diff.con commit Merge branch 'jc/maint-1.6.0-blank-at-eof' (early part) into jc/maint-blank-at-eof (61c6457)
   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
  17#ifdef NO_FAST_WORKING_DIRECTORY
  18#define FAST_WORKING_DIRECTORY 0
  19#else
  20#define FAST_WORKING_DIRECTORY 1
  21#endif
  22
  23static int diff_detect_rename_default;
  24static int diff_rename_limit_default = 200;
  25static int diff_suppress_blank_empty;
  26int diff_use_color_default = -1;
  27static const char *diff_word_regex_cfg;
  28static const char *external_diff_cmd_cfg;
  29int diff_auto_refresh_index = 1;
  30static int diff_mnemonic_prefix;
  31
  32static char diff_colors[][COLOR_MAXLEN] = {
  33        GIT_COLOR_RESET,
  34        GIT_COLOR_NORMAL,       /* PLAIN */
  35        GIT_COLOR_BOLD,         /* METAINFO */
  36        GIT_COLOR_CYAN,         /* FRAGINFO */
  37        GIT_COLOR_RED,          /* OLD */
  38        GIT_COLOR_GREEN,        /* NEW */
  39        GIT_COLOR_YELLOW,       /* COMMIT */
  40        GIT_COLOR_BG_RED,       /* WHITESPACE */
  41};
  42
  43static void diff_filespec_load_driver(struct diff_filespec *one);
  44static char *run_textconv(const char *, struct diff_filespec *, size_t *);
  45
  46static int parse_diff_color_slot(const char *var, int ofs)
  47{
  48        if (!strcasecmp(var+ofs, "plain"))
  49                return DIFF_PLAIN;
  50        if (!strcasecmp(var+ofs, "meta"))
  51                return DIFF_METAINFO;
  52        if (!strcasecmp(var+ofs, "frag"))
  53                return DIFF_FRAGINFO;
  54        if (!strcasecmp(var+ofs, "old"))
  55                return DIFF_FILE_OLD;
  56        if (!strcasecmp(var+ofs, "new"))
  57                return DIFF_FILE_NEW;
  58        if (!strcasecmp(var+ofs, "commit"))
  59                return DIFF_COMMIT;
  60        if (!strcasecmp(var+ofs, "whitespace"))
  61                return DIFF_WHITESPACE;
  62        die("bad config variable '%s'", var);
  63}
  64
  65static int git_config_rename(const char *var, const char *value)
  66{
  67        if (!value)
  68                return DIFF_DETECT_RENAME;
  69        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
  70                return  DIFF_DETECT_COPY;
  71        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
  72}
  73
  74/*
  75 * These are to give UI layer defaults.
  76 * The core-level commands such as git-diff-files should
  77 * never be affected by the setting of diff.renames
  78 * the user happens to have in the configuration file.
  79 */
  80int git_diff_ui_config(const char *var, const char *value, void *cb)
  81{
  82        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
  83                diff_use_color_default = git_config_colorbool(var, value, -1);
  84                return 0;
  85        }
  86        if (!strcmp(var, "diff.renames")) {
  87                diff_detect_rename_default = git_config_rename(var, value);
  88                return 0;
  89        }
  90        if (!strcmp(var, "diff.autorefreshindex")) {
  91                diff_auto_refresh_index = git_config_bool(var, value);
  92                return 0;
  93        }
  94        if (!strcmp(var, "diff.mnemonicprefix")) {
  95                diff_mnemonic_prefix = git_config_bool(var, value);
  96                return 0;
  97        }
  98        if (!strcmp(var, "diff.external"))
  99                return git_config_string(&external_diff_cmd_cfg, var, value);
 100        if (!strcmp(var, "diff.wordregex"))
 101                return git_config_string(&diff_word_regex_cfg, var, value);
 102
 103        return git_diff_basic_config(var, value, cb);
 104}
 105
 106int git_diff_basic_config(const char *var, const char *value, void *cb)
 107{
 108        if (!strcmp(var, "diff.renamelimit")) {
 109                diff_rename_limit_default = git_config_int(var, value);
 110                return 0;
 111        }
 112
 113        switch (userdiff_config(var, value)) {
 114                case 0: break;
 115                case -1: return -1;
 116                default: return 0;
 117        }
 118
 119        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 120                int slot = parse_diff_color_slot(var, 11);
 121                if (!value)
 122                        return config_error_nonbool(var);
 123                color_parse(value, var, diff_colors[slot]);
 124                return 0;
 125        }
 126
 127        /* like GNU diff's --suppress-blank-empty option  */
 128        if (!strcmp(var, "diff.suppressblankempty") ||
 129                        /* for backwards compatibility */
 130                        !strcmp(var, "diff.suppress-blank-empty")) {
 131                diff_suppress_blank_empty = git_config_bool(var, value);
 132                return 0;
 133        }
 134
 135        return git_color_default_config(var, value, cb);
 136}
 137
 138static char *quote_two(const char *one, const char *two)
 139{
 140        int need_one = quote_c_style(one, NULL, NULL, 1);
 141        int need_two = quote_c_style(two, NULL, NULL, 1);
 142        struct strbuf res = STRBUF_INIT;
 143
 144        if (need_one + need_two) {
 145                strbuf_addch(&res, '"');
 146                quote_c_style(one, &res, NULL, 1);
 147                quote_c_style(two, &res, NULL, 1);
 148                strbuf_addch(&res, '"');
 149        } else {
 150                strbuf_addstr(&res, one);
 151                strbuf_addstr(&res, two);
 152        }
 153        return strbuf_detach(&res, NULL);
 154}
 155
 156static const char *external_diff(void)
 157{
 158        static const char *external_diff_cmd = NULL;
 159        static int done_preparing = 0;
 160
 161        if (done_preparing)
 162                return external_diff_cmd;
 163        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 164        if (!external_diff_cmd)
 165                external_diff_cmd = external_diff_cmd_cfg;
 166        done_preparing = 1;
 167        return external_diff_cmd;
 168}
 169
 170static struct diff_tempfile {
 171        const char *name; /* filename external diff should read from */
 172        char hex[41];
 173        char mode[10];
 174        char tmp_path[PATH_MAX];
 175} diff_temp[2];
 176
 177typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 178
 179struct emit_callback {
 180        int color_diff;
 181        unsigned ws_rule;
 182        int blank_at_eof_in_preimage;
 183        int blank_at_eof_in_postimage;
 184        int lno_in_preimage;
 185        int lno_in_postimage;
 186        sane_truncate_fn truncate;
 187        const char **label_path;
 188        struct diff_words_data *diff_words;
 189        int *found_changesp;
 190        FILE *file;
 191};
 192
 193static int count_lines(const char *data, int size)
 194{
 195        int count, ch, completely_empty = 1, nl_just_seen = 0;
 196        count = 0;
 197        while (0 < size--) {
 198                ch = *data++;
 199                if (ch == '\n') {
 200                        count++;
 201                        nl_just_seen = 1;
 202                        completely_empty = 0;
 203                }
 204                else {
 205                        nl_just_seen = 0;
 206                        completely_empty = 0;
 207                }
 208        }
 209        if (completely_empty)
 210                return 0;
 211        if (!nl_just_seen)
 212                count++; /* no trailing newline */
 213        return count;
 214}
 215
 216static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 217{
 218        if (!DIFF_FILE_VALID(one)) {
 219                mf->ptr = (char *)""; /* does not matter */
 220                mf->size = 0;
 221                return 0;
 222        }
 223        else if (diff_populate_filespec(one, 0))
 224                return -1;
 225
 226        mf->ptr = one->data;
 227        mf->size = one->size;
 228        return 0;
 229}
 230
 231static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 232{
 233        char *ptr = mf->ptr;
 234        long size = mf->size;
 235        int cnt = 0;
 236
 237        if (!size)
 238                return cnt;
 239        ptr += size - 1; /* pointing at the very end */
 240        if (*ptr != '\n')
 241                ; /* incomplete line */
 242        else
 243                ptr--; /* skip the last LF */
 244        while (mf->ptr < ptr) {
 245                char *prev_eol;
 246                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 247                        if (*prev_eol == '\n')
 248                                break;
 249                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 250                        break;
 251                cnt++;
 252                ptr = prev_eol - 1;
 253        }
 254        return cnt;
 255}
 256
 257static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 258                               struct emit_callback *ecbdata)
 259{
 260        int l1, l2, at;
 261        unsigned ws_rule = ecbdata->ws_rule;
 262        l1 = count_trailing_blank(mf1, ws_rule);
 263        l2 = count_trailing_blank(mf2, ws_rule);
 264        if (l2 <= l1) {
 265                ecbdata->blank_at_eof_in_preimage = 0;
 266                ecbdata->blank_at_eof_in_postimage = 0;
 267                return;
 268        }
 269        at = count_lines(mf1->ptr, mf1->size);
 270        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 271
 272        at = count_lines(mf2->ptr, mf2->size);
 273        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 274}
 275
 276static void emit_line_0(FILE *file, const char *set, const char *reset,
 277                        int first, const char *line, int len)
 278{
 279        int has_trailing_newline, has_trailing_carriage_return;
 280        int nofirst;
 281
 282        if (len == 0) {
 283                has_trailing_newline = (first == '\n');
 284                has_trailing_carriage_return = (!has_trailing_newline &&
 285                                                (first == '\r'));
 286                nofirst = has_trailing_newline || has_trailing_carriage_return;
 287        } else {
 288                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 289                if (has_trailing_newline)
 290                        len--;
 291                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 292                if (has_trailing_carriage_return)
 293                        len--;
 294                nofirst = 0;
 295        }
 296
 297        fputs(set, file);
 298
 299        if (!nofirst)
 300                fputc(first, file);
 301        fwrite(line, len, 1, file);
 302        fputs(reset, file);
 303        if (has_trailing_carriage_return)
 304                fputc('\r', file);
 305        if (has_trailing_newline)
 306                fputc('\n', file);
 307}
 308
 309static void emit_line(FILE *file, const char *set, const char *reset,
 310                      const char *line, int len)
 311{
 312        emit_line_0(file, set, reset, line[0], line+1, len-1);
 313}
 314
 315static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 316{
 317        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 318              ecbdata->blank_at_eof_in_preimage &&
 319              ecbdata->blank_at_eof_in_postimage &&
 320              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 321              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 322                return 0;
 323        return ws_blank_line(line, len, ecbdata->ws_rule);
 324}
 325
 326static void emit_add_line(const char *reset,
 327                          struct emit_callback *ecbdata,
 328                          const char *line, int len)
 329{
 330        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 331        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 332
 333        if (!*ws)
 334                emit_line_0(ecbdata->file, set, reset, '+', line, len);
 335        else if (new_blank_line_at_eof(ecbdata, line, len))
 336                /* Blank line at EOF - paint '+' as well */
 337                emit_line_0(ecbdata->file, ws, reset, '+', line, len);
 338        else {
 339                /* Emit just the prefix, then the rest. */
 340                emit_line_0(ecbdata->file, set, reset, '+', "", 0);
 341                ws_check_emit(line, len, ecbdata->ws_rule,
 342                              ecbdata->file, set, reset, ws);
 343        }
 344}
 345
 346static struct diff_tempfile *claim_diff_tempfile(void) {
 347        int i;
 348        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 349                if (!diff_temp[i].name)
 350                        return diff_temp + i;
 351        die("BUG: diff is failing to clean up its tempfiles");
 352}
 353
 354static int remove_tempfile_installed;
 355
 356static void remove_tempfile(void)
 357{
 358        int i;
 359        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 360                if (diff_temp[i].name == diff_temp[i].tmp_path)
 361                        unlink_or_warn(diff_temp[i].name);
 362                diff_temp[i].name = NULL;
 363        }
 364}
 365
 366static void remove_tempfile_on_signal(int signo)
 367{
 368        remove_tempfile();
 369        sigchain_pop(signo);
 370        raise(signo);
 371}
 372
 373static void print_line_count(FILE *file, int count)
 374{
 375        switch (count) {
 376        case 0:
 377                fprintf(file, "0,0");
 378                break;
 379        case 1:
 380                fprintf(file, "1");
 381                break;
 382        default:
 383                fprintf(file, "1,%d", count);
 384                break;
 385        }
 386}
 387
 388static void copy_file_with_prefix(FILE *file,
 389                                  int prefix, const char *data, int size,
 390                                  const char *set, const char *reset)
 391{
 392        int ch, nl_just_seen = 1;
 393        while (0 < size--) {
 394                ch = *data++;
 395                if (nl_just_seen) {
 396                        fputs(set, file);
 397                        putc(prefix, file);
 398                }
 399                if (ch == '\n') {
 400                        nl_just_seen = 1;
 401                        fputs(reset, file);
 402                } else
 403                        nl_just_seen = 0;
 404                putc(ch, file);
 405        }
 406        if (!nl_just_seen)
 407                fprintf(file, "%s\n\\ No newline at end of file\n", reset);
 408}
 409
 410static void emit_rewrite_diff(const char *name_a,
 411                              const char *name_b,
 412                              struct diff_filespec *one,
 413                              struct diff_filespec *two,
 414                              const char *textconv_one,
 415                              const char *textconv_two,
 416                              struct diff_options *o)
 417{
 418        int lc_a, lc_b;
 419        int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
 420        const char *name_a_tab, *name_b_tab;
 421        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 422        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 423        const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
 424        const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
 425        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 426        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 427        const char *a_prefix, *b_prefix;
 428        const char *data_one, *data_two;
 429        size_t size_one, size_two;
 430
 431        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 432                a_prefix = o->b_prefix;
 433                b_prefix = o->a_prefix;
 434        } else {
 435                a_prefix = o->a_prefix;
 436                b_prefix = o->b_prefix;
 437        }
 438
 439        name_a += (*name_a == '/');
 440        name_b += (*name_b == '/');
 441        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 442        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 443
 444        strbuf_reset(&a_name);
 445        strbuf_reset(&b_name);
 446        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 447        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 448
 449        diff_populate_filespec(one, 0);
 450        diff_populate_filespec(two, 0);
 451        if (textconv_one) {
 452                data_one = run_textconv(textconv_one, one, &size_one);
 453                if (!data_one)
 454                        die("unable to read files to diff");
 455        }
 456        else {
 457                data_one = one->data;
 458                size_one = one->size;
 459        }
 460        if (textconv_two) {
 461                data_two = run_textconv(textconv_two, two, &size_two);
 462                if (!data_two)
 463                        die("unable to read files to diff");
 464        }
 465        else {
 466                data_two = two->data;
 467                size_two = two->size;
 468        }
 469
 470        lc_a = count_lines(data_one, size_one);
 471        lc_b = count_lines(data_two, size_two);
 472        fprintf(o->file,
 473                "%s--- %s%s%s\n%s+++ %s%s%s\n%s@@ -",
 474                metainfo, a_name.buf, name_a_tab, reset,
 475                metainfo, b_name.buf, name_b_tab, reset, fraginfo);
 476        print_line_count(o->file, lc_a);
 477        fprintf(o->file, " +");
 478        print_line_count(o->file, lc_b);
 479        fprintf(o->file, " @@%s\n", reset);
 480        if (lc_a)
 481                copy_file_with_prefix(o->file, '-', data_one, size_one, old, reset);
 482        if (lc_b)
 483                copy_file_with_prefix(o->file, '+', data_two, size_two, new, reset);
 484}
 485
 486struct diff_words_buffer {
 487        mmfile_t text;
 488        long alloc;
 489        struct diff_words_orig {
 490                const char *begin, *end;
 491        } *orig;
 492        int orig_nr, orig_alloc;
 493};
 494
 495static void diff_words_append(char *line, unsigned long len,
 496                struct diff_words_buffer *buffer)
 497{
 498        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 499        line++;
 500        len--;
 501        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 502        buffer->text.size += len;
 503        buffer->text.ptr[buffer->text.size] = '\0';
 504}
 505
 506struct diff_words_data {
 507        struct diff_words_buffer minus, plus;
 508        const char *current_plus;
 509        FILE *file;
 510        regex_t *word_regex;
 511};
 512
 513static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 514{
 515        struct diff_words_data *diff_words = priv;
 516        int minus_first, minus_len, plus_first, plus_len;
 517        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 518
 519        if (line[0] != '@' || parse_hunk_header(line, len,
 520                        &minus_first, &minus_len, &plus_first, &plus_len))
 521                return;
 522
 523        /* POSIX requires that first be decremented by one if len == 0... */
 524        if (minus_len) {
 525                minus_begin = diff_words->minus.orig[minus_first].begin;
 526                minus_end =
 527                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 528        } else
 529                minus_begin = minus_end =
 530                        diff_words->minus.orig[minus_first].end;
 531
 532        if (plus_len) {
 533                plus_begin = diff_words->plus.orig[plus_first].begin;
 534                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 535        } else
 536                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 537
 538        if (diff_words->current_plus != plus_begin)
 539                fwrite(diff_words->current_plus,
 540                                plus_begin - diff_words->current_plus, 1,
 541                                diff_words->file);
 542        if (minus_begin != minus_end)
 543                color_fwrite_lines(diff_words->file,
 544                                diff_get_color(1, DIFF_FILE_OLD),
 545                                minus_end - minus_begin, minus_begin);
 546        if (plus_begin != plus_end)
 547                color_fwrite_lines(diff_words->file,
 548                                diff_get_color(1, DIFF_FILE_NEW),
 549                                plus_end - plus_begin, plus_begin);
 550
 551        diff_words->current_plus = plus_end;
 552}
 553
 554/* This function starts looking at *begin, and returns 0 iff a word was found. */
 555static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 556                int *begin, int *end)
 557{
 558        if (word_regex && *begin < buffer->size) {
 559                regmatch_t match[1];
 560                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 561                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 562                                        '\n', match[0].rm_eo - match[0].rm_so);
 563                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 564                        *begin += match[0].rm_so;
 565                        return *begin >= *end;
 566                }
 567                return -1;
 568        }
 569
 570        /* find the next word */
 571        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 572                (*begin)++;
 573        if (*begin >= buffer->size)
 574                return -1;
 575
 576        /* find the end of the word */
 577        *end = *begin + 1;
 578        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 579                (*end)++;
 580
 581        return 0;
 582}
 583
 584/*
 585 * This function splits the words in buffer->text, stores the list with
 586 * newline separator into out, and saves the offsets of the original words
 587 * in buffer->orig.
 588 */
 589static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 590                regex_t *word_regex)
 591{
 592        int i, j;
 593        long alloc = 0;
 594
 595        out->size = 0;
 596        out->ptr = NULL;
 597
 598        /* fake an empty "0th" word */
 599        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 600        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 601        buffer->orig_nr = 1;
 602
 603        for (i = 0; i < buffer->text.size; i++) {
 604                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 605                        return;
 606
 607                /* store original boundaries */
 608                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 609                                buffer->orig_alloc);
 610                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 611                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 612                buffer->orig_nr++;
 613
 614                /* store one word */
 615                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 616                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 617                out->ptr[out->size + j - i] = '\n';
 618                out->size += j - i + 1;
 619
 620                i = j - 1;
 621        }
 622}
 623
 624/* this executes the word diff on the accumulated buffers */
 625static void diff_words_show(struct diff_words_data *diff_words)
 626{
 627        xpparam_t xpp;
 628        xdemitconf_t xecfg;
 629        xdemitcb_t ecb;
 630        mmfile_t minus, plus;
 631
 632        /* special case: only removal */
 633        if (!diff_words->plus.text.size) {
 634                color_fwrite_lines(diff_words->file,
 635                        diff_get_color(1, DIFF_FILE_OLD),
 636                        diff_words->minus.text.size, diff_words->minus.text.ptr);
 637                diff_words->minus.text.size = 0;
 638                return;
 639        }
 640
 641        diff_words->current_plus = diff_words->plus.text.ptr;
 642
 643        memset(&xpp, 0, sizeof(xpp));
 644        memset(&xecfg, 0, sizeof(xecfg));
 645        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 646        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 647        xpp.flags = XDF_NEED_MINIMAL;
 648        /* as only the hunk header will be parsed, we need a 0-context */
 649        xecfg.ctxlen = 0;
 650        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 651                      &xpp, &xecfg, &ecb);
 652        free(minus.ptr);
 653        free(plus.ptr);
 654        if (diff_words->current_plus != diff_words->plus.text.ptr +
 655                        diff_words->plus.text.size)
 656                fwrite(diff_words->current_plus,
 657                        diff_words->plus.text.ptr + diff_words->plus.text.size
 658                        - diff_words->current_plus, 1,
 659                        diff_words->file);
 660        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 661}
 662
 663static void free_diff_words_data(struct emit_callback *ecbdata)
 664{
 665        if (ecbdata->diff_words) {
 666                /* flush buffers */
 667                if (ecbdata->diff_words->minus.text.size ||
 668                                ecbdata->diff_words->plus.text.size)
 669                        diff_words_show(ecbdata->diff_words);
 670
 671                free (ecbdata->diff_words->minus.text.ptr);
 672                free (ecbdata->diff_words->minus.orig);
 673                free (ecbdata->diff_words->plus.text.ptr);
 674                free (ecbdata->diff_words->plus.orig);
 675                free(ecbdata->diff_words->word_regex);
 676                free(ecbdata->diff_words);
 677                ecbdata->diff_words = NULL;
 678        }
 679}
 680
 681const char *diff_get_color(int diff_use_color, enum color_diff ix)
 682{
 683        if (diff_use_color)
 684                return diff_colors[ix];
 685        return "";
 686}
 687
 688static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
 689{
 690        const char *cp;
 691        unsigned long allot;
 692        size_t l = len;
 693
 694        if (ecb->truncate)
 695                return ecb->truncate(line, len);
 696        cp = line;
 697        allot = l;
 698        while (0 < l) {
 699                (void) utf8_width(&cp, &l);
 700                if (!cp)
 701                        break; /* truncated in the middle? */
 702        }
 703        return allot - l;
 704}
 705
 706static void find_lno(const char *line, struct emit_callback *ecbdata)
 707{
 708        const char *p;
 709        ecbdata->lno_in_preimage = 0;
 710        ecbdata->lno_in_postimage = 0;
 711        p = strchr(line, '-');
 712        if (!p)
 713                return; /* cannot happen */
 714        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
 715        p = strchr(p, '+');
 716        if (!p)
 717                return; /* cannot happen */
 718        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
 719}
 720
 721static void fn_out_consume(void *priv, char *line, unsigned long len)
 722{
 723        struct emit_callback *ecbdata = priv;
 724        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 725        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 726        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 727
 728        *(ecbdata->found_changesp) = 1;
 729
 730        if (ecbdata->label_path[0]) {
 731                const char *name_a_tab, *name_b_tab;
 732
 733                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
 734                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
 735
 736                fprintf(ecbdata->file, "%s--- %s%s%s\n",
 737                        meta, ecbdata->label_path[0], reset, name_a_tab);
 738                fprintf(ecbdata->file, "%s+++ %s%s%s\n",
 739                        meta, ecbdata->label_path[1], reset, name_b_tab);
 740                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
 741        }
 742
 743        if (diff_suppress_blank_empty
 744            && len == 2 && line[0] == ' ' && line[1] == '\n') {
 745                line[0] = '\n';
 746                len = 1;
 747        }
 748
 749        if (line[0] == '@') {
 750                len = sane_truncate_line(ecbdata, line, len);
 751                find_lno(line, ecbdata);
 752                emit_line(ecbdata->file,
 753                          diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
 754                          reset, line, len);
 755                if (line[len-1] != '\n')
 756                        putc('\n', ecbdata->file);
 757                return;
 758        }
 759
 760        if (len < 1) {
 761                emit_line(ecbdata->file, reset, reset, line, len);
 762                return;
 763        }
 764
 765        if (ecbdata->diff_words) {
 766                if (line[0] == '-') {
 767                        diff_words_append(line, len,
 768                                          &ecbdata->diff_words->minus);
 769                        return;
 770                } else if (line[0] == '+') {
 771                        diff_words_append(line, len,
 772                                          &ecbdata->diff_words->plus);
 773                        return;
 774                }
 775                if (ecbdata->diff_words->minus.text.size ||
 776                    ecbdata->diff_words->plus.text.size)
 777                        diff_words_show(ecbdata->diff_words);
 778                line++;
 779                len--;
 780                emit_line(ecbdata->file, plain, reset, line, len);
 781                return;
 782        }
 783
 784        if (line[0] != '+') {
 785                const char *color =
 786                        diff_get_color(ecbdata->color_diff,
 787                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
 788                ecbdata->lno_in_preimage++;
 789                if (line[0] == ' ')
 790                        ecbdata->lno_in_postimage++;
 791                emit_line(ecbdata->file, color, reset, line, len);
 792        } else {
 793                ecbdata->lno_in_postimage++;
 794                emit_add_line(reset, ecbdata, line + 1, len - 1);
 795        }
 796}
 797
 798static char *pprint_rename(const char *a, const char *b)
 799{
 800        const char *old = a;
 801        const char *new = b;
 802        struct strbuf name = STRBUF_INIT;
 803        int pfx_length, sfx_length;
 804        int len_a = strlen(a);
 805        int len_b = strlen(b);
 806        int a_midlen, b_midlen;
 807        int qlen_a = quote_c_style(a, NULL, NULL, 0);
 808        int qlen_b = quote_c_style(b, NULL, NULL, 0);
 809
 810        if (qlen_a || qlen_b) {
 811                quote_c_style(a, &name, NULL, 0);
 812                strbuf_addstr(&name, " => ");
 813                quote_c_style(b, &name, NULL, 0);
 814                return strbuf_detach(&name, NULL);
 815        }
 816
 817        /* Find common prefix */
 818        pfx_length = 0;
 819        while (*old && *new && *old == *new) {
 820                if (*old == '/')
 821                        pfx_length = old - a + 1;
 822                old++;
 823                new++;
 824        }
 825
 826        /* Find common suffix */
 827        old = a + len_a;
 828        new = b + len_b;
 829        sfx_length = 0;
 830        while (a <= old && b <= new && *old == *new) {
 831                if (*old == '/')
 832                        sfx_length = len_a - (old - a);
 833                old--;
 834                new--;
 835        }
 836
 837        /*
 838         * pfx{mid-a => mid-b}sfx
 839         * {pfx-a => pfx-b}sfx
 840         * pfx{sfx-a => sfx-b}
 841         * name-a => name-b
 842         */
 843        a_midlen = len_a - pfx_length - sfx_length;
 844        b_midlen = len_b - pfx_length - sfx_length;
 845        if (a_midlen < 0)
 846                a_midlen = 0;
 847        if (b_midlen < 0)
 848                b_midlen = 0;
 849
 850        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
 851        if (pfx_length + sfx_length) {
 852                strbuf_add(&name, a, pfx_length);
 853                strbuf_addch(&name, '{');
 854        }
 855        strbuf_add(&name, a + pfx_length, a_midlen);
 856        strbuf_addstr(&name, " => ");
 857        strbuf_add(&name, b + pfx_length, b_midlen);
 858        if (pfx_length + sfx_length) {
 859                strbuf_addch(&name, '}');
 860                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
 861        }
 862        return strbuf_detach(&name, NULL);
 863}
 864
 865struct diffstat_t {
 866        int nr;
 867        int alloc;
 868        struct diffstat_file {
 869                char *from_name;
 870                char *name;
 871                char *print_name;
 872                unsigned is_unmerged:1;
 873                unsigned is_binary:1;
 874                unsigned is_renamed:1;
 875                unsigned int added, deleted;
 876        } **files;
 877};
 878
 879static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
 880                                          const char *name_a,
 881                                          const char *name_b)
 882{
 883        struct diffstat_file *x;
 884        x = xcalloc(sizeof (*x), 1);
 885        if (diffstat->nr == diffstat->alloc) {
 886                diffstat->alloc = alloc_nr(diffstat->alloc);
 887                diffstat->files = xrealloc(diffstat->files,
 888                                diffstat->alloc * sizeof(x));
 889        }
 890        diffstat->files[diffstat->nr++] = x;
 891        if (name_b) {
 892                x->from_name = xstrdup(name_a);
 893                x->name = xstrdup(name_b);
 894                x->is_renamed = 1;
 895        }
 896        else {
 897                x->from_name = NULL;
 898                x->name = xstrdup(name_a);
 899        }
 900        return x;
 901}
 902
 903static void diffstat_consume(void *priv, char *line, unsigned long len)
 904{
 905        struct diffstat_t *diffstat = priv;
 906        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
 907
 908        if (line[0] == '+')
 909                x->added++;
 910        else if (line[0] == '-')
 911                x->deleted++;
 912}
 913
 914const char mime_boundary_leader[] = "------------";
 915
 916static int scale_linear(int it, int width, int max_change)
 917{
 918        /*
 919         * make sure that at least one '-' is printed if there were deletions,
 920         * and likewise for '+'.
 921         */
 922        if (max_change < 2)
 923                return it;
 924        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
 925}
 926
 927static void show_name(FILE *file,
 928                      const char *prefix, const char *name, int len)
 929{
 930        fprintf(file, " %s%-*s |", prefix, len, name);
 931}
 932
 933static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
 934{
 935        if (cnt <= 0)
 936                return;
 937        fprintf(file, "%s", set);
 938        while (cnt--)
 939                putc(ch, file);
 940        fprintf(file, "%s", reset);
 941}
 942
 943static void fill_print_name(struct diffstat_file *file)
 944{
 945        char *pname;
 946
 947        if (file->print_name)
 948                return;
 949
 950        if (!file->is_renamed) {
 951                struct strbuf buf = STRBUF_INIT;
 952                if (quote_c_style(file->name, &buf, NULL, 0)) {
 953                        pname = strbuf_detach(&buf, NULL);
 954                } else {
 955                        pname = file->name;
 956                        strbuf_release(&buf);
 957                }
 958        } else {
 959                pname = pprint_rename(file->from_name, file->name);
 960        }
 961        file->print_name = pname;
 962}
 963
 964static void show_stats(struct diffstat_t *data, struct diff_options *options)
 965{
 966        int i, len, add, del, adds = 0, dels = 0;
 967        int max_change = 0, max_len = 0;
 968        int total_files = data->nr;
 969        int width, name_width;
 970        const char *reset, *set, *add_c, *del_c;
 971
 972        if (data->nr == 0)
 973                return;
 974
 975        width = options->stat_width ? options->stat_width : 80;
 976        name_width = options->stat_name_width ? options->stat_name_width : 50;
 977
 978        /* Sanity: give at least 5 columns to the graph,
 979         * but leave at least 10 columns for the name.
 980         */
 981        if (width < 25)
 982                width = 25;
 983        if (name_width < 10)
 984                name_width = 10;
 985        else if (width < name_width + 15)
 986                name_width = width - 15;
 987
 988        /* Find the longest filename and max number of changes */
 989        reset = diff_get_color_opt(options, DIFF_RESET);
 990        set   = diff_get_color_opt(options, DIFF_PLAIN);
 991        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
 992        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
 993
 994        for (i = 0; i < data->nr; i++) {
 995                struct diffstat_file *file = data->files[i];
 996                int change = file->added + file->deleted;
 997                fill_print_name(file);
 998                len = strlen(file->print_name);
 999                if (max_len < len)
1000                        max_len = len;
1001
1002                if (file->is_binary || file->is_unmerged)
1003                        continue;
1004                if (max_change < change)
1005                        max_change = change;
1006        }
1007
1008        /* Compute the width of the graph part;
1009         * 10 is for one blank at the beginning of the line plus
1010         * " | count " between the name and the graph.
1011         *
1012         * From here on, name_width is the width of the name area,
1013         * and width is the width of the graph area.
1014         */
1015        name_width = (name_width < max_len) ? name_width : max_len;
1016        if (width < (name_width + 10) + max_change)
1017                width = width - (name_width + 10);
1018        else
1019                width = max_change;
1020
1021        for (i = 0; i < data->nr; i++) {
1022                const char *prefix = "";
1023                char *name = data->files[i]->print_name;
1024                int added = data->files[i]->added;
1025                int deleted = data->files[i]->deleted;
1026                int name_len;
1027
1028                /*
1029                 * "scale" the filename
1030                 */
1031                len = name_width;
1032                name_len = strlen(name);
1033                if (name_width < name_len) {
1034                        char *slash;
1035                        prefix = "...";
1036                        len -= 3;
1037                        name += name_len - len;
1038                        slash = strchr(name, '/');
1039                        if (slash)
1040                                name = slash;
1041                }
1042
1043                if (data->files[i]->is_binary) {
1044                        show_name(options->file, prefix, name, len);
1045                        fprintf(options->file, "  Bin ");
1046                        fprintf(options->file, "%s%d%s", del_c, deleted, reset);
1047                        fprintf(options->file, " -> ");
1048                        fprintf(options->file, "%s%d%s", add_c, added, reset);
1049                        fprintf(options->file, " bytes");
1050                        fprintf(options->file, "\n");
1051                        continue;
1052                }
1053                else if (data->files[i]->is_unmerged) {
1054                        show_name(options->file, prefix, name, len);
1055                        fprintf(options->file, "  Unmerged\n");
1056                        continue;
1057                }
1058                else if (!data->files[i]->is_renamed &&
1059                         (added + deleted == 0)) {
1060                        total_files--;
1061                        continue;
1062                }
1063
1064                /*
1065                 * scale the add/delete
1066                 */
1067                add = added;
1068                del = deleted;
1069                adds += add;
1070                dels += del;
1071
1072                if (width <= max_change) {
1073                        add = scale_linear(add, width, max_change);
1074                        del = scale_linear(del, width, max_change);
1075                }
1076                show_name(options->file, prefix, name, len);
1077                fprintf(options->file, "%5d%s", added + deleted,
1078                                added + deleted ? " " : "");
1079                show_graph(options->file, '+', add, add_c, reset);
1080                show_graph(options->file, '-', del, del_c, reset);
1081                fprintf(options->file, "\n");
1082        }
1083        fprintf(options->file,
1084               " %d files changed, %d insertions(+), %d deletions(-)\n",
1085               total_files, adds, dels);
1086}
1087
1088static void show_shortstats(struct diffstat_t* data, struct diff_options *options)
1089{
1090        int i, adds = 0, dels = 0, total_files = data->nr;
1091
1092        if (data->nr == 0)
1093                return;
1094
1095        for (i = 0; i < data->nr; i++) {
1096                if (!data->files[i]->is_binary &&
1097                    !data->files[i]->is_unmerged) {
1098                        int added = data->files[i]->added;
1099                        int deleted= data->files[i]->deleted;
1100                        if (!data->files[i]->is_renamed &&
1101                            (added + deleted == 0)) {
1102                                total_files--;
1103                        } else {
1104                                adds += added;
1105                                dels += deleted;
1106                        }
1107                }
1108        }
1109        fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
1110               total_files, adds, dels);
1111}
1112
1113static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1114{
1115        int i;
1116
1117        if (data->nr == 0)
1118                return;
1119
1120        for (i = 0; i < data->nr; i++) {
1121                struct diffstat_file *file = data->files[i];
1122
1123                if (file->is_binary)
1124                        fprintf(options->file, "-\t-\t");
1125                else
1126                        fprintf(options->file,
1127                                "%d\t%d\t", file->added, file->deleted);
1128                if (options->line_termination) {
1129                        fill_print_name(file);
1130                        if (!file->is_renamed)
1131                                write_name_quoted(file->name, options->file,
1132                                                  options->line_termination);
1133                        else {
1134                                fputs(file->print_name, options->file);
1135                                putc(options->line_termination, options->file);
1136                        }
1137                } else {
1138                        if (file->is_renamed) {
1139                                putc('\0', options->file);
1140                                write_name_quoted(file->from_name, options->file, '\0');
1141                        }
1142                        write_name_quoted(file->name, options->file, '\0');
1143                }
1144        }
1145}
1146
1147struct dirstat_file {
1148        const char *name;
1149        unsigned long changed;
1150};
1151
1152struct dirstat_dir {
1153        struct dirstat_file *files;
1154        int alloc, nr, percent, cumulative;
1155};
1156
1157static long gather_dirstat(FILE *file, struct dirstat_dir *dir, unsigned long changed, const char *base, int baselen)
1158{
1159        unsigned long this_dir = 0;
1160        unsigned int sources = 0;
1161
1162        while (dir->nr) {
1163                struct dirstat_file *f = dir->files;
1164                int namelen = strlen(f->name);
1165                unsigned long this;
1166                char *slash;
1167
1168                if (namelen < baselen)
1169                        break;
1170                if (memcmp(f->name, base, baselen))
1171                        break;
1172                slash = strchr(f->name + baselen, '/');
1173                if (slash) {
1174                        int newbaselen = slash + 1 - f->name;
1175                        this = gather_dirstat(file, dir, changed, f->name, newbaselen);
1176                        sources++;
1177                } else {
1178                        this = f->changed;
1179                        dir->files++;
1180                        dir->nr--;
1181                        sources += 2;
1182                }
1183                this_dir += this;
1184        }
1185
1186        /*
1187         * We don't report dirstat's for
1188         *  - the top level
1189         *  - or cases where everything came from a single directory
1190         *    under this directory (sources == 1).
1191         */
1192        if (baselen && sources != 1) {
1193                int permille = this_dir * 1000 / changed;
1194                if (permille) {
1195                        int percent = permille / 10;
1196                        if (percent >= dir->percent) {
1197                                fprintf(file, "%4d.%01d%% %.*s\n", percent, permille % 10, baselen, base);
1198                                if (!dir->cumulative)
1199                                        return 0;
1200                        }
1201                }
1202        }
1203        return this_dir;
1204}
1205
1206static int dirstat_compare(const void *_a, const void *_b)
1207{
1208        const struct dirstat_file *a = _a;
1209        const struct dirstat_file *b = _b;
1210        return strcmp(a->name, b->name);
1211}
1212
1213static void show_dirstat(struct diff_options *options)
1214{
1215        int i;
1216        unsigned long changed;
1217        struct dirstat_dir dir;
1218        struct diff_queue_struct *q = &diff_queued_diff;
1219
1220        dir.files = NULL;
1221        dir.alloc = 0;
1222        dir.nr = 0;
1223        dir.percent = options->dirstat_percent;
1224        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1225
1226        changed = 0;
1227        for (i = 0; i < q->nr; i++) {
1228                struct diff_filepair *p = q->queue[i];
1229                const char *name;
1230                unsigned long copied, added, damage;
1231
1232                name = p->one->path ? p->one->path : p->two->path;
1233
1234                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1235                        diff_populate_filespec(p->one, 0);
1236                        diff_populate_filespec(p->two, 0);
1237                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1238                                               &copied, &added);
1239                        diff_free_filespec_data(p->one);
1240                        diff_free_filespec_data(p->two);
1241                } else if (DIFF_FILE_VALID(p->one)) {
1242                        diff_populate_filespec(p->one, 1);
1243                        copied = added = 0;
1244                        diff_free_filespec_data(p->one);
1245                } else if (DIFF_FILE_VALID(p->two)) {
1246                        diff_populate_filespec(p->two, 1);
1247                        copied = 0;
1248                        added = p->two->size;
1249                        diff_free_filespec_data(p->two);
1250                } else
1251                        continue;
1252
1253                /*
1254                 * Original minus copied is the removed material,
1255                 * added is the new material.  They are both damages
1256                 * made to the preimage. In --dirstat-by-file mode, count
1257                 * damaged files, not damaged lines. This is done by
1258                 * counting only a single damaged line per file.
1259                 */
1260                damage = (p->one->size - copied) + added;
1261                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE) && damage > 0)
1262                        damage = 1;
1263
1264                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1265                dir.files[dir.nr].name = name;
1266                dir.files[dir.nr].changed = damage;
1267                changed += damage;
1268                dir.nr++;
1269        }
1270
1271        /* This can happen even with many files, if everything was renames */
1272        if (!changed)
1273                return;
1274
1275        /* Show all directories with more than x% of the changes */
1276        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1277        gather_dirstat(options->file, &dir, changed, "", 0);
1278}
1279
1280static void free_diffstat_info(struct diffstat_t *diffstat)
1281{
1282        int i;
1283        for (i = 0; i < diffstat->nr; i++) {
1284                struct diffstat_file *f = diffstat->files[i];
1285                if (f->name != f->print_name)
1286                        free(f->print_name);
1287                free(f->name);
1288                free(f->from_name);
1289                free(f);
1290        }
1291        free(diffstat->files);
1292}
1293
1294struct checkdiff_t {
1295        const char *filename;
1296        int lineno;
1297        struct diff_options *o;
1298        unsigned ws_rule;
1299        unsigned status;
1300};
1301
1302static int is_conflict_marker(const char *line, unsigned long len)
1303{
1304        char firstchar;
1305        int cnt;
1306
1307        if (len < 8)
1308                return 0;
1309        firstchar = line[0];
1310        switch (firstchar) {
1311        case '=': case '>': case '<':
1312                break;
1313        default:
1314                return 0;
1315        }
1316        for (cnt = 1; cnt < 7; cnt++)
1317                if (line[cnt] != firstchar)
1318                        return 0;
1319        /* line[0] thru line[6] are same as firstchar */
1320        if (firstchar == '=') {
1321                /* divider between ours and theirs? */
1322                if (len != 8 || line[7] != '\n')
1323                        return 0;
1324        } else if (len < 8 || !isspace(line[7])) {
1325                /* not divider before ours nor after theirs */
1326                return 0;
1327        }
1328        return 1;
1329}
1330
1331static void checkdiff_consume(void *priv, char *line, unsigned long len)
1332{
1333        struct checkdiff_t *data = priv;
1334        int color_diff = DIFF_OPT_TST(data->o, COLOR_DIFF);
1335        const char *ws = diff_get_color(color_diff, DIFF_WHITESPACE);
1336        const char *reset = diff_get_color(color_diff, DIFF_RESET);
1337        const char *set = diff_get_color(color_diff, DIFF_FILE_NEW);
1338        char *err;
1339
1340        if (line[0] == '+') {
1341                unsigned bad;
1342                data->lineno++;
1343                if (is_conflict_marker(line + 1, len - 1)) {
1344                        data->status |= 1;
1345                        fprintf(data->o->file,
1346                                "%s:%d: leftover conflict marker\n",
1347                                data->filename, data->lineno);
1348                }
1349                bad = ws_check(line + 1, len - 1, data->ws_rule);
1350                if (!bad)
1351                        return;
1352                data->status |= bad;
1353                err = whitespace_error_string(bad);
1354                fprintf(data->o->file, "%s:%d: %s.\n",
1355                        data->filename, data->lineno, err);
1356                free(err);
1357                emit_line(data->o->file, set, reset, line, 1);
1358                ws_check_emit(line + 1, len - 1, data->ws_rule,
1359                              data->o->file, set, reset, ws);
1360        } else if (line[0] == ' ') {
1361                data->lineno++;
1362        } else if (line[0] == '@') {
1363                char *plus = strchr(line, '+');
1364                if (plus)
1365                        data->lineno = strtol(plus, NULL, 10) - 1;
1366                else
1367                        die("invalid diff");
1368        }
1369}
1370
1371static unsigned char *deflate_it(char *data,
1372                                 unsigned long size,
1373                                 unsigned long *result_size)
1374{
1375        int bound;
1376        unsigned char *deflated;
1377        z_stream stream;
1378
1379        memset(&stream, 0, sizeof(stream));
1380        deflateInit(&stream, zlib_compression_level);
1381        bound = deflateBound(&stream, size);
1382        deflated = xmalloc(bound);
1383        stream.next_out = deflated;
1384        stream.avail_out = bound;
1385
1386        stream.next_in = (unsigned char *)data;
1387        stream.avail_in = size;
1388        while (deflate(&stream, Z_FINISH) == Z_OK)
1389                ; /* nothing */
1390        deflateEnd(&stream);
1391        *result_size = stream.total_out;
1392        return deflated;
1393}
1394
1395static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two)
1396{
1397        void *cp;
1398        void *delta;
1399        void *deflated;
1400        void *data;
1401        unsigned long orig_size;
1402        unsigned long delta_size;
1403        unsigned long deflate_size;
1404        unsigned long data_size;
1405
1406        /* We could do deflated delta, or we could do just deflated two,
1407         * whichever is smaller.
1408         */
1409        delta = NULL;
1410        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1411        if (one->size && two->size) {
1412                delta = diff_delta(one->ptr, one->size,
1413                                   two->ptr, two->size,
1414                                   &delta_size, deflate_size);
1415                if (delta) {
1416                        void *to_free = delta;
1417                        orig_size = delta_size;
1418                        delta = deflate_it(delta, delta_size, &delta_size);
1419                        free(to_free);
1420                }
1421        }
1422
1423        if (delta && delta_size < deflate_size) {
1424                fprintf(file, "delta %lu\n", orig_size);
1425                free(deflated);
1426                data = delta;
1427                data_size = delta_size;
1428        }
1429        else {
1430                fprintf(file, "literal %lu\n", two->size);
1431                free(delta);
1432                data = deflated;
1433                data_size = deflate_size;
1434        }
1435
1436        /* emit data encoded in base85 */
1437        cp = data;
1438        while (data_size) {
1439                int bytes = (52 < data_size) ? 52 : data_size;
1440                char line[70];
1441                data_size -= bytes;
1442                if (bytes <= 26)
1443                        line[0] = bytes + 'A' - 1;
1444                else
1445                        line[0] = bytes - 26 + 'a' - 1;
1446                encode_85(line + 1, cp, bytes);
1447                cp = (char *) cp + bytes;
1448                fputs(line, file);
1449                fputc('\n', file);
1450        }
1451        fprintf(file, "\n");
1452        free(data);
1453}
1454
1455static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two)
1456{
1457        fprintf(file, "GIT binary patch\n");
1458        emit_binary_diff_body(file, one, two);
1459        emit_binary_diff_body(file, two, one);
1460}
1461
1462static void diff_filespec_load_driver(struct diff_filespec *one)
1463{
1464        if (!one->driver)
1465                one->driver = userdiff_find_by_path(one->path);
1466        if (!one->driver)
1467                one->driver = userdiff_find_by_name("default");
1468}
1469
1470int diff_filespec_is_binary(struct diff_filespec *one)
1471{
1472        if (one->is_binary == -1) {
1473                diff_filespec_load_driver(one);
1474                if (one->driver->binary != -1)
1475                        one->is_binary = one->driver->binary;
1476                else {
1477                        if (!one->data && DIFF_FILE_VALID(one))
1478                                diff_populate_filespec(one, 0);
1479                        if (one->data)
1480                                one->is_binary = buffer_is_binary(one->data,
1481                                                one->size);
1482                        if (one->is_binary == -1)
1483                                one->is_binary = 0;
1484                }
1485        }
1486        return one->is_binary;
1487}
1488
1489static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
1490{
1491        diff_filespec_load_driver(one);
1492        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
1493}
1494
1495static const char *userdiff_word_regex(struct diff_filespec *one)
1496{
1497        diff_filespec_load_driver(one);
1498        return one->driver->word_regex;
1499}
1500
1501void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
1502{
1503        if (!options->a_prefix)
1504                options->a_prefix = a;
1505        if (!options->b_prefix)
1506                options->b_prefix = b;
1507}
1508
1509static const char *get_textconv(struct diff_filespec *one)
1510{
1511        if (!DIFF_FILE_VALID(one))
1512                return NULL;
1513        if (!S_ISREG(one->mode))
1514                return NULL;
1515        diff_filespec_load_driver(one);
1516        return one->driver->textconv;
1517}
1518
1519static void builtin_diff(const char *name_a,
1520                         const char *name_b,
1521                         struct diff_filespec *one,
1522                         struct diff_filespec *two,
1523                         const char *xfrm_msg,
1524                         struct diff_options *o,
1525                         int complete_rewrite)
1526{
1527        mmfile_t mf1, mf2;
1528        const char *lbl[2];
1529        char *a_one, *b_two;
1530        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1531        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1532        const char *a_prefix, *b_prefix;
1533        const char *textconv_one = NULL, *textconv_two = NULL;
1534
1535        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
1536                textconv_one = get_textconv(one);
1537                textconv_two = get_textconv(two);
1538        }
1539
1540        diff_set_mnemonic_prefix(o, "a/", "b/");
1541        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
1542                a_prefix = o->b_prefix;
1543                b_prefix = o->a_prefix;
1544        } else {
1545                a_prefix = o->a_prefix;
1546                b_prefix = o->b_prefix;
1547        }
1548
1549        /* Never use a non-valid filename anywhere if at all possible */
1550        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
1551        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
1552
1553        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
1554        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
1555        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1556        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1557        fprintf(o->file, "%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1558        if (lbl[0][0] == '/') {
1559                /* /dev/null */
1560                fprintf(o->file, "%snew file mode %06o%s\n", set, two->mode, reset);
1561                if (xfrm_msg && xfrm_msg[0])
1562                        fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1563        }
1564        else if (lbl[1][0] == '/') {
1565                fprintf(o->file, "%sdeleted file mode %06o%s\n", set, one->mode, reset);
1566                if (xfrm_msg && xfrm_msg[0])
1567                        fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1568        }
1569        else {
1570                if (one->mode != two->mode) {
1571                        fprintf(o->file, "%sold mode %06o%s\n", set, one->mode, reset);
1572                        fprintf(o->file, "%snew mode %06o%s\n", set, two->mode, reset);
1573                }
1574                if (xfrm_msg && xfrm_msg[0])
1575                        fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1576                /*
1577                 * we do not run diff between different kind
1578                 * of objects.
1579                 */
1580                if ((one->mode ^ two->mode) & S_IFMT)
1581                        goto free_ab_and_return;
1582                if (complete_rewrite &&
1583                    (textconv_one || !diff_filespec_is_binary(one)) &&
1584                    (textconv_two || !diff_filespec_is_binary(two))) {
1585                        emit_rewrite_diff(name_a, name_b, one, two,
1586                                                textconv_one, textconv_two, o);
1587                        o->found_changes = 1;
1588                        goto free_ab_and_return;
1589                }
1590        }
1591
1592        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1593                die("unable to read files to diff");
1594
1595        if (!DIFF_OPT_TST(o, TEXT) &&
1596            ( (diff_filespec_is_binary(one) && !textconv_one) ||
1597              (diff_filespec_is_binary(two) && !textconv_two) )) {
1598                /* Quite common confusing case */
1599                if (mf1.size == mf2.size &&
1600                    !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1601                        goto free_ab_and_return;
1602                if (DIFF_OPT_TST(o, BINARY))
1603                        emit_binary_diff(o->file, &mf1, &mf2);
1604                else
1605                        fprintf(o->file, "Binary files %s and %s differ\n",
1606                                lbl[0], lbl[1]);
1607                o->found_changes = 1;
1608        }
1609        else {
1610                /* Crazy xdl interfaces.. */
1611                const char *diffopts = getenv("GIT_DIFF_OPTS");
1612                xpparam_t xpp;
1613                xdemitconf_t xecfg;
1614                xdemitcb_t ecb;
1615                struct emit_callback ecbdata;
1616                const struct userdiff_funcname *pe;
1617
1618                if (textconv_one) {
1619                        size_t size;
1620                        mf1.ptr = run_textconv(textconv_one, one, &size);
1621                        if (!mf1.ptr)
1622                                die("unable to read files to diff");
1623                        mf1.size = size;
1624                }
1625                if (textconv_two) {
1626                        size_t size;
1627                        mf2.ptr = run_textconv(textconv_two, two, &size);
1628                        if (!mf2.ptr)
1629                                die("unable to read files to diff");
1630                        mf2.size = size;
1631                }
1632
1633                pe = diff_funcname_pattern(one);
1634                if (!pe)
1635                        pe = diff_funcname_pattern(two);
1636
1637                memset(&xpp, 0, sizeof(xpp));
1638                memset(&xecfg, 0, sizeof(xecfg));
1639                memset(&ecbdata, 0, sizeof(ecbdata));
1640                ecbdata.label_path = lbl;
1641                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1642                ecbdata.found_changesp = &o->found_changes;
1643                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1644                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
1645                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
1646                ecbdata.file = o->file;
1647                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1648                xecfg.ctxlen = o->context;
1649                xecfg.interhunkctxlen = o->interhunkcontext;
1650                xecfg.flags = XDL_EMIT_FUNCNAMES;
1651                if (pe)
1652                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
1653                if (!diffopts)
1654                        ;
1655                else if (!prefixcmp(diffopts, "--unified="))
1656                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1657                else if (!prefixcmp(diffopts, "-u"))
1658                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1659                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS)) {
1660                        ecbdata.diff_words =
1661                                xcalloc(1, sizeof(struct diff_words_data));
1662                        ecbdata.diff_words->file = o->file;
1663                        if (!o->word_regex)
1664                                o->word_regex = userdiff_word_regex(one);
1665                        if (!o->word_regex)
1666                                o->word_regex = userdiff_word_regex(two);
1667                        if (!o->word_regex)
1668                                o->word_regex = diff_word_regex_cfg;
1669                        if (o->word_regex) {
1670                                ecbdata.diff_words->word_regex = (regex_t *)
1671                                        xmalloc(sizeof(regex_t));
1672                                if (regcomp(ecbdata.diff_words->word_regex,
1673                                                o->word_regex,
1674                                                REG_EXTENDED | REG_NEWLINE))
1675                                        die ("Invalid regular expression: %s",
1676                                                        o->word_regex);
1677                        }
1678                }
1679                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
1680                              &xpp, &xecfg, &ecb);
1681                if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1682                        free_diff_words_data(&ecbdata);
1683                if (textconv_one)
1684                        free(mf1.ptr);
1685                if (textconv_two)
1686                        free(mf2.ptr);
1687                xdiff_clear_find_func(&xecfg);
1688        }
1689
1690 free_ab_and_return:
1691        diff_free_filespec_data(one);
1692        diff_free_filespec_data(two);
1693        free(a_one);
1694        free(b_two);
1695        return;
1696}
1697
1698static void builtin_diffstat(const char *name_a, const char *name_b,
1699                             struct diff_filespec *one,
1700                             struct diff_filespec *two,
1701                             struct diffstat_t *diffstat,
1702                             struct diff_options *o,
1703                             int complete_rewrite)
1704{
1705        mmfile_t mf1, mf2;
1706        struct diffstat_file *data;
1707
1708        data = diffstat_add(diffstat, name_a, name_b);
1709
1710        if (!one || !two) {
1711                data->is_unmerged = 1;
1712                return;
1713        }
1714        if (complete_rewrite) {
1715                diff_populate_filespec(one, 0);
1716                diff_populate_filespec(two, 0);
1717                data->deleted = count_lines(one->data, one->size);
1718                data->added = count_lines(two->data, two->size);
1719                goto free_and_return;
1720        }
1721        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1722                die("unable to read files to diff");
1723
1724        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1725                data->is_binary = 1;
1726                data->added = mf2.size;
1727                data->deleted = mf1.size;
1728        } else {
1729                /* Crazy xdl interfaces.. */
1730                xpparam_t xpp;
1731                xdemitconf_t xecfg;
1732                xdemitcb_t ecb;
1733
1734                memset(&xpp, 0, sizeof(xpp));
1735                memset(&xecfg, 0, sizeof(xecfg));
1736                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1737                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
1738                              &xpp, &xecfg, &ecb);
1739        }
1740
1741 free_and_return:
1742        diff_free_filespec_data(one);
1743        diff_free_filespec_data(two);
1744}
1745
1746static void builtin_checkdiff(const char *name_a, const char *name_b,
1747                              const char *attr_path,
1748                              struct diff_filespec *one,
1749                              struct diff_filespec *two,
1750                              struct diff_options *o)
1751{
1752        mmfile_t mf1, mf2;
1753        struct checkdiff_t data;
1754
1755        if (!two)
1756                return;
1757
1758        memset(&data, 0, sizeof(data));
1759        data.filename = name_b ? name_b : name_a;
1760        data.lineno = 0;
1761        data.o = o;
1762        data.ws_rule = whitespace_rule(attr_path);
1763
1764        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1765                die("unable to read files to diff");
1766
1767        /*
1768         * All the other codepaths check both sides, but not checking
1769         * the "old" side here is deliberate.  We are checking the newly
1770         * introduced changes, and as long as the "new" side is text, we
1771         * can and should check what it introduces.
1772         */
1773        if (diff_filespec_is_binary(two))
1774                goto free_and_return;
1775        else {
1776                /* Crazy xdl interfaces.. */
1777                xpparam_t xpp;
1778                xdemitconf_t xecfg;
1779                xdemitcb_t ecb;
1780
1781                memset(&xpp, 0, sizeof(xpp));
1782                memset(&xecfg, 0, sizeof(xecfg));
1783                xecfg.ctxlen = 1; /* at least one context line */
1784                xpp.flags = XDF_NEED_MINIMAL;
1785                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
1786                              &xpp, &xecfg, &ecb);
1787
1788                if (data.ws_rule & WS_BLANK_AT_EOF) {
1789                        struct emit_callback ecbdata;
1790                        int blank_at_eof;
1791
1792                        ecbdata.ws_rule = data.ws_rule;
1793                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
1794                        blank_at_eof = ecbdata.blank_at_eof_in_preimage;
1795
1796                        if (blank_at_eof) {
1797                                static char *err;
1798                                if (!err)
1799                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
1800                                fprintf(o->file, "%s:%d: %s.\n",
1801                                        data.filename, blank_at_eof, err);
1802                                data.status = 1; /* report errors */
1803                        }
1804                }
1805        }
1806 free_and_return:
1807        diff_free_filespec_data(one);
1808        diff_free_filespec_data(two);
1809        if (data.status)
1810                DIFF_OPT_SET(o, CHECK_FAILED);
1811}
1812
1813struct diff_filespec *alloc_filespec(const char *path)
1814{
1815        int namelen = strlen(path);
1816        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1817
1818        memset(spec, 0, sizeof(*spec));
1819        spec->path = (char *)(spec + 1);
1820        memcpy(spec->path, path, namelen+1);
1821        spec->count = 1;
1822        spec->is_binary = -1;
1823        return spec;
1824}
1825
1826void free_filespec(struct diff_filespec *spec)
1827{
1828        if (!--spec->count) {
1829                diff_free_filespec_data(spec);
1830                free(spec);
1831        }
1832}
1833
1834void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1835                   unsigned short mode)
1836{
1837        if (mode) {
1838                spec->mode = canon_mode(mode);
1839                hashcpy(spec->sha1, sha1);
1840                spec->sha1_valid = !is_null_sha1(sha1);
1841        }
1842}
1843
1844/*
1845 * Given a name and sha1 pair, if the index tells us the file in
1846 * the work tree has that object contents, return true, so that
1847 * prepare_temp_file() does not have to inflate and extract.
1848 */
1849static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1850{
1851        struct cache_entry *ce;
1852        struct stat st;
1853        int pos, len;
1854
1855        /*
1856         * We do not read the cache ourselves here, because the
1857         * benchmark with my previous version that always reads cache
1858         * shows that it makes things worse for diff-tree comparing
1859         * two linux-2.6 kernel trees in an already checked out work
1860         * tree.  This is because most diff-tree comparisons deal with
1861         * only a small number of files, while reading the cache is
1862         * expensive for a large project, and its cost outweighs the
1863         * savings we get by not inflating the object to a temporary
1864         * file.  Practically, this code only helps when we are used
1865         * by diff-cache --cached, which does read the cache before
1866         * calling us.
1867         */
1868        if (!active_cache)
1869                return 0;
1870
1871        /* We want to avoid the working directory if our caller
1872         * doesn't need the data in a normal file, this system
1873         * is rather slow with its stat/open/mmap/close syscalls,
1874         * and the object is contained in a pack file.  The pack
1875         * is probably already open and will be faster to obtain
1876         * the data through than the working directory.  Loose
1877         * objects however would tend to be slower as they need
1878         * to be individually opened and inflated.
1879         */
1880        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
1881                return 0;
1882
1883        len = strlen(name);
1884        pos = cache_name_pos(name, len);
1885        if (pos < 0)
1886                return 0;
1887        ce = active_cache[pos];
1888
1889        /*
1890         * This is not the sha1 we are looking for, or
1891         * unreusable because it is not a regular file.
1892         */
1893        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
1894                return 0;
1895
1896        /*
1897         * If ce is marked as "assume unchanged", there is no
1898         * guarantee that work tree matches what we are looking for.
1899         */
1900        if (ce->ce_flags & CE_VALID)
1901                return 0;
1902
1903        /*
1904         * If ce matches the file in the work tree, we can reuse it.
1905         */
1906        if (ce_uptodate(ce) ||
1907            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
1908                return 1;
1909
1910        return 0;
1911}
1912
1913static int populate_from_stdin(struct diff_filespec *s)
1914{
1915        struct strbuf buf = STRBUF_INIT;
1916        size_t size = 0;
1917
1918        if (strbuf_read(&buf, 0, 0) < 0)
1919                return error("error while reading from stdin %s",
1920                                     strerror(errno));
1921
1922        s->should_munmap = 0;
1923        s->data = strbuf_detach(&buf, &size);
1924        s->size = size;
1925        s->should_free = 1;
1926        return 0;
1927}
1928
1929static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1930{
1931        int len;
1932        char *data = xmalloc(100);
1933        len = snprintf(data, 100,
1934                "Subproject commit %s\n", sha1_to_hex(s->sha1));
1935        s->data = data;
1936        s->size = len;
1937        s->should_free = 1;
1938        if (size_only) {
1939                s->data = NULL;
1940                free(data);
1941        }
1942        return 0;
1943}
1944
1945/*
1946 * While doing rename detection and pickaxe operation, we may need to
1947 * grab the data for the blob (or file) for our own in-core comparison.
1948 * diff_filespec has data and size fields for this purpose.
1949 */
1950int diff_populate_filespec(struct diff_filespec *s, int size_only)
1951{
1952        int err = 0;
1953        if (!DIFF_FILE_VALID(s))
1954                die("internal error: asking to populate invalid file.");
1955        if (S_ISDIR(s->mode))
1956                return -1;
1957
1958        if (s->data)
1959                return 0;
1960
1961        if (size_only && 0 < s->size)
1962                return 0;
1963
1964        if (S_ISGITLINK(s->mode))
1965                return diff_populate_gitlink(s, size_only);
1966
1967        if (!s->sha1_valid ||
1968            reuse_worktree_file(s->path, s->sha1, 0)) {
1969                struct strbuf buf = STRBUF_INIT;
1970                struct stat st;
1971                int fd;
1972
1973                if (!strcmp(s->path, "-"))
1974                        return populate_from_stdin(s);
1975
1976                if (lstat(s->path, &st) < 0) {
1977                        if (errno == ENOENT) {
1978                        err_empty:
1979                                err = -1;
1980                        empty:
1981                                s->data = (char *)"";
1982                                s->size = 0;
1983                                return err;
1984                        }
1985                }
1986                s->size = xsize_t(st.st_size);
1987                if (!s->size)
1988                        goto empty;
1989                if (S_ISLNK(st.st_mode)) {
1990                        struct strbuf sb = STRBUF_INIT;
1991
1992                        if (strbuf_readlink(&sb, s->path, s->size))
1993                                goto err_empty;
1994                        s->size = sb.len;
1995                        s->data = strbuf_detach(&sb, NULL);
1996                        s->should_free = 1;
1997                        return 0;
1998                }
1999                if (size_only)
2000                        return 0;
2001                fd = open(s->path, O_RDONLY);
2002                if (fd < 0)
2003                        goto err_empty;
2004                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2005                close(fd);
2006                s->should_munmap = 1;
2007
2008                /*
2009                 * Convert from working tree format to canonical git format
2010                 */
2011                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2012                        size_t size = 0;
2013                        munmap(s->data, s->size);
2014                        s->should_munmap = 0;
2015                        s->data = strbuf_detach(&buf, &size);
2016                        s->size = size;
2017                        s->should_free = 1;
2018                }
2019        }
2020        else {
2021                enum object_type type;
2022                if (size_only)
2023                        type = sha1_object_info(s->sha1, &s->size);
2024                else {
2025                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2026                        s->should_free = 1;
2027                }
2028        }
2029        return 0;
2030}
2031
2032void diff_free_filespec_blob(struct diff_filespec *s)
2033{
2034        if (s->should_free)
2035                free(s->data);
2036        else if (s->should_munmap)
2037                munmap(s->data, s->size);
2038
2039        if (s->should_free || s->should_munmap) {
2040                s->should_free = s->should_munmap = 0;
2041                s->data = NULL;
2042        }
2043}
2044
2045void diff_free_filespec_data(struct diff_filespec *s)
2046{
2047        diff_free_filespec_blob(s);
2048        free(s->cnt_data);
2049        s->cnt_data = NULL;
2050}
2051
2052static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2053                           void *blob,
2054                           unsigned long size,
2055                           const unsigned char *sha1,
2056                           int mode)
2057{
2058        int fd;
2059        struct strbuf buf = STRBUF_INIT;
2060        struct strbuf template = STRBUF_INIT;
2061        char *path_dup = xstrdup(path);
2062        const char *base = basename(path_dup);
2063
2064        /* Generate "XXXXXX_basename.ext" */
2065        strbuf_addstr(&template, "XXXXXX_");
2066        strbuf_addstr(&template, base);
2067
2068        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2069                        strlen(base) + 1);
2070        if (fd < 0)
2071                die_errno("unable to create temp-file");
2072        if (convert_to_working_tree(path,
2073                        (const char *)blob, (size_t)size, &buf)) {
2074                blob = buf.buf;
2075                size = buf.len;
2076        }
2077        if (write_in_full(fd, blob, size) != size)
2078                die_errno("unable to write temp-file");
2079        close(fd);
2080        temp->name = temp->tmp_path;
2081        strcpy(temp->hex, sha1_to_hex(sha1));
2082        temp->hex[40] = 0;
2083        sprintf(temp->mode, "%06o", mode);
2084        strbuf_release(&buf);
2085        strbuf_release(&template);
2086        free(path_dup);
2087}
2088
2089static struct diff_tempfile *prepare_temp_file(const char *name,
2090                struct diff_filespec *one)
2091{
2092        struct diff_tempfile *temp = claim_diff_tempfile();
2093
2094        if (!DIFF_FILE_VALID(one)) {
2095        not_a_valid_file:
2096                /* A '-' entry produces this for file-2, and
2097                 * a '+' entry produces this for file-1.
2098                 */
2099                temp->name = "/dev/null";
2100                strcpy(temp->hex, ".");
2101                strcpy(temp->mode, ".");
2102                return temp;
2103        }
2104
2105        if (!remove_tempfile_installed) {
2106                atexit(remove_tempfile);
2107                sigchain_push_common(remove_tempfile_on_signal);
2108                remove_tempfile_installed = 1;
2109        }
2110
2111        if (!one->sha1_valid ||
2112            reuse_worktree_file(name, one->sha1, 1)) {
2113                struct stat st;
2114                if (lstat(name, &st) < 0) {
2115                        if (errno == ENOENT)
2116                                goto not_a_valid_file;
2117                        die_errno("stat(%s)", name);
2118                }
2119                if (S_ISLNK(st.st_mode)) {
2120                        struct strbuf sb = STRBUF_INIT;
2121                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2122                                die_errno("readlink(%s)", name);
2123                        prep_temp_blob(name, temp, sb.buf, sb.len,
2124                                       (one->sha1_valid ?
2125                                        one->sha1 : null_sha1),
2126                                       (one->sha1_valid ?
2127                                        one->mode : S_IFLNK));
2128                        strbuf_release(&sb);
2129                }
2130                else {
2131                        /* we can borrow from the file in the work tree */
2132                        temp->name = name;
2133                        if (!one->sha1_valid)
2134                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2135                        else
2136                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2137                        /* Even though we may sometimes borrow the
2138                         * contents from the work tree, we always want
2139                         * one->mode.  mode is trustworthy even when
2140                         * !(one->sha1_valid), as long as
2141                         * DIFF_FILE_VALID(one).
2142                         */
2143                        sprintf(temp->mode, "%06o", one->mode);
2144                }
2145                return temp;
2146        }
2147        else {
2148                if (diff_populate_filespec(one, 0))
2149                        die("cannot read data blob for %s", one->path);
2150                prep_temp_blob(name, temp, one->data, one->size,
2151                               one->sha1, one->mode);
2152        }
2153        return temp;
2154}
2155
2156/* An external diff command takes:
2157 *
2158 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2159 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2160 *
2161 */
2162static void run_external_diff(const char *pgm,
2163                              const char *name,
2164                              const char *other,
2165                              struct diff_filespec *one,
2166                              struct diff_filespec *two,
2167                              const char *xfrm_msg,
2168                              int complete_rewrite)
2169{
2170        const char *spawn_arg[10];
2171        int retval;
2172        const char **arg = &spawn_arg[0];
2173
2174        if (one && two) {
2175                struct diff_tempfile *temp_one, *temp_two;
2176                const char *othername = (other ? other : name);
2177                temp_one = prepare_temp_file(name, one);
2178                temp_two = prepare_temp_file(othername, two);
2179                *arg++ = pgm;
2180                *arg++ = name;
2181                *arg++ = temp_one->name;
2182                *arg++ = temp_one->hex;
2183                *arg++ = temp_one->mode;
2184                *arg++ = temp_two->name;
2185                *arg++ = temp_two->hex;
2186                *arg++ = temp_two->mode;
2187                if (other) {
2188                        *arg++ = other;
2189                        *arg++ = xfrm_msg;
2190                }
2191        } else {
2192                *arg++ = pgm;
2193                *arg++ = name;
2194        }
2195        *arg = NULL;
2196        fflush(NULL);
2197        retval = run_command_v_opt(spawn_arg, 0);
2198        remove_tempfile();
2199        if (retval) {
2200                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2201                exit(1);
2202        }
2203}
2204
2205static int similarity_index(struct diff_filepair *p)
2206{
2207        return p->score * 100 / MAX_SCORE;
2208}
2209
2210static void fill_metainfo(struct strbuf *msg,
2211                          const char *name,
2212                          const char *other,
2213                          struct diff_filespec *one,
2214                          struct diff_filespec *two,
2215                          struct diff_options *o,
2216                          struct diff_filepair *p)
2217{
2218        strbuf_init(msg, PATH_MAX * 2 + 300);
2219        switch (p->status) {
2220        case DIFF_STATUS_COPIED:
2221                strbuf_addf(msg, "similarity index %d%%", similarity_index(p));
2222                strbuf_addstr(msg, "\ncopy from ");
2223                quote_c_style(name, msg, NULL, 0);
2224                strbuf_addstr(msg, "\ncopy to ");
2225                quote_c_style(other, msg, NULL, 0);
2226                strbuf_addch(msg, '\n');
2227                break;
2228        case DIFF_STATUS_RENAMED:
2229                strbuf_addf(msg, "similarity index %d%%", similarity_index(p));
2230                strbuf_addstr(msg, "\nrename from ");
2231                quote_c_style(name, msg, NULL, 0);
2232                strbuf_addstr(msg, "\nrename to ");
2233                quote_c_style(other, msg, NULL, 0);
2234                strbuf_addch(msg, '\n');
2235                break;
2236        case DIFF_STATUS_MODIFIED:
2237                if (p->score) {
2238                        strbuf_addf(msg, "dissimilarity index %d%%\n",
2239                                    similarity_index(p));
2240                        break;
2241                }
2242                /* fallthru */
2243        default:
2244                /* nothing */
2245                ;
2246        }
2247        if (one && two && hashcmp(one->sha1, two->sha1)) {
2248                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2249
2250                if (DIFF_OPT_TST(o, BINARY)) {
2251                        mmfile_t mf;
2252                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2253                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2254                                abbrev = 40;
2255                }
2256                strbuf_addf(msg, "index %.*s..%.*s",
2257                            abbrev, sha1_to_hex(one->sha1),
2258                            abbrev, sha1_to_hex(two->sha1));
2259                if (one->mode == two->mode)
2260                        strbuf_addf(msg, " %06o", one->mode);
2261                strbuf_addch(msg, '\n');
2262        }
2263        if (msg->len)
2264                strbuf_setlen(msg, msg->len - 1);
2265}
2266
2267static void run_diff_cmd(const char *pgm,
2268                         const char *name,
2269                         const char *other,
2270                         const char *attr_path,
2271                         struct diff_filespec *one,
2272                         struct diff_filespec *two,
2273                         struct strbuf *msg,
2274                         struct diff_options *o,
2275                         struct diff_filepair *p)
2276{
2277        const char *xfrm_msg = NULL;
2278        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2279
2280        if (msg) {
2281                fill_metainfo(msg, name, other, one, two, o, p);
2282                xfrm_msg = msg->len ? msg->buf : NULL;
2283        }
2284
2285        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2286                pgm = NULL;
2287        else {
2288                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2289                if (drv && drv->external)
2290                        pgm = drv->external;
2291        }
2292
2293        if (pgm) {
2294                run_external_diff(pgm, name, other, one, two, xfrm_msg,
2295                                  complete_rewrite);
2296                return;
2297        }
2298        if (one && two)
2299                builtin_diff(name, other ? other : name,
2300                             one, two, xfrm_msg, o, complete_rewrite);
2301        else
2302                fprintf(o->file, "* Unmerged path %s\n", name);
2303}
2304
2305static void diff_fill_sha1_info(struct diff_filespec *one)
2306{
2307        if (DIFF_FILE_VALID(one)) {
2308                if (!one->sha1_valid) {
2309                        struct stat st;
2310                        if (!strcmp(one->path, "-")) {
2311                                hashcpy(one->sha1, null_sha1);
2312                                return;
2313                        }
2314                        if (lstat(one->path, &st) < 0)
2315                                die_errno("stat '%s'", one->path);
2316                        if (index_path(one->sha1, one->path, &st, 0))
2317                                die("cannot hash %s", one->path);
2318                }
2319        }
2320        else
2321                hashclr(one->sha1);
2322}
2323
2324static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2325{
2326        /* Strip the prefix but do not molest /dev/null and absolute paths */
2327        if (*namep && **namep != '/')
2328                *namep += prefix_length;
2329        if (*otherp && **otherp != '/')
2330                *otherp += prefix_length;
2331}
2332
2333static void run_diff(struct diff_filepair *p, struct diff_options *o)
2334{
2335        const char *pgm = external_diff();
2336        struct strbuf msg;
2337        struct diff_filespec *one = p->one;
2338        struct diff_filespec *two = p->two;
2339        const char *name;
2340        const char *other;
2341        const char *attr_path;
2342
2343        name  = p->one->path;
2344        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2345        attr_path = name;
2346        if (o->prefix_length)
2347                strip_prefix(o->prefix_length, &name, &other);
2348
2349        if (DIFF_PAIR_UNMERGED(p)) {
2350                run_diff_cmd(pgm, name, NULL, attr_path,
2351                             NULL, NULL, NULL, o, p);
2352                return;
2353        }
2354
2355        diff_fill_sha1_info(one);
2356        diff_fill_sha1_info(two);
2357
2358        if (!pgm &&
2359            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2360            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2361                /*
2362                 * a filepair that changes between file and symlink
2363                 * needs to be split into deletion and creation.
2364                 */
2365                struct diff_filespec *null = alloc_filespec(two->path);
2366                run_diff_cmd(NULL, name, other, attr_path,
2367                             one, null, &msg, o, p);
2368                free(null);
2369                strbuf_release(&msg);
2370
2371                null = alloc_filespec(one->path);
2372                run_diff_cmd(NULL, name, other, attr_path,
2373                             null, two, &msg, o, p);
2374                free(null);
2375        }
2376        else
2377                run_diff_cmd(pgm, name, other, attr_path,
2378                             one, two, &msg, o, p);
2379
2380        strbuf_release(&msg);
2381}
2382
2383static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2384                         struct diffstat_t *diffstat)
2385{
2386        const char *name;
2387        const char *other;
2388        int complete_rewrite = 0;
2389
2390        if (DIFF_PAIR_UNMERGED(p)) {
2391                /* unmerged */
2392                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2393                return;
2394        }
2395
2396        name = p->one->path;
2397        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2398
2399        if (o->prefix_length)
2400                strip_prefix(o->prefix_length, &name, &other);
2401
2402        diff_fill_sha1_info(p->one);
2403        diff_fill_sha1_info(p->two);
2404
2405        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2406                complete_rewrite = 1;
2407        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2408}
2409
2410static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2411{
2412        const char *name;
2413        const char *other;
2414        const char *attr_path;
2415
2416        if (DIFF_PAIR_UNMERGED(p)) {
2417                /* unmerged */
2418                return;
2419        }
2420
2421        name = p->one->path;
2422        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2423        attr_path = other ? other : name;
2424
2425        if (o->prefix_length)
2426                strip_prefix(o->prefix_length, &name, &other);
2427
2428        diff_fill_sha1_info(p->one);
2429        diff_fill_sha1_info(p->two);
2430
2431        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
2432}
2433
2434void diff_setup(struct diff_options *options)
2435{
2436        memset(options, 0, sizeof(*options));
2437
2438        options->file = stdout;
2439
2440        options->line_termination = '\n';
2441        options->break_opt = -1;
2442        options->rename_limit = -1;
2443        options->dirstat_percent = 3;
2444        options->context = 3;
2445
2446        options->change = diff_change;
2447        options->add_remove = diff_addremove;
2448        if (diff_use_color_default > 0)
2449                DIFF_OPT_SET(options, COLOR_DIFF);
2450        options->detect_rename = diff_detect_rename_default;
2451
2452        if (!diff_mnemonic_prefix) {
2453                options->a_prefix = "a/";
2454                options->b_prefix = "b/";
2455        }
2456}
2457
2458int diff_setup_done(struct diff_options *options)
2459{
2460        int count = 0;
2461
2462        if (options->output_format & DIFF_FORMAT_NAME)
2463                count++;
2464        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2465                count++;
2466        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2467                count++;
2468        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2469                count++;
2470        if (count > 1)
2471                die("--name-only, --name-status, --check and -s are mutually exclusive");
2472
2473        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2474                options->detect_rename = DIFF_DETECT_COPY;
2475
2476        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
2477                options->prefix = NULL;
2478        if (options->prefix)
2479                options->prefix_length = strlen(options->prefix);
2480        else
2481                options->prefix_length = 0;
2482
2483        if (options->output_format & (DIFF_FORMAT_NAME |
2484                                      DIFF_FORMAT_NAME_STATUS |
2485                                      DIFF_FORMAT_CHECKDIFF |
2486                                      DIFF_FORMAT_NO_OUTPUT))
2487                options->output_format &= ~(DIFF_FORMAT_RAW |
2488                                            DIFF_FORMAT_NUMSTAT |
2489                                            DIFF_FORMAT_DIFFSTAT |
2490                                            DIFF_FORMAT_SHORTSTAT |
2491                                            DIFF_FORMAT_DIRSTAT |
2492                                            DIFF_FORMAT_SUMMARY |
2493                                            DIFF_FORMAT_PATCH);
2494
2495        /*
2496         * These cases always need recursive; we do not drop caller-supplied
2497         * recursive bits for other formats here.
2498         */
2499        if (options->output_format & (DIFF_FORMAT_PATCH |
2500                                      DIFF_FORMAT_NUMSTAT |
2501                                      DIFF_FORMAT_DIFFSTAT |
2502                                      DIFF_FORMAT_SHORTSTAT |
2503                                      DIFF_FORMAT_DIRSTAT |
2504                                      DIFF_FORMAT_SUMMARY |
2505                                      DIFF_FORMAT_CHECKDIFF))
2506                DIFF_OPT_SET(options, RECURSIVE);
2507        /*
2508         * Also pickaxe would not work very well if you do not say recursive
2509         */
2510        if (options->pickaxe)
2511                DIFF_OPT_SET(options, RECURSIVE);
2512
2513        if (options->detect_rename && options->rename_limit < 0)
2514                options->rename_limit = diff_rename_limit_default;
2515        if (options->setup & DIFF_SETUP_USE_CACHE) {
2516                if (!active_cache)
2517                        /* read-cache does not die even when it fails
2518                         * so it is safe for us to do this here.  Also
2519                         * it does not smudge active_cache or active_nr
2520                         * when it fails, so we do not have to worry about
2521                         * cleaning it up ourselves either.
2522                         */
2523                        read_cache();
2524        }
2525        if (options->abbrev <= 0 || 40 < options->abbrev)
2526                options->abbrev = 40; /* full */
2527
2528        /*
2529         * It does not make sense to show the first hit we happened
2530         * to have found.  It does not make sense not to return with
2531         * exit code in such a case either.
2532         */
2533        if (DIFF_OPT_TST(options, QUIET)) {
2534                options->output_format = DIFF_FORMAT_NO_OUTPUT;
2535                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2536        }
2537
2538        return 0;
2539}
2540
2541static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2542{
2543        char c, *eq;
2544        int len;
2545
2546        if (*arg != '-')
2547                return 0;
2548        c = *++arg;
2549        if (!c)
2550                return 0;
2551        if (c == arg_short) {
2552                c = *++arg;
2553                if (!c)
2554                        return 1;
2555                if (val && isdigit(c)) {
2556                        char *end;
2557                        int n = strtoul(arg, &end, 10);
2558                        if (*end)
2559                                return 0;
2560                        *val = n;
2561                        return 1;
2562                }
2563                return 0;
2564        }
2565        if (c != '-')
2566                return 0;
2567        arg++;
2568        eq = strchr(arg, '=');
2569        if (eq)
2570                len = eq - arg;
2571        else
2572                len = strlen(arg);
2573        if (!len || strncmp(arg, arg_long, len))
2574                return 0;
2575        if (eq) {
2576                int n;
2577                char *end;
2578                if (!isdigit(*++eq))
2579                        return 0;
2580                n = strtoul(eq, &end, 10);
2581                if (*end)
2582                        return 0;
2583                *val = n;
2584        }
2585        return 1;
2586}
2587
2588static int diff_scoreopt_parse(const char *opt);
2589
2590int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2591{
2592        const char *arg = av[0];
2593
2594        /* Output format options */
2595        if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2596                options->output_format |= DIFF_FORMAT_PATCH;
2597        else if (opt_arg(arg, 'U', "unified", &options->context))
2598                options->output_format |= DIFF_FORMAT_PATCH;
2599        else if (!strcmp(arg, "--raw"))
2600                options->output_format |= DIFF_FORMAT_RAW;
2601        else if (!strcmp(arg, "--patch-with-raw"))
2602                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2603        else if (!strcmp(arg, "--numstat"))
2604                options->output_format |= DIFF_FORMAT_NUMSTAT;
2605        else if (!strcmp(arg, "--shortstat"))
2606                options->output_format |= DIFF_FORMAT_SHORTSTAT;
2607        else if (opt_arg(arg, 'X', "dirstat", &options->dirstat_percent))
2608                options->output_format |= DIFF_FORMAT_DIRSTAT;
2609        else if (!strcmp(arg, "--cumulative")) {
2610                options->output_format |= DIFF_FORMAT_DIRSTAT;
2611                DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
2612        } else if (opt_arg(arg, 0, "dirstat-by-file",
2613                           &options->dirstat_percent)) {
2614                options->output_format |= DIFF_FORMAT_DIRSTAT;
2615                DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
2616        }
2617        else if (!strcmp(arg, "--check"))
2618                options->output_format |= DIFF_FORMAT_CHECKDIFF;
2619        else if (!strcmp(arg, "--summary"))
2620                options->output_format |= DIFF_FORMAT_SUMMARY;
2621        else if (!strcmp(arg, "--patch-with-stat"))
2622                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2623        else if (!strcmp(arg, "--name-only"))
2624                options->output_format |= DIFF_FORMAT_NAME;
2625        else if (!strcmp(arg, "--name-status"))
2626                options->output_format |= DIFF_FORMAT_NAME_STATUS;
2627        else if (!strcmp(arg, "-s"))
2628                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2629        else if (!prefixcmp(arg, "--stat")) {
2630                char *end;
2631                int width = options->stat_width;
2632                int name_width = options->stat_name_width;
2633                arg += 6;
2634                end = (char *)arg;
2635
2636                switch (*arg) {
2637                case '-':
2638                        if (!prefixcmp(arg, "-width="))
2639                                width = strtoul(arg + 7, &end, 10);
2640                        else if (!prefixcmp(arg, "-name-width="))
2641                                name_width = strtoul(arg + 12, &end, 10);
2642                        break;
2643                case '=':
2644                        width = strtoul(arg+1, &end, 10);
2645                        if (*end == ',')
2646                                name_width = strtoul(end+1, &end, 10);
2647                }
2648
2649                /* Important! This checks all the error cases! */
2650                if (*end)
2651                        return 0;
2652                options->output_format |= DIFF_FORMAT_DIFFSTAT;
2653                options->stat_name_width = name_width;
2654                options->stat_width = width;
2655        }
2656
2657        /* renames options */
2658        else if (!prefixcmp(arg, "-B")) {
2659                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2660                        return -1;
2661        }
2662        else if (!prefixcmp(arg, "-M")) {
2663                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2664                        return -1;
2665                options->detect_rename = DIFF_DETECT_RENAME;
2666        }
2667        else if (!prefixcmp(arg, "-C")) {
2668                if (options->detect_rename == DIFF_DETECT_COPY)
2669                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2670                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2671                        return -1;
2672                options->detect_rename = DIFF_DETECT_COPY;
2673        }
2674        else if (!strcmp(arg, "--no-renames"))
2675                options->detect_rename = 0;
2676        else if (!strcmp(arg, "--relative"))
2677                DIFF_OPT_SET(options, RELATIVE_NAME);
2678        else if (!prefixcmp(arg, "--relative=")) {
2679                DIFF_OPT_SET(options, RELATIVE_NAME);
2680                options->prefix = arg + 11;
2681        }
2682
2683        /* xdiff options */
2684        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2685                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
2686        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2687                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
2688        else if (!strcmp(arg, "--ignore-space-at-eol"))
2689                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
2690        else if (!strcmp(arg, "--patience"))
2691                DIFF_XDL_SET(options, PATIENCE_DIFF);
2692
2693        /* flags options */
2694        else if (!strcmp(arg, "--binary")) {
2695                options->output_format |= DIFF_FORMAT_PATCH;
2696                DIFF_OPT_SET(options, BINARY);
2697        }
2698        else if (!strcmp(arg, "--full-index"))
2699                DIFF_OPT_SET(options, FULL_INDEX);
2700        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2701                DIFF_OPT_SET(options, TEXT);
2702        else if (!strcmp(arg, "-R"))
2703                DIFF_OPT_SET(options, REVERSE_DIFF);
2704        else if (!strcmp(arg, "--find-copies-harder"))
2705                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2706        else if (!strcmp(arg, "--follow"))
2707                DIFF_OPT_SET(options, FOLLOW_RENAMES);
2708        else if (!strcmp(arg, "--color"))
2709                DIFF_OPT_SET(options, COLOR_DIFF);
2710        else if (!strcmp(arg, "--no-color"))
2711                DIFF_OPT_CLR(options, COLOR_DIFF);
2712        else if (!strcmp(arg, "--color-words")) {
2713                DIFF_OPT_SET(options, COLOR_DIFF);
2714                DIFF_OPT_SET(options, COLOR_DIFF_WORDS);
2715        }
2716        else if (!prefixcmp(arg, "--color-words=")) {
2717                DIFF_OPT_SET(options, COLOR_DIFF);
2718                DIFF_OPT_SET(options, COLOR_DIFF_WORDS);
2719                options->word_regex = arg + 14;
2720        }
2721        else if (!strcmp(arg, "--exit-code"))
2722                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2723        else if (!strcmp(arg, "--quiet"))
2724                DIFF_OPT_SET(options, QUIET);
2725        else if (!strcmp(arg, "--ext-diff"))
2726                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2727        else if (!strcmp(arg, "--no-ext-diff"))
2728                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2729        else if (!strcmp(arg, "--textconv"))
2730                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
2731        else if (!strcmp(arg, "--no-textconv"))
2732                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
2733        else if (!strcmp(arg, "--ignore-submodules"))
2734                DIFF_OPT_SET(options, IGNORE_SUBMODULES);
2735
2736        /* misc options */
2737        else if (!strcmp(arg, "-z"))
2738                options->line_termination = 0;
2739        else if (!prefixcmp(arg, "-l"))
2740                options->rename_limit = strtoul(arg+2, NULL, 10);
2741        else if (!prefixcmp(arg, "-S"))
2742                options->pickaxe = arg + 2;
2743        else if (!strcmp(arg, "--pickaxe-all"))
2744                options->pickaxe_opts = DIFF_PICKAXE_ALL;
2745        else if (!strcmp(arg, "--pickaxe-regex"))
2746                options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2747        else if (!prefixcmp(arg, "-O"))
2748                options->orderfile = arg + 2;
2749        else if (!prefixcmp(arg, "--diff-filter="))
2750                options->filter = arg + 14;
2751        else if (!strcmp(arg, "--abbrev"))
2752                options->abbrev = DEFAULT_ABBREV;
2753        else if (!prefixcmp(arg, "--abbrev=")) {
2754                options->abbrev = strtoul(arg + 9, NULL, 10);
2755                if (options->abbrev < MINIMUM_ABBREV)
2756                        options->abbrev = MINIMUM_ABBREV;
2757                else if (40 < options->abbrev)
2758                        options->abbrev = 40;
2759        }
2760        else if (!prefixcmp(arg, "--src-prefix="))
2761                options->a_prefix = arg + 13;
2762        else if (!prefixcmp(arg, "--dst-prefix="))
2763                options->b_prefix = arg + 13;
2764        else if (!strcmp(arg, "--no-prefix"))
2765                options->a_prefix = options->b_prefix = "";
2766        else if (opt_arg(arg, '\0', "inter-hunk-context",
2767                         &options->interhunkcontext))
2768                ;
2769        else if (!prefixcmp(arg, "--output=")) {
2770                options->file = fopen(arg + strlen("--output="), "w");
2771                options->close_file = 1;
2772        } else
2773                return 0;
2774        return 1;
2775}
2776
2777static int parse_num(const char **cp_p)
2778{
2779        unsigned long num, scale;
2780        int ch, dot;
2781        const char *cp = *cp_p;
2782
2783        num = 0;
2784        scale = 1;
2785        dot = 0;
2786        for(;;) {
2787                ch = *cp;
2788                if ( !dot && ch == '.' ) {
2789                        scale = 1;
2790                        dot = 1;
2791                } else if ( ch == '%' ) {
2792                        scale = dot ? scale*100 : 100;
2793                        cp++;   /* % is always at the end */
2794                        break;
2795                } else if ( ch >= '0' && ch <= '9' ) {
2796                        if ( scale < 100000 ) {
2797                                scale *= 10;
2798                                num = (num*10) + (ch-'0');
2799                        }
2800                } else {
2801                        break;
2802                }
2803                cp++;
2804        }
2805        *cp_p = cp;
2806
2807        /* user says num divided by scale and we say internally that
2808         * is MAX_SCORE * num / scale.
2809         */
2810        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2811}
2812
2813static int diff_scoreopt_parse(const char *opt)
2814{
2815        int opt1, opt2, cmd;
2816
2817        if (*opt++ != '-')
2818                return -1;
2819        cmd = *opt++;
2820        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2821                return -1; /* that is not a -M, -C nor -B option */
2822
2823        opt1 = parse_num(&opt);
2824        if (cmd != 'B')
2825                opt2 = 0;
2826        else {
2827                if (*opt == 0)
2828                        opt2 = 0;
2829                else if (*opt != '/')
2830                        return -1; /* we expect -B80/99 or -B80 */
2831                else {
2832                        opt++;
2833                        opt2 = parse_num(&opt);
2834                }
2835        }
2836        if (*opt != 0)
2837                return -1;
2838        return opt1 | (opt2 << 16);
2839}
2840
2841struct diff_queue_struct diff_queued_diff;
2842
2843void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2844{
2845        if (queue->alloc <= queue->nr) {
2846                queue->alloc = alloc_nr(queue->alloc);
2847                queue->queue = xrealloc(queue->queue,
2848                                        sizeof(dp) * queue->alloc);
2849        }
2850        queue->queue[queue->nr++] = dp;
2851}
2852
2853struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2854                                 struct diff_filespec *one,
2855                                 struct diff_filespec *two)
2856{
2857        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2858        dp->one = one;
2859        dp->two = two;
2860        if (queue)
2861                diff_q(queue, dp);
2862        return dp;
2863}
2864
2865void diff_free_filepair(struct diff_filepair *p)
2866{
2867        free_filespec(p->one);
2868        free_filespec(p->two);
2869        free(p);
2870}
2871
2872/* This is different from find_unique_abbrev() in that
2873 * it stuffs the result with dots for alignment.
2874 */
2875const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2876{
2877        int abblen;
2878        const char *abbrev;
2879        if (len == 40)
2880                return sha1_to_hex(sha1);
2881
2882        abbrev = find_unique_abbrev(sha1, len);
2883        abblen = strlen(abbrev);
2884        if (abblen < 37) {
2885                static char hex[41];
2886                if (len < abblen && abblen <= len + 2)
2887                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2888                else
2889                        sprintf(hex, "%s...", abbrev);
2890                return hex;
2891        }
2892        return sha1_to_hex(sha1);
2893}
2894
2895static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2896{
2897        int line_termination = opt->line_termination;
2898        int inter_name_termination = line_termination ? '\t' : '\0';
2899
2900        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2901                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
2902                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
2903                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2904        }
2905        if (p->score) {
2906                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
2907                        inter_name_termination);
2908        } else {
2909                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
2910        }
2911
2912        if (p->status == DIFF_STATUS_COPIED ||
2913            p->status == DIFF_STATUS_RENAMED) {
2914                const char *name_a, *name_b;
2915                name_a = p->one->path;
2916                name_b = p->two->path;
2917                strip_prefix(opt->prefix_length, &name_a, &name_b);
2918                write_name_quoted(name_a, opt->file, inter_name_termination);
2919                write_name_quoted(name_b, opt->file, line_termination);
2920        } else {
2921                const char *name_a, *name_b;
2922                name_a = p->one->mode ? p->one->path : p->two->path;
2923                name_b = NULL;
2924                strip_prefix(opt->prefix_length, &name_a, &name_b);
2925                write_name_quoted(name_a, opt->file, line_termination);
2926        }
2927}
2928
2929int diff_unmodified_pair(struct diff_filepair *p)
2930{
2931        /* This function is written stricter than necessary to support
2932         * the currently implemented transformers, but the idea is to
2933         * let transformers to produce diff_filepairs any way they want,
2934         * and filter and clean them up here before producing the output.
2935         */
2936        struct diff_filespec *one = p->one, *two = p->two;
2937
2938        if (DIFF_PAIR_UNMERGED(p))
2939                return 0; /* unmerged is interesting */
2940
2941        /* deletion, addition, mode or type change
2942         * and rename are all interesting.
2943         */
2944        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2945            DIFF_PAIR_MODE_CHANGED(p) ||
2946            strcmp(one->path, two->path))
2947                return 0;
2948
2949        /* both are valid and point at the same path.  that is, we are
2950         * dealing with a change.
2951         */
2952        if (one->sha1_valid && two->sha1_valid &&
2953            !hashcmp(one->sha1, two->sha1))
2954                return 1; /* no change */
2955        if (!one->sha1_valid && !two->sha1_valid)
2956                return 1; /* both look at the same file on the filesystem. */
2957        return 0;
2958}
2959
2960static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2961{
2962        if (diff_unmodified_pair(p))
2963                return;
2964
2965        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2966            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2967                return; /* no tree diffs in patch format */
2968
2969        run_diff(p, o);
2970}
2971
2972static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2973                            struct diffstat_t *diffstat)
2974{
2975        if (diff_unmodified_pair(p))
2976                return;
2977
2978        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2979            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2980                return; /* no tree diffs in patch format */
2981
2982        run_diffstat(p, o, diffstat);
2983}
2984
2985static void diff_flush_checkdiff(struct diff_filepair *p,
2986                struct diff_options *o)
2987{
2988        if (diff_unmodified_pair(p))
2989                return;
2990
2991        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2992            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2993                return; /* no tree diffs in patch format */
2994
2995        run_checkdiff(p, o);
2996}
2997
2998int diff_queue_is_empty(void)
2999{
3000        struct diff_queue_struct *q = &diff_queued_diff;
3001        int i;
3002        for (i = 0; i < q->nr; i++)
3003                if (!diff_unmodified_pair(q->queue[i]))
3004                        return 0;
3005        return 1;
3006}
3007
3008#if DIFF_DEBUG
3009void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3010{
3011        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3012                x, one ? one : "",
3013                s->path,
3014                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3015                s->mode,
3016                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3017        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3018                x, one ? one : "",
3019                s->size, s->xfrm_flags);
3020}
3021
3022void diff_debug_filepair(const struct diff_filepair *p, int i)
3023{
3024        diff_debug_filespec(p->one, i, "one");
3025        diff_debug_filespec(p->two, i, "two");
3026        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3027                p->score, p->status ? p->status : '?',
3028                p->one->rename_used, p->broken_pair);
3029}
3030
3031void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3032{
3033        int i;
3034        if (msg)
3035                fprintf(stderr, "%s\n", msg);
3036        fprintf(stderr, "q->nr = %d\n", q->nr);
3037        for (i = 0; i < q->nr; i++) {
3038                struct diff_filepair *p = q->queue[i];
3039                diff_debug_filepair(p, i);
3040        }
3041}
3042#endif
3043
3044static void diff_resolve_rename_copy(void)
3045{
3046        int i;
3047        struct diff_filepair *p;
3048        struct diff_queue_struct *q = &diff_queued_diff;
3049
3050        diff_debug_queue("resolve-rename-copy", q);
3051
3052        for (i = 0; i < q->nr; i++) {
3053                p = q->queue[i];
3054                p->status = 0; /* undecided */
3055                if (DIFF_PAIR_UNMERGED(p))
3056                        p->status = DIFF_STATUS_UNMERGED;
3057                else if (!DIFF_FILE_VALID(p->one))
3058                        p->status = DIFF_STATUS_ADDED;
3059                else if (!DIFF_FILE_VALID(p->two))
3060                        p->status = DIFF_STATUS_DELETED;
3061                else if (DIFF_PAIR_TYPE_CHANGED(p))
3062                        p->status = DIFF_STATUS_TYPE_CHANGED;
3063
3064                /* from this point on, we are dealing with a pair
3065                 * whose both sides are valid and of the same type, i.e.
3066                 * either in-place edit or rename/copy edit.
3067                 */
3068                else if (DIFF_PAIR_RENAME(p)) {
3069                        /*
3070                         * A rename might have re-connected a broken
3071                         * pair up, causing the pathnames to be the
3072                         * same again. If so, that's not a rename at
3073                         * all, just a modification..
3074                         *
3075                         * Otherwise, see if this source was used for
3076                         * multiple renames, in which case we decrement
3077                         * the count, and call it a copy.
3078                         */
3079                        if (!strcmp(p->one->path, p->two->path))
3080                                p->status = DIFF_STATUS_MODIFIED;
3081                        else if (--p->one->rename_used > 0)
3082                                p->status = DIFF_STATUS_COPIED;
3083                        else
3084                                p->status = DIFF_STATUS_RENAMED;
3085                }
3086                else if (hashcmp(p->one->sha1, p->two->sha1) ||
3087                         p->one->mode != p->two->mode ||
3088                         is_null_sha1(p->one->sha1))
3089                        p->status = DIFF_STATUS_MODIFIED;
3090                else {
3091                        /* This is a "no-change" entry and should not
3092                         * happen anymore, but prepare for broken callers.
3093                         */
3094                        error("feeding unmodified %s to diffcore",
3095                              p->one->path);
3096                        p->status = DIFF_STATUS_UNKNOWN;
3097                }
3098        }
3099        diff_debug_queue("resolve-rename-copy done", q);
3100}
3101
3102static int check_pair_status(struct diff_filepair *p)
3103{
3104        switch (p->status) {
3105        case DIFF_STATUS_UNKNOWN:
3106                return 0;
3107        case 0:
3108                die("internal error in diff-resolve-rename-copy");
3109        default:
3110                return 1;
3111        }
3112}
3113
3114static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3115{
3116        int fmt = opt->output_format;
3117
3118        if (fmt & DIFF_FORMAT_CHECKDIFF)
3119                diff_flush_checkdiff(p, opt);
3120        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
3121                diff_flush_raw(p, opt);
3122        else if (fmt & DIFF_FORMAT_NAME) {
3123                const char *name_a, *name_b;
3124                name_a = p->two->path;
3125                name_b = NULL;
3126                strip_prefix(opt->prefix_length, &name_a, &name_b);
3127                write_name_quoted(name_a, opt->file, opt->line_termination);
3128        }
3129}
3130
3131static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
3132{
3133        if (fs->mode)
3134                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
3135        else
3136                fprintf(file, " %s ", newdelete);
3137        write_name_quoted(fs->path, file, '\n');
3138}
3139
3140
3141static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name)
3142{
3143        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
3144                fprintf(file, " mode change %06o => %06o%c", p->one->mode, p->two->mode,
3145                        show_name ? ' ' : '\n');
3146                if (show_name) {
3147                        write_name_quoted(p->two->path, file, '\n');
3148                }
3149        }
3150}
3151
3152static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p)
3153{
3154        char *names = pprint_rename(p->one->path, p->two->path);
3155
3156        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
3157        free(names);
3158        show_mode_change(file, p, 0);
3159}
3160
3161static void diff_summary(FILE *file, struct diff_filepair *p)
3162{
3163        switch(p->status) {
3164        case DIFF_STATUS_DELETED:
3165                show_file_mode_name(file, "delete", p->one);
3166                break;
3167        case DIFF_STATUS_ADDED:
3168                show_file_mode_name(file, "create", p->two);
3169                break;
3170        case DIFF_STATUS_COPIED:
3171                show_rename_copy(file, "copy", p);
3172                break;
3173        case DIFF_STATUS_RENAMED:
3174                show_rename_copy(file, "rename", p);
3175                break;
3176        default:
3177                if (p->score) {
3178                        fputs(" rewrite ", file);
3179                        write_name_quoted(p->two->path, file, ' ');
3180                        fprintf(file, "(%d%%)\n", similarity_index(p));
3181                }
3182                show_mode_change(file, p, !p->score);
3183                break;
3184        }
3185}
3186
3187struct patch_id_t {
3188        git_SHA_CTX *ctx;
3189        int patchlen;
3190};
3191
3192static int remove_space(char *line, int len)
3193{
3194        int i;
3195        char *dst = line;
3196        unsigned char c;
3197
3198        for (i = 0; i < len; i++)
3199                if (!isspace((c = line[i])))
3200                        *dst++ = c;
3201
3202        return dst - line;
3203}
3204
3205static void patch_id_consume(void *priv, char *line, unsigned long len)
3206{
3207        struct patch_id_t *data = priv;
3208        int new_len;
3209
3210        /* Ignore line numbers when computing the SHA1 of the patch */
3211        if (!prefixcmp(line, "@@ -"))
3212                return;
3213
3214        new_len = remove_space(line, len);
3215
3216        git_SHA1_Update(data->ctx, line, new_len);
3217        data->patchlen += new_len;
3218}
3219
3220/* returns 0 upon success, and writes result into sha1 */
3221static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
3222{
3223        struct diff_queue_struct *q = &diff_queued_diff;
3224        int i;
3225        git_SHA_CTX ctx;
3226        struct patch_id_t data;
3227        char buffer[PATH_MAX * 4 + 20];
3228
3229        git_SHA1_Init(&ctx);
3230        memset(&data, 0, sizeof(struct patch_id_t));
3231        data.ctx = &ctx;
3232
3233        for (i = 0; i < q->nr; i++) {
3234                xpparam_t xpp;
3235                xdemitconf_t xecfg;
3236                xdemitcb_t ecb;
3237                mmfile_t mf1, mf2;
3238                struct diff_filepair *p = q->queue[i];
3239                int len1, len2;
3240
3241                memset(&xpp, 0, sizeof(xpp));
3242                memset(&xecfg, 0, sizeof(xecfg));
3243                if (p->status == 0)
3244                        return error("internal diff status error");
3245                if (p->status == DIFF_STATUS_UNKNOWN)
3246                        continue;
3247                if (diff_unmodified_pair(p))
3248                        continue;
3249                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3250                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3251                        continue;
3252                if (DIFF_PAIR_UNMERGED(p))
3253                        continue;
3254
3255                diff_fill_sha1_info(p->one);
3256                diff_fill_sha1_info(p->two);
3257                if (fill_mmfile(&mf1, p->one) < 0 ||
3258                                fill_mmfile(&mf2, p->two) < 0)
3259                        return error("unable to read files to diff");
3260
3261                len1 = remove_space(p->one->path, strlen(p->one->path));
3262                len2 = remove_space(p->two->path, strlen(p->two->path));
3263                if (p->one->mode == 0)
3264                        len1 = snprintf(buffer, sizeof(buffer),
3265                                        "diff--gita/%.*sb/%.*s"
3266                                        "newfilemode%06o"
3267                                        "---/dev/null"
3268                                        "+++b/%.*s",
3269                                        len1, p->one->path,
3270                                        len2, p->two->path,
3271                                        p->two->mode,
3272                                        len2, p->two->path);
3273                else if (p->two->mode == 0)
3274                        len1 = snprintf(buffer, sizeof(buffer),
3275                                        "diff--gita/%.*sb/%.*s"
3276                                        "deletedfilemode%06o"
3277                                        "---a/%.*s"
3278                                        "+++/dev/null",
3279                                        len1, p->one->path,
3280                                        len2, p->two->path,
3281                                        p->one->mode,
3282                                        len1, p->one->path);
3283                else
3284                        len1 = snprintf(buffer, sizeof(buffer),
3285                                        "diff--gita/%.*sb/%.*s"
3286                                        "---a/%.*s"
3287                                        "+++b/%.*s",
3288                                        len1, p->one->path,
3289                                        len2, p->two->path,
3290                                        len1, p->one->path,
3291                                        len2, p->two->path);
3292                git_SHA1_Update(&ctx, buffer, len1);
3293
3294                xpp.flags = XDF_NEED_MINIMAL;
3295                xecfg.ctxlen = 3;
3296                xecfg.flags = XDL_EMIT_FUNCNAMES;
3297                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
3298                              &xpp, &xecfg, &ecb);
3299        }
3300
3301        git_SHA1_Final(sha1, &ctx);
3302        return 0;
3303}
3304
3305int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
3306{
3307        struct diff_queue_struct *q = &diff_queued_diff;
3308        int i;
3309        int result = diff_get_patch_id(options, sha1);
3310
3311        for (i = 0; i < q->nr; i++)
3312                diff_free_filepair(q->queue[i]);
3313
3314        free(q->queue);
3315        q->queue = NULL;
3316        q->nr = q->alloc = 0;
3317
3318        return result;
3319}
3320
3321static int is_summary_empty(const struct diff_queue_struct *q)
3322{
3323        int i;
3324
3325        for (i = 0; i < q->nr; i++) {
3326                const struct diff_filepair *p = q->queue[i];
3327
3328                switch (p->status) {
3329                case DIFF_STATUS_DELETED:
3330                case DIFF_STATUS_ADDED:
3331                case DIFF_STATUS_COPIED:
3332                case DIFF_STATUS_RENAMED:
3333                        return 0;
3334                default:
3335                        if (p->score)
3336                                return 0;
3337                        if (p->one->mode && p->two->mode &&
3338                            p->one->mode != p->two->mode)
3339                                return 0;
3340                        break;
3341                }
3342        }
3343        return 1;
3344}
3345
3346void diff_flush(struct diff_options *options)
3347{
3348        struct diff_queue_struct *q = &diff_queued_diff;
3349        int i, output_format = options->output_format;
3350        int separator = 0;
3351
3352        /*
3353         * Order: raw, stat, summary, patch
3354         * or:    name/name-status/checkdiff (other bits clear)
3355         */
3356        if (!q->nr)
3357                goto free_queue;
3358
3359        if (output_format & (DIFF_FORMAT_RAW |
3360                             DIFF_FORMAT_NAME |
3361                             DIFF_FORMAT_NAME_STATUS |
3362                             DIFF_FORMAT_CHECKDIFF)) {
3363                for (i = 0; i < q->nr; i++) {
3364                        struct diff_filepair *p = q->queue[i];
3365                        if (check_pair_status(p))
3366                                flush_one_pair(p, options);
3367                }
3368                separator++;
3369        }
3370
3371        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
3372                struct diffstat_t diffstat;
3373
3374                memset(&diffstat, 0, sizeof(struct diffstat_t));
3375                for (i = 0; i < q->nr; i++) {
3376                        struct diff_filepair *p = q->queue[i];
3377                        if (check_pair_status(p))
3378                                diff_flush_stat(p, options, &diffstat);
3379                }
3380                if (output_format & DIFF_FORMAT_NUMSTAT)
3381                        show_numstat(&diffstat, options);
3382                if (output_format & DIFF_FORMAT_DIFFSTAT)
3383                        show_stats(&diffstat, options);
3384                if (output_format & DIFF_FORMAT_SHORTSTAT)
3385                        show_shortstats(&diffstat, options);
3386                free_diffstat_info(&diffstat);
3387                separator++;
3388        }
3389        if (output_format & DIFF_FORMAT_DIRSTAT)
3390                show_dirstat(options);
3391
3392        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
3393                for (i = 0; i < q->nr; i++)
3394                        diff_summary(options->file, q->queue[i]);
3395                separator++;
3396        }
3397
3398        if (output_format & DIFF_FORMAT_PATCH) {
3399                if (separator) {
3400                        putc(options->line_termination, options->file);
3401                        if (options->stat_sep) {
3402                                /* attach patch instead of inline */
3403                                fputs(options->stat_sep, options->file);
3404                        }
3405                }
3406
3407                for (i = 0; i < q->nr; i++) {
3408                        struct diff_filepair *p = q->queue[i];
3409                        if (check_pair_status(p))
3410                                diff_flush_patch(p, options);
3411                }
3412        }
3413
3414        if (output_format & DIFF_FORMAT_CALLBACK)
3415                options->format_callback(q, options, options->format_callback_data);
3416
3417        for (i = 0; i < q->nr; i++)
3418                diff_free_filepair(q->queue[i]);
3419free_queue:
3420        free(q->queue);
3421        q->queue = NULL;
3422        q->nr = q->alloc = 0;
3423        if (options->close_file)
3424                fclose(options->file);
3425}
3426
3427static void diffcore_apply_filter(const char *filter)
3428{
3429        int i;
3430        struct diff_queue_struct *q = &diff_queued_diff;
3431        struct diff_queue_struct outq;
3432        outq.queue = NULL;
3433        outq.nr = outq.alloc = 0;
3434
3435        if (!filter)
3436                return;
3437
3438        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3439                int found;
3440                for (i = found = 0; !found && i < q->nr; i++) {
3441                        struct diff_filepair *p = q->queue[i];
3442                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3443                             ((p->score &&
3444                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3445                              (!p->score &&
3446                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3447                            ((p->status != DIFF_STATUS_MODIFIED) &&
3448                             strchr(filter, p->status)))
3449                                found++;
3450                }
3451                if (found)
3452                        return;
3453
3454                /* otherwise we will clear the whole queue
3455                 * by copying the empty outq at the end of this
3456                 * function, but first clear the current entries
3457                 * in the queue.
3458                 */
3459                for (i = 0; i < q->nr; i++)
3460                        diff_free_filepair(q->queue[i]);
3461        }
3462        else {
3463                /* Only the matching ones */
3464                for (i = 0; i < q->nr; i++) {
3465                        struct diff_filepair *p = q->queue[i];
3466
3467                        if (((p->status == DIFF_STATUS_MODIFIED) &&
3468                             ((p->score &&
3469                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3470                              (!p->score &&
3471                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3472                            ((p->status != DIFF_STATUS_MODIFIED) &&
3473                             strchr(filter, p->status)))
3474                                diff_q(&outq, p);
3475                        else
3476                                diff_free_filepair(p);
3477                }
3478        }
3479        free(q->queue);
3480        *q = outq;
3481}
3482
3483/* Check whether two filespecs with the same mode and size are identical */
3484static int diff_filespec_is_identical(struct diff_filespec *one,
3485                                      struct diff_filespec *two)
3486{
3487        if (S_ISGITLINK(one->mode))
3488                return 0;
3489        if (diff_populate_filespec(one, 0))
3490                return 0;
3491        if (diff_populate_filespec(two, 0))
3492                return 0;
3493        return !memcmp(one->data, two->data, one->size);
3494}
3495
3496static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3497{
3498        int i;
3499        struct diff_queue_struct *q = &diff_queued_diff;
3500        struct diff_queue_struct outq;
3501        outq.queue = NULL;
3502        outq.nr = outq.alloc = 0;
3503
3504        for (i = 0; i < q->nr; i++) {
3505                struct diff_filepair *p = q->queue[i];
3506
3507                /*
3508                 * 1. Entries that come from stat info dirtyness
3509                 *    always have both sides (iow, not create/delete),
3510                 *    one side of the object name is unknown, with
3511                 *    the same mode and size.  Keep the ones that
3512                 *    do not match these criteria.  They have real
3513                 *    differences.
3514                 *
3515                 * 2. At this point, the file is known to be modified,
3516                 *    with the same mode and size, and the object
3517                 *    name of one side is unknown.  Need to inspect
3518                 *    the identical contents.
3519                 */
3520                if (!DIFF_FILE_VALID(p->one) || /* (1) */
3521                    !DIFF_FILE_VALID(p->two) ||
3522                    (p->one->sha1_valid && p->two->sha1_valid) ||
3523                    (p->one->mode != p->two->mode) ||
3524                    diff_populate_filespec(p->one, 1) ||
3525                    diff_populate_filespec(p->two, 1) ||
3526                    (p->one->size != p->two->size) ||
3527                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3528                        diff_q(&outq, p);
3529                else {
3530                        /*
3531                         * The caller can subtract 1 from skip_stat_unmatch
3532                         * to determine how many paths were dirty only
3533                         * due to stat info mismatch.
3534                         */
3535                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3536                                diffopt->skip_stat_unmatch++;
3537                        diff_free_filepair(p);
3538                }
3539        }
3540        free(q->queue);
3541        *q = outq;
3542}
3543
3544void diffcore_std(struct diff_options *options)
3545{
3546        if (options->skip_stat_unmatch)
3547                diffcore_skip_stat_unmatch(options);
3548        if (options->break_opt != -1)
3549                diffcore_break(options->break_opt);
3550        if (options->detect_rename)
3551                diffcore_rename(options);
3552        if (options->break_opt != -1)
3553                diffcore_merge_broken();
3554        if (options->pickaxe)
3555                diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3556        if (options->orderfile)
3557                diffcore_order(options->orderfile);
3558        diff_resolve_rename_copy();
3559        diffcore_apply_filter(options->filter);
3560
3561        if (diff_queued_diff.nr)
3562                DIFF_OPT_SET(options, HAS_CHANGES);
3563        else
3564                DIFF_OPT_CLR(options, HAS_CHANGES);
3565}
3566
3567int diff_result_code(struct diff_options *opt, int status)
3568{
3569        int result = 0;
3570        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3571            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3572                return status;
3573        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3574            DIFF_OPT_TST(opt, HAS_CHANGES))
3575                result |= 01;
3576        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3577            DIFF_OPT_TST(opt, CHECK_FAILED))
3578                result |= 02;
3579        return result;
3580}
3581
3582void diff_addremove(struct diff_options *options,
3583                    int addremove, unsigned mode,
3584                    const unsigned char *sha1,
3585                    const char *concatpath)
3586{
3587        struct diff_filespec *one, *two;
3588
3589        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(mode))
3590                return;
3591
3592        /* This may look odd, but it is a preparation for
3593         * feeding "there are unchanged files which should
3594         * not produce diffs, but when you are doing copy
3595         * detection you would need them, so here they are"
3596         * entries to the diff-core.  They will be prefixed
3597         * with something like '=' or '*' (I haven't decided
3598         * which but should not make any difference).
3599         * Feeding the same new and old to diff_change()
3600         * also has the same effect.
3601         * Before the final output happens, they are pruned after
3602         * merged into rename/copy pairs as appropriate.
3603         */
3604        if (DIFF_OPT_TST(options, REVERSE_DIFF))
3605                addremove = (addremove == '+' ? '-' :
3606                             addremove == '-' ? '+' : addremove);
3607
3608        if (options->prefix &&
3609            strncmp(concatpath, options->prefix, options->prefix_length))
3610                return;
3611
3612        one = alloc_filespec(concatpath);
3613        two = alloc_filespec(concatpath);
3614
3615        if (addremove != '+')
3616                fill_filespec(one, sha1, mode);
3617        if (addremove != '-')
3618                fill_filespec(two, sha1, mode);
3619
3620        diff_queue(&diff_queued_diff, one, two);
3621        DIFF_OPT_SET(options, HAS_CHANGES);
3622}
3623
3624void diff_change(struct diff_options *options,
3625                 unsigned old_mode, unsigned new_mode,
3626                 const unsigned char *old_sha1,
3627                 const unsigned char *new_sha1,
3628                 const char *concatpath)
3629{
3630        struct diff_filespec *one, *two;
3631
3632        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(old_mode)
3633                        && S_ISGITLINK(new_mode))
3634                return;
3635
3636        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3637                unsigned tmp;
3638                const unsigned char *tmp_c;
3639                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3640                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3641        }
3642
3643        if (options->prefix &&
3644            strncmp(concatpath, options->prefix, options->prefix_length))
3645                return;
3646
3647        one = alloc_filespec(concatpath);
3648        two = alloc_filespec(concatpath);
3649        fill_filespec(one, old_sha1, old_mode);
3650        fill_filespec(two, new_sha1, new_mode);
3651
3652        diff_queue(&diff_queued_diff, one, two);
3653        DIFF_OPT_SET(options, HAS_CHANGES);
3654}
3655
3656void diff_unmerge(struct diff_options *options,
3657                  const char *path,
3658                  unsigned mode, const unsigned char *sha1)
3659{
3660        struct diff_filespec *one, *two;
3661
3662        if (options->prefix &&
3663            strncmp(path, options->prefix, options->prefix_length))
3664                return;
3665
3666        one = alloc_filespec(path);
3667        two = alloc_filespec(path);
3668        fill_filespec(one, sha1, mode);
3669        diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3670}
3671
3672static char *run_textconv(const char *pgm, struct diff_filespec *spec,
3673                size_t *outsize)
3674{
3675        struct diff_tempfile *temp;
3676        const char *argv[3];
3677        const char **arg = argv;
3678        struct child_process child;
3679        struct strbuf buf = STRBUF_INIT;
3680
3681        temp = prepare_temp_file(spec->path, spec);
3682        *arg++ = pgm;
3683        *arg++ = temp->name;
3684        *arg = NULL;
3685
3686        memset(&child, 0, sizeof(child));
3687        child.argv = argv;
3688        child.out = -1;
3689        if (start_command(&child) != 0 ||
3690            strbuf_read(&buf, child.out, 0) < 0 ||
3691            finish_command(&child) != 0) {
3692                strbuf_release(&buf);
3693                remove_tempfile();
3694                error("error running textconv command '%s'", pgm);
3695                return NULL;
3696        }
3697        remove_tempfile();
3698
3699        return strbuf_detach(&buf, outsize);
3700}