
3543
3544                /* otherwise we will clear the whole queue
3545                 * by copying the empty outq at the end of this
3546                 * function, but first clear the current entries
3547                 * in the queue.
3548                 */
3549                for (i = 0; i < q->nr; i++)
3550                        diff_free_filepair(q->queue[i]);
3551        }
3552        else {
3553                /* Only the matching ones */
3554                for (i = 0; i < q->nr; i++) {
3555                        struct diff_filepair *p = q->queue[i];
3556
3557                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3558                             ((p->score &&
3559                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3560                              (!p->score &&
3561                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3562                            ((p->status != DIFF_STATUS_MODIFIED) &&
3563                             strchr(filter, p->status)))
3564                                diff_q(&outq, p);
3565                        else
3566                                diff_free_filepair(p);
3567                }
3568        }
3569        free(q->queue);
3570        *q = outq;
3571}
3572
3573/* Check whether two filespecs with the same mode and size are identical */
3574static int diff_filespec_is_identical(struct diff_filespec *one,
3575                                      struct diff_filespec *two)
3576{
3577        if (S_ISGITLINK(one->mode))
3578                return 0;
3579        if (diff_populate_filespec(one, 0))
3580                return 0;
3581        if (diff_populate_filespec(two, 0))
3582                return 0;
3583        return !memcmp(one->data, two->data, one->size);
3584}
3585
3586static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3587{
3588        int i;
3589        struct diff_queue_struct *q = &diff_queued_diff;
3590        struct diff_queue_struct outq;
3591        outq.queue = NULL;
3592        outq.nr = outq.alloc = 0;
3593
3594        for (i = 0; i < q->nr; i++) {
3595                struct diff_filepair *p = q->queue[i];
3596
3597                /*
3598                 * 1. Entries that come from stat info dirtyness
3599                 *    always have both sides (iow, not create/delete),
3600                 *    one side of the object name is unknown, with
3601                 *    the same mode and size.  Keep the ones that
3602                 *    do not match these criteria.  They have real
3603                 *    differences.
3604                 *
3605                 * 2. At this point, the file is known to be modified,
3606                 *    with the same mode and size, and the object
3607                 *    name of one side is unknown.  Need to inspect
3608                 *    the identical contents.
3609                 */
3610                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3611                    !DIFF_FILE_VALID(p->two) ||
3612                    (p->one->sha1_valid && p->two->sha1_valid) ||
3613                    (p->one->mode != p->two->mode) ||
3614                    diff_populate_filespec(p->one, 1) ||
3615                    diff_populate_filespec(p->two, 1) ||
3616                    (p->one->size != p->two->size) ||
3617                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3618                        diff_q(&outq, p);
3619                else {
3620                        /*
3621                         * The caller can subtract 1 from skip_stat_unmatch
3622                         * to determine how many paths were dirty only
3623                         * due to stat info mismatch.
3624                         */
3625                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3626                                diffopt->skip_stat_unmatch++;
3627                        diff_free_filepair(p);
3628                }
3629        }
3630        free(q->queue);
3631        *q = outq;
3632}
3633
3634void diffcore_std(struct diff_options *options)
3635{
3636        if (options->skip_stat_unmatch)
3637                diffcore_skip_stat_unmatch(options);
3638        if (options->break_opt != -1)
3639                diffcore_break(options->break_opt);
3640        if (options->detect_rename)
3641                diffcore_rename(options);
3642        if (options->break_opt != -1)
3643                diffcore_merge_broken();
3644        if (options->pickaxe)
3645                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3646        if (options->orderfile)
3647                diffcore_order(options->orderfile);
3648        diff_resolve_rename_copy();
3649        diffcore_apply_filter(options->filter);
3650
3651        if (diff_queued_diff.nr)
3652                DIFF_OPT_SET(options, HAS_CHANGES);
3653        else
3654                DIFF_OPT_CLR(options, HAS_CHANGES);
3655}
3656
3657int diff_result_code(struct diff_options *opt, int status)
3658{
3659        int result = 0;
3660        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3661            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3662                return status;
3663        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3664            DIFF_OPT_TST(opt, HAS_CHANGES))
3665                result |= 01;
3666        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3667            DIFF_OPT_TST(opt, CHECK_FAILED))
3668                result |= 02;
3669        return result;
3670}
3671
3672void diff_addremove(struct diff_options *options,
3673                    int addremove, unsigned mode,
3674                    const unsigned char *sha1,
3675                    const char *concatpath)
3676{
3677        struct diff_filespec *one, *two;
3678
3679        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(mode))
3680                return;
3681
3682        /* This may look odd, but it is a preparation for
3683         * feeding "there are unchanged files which should
3684         * not produce diffs, but when you are doing copy
3685         * detection you would need them, so here they are"
3686         * entries to the diff-core.  They will be prefixed
3687         * with something like '=' or '*' (I haven't decided
3688         * which but should not make any difference).
3689         * Feeding the same new and old to diff_change()
3690         * also has the same effect.
3691         * Before the final output happens, they are pruned after
3692         * merged into rename/copy pairs as appropriate.
3693         */
3694        if (DIFF_OPT_TST(options, REVERSE_DIFF))
3695                addremove = (addremove == '+' ? '-' :
3696                             addremove == '-' ? '+' : addremove);
3697
3698        if (options->prefix &&
3699            strncmp(concatpath, options->prefix, options->prefix_length))
3700                return;
3701
3702        one = alloc_filespec(concatpath);
3703        two = alloc_filespec(concatpath);
3704
3705        if (addremove != '+')
3706                fill_filespec(one, sha1, mode);
3707        if (addremove != '-')
3708                fill_filespec(two, sha1, mode);
3709
3710        diff_queue(&diff_queued_diff, one, two);
3711        DIFF_OPT_SET(options, HAS_CHANGES);
3712}
3713
3714void diff_change(struct diff_options *options,
3715                 unsigned old_mode, unsigned new_mode,
3716                 const unsigned char *old_sha1,
3717                 const unsigned char *new_sha1,
3718                 const char *concatpath)
3719{
3720        struct diff_filespec *one, *two;
3721
3722        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(old_mode)
3723                        && S_ISGITLINK(new_mode))
3724                return;
3725
3726        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3727                unsigned tmp;
3728                const unsigned char *tmp_c;
3729                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3730                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3731        }
3732
3733        if (options->prefix &&
3734            strncmp(concatpath, options->prefix, options->prefix_length))
3735                return;
3736
3737        one = alloc_filespec(concatpath);
3738        two = alloc_filespec(concatpath);
3739        fill_filespec(one, old_sha1, old_mode);
3740        fill_filespec(two, new_sha1, new_mode);
3741
3742        diff_queue(&diff_queued_diff, one, two);
3743        DIFF_OPT_SET(options, HAS_CHANGES);
3744}
3745
3746void diff_unmerge(struct diff_options *options,
3747                  const char *path,
3748                  unsigned mode, const unsigned char *sha1)
3749{
3750        struct diff_filespec *one, *two;
3751
3752        if (options->prefix &&
3753            strncmp(path, options->prefix, options->prefix_length))
3754                return;
3755
3756        one = alloc_filespec(path);
3757        two = alloc_filespec(path);
3758        fill_filespec(one, sha1, mode);
3759        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3760}
3761
3762static char *run_textconv(const char *pgm, struct diff_filespec *spec,
3763                size_t *outsize)
3764{
3765        struct diff_tempfile *temp;
3766        const char *argv[3];
3767        const char **arg = argv;
3768        struct child_process child;
3769        struct strbuf buf = STRBUF_INIT;
3770
3771        temp = prepare_temp_file(spec->path, spec);
3772        *arg++ = pgm;
3773        *arg++ = temp->name;
3774        *arg = NULL;
3775
3776        memset(&child, 0, sizeof(child));
3777        child.argv = argv;
3778        child.out = -1;
3779        if (start_command(&child) != 0 ||
3780            strbuf_read(&buf, child.out, 0) < 0 ||
3781            finish_command(&child) != 0) {
3782                strbuf_release(&buf);
3783                remove_tempfile();
3784                error("error running textconv command '%s'", pgm);
3785                return NULL;
3786        }
3787        remove_tempfile();
3788
3789        return strbuf_detach(&buf, outsize);
3790}