diff.con commit add tests for various blame formats (838466b)
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "quote.h"
   6#include "diff.h"
   7#include "diffcore.h"
   8#include "delta.h"
   9#include "xdiff-interface.h"
  10#include "color.h"
  11#include "attr.h"
  12#include "run-command.h"
  13#include "utf8.h"
  14#include "userdiff.h"
  15#include "sigchain.h"
  16#include "submodule.h"
  17#include "ll-merge.h"
  18
  19#ifdef NO_FAST_WORKING_DIRECTORY
  20#define FAST_WORKING_DIRECTORY 0
  21#else
  22#define FAST_WORKING_DIRECTORY 1
  23#endif
  24
  25static int diff_detect_rename_default;
  26static int diff_rename_limit_default = 400;
  27static int diff_suppress_blank_empty;
  28int diff_use_color_default = -1;
  29static const char *diff_word_regex_cfg;
  30static const char *external_diff_cmd_cfg;
  31int diff_auto_refresh_index = 1;
  32static int diff_mnemonic_prefix;
  33static int diff_no_prefix;
  34static struct diff_options default_diff_options;
  35
  36static char diff_colors[][COLOR_MAXLEN] = {
  37        GIT_COLOR_RESET,
  38        GIT_COLOR_NORMAL,       /* PLAIN */
  39        GIT_COLOR_BOLD,         /* METAINFO */
  40        GIT_COLOR_CYAN,         /* FRAGINFO */
  41        GIT_COLOR_RED,          /* OLD */
  42        GIT_COLOR_GREEN,        /* NEW */
  43        GIT_COLOR_YELLOW,       /* COMMIT */
  44        GIT_COLOR_BG_RED,       /* WHITESPACE */
  45        GIT_COLOR_NORMAL,       /* FUNCINFO */
  46};
  47
  48static int parse_diff_color_slot(const char *var, int ofs)
  49{
  50        if (!strcasecmp(var+ofs, "plain"))
  51                return DIFF_PLAIN;
  52        if (!strcasecmp(var+ofs, "meta"))
  53                return DIFF_METAINFO;
  54        if (!strcasecmp(var+ofs, "frag"))
  55                return DIFF_FRAGINFO;
  56        if (!strcasecmp(var+ofs, "old"))
  57                return DIFF_FILE_OLD;
  58        if (!strcasecmp(var+ofs, "new"))
  59                return DIFF_FILE_NEW;
  60        if (!strcasecmp(var+ofs, "commit"))
  61                return DIFF_COMMIT;
  62        if (!strcasecmp(var+ofs, "whitespace"))
  63                return DIFF_WHITESPACE;
  64        if (!strcasecmp(var+ofs, "func"))
  65                return DIFF_FUNCINFO;
  66        return -1;
  67}
  68
  69static int git_config_rename(const char *var, const char *value)
  70{
  71        if (!value)
  72                return DIFF_DETECT_RENAME;
  73        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
  74                return  DIFF_DETECT_COPY;
  75        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
  76}
  77
  78/*
  79 * These are to give UI layer defaults.
  80 * The core-level commands such as git-diff-files should
  81 * never be affected by the setting of diff.renames
  82 * the user happens to have in the configuration file.
  83 */
  84int git_diff_ui_config(const char *var, const char *value, void *cb)
  85{
  86        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
  87                diff_use_color_default = git_config_colorbool(var, value, -1);
  88                return 0;
  89        }
  90        if (!strcmp(var, "diff.renames")) {
  91                diff_detect_rename_default = git_config_rename(var, value);
  92                return 0;
  93        }
  94        if (!strcmp(var, "diff.autorefreshindex")) {
  95                diff_auto_refresh_index = git_config_bool(var, value);
  96                return 0;
  97        }
  98        if (!strcmp(var, "diff.mnemonicprefix")) {
  99                diff_mnemonic_prefix = git_config_bool(var, value);
 100                return 0;
 101        }
 102        if (!strcmp(var, "diff.noprefix")) {
 103                diff_no_prefix = git_config_bool(var, value);
 104                return 0;
 105        }
 106        if (!strcmp(var, "diff.external"))
 107                return git_config_string(&external_diff_cmd_cfg, var, value);
 108        if (!strcmp(var, "diff.wordregex"))
 109                return git_config_string(&diff_word_regex_cfg, var, value);
 110
 111        if (!strcmp(var, "diff.ignoresubmodules"))
 112                handle_ignore_submodules_arg(&default_diff_options, value);
 113
 114        return git_diff_basic_config(var, value, cb);
 115}
 116
 117int git_diff_basic_config(const char *var, const char *value, void *cb)
 118{
 119        if (!strcmp(var, "diff.renamelimit")) {
 120                diff_rename_limit_default = git_config_int(var, value);
 121                return 0;
 122        }
 123
 124        switch (userdiff_config(var, value)) {
 125                case 0: break;
 126                case -1: return -1;
 127                default: return 0;
 128        }
 129
 130        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 131                int slot = parse_diff_color_slot(var, 11);
 132                if (slot < 0)
 133                        return 0;
 134                if (!value)
 135                        return config_error_nonbool(var);
 136                color_parse(value, var, diff_colors[slot]);
 137                return 0;
 138        }
 139
 140        /* like GNU diff's --suppress-blank-empty option  */
 141        if (!strcmp(var, "diff.suppressblankempty") ||
 142                        /* for backwards compatibility */
 143                        !strcmp(var, "diff.suppress-blank-empty")) {
 144                diff_suppress_blank_empty = git_config_bool(var, value);
 145                return 0;
 146        }
 147
 148        if (!prefixcmp(var, "submodule."))
 149                return parse_submodule_config_option(var, value);
 150
 151        return git_color_default_config(var, value, cb);
 152}
 153
 154static char *quote_two(const char *one, const char *two)
 155{
 156        int need_one = quote_c_style(one, NULL, NULL, 1);
 157        int need_two = quote_c_style(two, NULL, NULL, 1);
 158        struct strbuf res = STRBUF_INIT;
 159
 160        if (need_one + need_two) {
 161                strbuf_addch(&res, '"');
 162                quote_c_style(one, &res, NULL, 1);
 163                quote_c_style(two, &res, NULL, 1);
 164                strbuf_addch(&res, '"');
 165        } else {
 166                strbuf_addstr(&res, one);
 167                strbuf_addstr(&res, two);
 168        }
 169        return strbuf_detach(&res, NULL);
 170}
 171
 172static const char *external_diff(void)
 173{
 174        static const char *external_diff_cmd = NULL;
 175        static int done_preparing = 0;
 176
 177        if (done_preparing)
 178                return external_diff_cmd;
 179        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 180        if (!external_diff_cmd)
 181                external_diff_cmd = external_diff_cmd_cfg;
 182        done_preparing = 1;
 183        return external_diff_cmd;
 184}
 185
 186static struct diff_tempfile {
 187        const char *name; /* filename external diff should read from */
 188        char hex[41];
 189        char mode[10];
 190        char tmp_path[PATH_MAX];
 191} diff_temp[2];
 192
 193typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 194
 195struct emit_callback {
 196        int color_diff;
 197        unsigned ws_rule;
 198        int blank_at_eof_in_preimage;
 199        int blank_at_eof_in_postimage;
 200        int lno_in_preimage;
 201        int lno_in_postimage;
 202        sane_truncate_fn truncate;
 203        const char **label_path;
 204        struct diff_words_data *diff_words;
 205        struct diff_options *opt;
 206        int *found_changesp;
 207        struct strbuf *header;
 208};
 209
 210static int count_lines(const char *data, int size)
 211{
 212        int count, ch, completely_empty = 1, nl_just_seen = 0;
 213        count = 0;
 214        while (0 < size--) {
 215                ch = *data++;
 216                if (ch == '\n') {
 217                        count++;
 218                        nl_just_seen = 1;
 219                        completely_empty = 0;
 220                }
 221                else {
 222                        nl_just_seen = 0;
 223                        completely_empty = 0;
 224                }
 225        }
 226        if (completely_empty)
 227                return 0;
 228        if (!nl_just_seen)
 229                count++; /* no trailing newline */
 230        return count;
 231}
 232
 233static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 234{
 235        if (!DIFF_FILE_VALID(one)) {
 236                mf->ptr = (char *)""; /* does not matter */
 237                mf->size = 0;
 238                return 0;
 239        }
 240        else if (diff_populate_filespec(one, 0))
 241                return -1;
 242
 243        mf->ptr = one->data;
 244        mf->size = one->size;
 245        return 0;
 246}
 247
 248/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 249static unsigned long diff_filespec_size(struct diff_filespec *one)
 250{
 251        if (!DIFF_FILE_VALID(one))
 252                return 0;
 253        diff_populate_filespec(one, 1);
 254        return one->size;
 255}
 256
 257static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 258{
 259        char *ptr = mf->ptr;
 260        long size = mf->size;
 261        int cnt = 0;
 262
 263        if (!size)
 264                return cnt;
 265        ptr += size - 1; /* pointing at the very end */
 266        if (*ptr != '\n')
 267                ; /* incomplete line */
 268        else
 269                ptr--; /* skip the last LF */
 270        while (mf->ptr < ptr) {
 271                char *prev_eol;
 272                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 273                        if (*prev_eol == '\n')
 274                                break;
 275                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 276                        break;
 277                cnt++;
 278                ptr = prev_eol - 1;
 279        }
 280        return cnt;
 281}
 282
 283static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 284                               struct emit_callback *ecbdata)
 285{
 286        int l1, l2, at;
 287        unsigned ws_rule = ecbdata->ws_rule;
 288        l1 = count_trailing_blank(mf1, ws_rule);
 289        l2 = count_trailing_blank(mf2, ws_rule);
 290        if (l2 <= l1) {
 291                ecbdata->blank_at_eof_in_preimage = 0;
 292                ecbdata->blank_at_eof_in_postimage = 0;
 293                return;
 294        }
 295        at = count_lines(mf1->ptr, mf1->size);
 296        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 297
 298        at = count_lines(mf2->ptr, mf2->size);
 299        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 300}
 301
 302static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 303                        int first, const char *line, int len)
 304{
 305        int has_trailing_newline, has_trailing_carriage_return;
 306        int nofirst;
 307        FILE *file = o->file;
 308
 309        if (o->output_prefix) {
 310                struct strbuf *msg = NULL;
 311                msg = o->output_prefix(o, o->output_prefix_data);
 312                assert(msg);
 313                fwrite(msg->buf, msg->len, 1, file);
 314        }
 315
 316        if (len == 0) {
 317                has_trailing_newline = (first == '\n');
 318                has_trailing_carriage_return = (!has_trailing_newline &&
 319                                                (first == '\r'));
 320                nofirst = has_trailing_newline || has_trailing_carriage_return;
 321        } else {
 322                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 323                if (has_trailing_newline)
 324                        len--;
 325                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 326                if (has_trailing_carriage_return)
 327                        len--;
 328                nofirst = 0;
 329        }
 330
 331        if (len || !nofirst) {
 332                fputs(set, file);
 333                if (!nofirst)
 334                        fputc(first, file);
 335                fwrite(line, len, 1, file);
 336                fputs(reset, file);
 337        }
 338        if (has_trailing_carriage_return)
 339                fputc('\r', file);
 340        if (has_trailing_newline)
 341                fputc('\n', file);
 342}
 343
 344static void emit_line(struct diff_options *o, const char *set, const char *reset,
 345                      const char *line, int len)
 346{
 347        emit_line_0(o, set, reset, line[0], line+1, len-1);
 348}
 349
 350static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 351{
 352        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 353              ecbdata->blank_at_eof_in_preimage &&
 354              ecbdata->blank_at_eof_in_postimage &&
 355              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 356              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 357                return 0;
 358        return ws_blank_line(line, len, ecbdata->ws_rule);
 359}
 360
 361static void emit_add_line(const char *reset,
 362                          struct emit_callback *ecbdata,
 363                          const char *line, int len)
 364{
 365        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 366        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 367
 368        if (!*ws)
 369                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 370        else if (new_blank_line_at_eof(ecbdata, line, len))
 371                /* Blank line at EOF - paint '+' as well */
 372                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 373        else {
 374                /* Emit just the prefix, then the rest. */
 375                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 376                ws_check_emit(line, len, ecbdata->ws_rule,
 377                              ecbdata->opt->file, set, reset, ws);
 378        }
 379}
 380
 381static void emit_hunk_header(struct emit_callback *ecbdata,
 382                             const char *line, int len)
 383{
 384        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 385        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 386        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 387        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 388        static const char atat[2] = { '@', '@' };
 389        const char *cp, *ep;
 390        struct strbuf msgbuf = STRBUF_INIT;
 391        int org_len = len;
 392        int i = 1;
 393
 394        /*
 395         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 396         * it always is at least 10 bytes long.
 397         */
 398        if (len < 10 ||
 399            memcmp(line, atat, 2) ||
 400            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 401                emit_line(ecbdata->opt, plain, reset, line, len);
 402                return;
 403        }
 404        ep += 2; /* skip over @@ */
 405
 406        /* The hunk header in fraginfo color */
 407        strbuf_add(&msgbuf, frag, strlen(frag));
 408        strbuf_add(&msgbuf, line, ep - line);
 409        strbuf_add(&msgbuf, reset, strlen(reset));
 410
 411        /*
 412         * trailing "\r\n"
 413         */
 414        for ( ; i < 3; i++)
 415                if (line[len - i] == '\r' || line[len - i] == '\n')
 416                        len--;
 417
 418        /* blank before the func header */
 419        for (cp = ep; ep - line < len; ep++)
 420                if (*ep != ' ' && *ep != '\t')
 421                        break;
 422        if (ep != cp) {
 423                strbuf_add(&msgbuf, plain, strlen(plain));
 424                strbuf_add(&msgbuf, cp, ep - cp);
 425                strbuf_add(&msgbuf, reset, strlen(reset));
 426        }
 427
 428        if (ep < line + len) {
 429                strbuf_add(&msgbuf, func, strlen(func));
 430                strbuf_add(&msgbuf, ep, line + len - ep);
 431                strbuf_add(&msgbuf, reset, strlen(reset));
 432        }
 433
 434        strbuf_add(&msgbuf, line + len, org_len - len);
 435        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 436        strbuf_release(&msgbuf);
 437}
 438
 439static struct diff_tempfile *claim_diff_tempfile(void) {
 440        int i;
 441        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 442                if (!diff_temp[i].name)
 443                        return diff_temp + i;
 444        die("BUG: diff is failing to clean up its tempfiles");
 445}
 446
 447static int remove_tempfile_installed;
 448
 449static void remove_tempfile(void)
 450{
 451        int i;
 452        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 453                if (diff_temp[i].name == diff_temp[i].tmp_path)
 454                        unlink_or_warn(diff_temp[i].name);
 455                diff_temp[i].name = NULL;
 456        }
 457}
 458
 459static void remove_tempfile_on_signal(int signo)
 460{
 461        remove_tempfile();
 462        sigchain_pop(signo);
 463        raise(signo);
 464}
 465
 466static void print_line_count(FILE *file, int count)
 467{
 468        switch (count) {
 469        case 0:
 470                fprintf(file, "0,0");
 471                break;
 472        case 1:
 473                fprintf(file, "1");
 474                break;
 475        default:
 476                fprintf(file, "1,%d", count);
 477                break;
 478        }
 479}
 480
 481static void emit_rewrite_lines(struct emit_callback *ecb,
 482                               int prefix, const char *data, int size)
 483{
 484        const char *endp = NULL;
 485        static const char *nneof = " No newline at end of file\n";
 486        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 487        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 488
 489        while (0 < size) {
 490                int len;
 491
 492                endp = memchr(data, '\n', size);
 493                len = endp ? (endp - data + 1) : size;
 494                if (prefix != '+') {
 495                        ecb->lno_in_preimage++;
 496                        emit_line_0(ecb->opt, old, reset, '-',
 497                                    data, len);
 498                } else {
 499                        ecb->lno_in_postimage++;
 500                        emit_add_line(reset, ecb, data, len);
 501                }
 502                size -= len;
 503                data += len;
 504        }
 505        if (!endp) {
 506                const char *plain = diff_get_color(ecb->color_diff,
 507                                                   DIFF_PLAIN);
 508                emit_line_0(ecb->opt, plain, reset, '\\',
 509                            nneof, strlen(nneof));
 510        }
 511}
 512
 513static void emit_rewrite_diff(const char *name_a,
 514                              const char *name_b,
 515                              struct diff_filespec *one,
 516                              struct diff_filespec *two,
 517                              struct userdiff_driver *textconv_one,
 518                              struct userdiff_driver *textconv_two,
 519                              struct diff_options *o)
 520{
 521        int lc_a, lc_b;
 522        int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
 523        const char *name_a_tab, *name_b_tab;
 524        const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
 525        const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
 526        const char *reset = diff_get_color(color_diff, DIFF_RESET);
 527        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 528        const char *a_prefix, *b_prefix;
 529        char *data_one, *data_two;
 530        size_t size_one, size_two;
 531        struct emit_callback ecbdata;
 532        char *line_prefix = "";
 533        struct strbuf *msgbuf;
 534
 535        if (o && o->output_prefix) {
 536                msgbuf = o->output_prefix(o, o->output_prefix_data);
 537                line_prefix = msgbuf->buf;
 538        }
 539
 540        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 541                a_prefix = o->b_prefix;
 542                b_prefix = o->a_prefix;
 543        } else {
 544                a_prefix = o->a_prefix;
 545                b_prefix = o->b_prefix;
 546        }
 547
 548        name_a += (*name_a == '/');
 549        name_b += (*name_b == '/');
 550        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 551        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 552
 553        strbuf_reset(&a_name);
 554        strbuf_reset(&b_name);
 555        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 556        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 557
 558        size_one = fill_textconv(textconv_one, one, &data_one);
 559        size_two = fill_textconv(textconv_two, two, &data_two);
 560
 561        memset(&ecbdata, 0, sizeof(ecbdata));
 562        ecbdata.color_diff = color_diff;
 563        ecbdata.found_changesp = &o->found_changes;
 564        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 565        ecbdata.opt = o;
 566        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 567                mmfile_t mf1, mf2;
 568                mf1.ptr = (char *)data_one;
 569                mf2.ptr = (char *)data_two;
 570                mf1.size = size_one;
 571                mf2.size = size_two;
 572                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 573        }
 574        ecbdata.lno_in_preimage = 1;
 575        ecbdata.lno_in_postimage = 1;
 576
 577        lc_a = count_lines(data_one, size_one);
 578        lc_b = count_lines(data_two, size_two);
 579        fprintf(o->file,
 580                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 581                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 582                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 583                line_prefix, fraginfo);
 584        if (!o->irreversible_delete)
 585                print_line_count(o->file, lc_a);
 586        else
 587                fprintf(o->file, "?,?");
 588        fprintf(o->file, " +");
 589        print_line_count(o->file, lc_b);
 590        fprintf(o->file, " @@%s\n", reset);
 591        if (lc_a && !o->irreversible_delete)
 592                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 593        if (lc_b)
 594                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 595        if (textconv_one)
 596                free((char *)data_one);
 597        if (textconv_two)
 598                free((char *)data_two);
 599}
 600
 601struct diff_words_buffer {
 602        mmfile_t text;
 603        long alloc;
 604        struct diff_words_orig {
 605                const char *begin, *end;
 606        } *orig;
 607        int orig_nr, orig_alloc;
 608};
 609
 610static void diff_words_append(char *line, unsigned long len,
 611                struct diff_words_buffer *buffer)
 612{
 613        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 614        line++;
 615        len--;
 616        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 617        buffer->text.size += len;
 618        buffer->text.ptr[buffer->text.size] = '\0';
 619}
 620
 621struct diff_words_style_elem {
 622        const char *prefix;
 623        const char *suffix;
 624        const char *color; /* NULL; filled in by the setup code if
 625                            * color is enabled */
 626};
 627
 628struct diff_words_style {
 629        enum diff_words_type type;
 630        struct diff_words_style_elem new, old, ctx;
 631        const char *newline;
 632};
 633
 634static struct diff_words_style diff_words_styles[] = {
 635        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 636        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 637        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 638};
 639
 640struct diff_words_data {
 641        struct diff_words_buffer minus, plus;
 642        const char *current_plus;
 643        int last_minus;
 644        struct diff_options *opt;
 645        regex_t *word_regex;
 646        enum diff_words_type type;
 647        struct diff_words_style *style;
 648};
 649
 650static int fn_out_diff_words_write_helper(FILE *fp,
 651                                          struct diff_words_style_elem *st_el,
 652                                          const char *newline,
 653                                          size_t count, const char *buf,
 654                                          const char *line_prefix)
 655{
 656        int print = 0;
 657
 658        while (count) {
 659                char *p = memchr(buf, '\n', count);
 660                if (print)
 661                        fputs(line_prefix, fp);
 662                if (p != buf) {
 663                        if (st_el->color && fputs(st_el->color, fp) < 0)
 664                                return -1;
 665                        if (fputs(st_el->prefix, fp) < 0 ||
 666                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 667                            fputs(st_el->suffix, fp) < 0)
 668                                return -1;
 669                        if (st_el->color && *st_el->color
 670                            && fputs(GIT_COLOR_RESET, fp) < 0)
 671                                return -1;
 672                }
 673                if (!p)
 674                        return 0;
 675                if (fputs(newline, fp) < 0)
 676                        return -1;
 677                count -= p + 1 - buf;
 678                buf = p + 1;
 679                print = 1;
 680        }
 681        return 0;
 682}
 683
 684/*
 685 * '--color-words' algorithm can be described as:
 686 *
 687 *   1. collect a the minus/plus lines of a diff hunk, divided into
 688 *      minus-lines and plus-lines;
 689 *
 690 *   2. break both minus-lines and plus-lines into words and
 691 *      place them into two mmfile_t with one word for each line;
 692 *
 693 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 694 *
 695 * And for the common parts of the both file, we output the plus side text.
 696 * diff_words->current_plus is used to trace the current position of the plus file
 697 * which printed. diff_words->last_minus is used to trace the last minus word
 698 * printed.
 699 *
 700 * For '--graph' to work with '--color-words', we need to output the graph prefix
 701 * on each line of color words output. Generally, there are two conditions on
 702 * which we should output the prefix.
 703 *
 704 *   1. diff_words->last_minus == 0 &&
 705 *      diff_words->current_plus == diff_words->plus.text.ptr
 706 *
 707 *      that is: the plus text must start as a new line, and if there is no minus
 708 *      word printed, a graph prefix must be printed.
 709 *
 710 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 711 *      *(diff_words->current_plus - 1) == '\n'
 712 *
 713 *      that is: a graph prefix must be printed following a '\n'
 714 */
 715static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 716{
 717        if ((diff_words->last_minus == 0 &&
 718                diff_words->current_plus == diff_words->plus.text.ptr) ||
 719                (diff_words->current_plus > diff_words->plus.text.ptr &&
 720                *(diff_words->current_plus - 1) == '\n')) {
 721                return 1;
 722        } else {
 723                return 0;
 724        }
 725}
 726
 727static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 728{
 729        struct diff_words_data *diff_words = priv;
 730        struct diff_words_style *style = diff_words->style;
 731        int minus_first, minus_len, plus_first, plus_len;
 732        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 733        struct diff_options *opt = diff_words->opt;
 734        struct strbuf *msgbuf;
 735        char *line_prefix = "";
 736
 737        if (line[0] != '@' || parse_hunk_header(line, len,
 738                        &minus_first, &minus_len, &plus_first, &plus_len))
 739                return;
 740
 741        assert(opt);
 742        if (opt->output_prefix) {
 743                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 744                line_prefix = msgbuf->buf;
 745        }
 746
 747        /* POSIX requires that first be decremented by one if len == 0... */
 748        if (minus_len) {
 749                minus_begin = diff_words->minus.orig[minus_first].begin;
 750                minus_end =
 751                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 752        } else
 753                minus_begin = minus_end =
 754                        diff_words->minus.orig[minus_first].end;
 755
 756        if (plus_len) {
 757                plus_begin = diff_words->plus.orig[plus_first].begin;
 758                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 759        } else
 760                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 761
 762        if (color_words_output_graph_prefix(diff_words)) {
 763                fputs(line_prefix, diff_words->opt->file);
 764        }
 765        if (diff_words->current_plus != plus_begin) {
 766                fn_out_diff_words_write_helper(diff_words->opt->file,
 767                                &style->ctx, style->newline,
 768                                plus_begin - diff_words->current_plus,
 769                                diff_words->current_plus, line_prefix);
 770                if (*(plus_begin - 1) == '\n')
 771                        fputs(line_prefix, diff_words->opt->file);
 772        }
 773        if (minus_begin != minus_end) {
 774                fn_out_diff_words_write_helper(diff_words->opt->file,
 775                                &style->old, style->newline,
 776                                minus_end - minus_begin, minus_begin,
 777                                line_prefix);
 778        }
 779        if (plus_begin != plus_end) {
 780                fn_out_diff_words_write_helper(diff_words->opt->file,
 781                                &style->new, style->newline,
 782                                plus_end - plus_begin, plus_begin,
 783                                line_prefix);
 784        }
 785
 786        diff_words->current_plus = plus_end;
 787        diff_words->last_minus = minus_first;
 788}
 789
 790/* This function starts looking at *begin, and returns 0 iff a word was found. */
 791static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 792                int *begin, int *end)
 793{
 794        if (word_regex && *begin < buffer->size) {
 795                regmatch_t match[1];
 796                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 797                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 798                                        '\n', match[0].rm_eo - match[0].rm_so);
 799                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 800                        *begin += match[0].rm_so;
 801                        return *begin >= *end;
 802                }
 803                return -1;
 804        }
 805
 806        /* find the next word */
 807        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 808                (*begin)++;
 809        if (*begin >= buffer->size)
 810                return -1;
 811
 812        /* find the end of the word */
 813        *end = *begin + 1;
 814        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 815                (*end)++;
 816
 817        return 0;
 818}
 819
 820/*
 821 * This function splits the words in buffer->text, stores the list with
 822 * newline separator into out, and saves the offsets of the original words
 823 * in buffer->orig.
 824 */
 825static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 826                regex_t *word_regex)
 827{
 828        int i, j;
 829        long alloc = 0;
 830
 831        out->size = 0;
 832        out->ptr = NULL;
 833
 834        /* fake an empty "0th" word */
 835        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 836        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 837        buffer->orig_nr = 1;
 838
 839        for (i = 0; i < buffer->text.size; i++) {
 840                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 841                        return;
 842
 843                /* store original boundaries */
 844                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 845                                buffer->orig_alloc);
 846                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 847                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 848                buffer->orig_nr++;
 849
 850                /* store one word */
 851                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 852                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 853                out->ptr[out->size + j - i] = '\n';
 854                out->size += j - i + 1;
 855
 856                i = j - 1;
 857        }
 858}
 859
 860/* this executes the word diff on the accumulated buffers */
 861static void diff_words_show(struct diff_words_data *diff_words)
 862{
 863        xpparam_t xpp;
 864        xdemitconf_t xecfg;
 865        mmfile_t minus, plus;
 866        struct diff_words_style *style = diff_words->style;
 867
 868        struct diff_options *opt = diff_words->opt;
 869        struct strbuf *msgbuf;
 870        char *line_prefix = "";
 871
 872        assert(opt);
 873        if (opt->output_prefix) {
 874                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 875                line_prefix = msgbuf->buf;
 876        }
 877
 878        /* special case: only removal */
 879        if (!diff_words->plus.text.size) {
 880                fputs(line_prefix, diff_words->opt->file);
 881                fn_out_diff_words_write_helper(diff_words->opt->file,
 882                        &style->old, style->newline,
 883                        diff_words->minus.text.size,
 884                        diff_words->minus.text.ptr, line_prefix);
 885                diff_words->minus.text.size = 0;
 886                return;
 887        }
 888
 889        diff_words->current_plus = diff_words->plus.text.ptr;
 890        diff_words->last_minus = 0;
 891
 892        memset(&xpp, 0, sizeof(xpp));
 893        memset(&xecfg, 0, sizeof(xecfg));
 894        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 895        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 896        xpp.flags = 0;
 897        /* as only the hunk header will be parsed, we need a 0-context */
 898        xecfg.ctxlen = 0;
 899        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 900                      &xpp, &xecfg);
 901        free(minus.ptr);
 902        free(plus.ptr);
 903        if (diff_words->current_plus != diff_words->plus.text.ptr +
 904                        diff_words->plus.text.size) {
 905                if (color_words_output_graph_prefix(diff_words))
 906                        fputs(line_prefix, diff_words->opt->file);
 907                fn_out_diff_words_write_helper(diff_words->opt->file,
 908                        &style->ctx, style->newline,
 909                        diff_words->plus.text.ptr + diff_words->plus.text.size
 910                        - diff_words->current_plus, diff_words->current_plus,
 911                        line_prefix);
 912        }
 913        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 914}
 915
 916/* In "color-words" mode, show word-diff of words accumulated in the buffer */
 917static void diff_words_flush(struct emit_callback *ecbdata)
 918{
 919        if (ecbdata->diff_words->minus.text.size ||
 920            ecbdata->diff_words->plus.text.size)
 921                diff_words_show(ecbdata->diff_words);
 922}
 923
 924static void free_diff_words_data(struct emit_callback *ecbdata)
 925{
 926        if (ecbdata->diff_words) {
 927                diff_words_flush(ecbdata);
 928                free (ecbdata->diff_words->minus.text.ptr);
 929                free (ecbdata->diff_words->minus.orig);
 930                free (ecbdata->diff_words->plus.text.ptr);
 931                free (ecbdata->diff_words->plus.orig);
 932                if (ecbdata->diff_words->word_regex) {
 933                        regfree(ecbdata->diff_words->word_regex);
 934                        free(ecbdata->diff_words->word_regex);
 935                }
 936                free(ecbdata->diff_words);
 937                ecbdata->diff_words = NULL;
 938        }
 939}
 940
 941const char *diff_get_color(int diff_use_color, enum color_diff ix)
 942{
 943        if (diff_use_color)
 944                return diff_colors[ix];
 945        return "";
 946}
 947
 948static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
 949{
 950        const char *cp;
 951        unsigned long allot;
 952        size_t l = len;
 953
 954        if (ecb->truncate)
 955                return ecb->truncate(line, len);
 956        cp = line;
 957        allot = l;
 958        while (0 < l) {
 959                (void) utf8_width(&cp, &l);
 960                if (!cp)
 961                        break; /* truncated in the middle? */
 962        }
 963        return allot - l;
 964}
 965
 966static void find_lno(const char *line, struct emit_callback *ecbdata)
 967{
 968        const char *p;
 969        ecbdata->lno_in_preimage = 0;
 970        ecbdata->lno_in_postimage = 0;
 971        p = strchr(line, '-');
 972        if (!p)
 973                return; /* cannot happen */
 974        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
 975        p = strchr(p, '+');
 976        if (!p)
 977                return; /* cannot happen */
 978        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
 979}
 980
 981static void fn_out_consume(void *priv, char *line, unsigned long len)
 982{
 983        struct emit_callback *ecbdata = priv;
 984        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
 985        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 986        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 987        struct diff_options *o = ecbdata->opt;
 988        char *line_prefix = "";
 989        struct strbuf *msgbuf;
 990
 991        if (o && o->output_prefix) {
 992                msgbuf = o->output_prefix(o, o->output_prefix_data);
 993                line_prefix = msgbuf->buf;
 994        }
 995
 996        if (ecbdata->header) {
 997                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
 998                strbuf_reset(ecbdata->header);
 999                ecbdata->header = NULL;
1000        }
1001        *(ecbdata->found_changesp) = 1;
1002
1003        if (ecbdata->label_path[0]) {
1004                const char *name_a_tab, *name_b_tab;
1005
1006                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1007                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1008
1009                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1010                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1011                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1012                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1013                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1014        }
1015
1016        if (diff_suppress_blank_empty
1017            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1018                line[0] = '\n';
1019                len = 1;
1020        }
1021
1022        if (line[0] == '@') {
1023                if (ecbdata->diff_words)
1024                        diff_words_flush(ecbdata);
1025                len = sane_truncate_line(ecbdata, line, len);
1026                find_lno(line, ecbdata);
1027                emit_hunk_header(ecbdata, line, len);
1028                if (line[len-1] != '\n')
1029                        putc('\n', ecbdata->opt->file);
1030                return;
1031        }
1032
1033        if (len < 1) {
1034                emit_line(ecbdata->opt, reset, reset, line, len);
1035                if (ecbdata->diff_words
1036                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1037                        fputs("~\n", ecbdata->opt->file);
1038                return;
1039        }
1040
1041        if (ecbdata->diff_words) {
1042                if (line[0] == '-') {
1043                        diff_words_append(line, len,
1044                                          &ecbdata->diff_words->minus);
1045                        return;
1046                } else if (line[0] == '+') {
1047                        diff_words_append(line, len,
1048                                          &ecbdata->diff_words->plus);
1049                        return;
1050                }
1051                diff_words_flush(ecbdata);
1052                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1053                        emit_line(ecbdata->opt, plain, reset, line, len);
1054                        fputs("~\n", ecbdata->opt->file);
1055                } else {
1056                        /* don't print the prefix character */
1057                        emit_line(ecbdata->opt, plain, reset, line+1, len-1);
1058                }
1059                return;
1060        }
1061
1062        if (line[0] != '+') {
1063                const char *color =
1064                        diff_get_color(ecbdata->color_diff,
1065                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1066                ecbdata->lno_in_preimage++;
1067                if (line[0] == ' ')
1068                        ecbdata->lno_in_postimage++;
1069                emit_line(ecbdata->opt, color, reset, line, len);
1070        } else {
1071                ecbdata->lno_in_postimage++;
1072                emit_add_line(reset, ecbdata, line + 1, len - 1);
1073        }
1074}
1075
1076static char *pprint_rename(const char *a, const char *b)
1077{
1078        const char *old = a;
1079        const char *new = b;
1080        struct strbuf name = STRBUF_INIT;
1081        int pfx_length, sfx_length;
1082        int len_a = strlen(a);
1083        int len_b = strlen(b);
1084        int a_midlen, b_midlen;
1085        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1086        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1087
1088        if (qlen_a || qlen_b) {
1089                quote_c_style(a, &name, NULL, 0);
1090                strbuf_addstr(&name, " => ");
1091                quote_c_style(b, &name, NULL, 0);
1092                return strbuf_detach(&name, NULL);
1093        }
1094
1095        /* Find common prefix */
1096        pfx_length = 0;
1097        while (*old && *new && *old == *new) {
1098                if (*old == '/')
1099                        pfx_length = old - a + 1;
1100                old++;
1101                new++;
1102        }
1103
1104        /* Find common suffix */
1105        old = a + len_a;
1106        new = b + len_b;
1107        sfx_length = 0;
1108        while (a <= old && b <= new && *old == *new) {
1109                if (*old == '/')
1110                        sfx_length = len_a - (old - a);
1111                old--;
1112                new--;
1113        }
1114
1115        /*
1116         * pfx{mid-a => mid-b}sfx
1117         * {pfx-a => pfx-b}sfx
1118         * pfx{sfx-a => sfx-b}
1119         * name-a => name-b
1120         */
1121        a_midlen = len_a - pfx_length - sfx_length;
1122        b_midlen = len_b - pfx_length - sfx_length;
1123        if (a_midlen < 0)
1124                a_midlen = 0;
1125        if (b_midlen < 0)
1126                b_midlen = 0;
1127
1128        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1129        if (pfx_length + sfx_length) {
1130                strbuf_add(&name, a, pfx_length);
1131                strbuf_addch(&name, '{');
1132        }
1133        strbuf_add(&name, a + pfx_length, a_midlen);
1134        strbuf_addstr(&name, " => ");
1135        strbuf_add(&name, b + pfx_length, b_midlen);
1136        if (pfx_length + sfx_length) {
1137                strbuf_addch(&name, '}');
1138                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1139        }
1140        return strbuf_detach(&name, NULL);
1141}
1142
1143struct diffstat_t {
1144        int nr;
1145        int alloc;
1146        struct diffstat_file {
1147                char *from_name;
1148                char *name;
1149                char *print_name;
1150                unsigned is_unmerged:1;
1151                unsigned is_binary:1;
1152                unsigned is_renamed:1;
1153                uintmax_t added, deleted;
1154        } **files;
1155};
1156
1157static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1158                                          const char *name_a,
1159                                          const char *name_b)
1160{
1161        struct diffstat_file *x;
1162        x = xcalloc(sizeof (*x), 1);
1163        if (diffstat->nr == diffstat->alloc) {
1164                diffstat->alloc = alloc_nr(diffstat->alloc);
1165                diffstat->files = xrealloc(diffstat->files,
1166                                diffstat->alloc * sizeof(x));
1167        }
1168        diffstat->files[diffstat->nr++] = x;
1169        if (name_b) {
1170                x->from_name = xstrdup(name_a);
1171                x->name = xstrdup(name_b);
1172                x->is_renamed = 1;
1173        }
1174        else {
1175                x->from_name = NULL;
1176                x->name = xstrdup(name_a);
1177        }
1178        return x;
1179}
1180
1181static void diffstat_consume(void *priv, char *line, unsigned long len)
1182{
1183        struct diffstat_t *diffstat = priv;
1184        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1185
1186        if (line[0] == '+')
1187                x->added++;
1188        else if (line[0] == '-')
1189                x->deleted++;
1190}
1191
1192const char mime_boundary_leader[] = "------------";
1193
1194static int scale_linear(int it, int width, int max_change)
1195{
1196        /*
1197         * make sure that at least one '-' is printed if there were deletions,
1198         * and likewise for '+'.
1199         */
1200        if (max_change < 2)
1201                return it;
1202        return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
1203}
1204
1205static void show_name(FILE *file,
1206                      const char *prefix, const char *name, int len)
1207{
1208        fprintf(file, " %s%-*s |", prefix, len, name);
1209}
1210
1211static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1212{
1213        if (cnt <= 0)
1214                return;
1215        fprintf(file, "%s", set);
1216        while (cnt--)
1217                putc(ch, file);
1218        fprintf(file, "%s", reset);
1219}
1220
1221static void fill_print_name(struct diffstat_file *file)
1222{
1223        char *pname;
1224
1225        if (file->print_name)
1226                return;
1227
1228        if (!file->is_renamed) {
1229                struct strbuf buf = STRBUF_INIT;
1230                if (quote_c_style(file->name, &buf, NULL, 0)) {
1231                        pname = strbuf_detach(&buf, NULL);
1232                } else {
1233                        pname = file->name;
1234                        strbuf_release(&buf);
1235                }
1236        } else {
1237                pname = pprint_rename(file->from_name, file->name);
1238        }
1239        file->print_name = pname;
1240}
1241
1242static void show_stats(struct diffstat_t *data, struct diff_options *options)
1243{
1244        int i, len, add, del, adds = 0, dels = 0;
1245        uintmax_t max_change = 0, max_len = 0;
1246        int total_files = data->nr;
1247        int width, name_width;
1248        const char *reset, *add_c, *del_c;
1249        const char *line_prefix = "";
1250        struct strbuf *msg = NULL;
1251
1252        if (data->nr == 0)
1253                return;
1254
1255        if (options->output_prefix) {
1256                msg = options->output_prefix(options, options->output_prefix_data);
1257                line_prefix = msg->buf;
1258        }
1259
1260        width = options->stat_width ? options->stat_width : 80;
1261        name_width = options->stat_name_width ? options->stat_name_width : 50;
1262
1263        /* Sanity: give at least 5 columns to the graph,
1264         * but leave at least 10 columns for the name.
1265         */
1266        if (width < 25)
1267                width = 25;
1268        if (name_width < 10)
1269                name_width = 10;
1270        else if (width < name_width + 15)
1271                name_width = width - 15;
1272
1273        /* Find the longest filename and max number of changes */
1274        reset = diff_get_color_opt(options, DIFF_RESET);
1275        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1276        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1277
1278        for (i = 0; i < data->nr; i++) {
1279                struct diffstat_file *file = data->files[i];
1280                uintmax_t change = file->added + file->deleted;
1281                fill_print_name(file);
1282                len = strlen(file->print_name);
1283                if (max_len < len)
1284                        max_len = len;
1285
1286                if (file->is_binary || file->is_unmerged)
1287                        continue;
1288                if (max_change < change)
1289                        max_change = change;
1290        }
1291
1292        /* Compute the width of the graph part;
1293         * 10 is for one blank at the beginning of the line plus
1294         * " | count " between the name and the graph.
1295         *
1296         * From here on, name_width is the width of the name area,
1297         * and width is the width of the graph area.
1298         */
1299        name_width = (name_width < max_len) ? name_width : max_len;
1300        if (width < (name_width + 10) + max_change)
1301                width = width - (name_width + 10);
1302        else
1303                width = max_change;
1304
1305        for (i = 0; i < data->nr; i++) {
1306                const char *prefix = "";
1307                char *name = data->files[i]->print_name;
1308                uintmax_t added = data->files[i]->added;
1309                uintmax_t deleted = data->files[i]->deleted;
1310                int name_len;
1311
1312                /*
1313                 * "scale" the filename
1314                 */
1315                len = name_width;
1316                name_len = strlen(name);
1317                if (name_width < name_len) {
1318                        char *slash;
1319                        prefix = "...";
1320                        len -= 3;
1321                        name += name_len - len;
1322                        slash = strchr(name, '/');
1323                        if (slash)
1324                                name = slash;
1325                }
1326
1327                if (data->files[i]->is_binary) {
1328                        fprintf(options->file, "%s", line_prefix);
1329                        show_name(options->file, prefix, name, len);
1330                        fprintf(options->file, "  Bin ");
1331                        fprintf(options->file, "%s%"PRIuMAX"%s",
1332                                del_c, deleted, reset);
1333                        fprintf(options->file, " -> ");
1334                        fprintf(options->file, "%s%"PRIuMAX"%s",
1335                                add_c, added, reset);
1336                        fprintf(options->file, " bytes");
1337                        fprintf(options->file, "\n");
1338                        continue;
1339                }
1340                else if (data->files[i]->is_unmerged) {
1341                        fprintf(options->file, "%s", line_prefix);
1342                        show_name(options->file, prefix, name, len);
1343                        fprintf(options->file, "  Unmerged\n");
1344                        continue;
1345                }
1346                else if (!data->files[i]->is_renamed &&
1347                         (added + deleted == 0)) {
1348                        total_files--;
1349                        continue;
1350                }
1351
1352                /*
1353                 * scale the add/delete
1354                 */
1355                add = added;
1356                del = deleted;
1357                adds += add;
1358                dels += del;
1359
1360                if (width <= max_change) {
1361                        add = scale_linear(add, width, max_change);
1362                        del = scale_linear(del, width, max_change);
1363                }
1364                fprintf(options->file, "%s", line_prefix);
1365                show_name(options->file, prefix, name, len);
1366                fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1367                                added + deleted ? " " : "");
1368                show_graph(options->file, '+', add, add_c, reset);
1369                show_graph(options->file, '-', del, del_c, reset);
1370                fprintf(options->file, "\n");
1371        }
1372        fprintf(options->file, "%s", line_prefix);
1373        fprintf(options->file,
1374               " %d files changed, %d insertions(+), %d deletions(-)\n",
1375               total_files, adds, dels);
1376}
1377
1378static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1379{
1380        int i, adds = 0, dels = 0, total_files = data->nr;
1381
1382        if (data->nr == 0)
1383                return;
1384
1385        for (i = 0; i < data->nr; i++) {
1386                if (!data->files[i]->is_binary &&
1387                    !data->files[i]->is_unmerged) {
1388                        int added = data->files[i]->added;
1389                        int deleted= data->files[i]->deleted;
1390                        if (!data->files[i]->is_renamed &&
1391                            (added + deleted == 0)) {
1392                                total_files--;
1393                        } else {
1394                                adds += added;
1395                                dels += deleted;
1396                        }
1397                }
1398        }
1399        if (options->output_prefix) {
1400                struct strbuf *msg = NULL;
1401                msg = options->output_prefix(options,
1402                                options->output_prefix_data);
1403                fprintf(options->file, "%s", msg->buf);
1404        }
1405        fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
1406               total_files, adds, dels);
1407}
1408
1409static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1410{
1411        int i;
1412
1413        if (data->nr == 0)
1414                return;
1415
1416        for (i = 0; i < data->nr; i++) {
1417                struct diffstat_file *file = data->files[i];
1418
1419                if (options->output_prefix) {
1420                        struct strbuf *msg = NULL;
1421                        msg = options->output_prefix(options,
1422                                        options->output_prefix_data);
1423                        fprintf(options->file, "%s", msg->buf);
1424                }
1425
1426                if (file->is_binary)
1427                        fprintf(options->file, "-\t-\t");
1428                else
1429                        fprintf(options->file,
1430                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1431                                file->added, file->deleted);
1432                if (options->line_termination) {
1433                        fill_print_name(file);
1434                        if (!file->is_renamed)
1435                                write_name_quoted(file->name, options->file,
1436                                                  options->line_termination);
1437                        else {
1438                                fputs(file->print_name, options->file);
1439                                putc(options->line_termination, options->file);
1440                        }
1441                } else {
1442                        if (file->is_renamed) {
1443                                putc('\0', options->file);
1444                                write_name_quoted(file->from_name, options->file, '\0');
1445                        }
1446                        write_name_quoted(file->name, options->file, '\0');
1447                }
1448        }
1449}
1450
1451struct dirstat_file {
1452        const char *name;
1453        unsigned long changed;
1454};
1455
1456struct dirstat_dir {
1457        struct dirstat_file *files;
1458        int alloc, nr, percent, cumulative;
1459};
1460
1461static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1462                unsigned long changed, const char *base, int baselen)
1463{
1464        unsigned long this_dir = 0;
1465        unsigned int sources = 0;
1466        const char *line_prefix = "";
1467        struct strbuf *msg = NULL;
1468
1469        if (opt->output_prefix) {
1470                msg = opt->output_prefix(opt, opt->output_prefix_data);
1471                line_prefix = msg->buf;
1472        }
1473
1474        while (dir->nr) {
1475                struct dirstat_file *f = dir->files;
1476                int namelen = strlen(f->name);
1477                unsigned long this;
1478                char *slash;
1479
1480                if (namelen < baselen)
1481                        break;
1482                if (memcmp(f->name, base, baselen))
1483                        break;
1484                slash = strchr(f->name + baselen, '/');
1485                if (slash) {
1486                        int newbaselen = slash + 1 - f->name;
1487                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1488                        sources++;
1489                } else {
1490                        this = f->changed;
1491                        dir->files++;
1492                        dir->nr--;
1493                        sources += 2;
1494                }
1495                this_dir += this;
1496        }
1497
1498        /*
1499         * We don't report dirstat's for
1500         *  - the top level
1501         *  - or cases where everything came from a single directory
1502         *    under this directory (sources == 1).
1503         */
1504        if (baselen && sources != 1) {
1505                int permille = this_dir * 1000 / changed;
1506                if (permille) {
1507                        int percent = permille / 10;
1508                        if (percent >= dir->percent) {
1509                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1510                                        percent, permille % 10, baselen, base);
1511                                if (!dir->cumulative)
1512                                        return 0;
1513                        }
1514                }
1515        }
1516        return this_dir;
1517}
1518
1519static int dirstat_compare(const void *_a, const void *_b)
1520{
1521        const struct dirstat_file *a = _a;
1522        const struct dirstat_file *b = _b;
1523        return strcmp(a->name, b->name);
1524}
1525
1526static void show_dirstat(struct diff_options *options)
1527{
1528        int i;
1529        unsigned long changed;
1530        struct dirstat_dir dir;
1531        struct diff_queue_struct *q = &diff_queued_diff;
1532
1533        dir.files = NULL;
1534        dir.alloc = 0;
1535        dir.nr = 0;
1536        dir.percent = options->dirstat_percent;
1537        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1538
1539        changed = 0;
1540        for (i = 0; i < q->nr; i++) {
1541                struct diff_filepair *p = q->queue[i];
1542                const char *name;
1543                unsigned long copied, added, damage;
1544                int content_changed;
1545
1546                name = p->two->path ? p->two->path : p->one->path;
1547
1548                if (p->one->sha1_valid && p->two->sha1_valid)
1549                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1550                else
1551                        content_changed = 1;
1552
1553                if (!content_changed) {
1554                        /*
1555                         * The SHA1 has not changed, so pre-/post-content is
1556                         * identical. We can therefore skip looking at the
1557                         * file contents altogether.
1558                         */
1559                        damage = 0;
1560                        goto found_damage;
1561                }
1562
1563                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1564                        /*
1565                         * In --dirstat-by-file mode, we don't really need to
1566                         * look at the actual file contents at all.
1567                         * The fact that the SHA1 changed is enough for us to
1568                         * add this file to the list of results
1569                         * (with each file contributing equal damage).
1570                         */
1571                        damage = 1;
1572                        goto found_damage;
1573                }
1574
1575                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1576                        diff_populate_filespec(p->one, 0);
1577                        diff_populate_filespec(p->two, 0);
1578                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1579                                               &copied, &added);
1580                        diff_free_filespec_data(p->one);
1581                        diff_free_filespec_data(p->two);
1582                } else if (DIFF_FILE_VALID(p->one)) {
1583                        diff_populate_filespec(p->one, 1);
1584                        copied = added = 0;
1585                        diff_free_filespec_data(p->one);
1586                } else if (DIFF_FILE_VALID(p->two)) {
1587                        diff_populate_filespec(p->two, 1);
1588                        copied = 0;
1589                        added = p->two->size;
1590                        diff_free_filespec_data(p->two);
1591                } else
1592                        continue;
1593
1594                /*
1595                 * Original minus copied is the removed material,
1596                 * added is the new material.  They are both damages
1597                 * made to the preimage.
1598                 * If the resulting damage is zero, we know that
1599                 * diffcore_count_changes() considers the two entries to
1600                 * be identical, but since content_changed is true, we
1601                 * know that there must have been _some_ kind of change,
1602                 * so we force all entries to have damage > 0.
1603                 */
1604                damage = (p->one->size - copied) + added;
1605                if (!damage)
1606                        damage = 1;
1607
1608found_damage:
1609                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1610                dir.files[dir.nr].name = name;
1611                dir.files[dir.nr].changed = damage;
1612                changed += damage;
1613                dir.nr++;
1614        }
1615
1616        /* This can happen even with many files, if everything was renames */
1617        if (!changed)
1618                return;
1619
1620        /* Show all directories with more than x% of the changes */
1621        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1622        gather_dirstat(options, &dir, changed, "", 0);
1623}
1624
1625static void free_diffstat_info(struct diffstat_t *diffstat)
1626{
1627        int i;
1628        for (i = 0; i < diffstat->nr; i++) {
1629                struct diffstat_file *f = diffstat->files[i];
1630                if (f->name != f->print_name)
1631                        free(f->print_name);
1632                free(f->name);
1633                free(f->from_name);
1634                free(f);
1635        }
1636        free(diffstat->files);
1637}
1638
1639struct checkdiff_t {
1640        const char *filename;
1641        int lineno;
1642        int conflict_marker_size;
1643        struct diff_options *o;
1644        unsigned ws_rule;
1645        unsigned status;
1646};
1647
1648static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1649{
1650        char firstchar;
1651        int cnt;
1652
1653        if (len < marker_size + 1)
1654                return 0;
1655        firstchar = line[0];
1656        switch (firstchar) {
1657        case '=': case '>': case '<': case '|':
1658                break;
1659        default:
1660                return 0;
1661        }
1662        for (cnt = 1; cnt < marker_size; cnt++)
1663                if (line[cnt] != firstchar)
1664                        return 0;
1665        /* line[1] thru line[marker_size-1] are same as firstchar */
1666        if (len < marker_size + 1 || !isspace(line[marker_size]))
1667                return 0;
1668        return 1;
1669}
1670
1671static void checkdiff_consume(void *priv, char *line, unsigned long len)
1672{
1673        struct checkdiff_t *data = priv;
1674        int color_diff = DIFF_OPT_TST(data->o, COLOR_DIFF);
1675        int marker_size = data->conflict_marker_size;
1676        const char *ws = diff_get_color(color_diff, DIFF_WHITESPACE);
1677        const char *reset = diff_get_color(color_diff, DIFF_RESET);
1678        const char *set = diff_get_color(color_diff, DIFF_FILE_NEW);
1679        char *err;
1680        char *line_prefix = "";
1681        struct strbuf *msgbuf;
1682
1683        assert(data->o);
1684        if (data->o->output_prefix) {
1685                msgbuf = data->o->output_prefix(data->o,
1686                        data->o->output_prefix_data);
1687                line_prefix = msgbuf->buf;
1688        }
1689
1690        if (line[0] == '+') {
1691                unsigned bad;
1692                data->lineno++;
1693                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1694                        data->status |= 1;
1695                        fprintf(data->o->file,
1696                                "%s%s:%d: leftover conflict marker\n",
1697                                line_prefix, data->filename, data->lineno);
1698                }
1699                bad = ws_check(line + 1, len - 1, data->ws_rule);
1700                if (!bad)
1701                        return;
1702                data->status |= bad;
1703                err = whitespace_error_string(bad);
1704                fprintf(data->o->file, "%s%s:%d: %s.\n",
1705                        line_prefix, data->filename, data->lineno, err);
1706                free(err);
1707                emit_line(data->o, set, reset, line, 1);
1708                ws_check_emit(line + 1, len - 1, data->ws_rule,
1709                              data->o->file, set, reset, ws);
1710        } else if (line[0] == ' ') {
1711                data->lineno++;
1712        } else if (line[0] == '@') {
1713                char *plus = strchr(line, '+');
1714                if (plus)
1715                        data->lineno = strtol(plus, NULL, 10) - 1;
1716                else
1717                        die("invalid diff");
1718        }
1719}
1720
1721static unsigned char *deflate_it(char *data,
1722                                 unsigned long size,
1723                                 unsigned long *result_size)
1724{
1725        int bound;
1726        unsigned char *deflated;
1727        z_stream stream;
1728
1729        memset(&stream, 0, sizeof(stream));
1730        deflateInit(&stream, zlib_compression_level);
1731        bound = deflateBound(&stream, size);
1732        deflated = xmalloc(bound);
1733        stream.next_out = deflated;
1734        stream.avail_out = bound;
1735
1736        stream.next_in = (unsigned char *)data;
1737        stream.avail_in = size;
1738        while (deflate(&stream, Z_FINISH) == Z_OK)
1739                ; /* nothing */
1740        deflateEnd(&stream);
1741        *result_size = stream.total_out;
1742        return deflated;
1743}
1744
1745static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1746{
1747        void *cp;
1748        void *delta;
1749        void *deflated;
1750        void *data;
1751        unsigned long orig_size;
1752        unsigned long delta_size;
1753        unsigned long deflate_size;
1754        unsigned long data_size;
1755
1756        /* We could do deflated delta, or we could do just deflated two,
1757         * whichever is smaller.
1758         */
1759        delta = NULL;
1760        deflated = deflate_it(two->ptr, two->size, &deflate_size);
1761        if (one->size && two->size) {
1762                delta = diff_delta(one->ptr, one->size,
1763                                   two->ptr, two->size,
1764                                   &delta_size, deflate_size);
1765                if (delta) {
1766                        void *to_free = delta;
1767                        orig_size = delta_size;
1768                        delta = deflate_it(delta, delta_size, &delta_size);
1769                        free(to_free);
1770                }
1771        }
1772
1773        if (delta && delta_size < deflate_size) {
1774                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
1775                free(deflated);
1776                data = delta;
1777                data_size = delta_size;
1778        }
1779        else {
1780                fprintf(file, "%sliteral %lu\n", prefix, two->size);
1781                free(delta);
1782                data = deflated;
1783                data_size = deflate_size;
1784        }
1785
1786        /* emit data encoded in base85 */
1787        cp = data;
1788        while (data_size) {
1789                int bytes = (52 < data_size) ? 52 : data_size;
1790                char line[70];
1791                data_size -= bytes;
1792                if (bytes <= 26)
1793                        line[0] = bytes + 'A' - 1;
1794                else
1795                        line[0] = bytes - 26 + 'a' - 1;
1796                encode_85(line + 1, cp, bytes);
1797                cp = (char *) cp + bytes;
1798                fprintf(file, "%s", prefix);
1799                fputs(line, file);
1800                fputc('\n', file);
1801        }
1802        fprintf(file, "%s\n", prefix);
1803        free(data);
1804}
1805
1806static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1807{
1808        fprintf(file, "%sGIT binary patch\n", prefix);
1809        emit_binary_diff_body(file, one, two, prefix);
1810        emit_binary_diff_body(file, two, one, prefix);
1811}
1812
1813static void diff_filespec_load_driver(struct diff_filespec *one)
1814{
1815        /* Use already-loaded driver */
1816        if (one->driver)
1817                return;
1818
1819        if (S_ISREG(one->mode))
1820                one->driver = userdiff_find_by_path(one->path);
1821
1822        /* Fallback to default settings */
1823        if (!one->driver)
1824                one->driver = userdiff_find_by_name("default");
1825}
1826
1827int diff_filespec_is_binary(struct diff_filespec *one)
1828{
1829        if (one->is_binary == -1) {
1830                diff_filespec_load_driver(one);
1831                if (one->driver->binary != -1)
1832                        one->is_binary = one->driver->binary;
1833                else {
1834                        if (!one->data && DIFF_FILE_VALID(one))
1835                                diff_populate_filespec(one, 0);
1836                        if (one->data)
1837                                one->is_binary = buffer_is_binary(one->data,
1838                                                one->size);
1839                        if (one->is_binary == -1)
1840                                one->is_binary = 0;
1841                }
1842        }
1843        return one->is_binary;
1844}
1845
1846static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
1847{
1848        diff_filespec_load_driver(one);
1849        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
1850}
1851
1852static const char *userdiff_word_regex(struct diff_filespec *one)
1853{
1854        diff_filespec_load_driver(one);
1855        return one->driver->word_regex;
1856}
1857
1858void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
1859{
1860        if (!options->a_prefix)
1861                options->a_prefix = a;
1862        if (!options->b_prefix)
1863                options->b_prefix = b;
1864}
1865
1866struct userdiff_driver *get_textconv(struct diff_filespec *one)
1867{
1868        if (!DIFF_FILE_VALID(one))
1869                return NULL;
1870
1871        diff_filespec_load_driver(one);
1872        if (!one->driver->textconv)
1873                return NULL;
1874
1875        if (one->driver->textconv_want_cache && !one->driver->textconv_cache) {
1876                struct notes_cache *c = xmalloc(sizeof(*c));
1877                struct strbuf name = STRBUF_INIT;
1878
1879                strbuf_addf(&name, "textconv/%s", one->driver->name);
1880                notes_cache_init(c, name.buf, one->driver->textconv);
1881                one->driver->textconv_cache = c;
1882        }
1883
1884        return one->driver;
1885}
1886
1887static void builtin_diff(const char *name_a,
1888                         const char *name_b,
1889                         struct diff_filespec *one,
1890                         struct diff_filespec *two,
1891                         const char *xfrm_msg,
1892                         int must_show_header,
1893                         struct diff_options *o,
1894                         int complete_rewrite)
1895{
1896        mmfile_t mf1, mf2;
1897        const char *lbl[2];
1898        char *a_one, *b_two;
1899        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1900        const char *reset = diff_get_color_opt(o, DIFF_RESET);
1901        const char *a_prefix, *b_prefix;
1902        struct userdiff_driver *textconv_one = NULL;
1903        struct userdiff_driver *textconv_two = NULL;
1904        struct strbuf header = STRBUF_INIT;
1905        struct strbuf *msgbuf;
1906        char *line_prefix = "";
1907
1908        if (o->output_prefix) {
1909                msgbuf = o->output_prefix(o, o->output_prefix_data);
1910                line_prefix = msgbuf->buf;
1911        }
1912
1913        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
1914                        (!one->mode || S_ISGITLINK(one->mode)) &&
1915                        (!two->mode || S_ISGITLINK(two->mode))) {
1916                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
1917                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
1918                show_submodule_summary(o->file, one ? one->path : two->path,
1919                                one->sha1, two->sha1, two->dirty_submodule,
1920                                del, add, reset);
1921                return;
1922        }
1923
1924        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
1925                textconv_one = get_textconv(one);
1926                textconv_two = get_textconv(two);
1927        }
1928
1929        diff_set_mnemonic_prefix(o, "a/", "b/");
1930        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
1931                a_prefix = o->b_prefix;
1932                b_prefix = o->a_prefix;
1933        } else {
1934                a_prefix = o->a_prefix;
1935                b_prefix = o->b_prefix;
1936        }
1937
1938        /* Never use a non-valid filename anywhere if at all possible */
1939        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
1940        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
1941
1942        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
1943        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
1944        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1945        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1946        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
1947        if (lbl[0][0] == '/') {
1948                /* /dev/null */
1949                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
1950                if (xfrm_msg)
1951                        strbuf_addstr(&header, xfrm_msg);
1952                must_show_header = 1;
1953        }
1954        else if (lbl[1][0] == '/') {
1955                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
1956                if (xfrm_msg)
1957                        strbuf_addstr(&header, xfrm_msg);
1958                must_show_header = 1;
1959        }
1960        else {
1961                if (one->mode != two->mode) {
1962                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
1963                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
1964                        must_show_header = 1;
1965                }
1966                if (xfrm_msg)
1967                        strbuf_addstr(&header, xfrm_msg);
1968
1969                /*
1970                 * we do not run diff between different kind
1971                 * of objects.
1972                 */
1973                if ((one->mode ^ two->mode) & S_IFMT)
1974                        goto free_ab_and_return;
1975                if (complete_rewrite &&
1976                    (textconv_one || !diff_filespec_is_binary(one)) &&
1977                    (textconv_two || !diff_filespec_is_binary(two))) {
1978                        fprintf(o->file, "%s", header.buf);
1979                        strbuf_reset(&header);
1980                        emit_rewrite_diff(name_a, name_b, one, two,
1981                                                textconv_one, textconv_two, o);
1982                        o->found_changes = 1;
1983                        goto free_ab_and_return;
1984                }
1985        }
1986
1987        if (o->irreversible_delete && lbl[1][0] == '/') {
1988                fprintf(o->file, "%s", header.buf);
1989                strbuf_reset(&header);
1990                goto free_ab_and_return;
1991        } else if (!DIFF_OPT_TST(o, TEXT) &&
1992            ( (!textconv_one && diff_filespec_is_binary(one)) ||
1993              (!textconv_two && diff_filespec_is_binary(two)) )) {
1994                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1995                        die("unable to read files to diff");
1996                /* Quite common confusing case */
1997                if (mf1.size == mf2.size &&
1998                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
1999                        if (must_show_header)
2000                                fprintf(o->file, "%s", header.buf);
2001                        goto free_ab_and_return;
2002                }
2003                fprintf(o->file, "%s", header.buf);
2004                strbuf_reset(&header);
2005                if (DIFF_OPT_TST(o, BINARY))
2006                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2007                else
2008                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2009                                line_prefix, lbl[0], lbl[1]);
2010                o->found_changes = 1;
2011        } else {
2012                /* Crazy xdl interfaces.. */
2013                const char *diffopts = getenv("GIT_DIFF_OPTS");
2014                xpparam_t xpp;
2015                xdemitconf_t xecfg;
2016                struct emit_callback ecbdata;
2017                const struct userdiff_funcname *pe;
2018
2019                if (!DIFF_XDL_TST(o, WHITESPACE_FLAGS) || must_show_header) {
2020                        fprintf(o->file, "%s", header.buf);
2021                        strbuf_reset(&header);
2022                }
2023
2024                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2025                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2026
2027                pe = diff_funcname_pattern(one);
2028                if (!pe)
2029                        pe = diff_funcname_pattern(two);
2030
2031                memset(&xpp, 0, sizeof(xpp));
2032                memset(&xecfg, 0, sizeof(xecfg));
2033                memset(&ecbdata, 0, sizeof(ecbdata));
2034                ecbdata.label_path = lbl;
2035                ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
2036                ecbdata.found_changesp = &o->found_changes;
2037                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2038                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2039                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2040                ecbdata.opt = o;
2041                ecbdata.header = header.len ? &header : NULL;
2042                xpp.flags = o->xdl_opts;
2043                xecfg.ctxlen = o->context;
2044                xecfg.interhunkctxlen = o->interhunkcontext;
2045                xecfg.flags = XDL_EMIT_FUNCNAMES;
2046                if (pe)
2047                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2048                if (!diffopts)
2049                        ;
2050                else if (!prefixcmp(diffopts, "--unified="))
2051                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2052                else if (!prefixcmp(diffopts, "-u"))
2053                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2054                if (o->word_diff) {
2055                        int i;
2056
2057                        ecbdata.diff_words =
2058                                xcalloc(1, sizeof(struct diff_words_data));
2059                        ecbdata.diff_words->type = o->word_diff;
2060                        ecbdata.diff_words->opt = o;
2061                        if (!o->word_regex)
2062                                o->word_regex = userdiff_word_regex(one);
2063                        if (!o->word_regex)
2064                                o->word_regex = userdiff_word_regex(two);
2065                        if (!o->word_regex)
2066                                o->word_regex = diff_word_regex_cfg;
2067                        if (o->word_regex) {
2068                                ecbdata.diff_words->word_regex = (regex_t *)
2069                                        xmalloc(sizeof(regex_t));
2070                                if (regcomp(ecbdata.diff_words->word_regex,
2071                                                o->word_regex,
2072                                                REG_EXTENDED | REG_NEWLINE))
2073                                        die ("Invalid regular expression: %s",
2074                                                        o->word_regex);
2075                        }
2076                        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2077                                if (o->word_diff == diff_words_styles[i].type) {
2078                                        ecbdata.diff_words->style =
2079                                                &diff_words_styles[i];
2080                                        break;
2081                                }
2082                        }
2083                        if (DIFF_OPT_TST(o, COLOR_DIFF)) {
2084                                struct diff_words_style *st = ecbdata.diff_words->style;
2085                                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2086                                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2087                                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
2088                        }
2089                }
2090                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2091                              &xpp, &xecfg);
2092                if (o->word_diff)
2093                        free_diff_words_data(&ecbdata);
2094                if (textconv_one)
2095                        free(mf1.ptr);
2096                if (textconv_two)
2097                        free(mf2.ptr);
2098                xdiff_clear_find_func(&xecfg);
2099        }
2100
2101 free_ab_and_return:
2102        strbuf_release(&header);
2103        diff_free_filespec_data(one);
2104        diff_free_filespec_data(two);
2105        free(a_one);
2106        free(b_two);
2107        return;
2108}
2109
2110static void builtin_diffstat(const char *name_a, const char *name_b,
2111                             struct diff_filespec *one,
2112                             struct diff_filespec *two,
2113                             struct diffstat_t *diffstat,
2114                             struct diff_options *o,
2115                             int complete_rewrite)
2116{
2117        mmfile_t mf1, mf2;
2118        struct diffstat_file *data;
2119
2120        data = diffstat_add(diffstat, name_a, name_b);
2121
2122        if (!one || !two) {
2123                data->is_unmerged = 1;
2124                return;
2125        }
2126
2127        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2128                data->is_binary = 1;
2129                data->added = diff_filespec_size(two);
2130                data->deleted = diff_filespec_size(one);
2131        }
2132
2133        else if (complete_rewrite) {
2134                diff_populate_filespec(one, 0);
2135                diff_populate_filespec(two, 0);
2136                data->deleted = count_lines(one->data, one->size);
2137                data->added = count_lines(two->data, two->size);
2138        }
2139
2140        else {
2141                /* Crazy xdl interfaces.. */
2142                xpparam_t xpp;
2143                xdemitconf_t xecfg;
2144
2145                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2146                        die("unable to read files to diff");
2147
2148                memset(&xpp, 0, sizeof(xpp));
2149                memset(&xecfg, 0, sizeof(xecfg));
2150                xpp.flags = o->xdl_opts;
2151                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2152                              &xpp, &xecfg);
2153        }
2154
2155        diff_free_filespec_data(one);
2156        diff_free_filespec_data(two);
2157}
2158
2159static void builtin_checkdiff(const char *name_a, const char *name_b,
2160                              const char *attr_path,
2161                              struct diff_filespec *one,
2162                              struct diff_filespec *two,
2163                              struct diff_options *o)
2164{
2165        mmfile_t mf1, mf2;
2166        struct checkdiff_t data;
2167
2168        if (!two)
2169                return;
2170
2171        memset(&data, 0, sizeof(data));
2172        data.filename = name_b ? name_b : name_a;
2173        data.lineno = 0;
2174        data.o = o;
2175        data.ws_rule = whitespace_rule(attr_path);
2176        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2177
2178        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2179                die("unable to read files to diff");
2180
2181        /*
2182         * All the other codepaths check both sides, but not checking
2183         * the "old" side here is deliberate.  We are checking the newly
2184         * introduced changes, and as long as the "new" side is text, we
2185         * can and should check what it introduces.
2186         */
2187        if (diff_filespec_is_binary(two))
2188                goto free_and_return;
2189        else {
2190                /* Crazy xdl interfaces.. */
2191                xpparam_t xpp;
2192                xdemitconf_t xecfg;
2193
2194                memset(&xpp, 0, sizeof(xpp));
2195                memset(&xecfg, 0, sizeof(xecfg));
2196                xecfg.ctxlen = 1; /* at least one context line */
2197                xpp.flags = 0;
2198                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2199                              &xpp, &xecfg);
2200
2201                if (data.ws_rule & WS_BLANK_AT_EOF) {
2202                        struct emit_callback ecbdata;
2203                        int blank_at_eof;
2204
2205                        ecbdata.ws_rule = data.ws_rule;
2206                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2207                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2208
2209                        if (blank_at_eof) {
2210                                static char *err;
2211                                if (!err)
2212                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2213                                fprintf(o->file, "%s:%d: %s.\n",
2214                                        data.filename, blank_at_eof, err);
2215                                data.status = 1; /* report errors */
2216                        }
2217                }
2218        }
2219 free_and_return:
2220        diff_free_filespec_data(one);
2221        diff_free_filespec_data(two);
2222        if (data.status)
2223                DIFF_OPT_SET(o, CHECK_FAILED);
2224}
2225
2226struct diff_filespec *alloc_filespec(const char *path)
2227{
2228        int namelen = strlen(path);
2229        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2230
2231        memset(spec, 0, sizeof(*spec));
2232        spec->path = (char *)(spec + 1);
2233        memcpy(spec->path, path, namelen+1);
2234        spec->count = 1;
2235        spec->is_binary = -1;
2236        return spec;
2237}
2238
2239void free_filespec(struct diff_filespec *spec)
2240{
2241        if (!--spec->count) {
2242                diff_free_filespec_data(spec);
2243                free(spec);
2244        }
2245}
2246
2247void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2248                   unsigned short mode)
2249{
2250        if (mode) {
2251                spec->mode = canon_mode(mode);
2252                hashcpy(spec->sha1, sha1);
2253                spec->sha1_valid = !is_null_sha1(sha1);
2254        }
2255}
2256
2257/*
2258 * Given a name and sha1 pair, if the index tells us the file in
2259 * the work tree has that object contents, return true, so that
2260 * prepare_temp_file() does not have to inflate and extract.
2261 */
2262static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2263{
2264        struct cache_entry *ce;
2265        struct stat st;
2266        int pos, len;
2267
2268        /*
2269         * We do not read the cache ourselves here, because the
2270         * benchmark with my previous version that always reads cache
2271         * shows that it makes things worse for diff-tree comparing
2272         * two linux-2.6 kernel trees in an already checked out work
2273         * tree.  This is because most diff-tree comparisons deal with
2274         * only a small number of files, while reading the cache is
2275         * expensive for a large project, and its cost outweighs the
2276         * savings we get by not inflating the object to a temporary
2277         * file.  Practically, this code only helps when we are used
2278         * by diff-cache --cached, which does read the cache before
2279         * calling us.
2280         */
2281        if (!active_cache)
2282                return 0;
2283
2284        /* We want to avoid the working directory if our caller
2285         * doesn't need the data in a normal file, this system
2286         * is rather slow with its stat/open/mmap/close syscalls,
2287         * and the object is contained in a pack file.  The pack
2288         * is probably already open and will be faster to obtain
2289         * the data through than the working directory.  Loose
2290         * objects however would tend to be slower as they need
2291         * to be individually opened and inflated.
2292         */
2293        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2294                return 0;
2295
2296        len = strlen(name);
2297        pos = cache_name_pos(name, len);
2298        if (pos < 0)
2299                return 0;
2300        ce = active_cache[pos];
2301
2302        /*
2303         * This is not the sha1 we are looking for, or
2304         * unreusable because it is not a regular file.
2305         */
2306        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2307                return 0;
2308
2309        /*
2310         * If ce is marked as "assume unchanged", there is no
2311         * guarantee that work tree matches what we are looking for.
2312         */
2313        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2314                return 0;
2315
2316        /*
2317         * If ce matches the file in the work tree, we can reuse it.
2318         */
2319        if (ce_uptodate(ce) ||
2320            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2321                return 1;
2322
2323        return 0;
2324}
2325
2326static int populate_from_stdin(struct diff_filespec *s)
2327{
2328        struct strbuf buf = STRBUF_INIT;
2329        size_t size = 0;
2330
2331        if (strbuf_read(&buf, 0, 0) < 0)
2332                return error("error while reading from stdin %s",
2333                                     strerror(errno));
2334
2335        s->should_munmap = 0;
2336        s->data = strbuf_detach(&buf, &size);
2337        s->size = size;
2338        s->should_free = 1;
2339        return 0;
2340}
2341
2342static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2343{
2344        int len;
2345        char *data = xmalloc(100), *dirty = "";
2346
2347        /* Are we looking at the work tree? */
2348        if (s->dirty_submodule)
2349                dirty = "-dirty";
2350
2351        len = snprintf(data, 100,
2352                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2353        s->data = data;
2354        s->size = len;
2355        s->should_free = 1;
2356        if (size_only) {
2357                s->data = NULL;
2358                free(data);
2359        }
2360        return 0;
2361}
2362
2363/*
2364 * While doing rename detection and pickaxe operation, we may need to
2365 * grab the data for the blob (or file) for our own in-core comparison.
2366 * diff_filespec has data and size fields for this purpose.
2367 */
2368int diff_populate_filespec(struct diff_filespec *s, int size_only)
2369{
2370        int err = 0;
2371        if (!DIFF_FILE_VALID(s))
2372                die("internal error: asking to populate invalid file.");
2373        if (S_ISDIR(s->mode))
2374                return -1;
2375
2376        if (s->data)
2377                return 0;
2378
2379        if (size_only && 0 < s->size)
2380                return 0;
2381
2382        if (S_ISGITLINK(s->mode))
2383                return diff_populate_gitlink(s, size_only);
2384
2385        if (!s->sha1_valid ||
2386            reuse_worktree_file(s->path, s->sha1, 0)) {
2387                struct strbuf buf = STRBUF_INIT;
2388                struct stat st;
2389                int fd;
2390
2391                if (!strcmp(s->path, "-"))
2392                        return populate_from_stdin(s);
2393
2394                if (lstat(s->path, &st) < 0) {
2395                        if (errno == ENOENT) {
2396                        err_empty:
2397                                err = -1;
2398                        empty:
2399                                s->data = (char *)"";
2400                                s->size = 0;
2401                                return err;
2402                        }
2403                }
2404                s->size = xsize_t(st.st_size);
2405                if (!s->size)
2406                        goto empty;
2407                if (S_ISLNK(st.st_mode)) {
2408                        struct strbuf sb = STRBUF_INIT;
2409
2410                        if (strbuf_readlink(&sb, s->path, s->size))
2411                                goto err_empty;
2412                        s->size = sb.len;
2413                        s->data = strbuf_detach(&sb, NULL);
2414                        s->should_free = 1;
2415                        return 0;
2416                }
2417                if (size_only)
2418                        return 0;
2419                fd = open(s->path, O_RDONLY);
2420                if (fd < 0)
2421                        goto err_empty;
2422                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2423                close(fd);
2424                s->should_munmap = 1;
2425
2426                /*
2427                 * Convert from working tree format to canonical git format
2428                 */
2429                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2430                        size_t size = 0;
2431                        munmap(s->data, s->size);
2432                        s->should_munmap = 0;
2433                        s->data = strbuf_detach(&buf, &size);
2434                        s->size = size;
2435                        s->should_free = 1;
2436                }
2437        }
2438        else {
2439                enum object_type type;
2440                if (size_only) {
2441                        type = sha1_object_info(s->sha1, &s->size);
2442                        if (type < 0)
2443                                die("unable to read %s", sha1_to_hex(s->sha1));
2444                } else {
2445                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2446                        if (!s->data)
2447                                die("unable to read %s", sha1_to_hex(s->sha1));
2448                        s->should_free = 1;
2449                }
2450        }
2451        return 0;
2452}
2453
2454void diff_free_filespec_blob(struct diff_filespec *s)
2455{
2456        if (s->should_free)
2457                free(s->data);
2458        else if (s->should_munmap)
2459                munmap(s->data, s->size);
2460
2461        if (s->should_free || s->should_munmap) {
2462                s->should_free = s->should_munmap = 0;
2463                s->data = NULL;
2464        }
2465}
2466
2467void diff_free_filespec_data(struct diff_filespec *s)
2468{
2469        diff_free_filespec_blob(s);
2470        free(s->cnt_data);
2471        s->cnt_data = NULL;
2472}
2473
2474static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2475                           void *blob,
2476                           unsigned long size,
2477                           const unsigned char *sha1,
2478                           int mode)
2479{
2480        int fd;
2481        struct strbuf buf = STRBUF_INIT;
2482        struct strbuf template = STRBUF_INIT;
2483        char *path_dup = xstrdup(path);
2484        const char *base = basename(path_dup);
2485
2486        /* Generate "XXXXXX_basename.ext" */
2487        strbuf_addstr(&template, "XXXXXX_");
2488        strbuf_addstr(&template, base);
2489
2490        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2491                        strlen(base) + 1);
2492        if (fd < 0)
2493                die_errno("unable to create temp-file");
2494        if (convert_to_working_tree(path,
2495                        (const char *)blob, (size_t)size, &buf)) {
2496                blob = buf.buf;
2497                size = buf.len;
2498        }
2499        if (write_in_full(fd, blob, size) != size)
2500                die_errno("unable to write temp-file");
2501        close(fd);
2502        temp->name = temp->tmp_path;
2503        strcpy(temp->hex, sha1_to_hex(sha1));
2504        temp->hex[40] = 0;
2505        sprintf(temp->mode, "%06o", mode);
2506        strbuf_release(&buf);
2507        strbuf_release(&template);
2508        free(path_dup);
2509}
2510
2511static struct diff_tempfile *prepare_temp_file(const char *name,
2512                struct diff_filespec *one)
2513{
2514        struct diff_tempfile *temp = claim_diff_tempfile();
2515
2516        if (!DIFF_FILE_VALID(one)) {
2517        not_a_valid_file:
2518                /* A '-' entry produces this for file-2, and
2519                 * a '+' entry produces this for file-1.
2520                 */
2521                temp->name = "/dev/null";
2522                strcpy(temp->hex, ".");
2523                strcpy(temp->mode, ".");
2524                return temp;
2525        }
2526
2527        if (!remove_tempfile_installed) {
2528                atexit(remove_tempfile);
2529                sigchain_push_common(remove_tempfile_on_signal);
2530                remove_tempfile_installed = 1;
2531        }
2532
2533        if (!one->sha1_valid ||
2534            reuse_worktree_file(name, one->sha1, 1)) {
2535                struct stat st;
2536                if (lstat(name, &st) < 0) {
2537                        if (errno == ENOENT)
2538                                goto not_a_valid_file;
2539                        die_errno("stat(%s)", name);
2540                }
2541                if (S_ISLNK(st.st_mode)) {
2542                        struct strbuf sb = STRBUF_INIT;
2543                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2544                                die_errno("readlink(%s)", name);
2545                        prep_temp_blob(name, temp, sb.buf, sb.len,
2546                                       (one->sha1_valid ?
2547                                        one->sha1 : null_sha1),
2548                                       (one->sha1_valid ?
2549                                        one->mode : S_IFLNK));
2550                        strbuf_release(&sb);
2551                }
2552                else {
2553                        /* we can borrow from the file in the work tree */
2554                        temp->name = name;
2555                        if (!one->sha1_valid)
2556                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2557                        else
2558                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2559                        /* Even though we may sometimes borrow the
2560                         * contents from the work tree, we always want
2561                         * one->mode.  mode is trustworthy even when
2562                         * !(one->sha1_valid), as long as
2563                         * DIFF_FILE_VALID(one).
2564                         */
2565                        sprintf(temp->mode, "%06o", one->mode);
2566                }
2567                return temp;
2568        }
2569        else {
2570                if (diff_populate_filespec(one, 0))
2571                        die("cannot read data blob for %s", one->path);
2572                prep_temp_blob(name, temp, one->data, one->size,
2573                               one->sha1, one->mode);
2574        }
2575        return temp;
2576}
2577
2578/* An external diff command takes:
2579 *
2580 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2581 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2582 *
2583 */
2584static void run_external_diff(const char *pgm,
2585                              const char *name,
2586                              const char *other,
2587                              struct diff_filespec *one,
2588                              struct diff_filespec *two,
2589                              const char *xfrm_msg,
2590                              int complete_rewrite)
2591{
2592        const char *spawn_arg[10];
2593        int retval;
2594        const char **arg = &spawn_arg[0];
2595
2596        if (one && two) {
2597                struct diff_tempfile *temp_one, *temp_two;
2598                const char *othername = (other ? other : name);
2599                temp_one = prepare_temp_file(name, one);
2600                temp_two = prepare_temp_file(othername, two);
2601                *arg++ = pgm;
2602                *arg++ = name;
2603                *arg++ = temp_one->name;
2604                *arg++ = temp_one->hex;
2605                *arg++ = temp_one->mode;
2606                *arg++ = temp_two->name;
2607                *arg++ = temp_two->hex;
2608                *arg++ = temp_two->mode;
2609                if (other) {
2610                        *arg++ = other;
2611                        *arg++ = xfrm_msg;
2612                }
2613        } else {
2614                *arg++ = pgm;
2615                *arg++ = name;
2616        }
2617        *arg = NULL;
2618        fflush(NULL);
2619        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2620        remove_tempfile();
2621        if (retval) {
2622                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2623                exit(1);
2624        }
2625}
2626
2627static int similarity_index(struct diff_filepair *p)
2628{
2629        return p->score * 100 / MAX_SCORE;
2630}
2631
2632static void fill_metainfo(struct strbuf *msg,
2633                          const char *name,
2634                          const char *other,
2635                          struct diff_filespec *one,
2636                          struct diff_filespec *two,
2637                          struct diff_options *o,
2638                          struct diff_filepair *p,
2639                          int *must_show_header,
2640                          int use_color)
2641{
2642        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2643        const char *reset = diff_get_color(use_color, DIFF_RESET);
2644        struct strbuf *msgbuf;
2645        char *line_prefix = "";
2646
2647        *must_show_header = 1;
2648        if (o->output_prefix) {
2649                msgbuf = o->output_prefix(o, o->output_prefix_data);
2650                line_prefix = msgbuf->buf;
2651        }
2652        strbuf_init(msg, PATH_MAX * 2 + 300);
2653        switch (p->status) {
2654        case DIFF_STATUS_COPIED:
2655                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2656                            line_prefix, set, similarity_index(p));
2657                strbuf_addf(msg, "%s\n%s%scopy from ",
2658                            reset,  line_prefix, set);
2659                quote_c_style(name, msg, NULL, 0);
2660                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2661                quote_c_style(other, msg, NULL, 0);
2662                strbuf_addf(msg, "%s\n", reset);
2663                break;
2664        case DIFF_STATUS_RENAMED:
2665                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2666                            line_prefix, set, similarity_index(p));
2667                strbuf_addf(msg, "%s\n%s%srename from ",
2668                            reset, line_prefix, set);
2669                quote_c_style(name, msg, NULL, 0);
2670                strbuf_addf(msg, "%s\n%s%srename to ",
2671                            reset, line_prefix, set);
2672                quote_c_style(other, msg, NULL, 0);
2673                strbuf_addf(msg, "%s\n", reset);
2674                break;
2675        case DIFF_STATUS_MODIFIED:
2676                if (p->score) {
2677                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2678                                    line_prefix,
2679                                    set, similarity_index(p), reset);
2680                        break;
2681                }
2682                /* fallthru */
2683        default:
2684                *must_show_header = 0;
2685        }
2686        if (one && two && hashcmp(one->sha1, two->sha1)) {
2687                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2688
2689                if (DIFF_OPT_TST(o, BINARY)) {
2690                        mmfile_t mf;
2691                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2692                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2693                                abbrev = 40;
2694                }
2695                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2696                            find_unique_abbrev(one->sha1, abbrev));
2697                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2698                if (one->mode == two->mode)
2699                        strbuf_addf(msg, " %06o", one->mode);
2700                strbuf_addf(msg, "%s\n", reset);
2701        }
2702}
2703
2704static void run_diff_cmd(const char *pgm,
2705                         const char *name,
2706                         const char *other,
2707                         const char *attr_path,
2708                         struct diff_filespec *one,
2709                         struct diff_filespec *two,
2710                         struct strbuf *msg,
2711                         struct diff_options *o,
2712                         struct diff_filepair *p)
2713{
2714        const char *xfrm_msg = NULL;
2715        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2716        int must_show_header = 0;
2717
2718        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2719                pgm = NULL;
2720        else {
2721                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2722                if (drv && drv->external)
2723                        pgm = drv->external;
2724        }
2725
2726        if (msg) {
2727                /*
2728                 * don't use colors when the header is intended for an
2729                 * external diff driver
2730                 */
2731                fill_metainfo(msg, name, other, one, two, o, p,
2732                              &must_show_header,
2733                              DIFF_OPT_TST(o, COLOR_DIFF) && !pgm);
2734                xfrm_msg = msg->len ? msg->buf : NULL;
2735        }
2736
2737        if (pgm) {
2738                run_external_diff(pgm, name, other, one, two, xfrm_msg,
2739                                  complete_rewrite);
2740                return;
2741        }
2742        if (one && two)
2743                builtin_diff(name, other ? other : name,
2744                             one, two, xfrm_msg, must_show_header,
2745                             o, complete_rewrite);
2746        else
2747                fprintf(o->file, "* Unmerged path %s\n", name);
2748}
2749
2750static void diff_fill_sha1_info(struct diff_filespec *one)
2751{
2752        if (DIFF_FILE_VALID(one)) {
2753                if (!one->sha1_valid) {
2754                        struct stat st;
2755                        if (!strcmp(one->path, "-")) {
2756                                hashcpy(one->sha1, null_sha1);
2757                                return;
2758                        }
2759                        if (lstat(one->path, &st) < 0)
2760                                die_errno("stat '%s'", one->path);
2761                        if (index_path(one->sha1, one->path, &st, 0))
2762                                die("cannot hash %s", one->path);
2763                }
2764        }
2765        else
2766                hashclr(one->sha1);
2767}
2768
2769static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2770{
2771        /* Strip the prefix but do not molest /dev/null and absolute paths */
2772        if (*namep && **namep != '/') {
2773                *namep += prefix_length;
2774                if (**namep == '/')
2775                        ++*namep;
2776        }
2777        if (*otherp && **otherp != '/') {
2778                *otherp += prefix_length;
2779                if (**otherp == '/')
2780                        ++*otherp;
2781        }
2782}
2783
2784static void run_diff(struct diff_filepair *p, struct diff_options *o)
2785{
2786        const char *pgm = external_diff();
2787        struct strbuf msg;
2788        struct diff_filespec *one = p->one;
2789        struct diff_filespec *two = p->two;
2790        const char *name;
2791        const char *other;
2792        const char *attr_path;
2793
2794        name  = p->one->path;
2795        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2796        attr_path = name;
2797        if (o->prefix_length)
2798                strip_prefix(o->prefix_length, &name, &other);
2799
2800        if (DIFF_PAIR_UNMERGED(p)) {
2801                run_diff_cmd(pgm, name, NULL, attr_path,
2802                             NULL, NULL, NULL, o, p);
2803                return;
2804        }
2805
2806        diff_fill_sha1_info(one);
2807        diff_fill_sha1_info(two);
2808
2809        if (!pgm &&
2810            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2811            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2812                /*
2813                 * a filepair that changes between file and symlink
2814                 * needs to be split into deletion and creation.
2815                 */
2816                struct diff_filespec *null = alloc_filespec(two->path);
2817                run_diff_cmd(NULL, name, other, attr_path,
2818                             one, null, &msg, o, p);
2819                free(null);
2820                strbuf_release(&msg);
2821
2822                null = alloc_filespec(one->path);
2823                run_diff_cmd(NULL, name, other, attr_path,
2824                             null, two, &msg, o, p);
2825                free(null);
2826        }
2827        else
2828                run_diff_cmd(pgm, name, other, attr_path,
2829                             one, two, &msg, o, p);
2830
2831        strbuf_release(&msg);
2832}
2833
2834static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2835                         struct diffstat_t *diffstat)
2836{
2837        const char *name;
2838        const char *other;
2839        int complete_rewrite = 0;
2840
2841        if (DIFF_PAIR_UNMERGED(p)) {
2842                /* unmerged */
2843                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2844                return;
2845        }
2846
2847        name = p->one->path;
2848        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2849
2850        if (o->prefix_length)
2851                strip_prefix(o->prefix_length, &name, &other);
2852
2853        diff_fill_sha1_info(p->one);
2854        diff_fill_sha1_info(p->two);
2855
2856        if (p->status == DIFF_STATUS_MODIFIED && p->score)
2857                complete_rewrite = 1;
2858        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2859}
2860
2861static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2862{
2863        const char *name;
2864        const char *other;
2865        const char *attr_path;
2866
2867        if (DIFF_PAIR_UNMERGED(p)) {
2868                /* unmerged */
2869                return;
2870        }
2871
2872        name = p->one->path;
2873        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2874        attr_path = other ? other : name;
2875
2876        if (o->prefix_length)
2877                strip_prefix(o->prefix_length, &name, &other);
2878
2879        diff_fill_sha1_info(p->one);
2880        diff_fill_sha1_info(p->two);
2881
2882        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
2883}
2884
2885void diff_setup(struct diff_options *options)
2886{
2887        memcpy(options, &default_diff_options, sizeof(*options));
2888
2889        options->file = stdout;
2890
2891        options->line_termination = '\n';
2892        options->break_opt = -1;
2893        options->rename_limit = -1;
2894        options->dirstat_percent = 3;
2895        options->context = 3;
2896
2897        options->change = diff_change;
2898        options->add_remove = diff_addremove;
2899        if (diff_use_color_default > 0)
2900                DIFF_OPT_SET(options, COLOR_DIFF);
2901        options->detect_rename = diff_detect_rename_default;
2902
2903        if (diff_no_prefix) {
2904                options->a_prefix = options->b_prefix = "";
2905        } else if (!diff_mnemonic_prefix) {
2906                options->a_prefix = "a/";
2907                options->b_prefix = "b/";
2908        }
2909}
2910
2911int diff_setup_done(struct diff_options *options)
2912{
2913        int count = 0;
2914
2915        if (options->output_format & DIFF_FORMAT_NAME)
2916                count++;
2917        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2918                count++;
2919        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2920                count++;
2921        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2922                count++;
2923        if (count > 1)
2924                die("--name-only, --name-status, --check and -s are mutually exclusive");
2925
2926        /*
2927         * Most of the time we can say "there are changes"
2928         * only by checking if there are changed paths, but
2929         * --ignore-whitespace* options force us to look
2930         * inside contents.
2931         */
2932
2933        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
2934            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
2935            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
2936                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
2937        else
2938                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
2939
2940        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2941                options->detect_rename = DIFF_DETECT_COPY;
2942
2943        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
2944                options->prefix = NULL;
2945        if (options->prefix)
2946                options->prefix_length = strlen(options->prefix);
2947        else
2948                options->prefix_length = 0;
2949
2950        if (options->output_format & (DIFF_FORMAT_NAME |
2951                                      DIFF_FORMAT_NAME_STATUS |
2952                                      DIFF_FORMAT_CHECKDIFF |
2953                                      DIFF_FORMAT_NO_OUTPUT))
2954                options->output_format &= ~(DIFF_FORMAT_RAW |
2955                                            DIFF_FORMAT_NUMSTAT |
2956                                            DIFF_FORMAT_DIFFSTAT |
2957                                            DIFF_FORMAT_SHORTSTAT |
2958                                            DIFF_FORMAT_DIRSTAT |
2959                                            DIFF_FORMAT_SUMMARY |
2960                                            DIFF_FORMAT_PATCH);
2961
2962        /*
2963         * These cases always need recursive; we do not drop caller-supplied
2964         * recursive bits for other formats here.
2965         */
2966        if (options->output_format & (DIFF_FORMAT_PATCH |
2967                                      DIFF_FORMAT_NUMSTAT |
2968                                      DIFF_FORMAT_DIFFSTAT |
2969                                      DIFF_FORMAT_SHORTSTAT |
2970                                      DIFF_FORMAT_DIRSTAT |
2971                                      DIFF_FORMAT_SUMMARY |
2972                                      DIFF_FORMAT_CHECKDIFF))
2973                DIFF_OPT_SET(options, RECURSIVE);
2974        /*
2975         * Also pickaxe would not work very well if you do not say recursive
2976         */
2977        if (options->pickaxe)
2978                DIFF_OPT_SET(options, RECURSIVE);
2979        /*
2980         * When patches are generated, submodules diffed against the work tree
2981         * must be checked for dirtiness too so it can be shown in the output
2982         */
2983        if (options->output_format & DIFF_FORMAT_PATCH)
2984                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
2985
2986        if (options->detect_rename && options->rename_limit < 0)
2987                options->rename_limit = diff_rename_limit_default;
2988        if (options->setup & DIFF_SETUP_USE_CACHE) {
2989                if (!active_cache)
2990                        /* read-cache does not die even when it fails
2991                         * so it is safe for us to do this here.  Also
2992                         * it does not smudge active_cache or active_nr
2993                         * when it fails, so we do not have to worry about
2994                         * cleaning it up ourselves either.
2995                         */
2996                        read_cache();
2997        }
2998        if (options->abbrev <= 0 || 40 < options->abbrev)
2999                options->abbrev = 40; /* full */
3000
3001        /*
3002         * It does not make sense to show the first hit we happened
3003         * to have found.  It does not make sense not to return with
3004         * exit code in such a case either.
3005         */
3006        if (DIFF_OPT_TST(options, QUICK)) {
3007                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3008                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3009        }
3010
3011        return 0;
3012}
3013
3014static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3015{
3016        char c, *eq;
3017        int len;
3018
3019        if (*arg != '-')
3020                return 0;
3021        c = *++arg;
3022        if (!c)
3023                return 0;
3024        if (c == arg_short) {
3025                c = *++arg;
3026                if (!c)
3027                        return 1;
3028                if (val && isdigit(c)) {
3029                        char *end;
3030                        int n = strtoul(arg, &end, 10);
3031                        if (*end)
3032                                return 0;
3033                        *val = n;
3034                        return 1;
3035                }
3036                return 0;
3037        }
3038        if (c != '-')
3039                return 0;
3040        arg++;
3041        eq = strchr(arg, '=');
3042        if (eq)
3043                len = eq - arg;
3044        else
3045                len = strlen(arg);
3046        if (!len || strncmp(arg, arg_long, len))
3047                return 0;
3048        if (eq) {
3049                int n;
3050                char *end;
3051                if (!isdigit(*++eq))
3052                        return 0;
3053                n = strtoul(eq, &end, 10);
3054                if (*end)
3055                        return 0;
3056                *val = n;
3057        }
3058        return 1;
3059}
3060
3061static int diff_scoreopt_parse(const char *opt);
3062
3063static inline int short_opt(char opt, const char **argv,
3064                            const char **optarg)
3065{
3066        const char *arg = argv[0];
3067        if (arg[0] != '-' || arg[1] != opt)
3068                return 0;
3069        if (arg[2] != '\0') {
3070                *optarg = arg + 2;
3071                return 1;
3072        }
3073        if (!argv[1])
3074                die("Option '%c' requires a value", opt);
3075        *optarg = argv[1];
3076        return 2;
3077}
3078
3079int parse_long_opt(const char *opt, const char **argv,
3080                   const char **optarg)
3081{
3082        const char *arg = argv[0];
3083        if (arg[0] != '-' || arg[1] != '-')
3084                return 0;
3085        arg += strlen("--");
3086        if (prefixcmp(arg, opt))
3087                return 0;
3088        arg += strlen(opt);
3089        if (*arg == '=') { /* sticked form: --option=value */
3090                *optarg = arg + 1;
3091                return 1;
3092        }
3093        if (*arg != '\0')
3094                return 0;
3095        /* separate form: --option value */
3096        if (!argv[1])
3097                die("Option '--%s' requires a value", opt);
3098        *optarg = argv[1];
3099        return 2;
3100}
3101
3102static int stat_opt(struct diff_options *options, const char **av)
3103{
3104        const char *arg = av[0];
3105        char *end;
3106        int width = options->stat_width;
3107        int name_width = options->stat_name_width;
3108        int argcount = 1;
3109
3110        arg += strlen("--stat");
3111        end = (char *)arg;
3112
3113        switch (*arg) {
3114        case '-':
3115                if (!prefixcmp(arg, "-width")) {
3116                        arg += strlen("-width");
3117                        if (*arg == '=')
3118                                width = strtoul(arg + 1, &end, 10);
3119                        else if (!*arg && !av[1])
3120                                die("Option '--stat-width' requires a value");
3121                        else if (!*arg) {
3122                                width = strtoul(av[1], &end, 10);
3123                                argcount = 2;
3124                        }
3125                } else if (!prefixcmp(arg, "-name-width")) {
3126                        arg += strlen("-name-width");
3127                        if (*arg == '=')
3128                                name_width = strtoul(arg + 1, &end, 10);
3129                        else if (!*arg && !av[1])
3130                                die("Option '--stat-name-width' requires a value");
3131                        else if (!*arg) {
3132                                name_width = strtoul(av[1], &end, 10);
3133                                argcount = 2;
3134                        }
3135                }
3136                break;
3137        case '=':
3138                width = strtoul(arg+1, &end, 10);
3139                if (*end == ',')
3140                        name_width = strtoul(end+1, &end, 10);
3141        }
3142
3143        /* Important! This checks all the error cases! */
3144        if (*end)
3145                return 0;
3146        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3147        options->stat_name_width = name_width;
3148        options->stat_width = width;
3149        return argcount;
3150}
3151
3152int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3153{
3154        const char *arg = av[0];
3155        const char *optarg;
3156        int argcount;
3157
3158        /* Output format options */
3159        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3160                options->output_format |= DIFF_FORMAT_PATCH;
3161        else if (opt_arg(arg, 'U', "unified", &options->context))
3162                options->output_format |= DIFF_FORMAT_PATCH;
3163        else if (!strcmp(arg, "--raw"))
3164                options->output_format |= DIFF_FORMAT_RAW;
3165        else if (!strcmp(arg, "--patch-with-raw"))
3166                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3167        else if (!strcmp(arg, "--numstat"))
3168                options->output_format |= DIFF_FORMAT_NUMSTAT;
3169        else if (!strcmp(arg, "--shortstat"))
3170                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3171        else if (opt_arg(arg, 'X', "dirstat", &options->dirstat_percent))
3172                options->output_format |= DIFF_FORMAT_DIRSTAT;
3173        else if (!strcmp(arg, "--cumulative")) {
3174                options->output_format |= DIFF_FORMAT_DIRSTAT;
3175                DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
3176        } else if (opt_arg(arg, 0, "dirstat-by-file",
3177                           &options->dirstat_percent)) {
3178                options->output_format |= DIFF_FORMAT_DIRSTAT;
3179                DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
3180        }
3181        else if (!strcmp(arg, "--check"))
3182                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3183        else if (!strcmp(arg, "--summary"))
3184                options->output_format |= DIFF_FORMAT_SUMMARY;
3185        else if (!strcmp(arg, "--patch-with-stat"))
3186                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3187        else if (!strcmp(arg, "--name-only"))
3188                options->output_format |= DIFF_FORMAT_NAME;
3189        else if (!strcmp(arg, "--name-status"))
3190                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3191        else if (!strcmp(arg, "-s"))
3192                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3193        else if (!prefixcmp(arg, "--stat"))
3194                /* --stat, --stat-width, or --stat-name-width */
3195                return stat_opt(options, av);
3196
3197        /* renames options */
3198        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3199                 !strcmp(arg, "--break-rewrites")) {
3200                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3201                        return error("invalid argument to -B: %s", arg+2);
3202        }
3203        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3204                 !strcmp(arg, "--find-renames")) {
3205                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3206                        return error("invalid argument to -M: %s", arg+2);
3207                options->detect_rename = DIFF_DETECT_RENAME;
3208        }
3209        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3210                options->irreversible_delete = 1;
3211        }
3212        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3213                 !strcmp(arg, "--find-copies")) {
3214                if (options->detect_rename == DIFF_DETECT_COPY)
3215                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3216                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3217                        return error("invalid argument to -C: %s", arg+2);
3218                options->detect_rename = DIFF_DETECT_COPY;
3219        }
3220        else if (!strcmp(arg, "--no-renames"))
3221                options->detect_rename = 0;
3222        else if (!strcmp(arg, "--relative"))
3223                DIFF_OPT_SET(options, RELATIVE_NAME);
3224        else if (!prefixcmp(arg, "--relative=")) {
3225                DIFF_OPT_SET(options, RELATIVE_NAME);
3226                options->prefix = arg + 11;
3227        }
3228
3229        /* xdiff options */
3230        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3231                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3232        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3233                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3234        else if (!strcmp(arg, "--ignore-space-at-eol"))
3235                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3236        else if (!strcmp(arg, "--patience"))
3237                DIFF_XDL_SET(options, PATIENCE_DIFF);
3238
3239        /* flags options */
3240        else if (!strcmp(arg, "--binary")) {
3241                options->output_format |= DIFF_FORMAT_PATCH;
3242                DIFF_OPT_SET(options, BINARY);
3243        }
3244        else if (!strcmp(arg, "--full-index"))
3245                DIFF_OPT_SET(options, FULL_INDEX);
3246        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3247                DIFF_OPT_SET(options, TEXT);
3248        else if (!strcmp(arg, "-R"))
3249                DIFF_OPT_SET(options, REVERSE_DIFF);
3250        else if (!strcmp(arg, "--find-copies-harder"))
3251                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3252        else if (!strcmp(arg, "--follow"))
3253                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3254        else if (!strcmp(arg, "--color"))
3255                DIFF_OPT_SET(options, COLOR_DIFF);
3256        else if (!prefixcmp(arg, "--color=")) {
3257                int value = git_config_colorbool(NULL, arg+8, -1);
3258                if (value == 0)
3259                        DIFF_OPT_CLR(options, COLOR_DIFF);
3260                else if (value > 0)
3261                        DIFF_OPT_SET(options, COLOR_DIFF);
3262                else
3263                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3264        }
3265        else if (!strcmp(arg, "--no-color"))
3266                DIFF_OPT_CLR(options, COLOR_DIFF);
3267        else if (!strcmp(arg, "--color-words")) {
3268                DIFF_OPT_SET(options, COLOR_DIFF);
3269                options->word_diff = DIFF_WORDS_COLOR;
3270        }
3271        else if (!prefixcmp(arg, "--color-words=")) {
3272                DIFF_OPT_SET(options, COLOR_DIFF);
3273                options->word_diff = DIFF_WORDS_COLOR;
3274                options->word_regex = arg + 14;
3275        }
3276        else if (!strcmp(arg, "--word-diff")) {
3277                if (options->word_diff == DIFF_WORDS_NONE)
3278                        options->word_diff = DIFF_WORDS_PLAIN;
3279        }
3280        else if (!prefixcmp(arg, "--word-diff=")) {
3281                const char *type = arg + 12;
3282                if (!strcmp(type, "plain"))
3283                        options->word_diff = DIFF_WORDS_PLAIN;
3284                else if (!strcmp(type, "color")) {
3285                        DIFF_OPT_SET(options, COLOR_DIFF);
3286                        options->word_diff = DIFF_WORDS_COLOR;
3287                }
3288                else if (!strcmp(type, "porcelain"))
3289                        options->word_diff = DIFF_WORDS_PORCELAIN;
3290                else if (!strcmp(type, "none"))
3291                        options->word_diff = DIFF_WORDS_NONE;
3292                else
3293                        die("bad --word-diff argument: %s", type);
3294        }
3295        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3296                if (options->word_diff == DIFF_WORDS_NONE)
3297                        options->word_diff = DIFF_WORDS_PLAIN;
3298                options->word_regex = optarg;
3299                return argcount;
3300        }
3301        else if (!strcmp(arg, "--exit-code"))
3302                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3303        else if (!strcmp(arg, "--quiet"))
3304                DIFF_OPT_SET(options, QUICK);
3305        else if (!strcmp(arg, "--ext-diff"))
3306                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3307        else if (!strcmp(arg, "--no-ext-diff"))
3308                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3309        else if (!strcmp(arg, "--textconv"))
3310                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3311        else if (!strcmp(arg, "--no-textconv"))
3312                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3313        else if (!strcmp(arg, "--ignore-submodules")) {
3314                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3315                handle_ignore_submodules_arg(options, "all");
3316        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3317                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3318                handle_ignore_submodules_arg(options, arg + 20);
3319        } else if (!strcmp(arg, "--submodule"))
3320                DIFF_OPT_SET(options, SUBMODULE_LOG);
3321        else if (!prefixcmp(arg, "--submodule=")) {
3322                if (!strcmp(arg + 12, "log"))
3323                        DIFF_OPT_SET(options, SUBMODULE_LOG);
3324        }
3325
3326        /* misc options */
3327        else if (!strcmp(arg, "-z"))
3328                options->line_termination = 0;
3329        else if ((argcount = short_opt('l', av, &optarg))) {
3330                options->rename_limit = strtoul(optarg, NULL, 10);
3331                return argcount;
3332        }
3333        else if ((argcount = short_opt('S', av, &optarg))) {
3334                options->pickaxe = optarg;
3335                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3336                return argcount;
3337        } else if ((argcount = short_opt('G', av, &optarg))) {
3338                options->pickaxe = optarg;
3339                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3340                return argcount;
3341        }
3342        else if (!strcmp(arg, "--pickaxe-all"))
3343                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3344        else if (!strcmp(arg, "--pickaxe-regex"))
3345                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3346        else if ((argcount = short_opt('O', av, &optarg))) {
3347                options->orderfile = optarg;
3348                return argcount;
3349        }
3350        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3351                options->filter = optarg;
3352                return argcount;
3353        }
3354        else if (!strcmp(arg, "--abbrev"))
3355                options->abbrev = DEFAULT_ABBREV;
3356        else if (!prefixcmp(arg, "--abbrev=")) {
3357                options->abbrev = strtoul(arg + 9, NULL, 10);
3358                if (options->abbrev < MINIMUM_ABBREV)
3359                        options->abbrev = MINIMUM_ABBREV;
3360                else if (40 < options->abbrev)
3361                        options->abbrev = 40;
3362        }
3363        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3364                options->a_prefix = optarg;
3365                return argcount;
3366        }
3367        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3368                options->b_prefix = optarg;
3369                return argcount;
3370        }
3371        else if (!strcmp(arg, "--no-prefix"))
3372                options->a_prefix = options->b_prefix = "";
3373        else if (opt_arg(arg, '\0', "inter-hunk-context",
3374                         &options->interhunkcontext))
3375                ;
3376        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3377                options->file = fopen(optarg, "w");
3378                if (!options->file)
3379                        die_errno("Could not open '%s'", optarg);
3380                options->close_file = 1;
3381                return argcount;
3382        } else
3383                return 0;
3384        return 1;
3385}
3386
3387int parse_rename_score(const char **cp_p)
3388{
3389        unsigned long num, scale;
3390        int ch, dot;
3391        const char *cp = *cp_p;
3392
3393        num = 0;
3394        scale = 1;
3395        dot = 0;
3396        for (;;) {
3397                ch = *cp;
3398                if ( !dot && ch == '.' ) {
3399                        scale = 1;
3400                        dot = 1;
3401                } else if ( ch == '%' ) {
3402                        scale = dot ? scale*100 : 100;
3403                        cp++;   /* % is always at the end */
3404                        break;
3405                } else if ( ch >= '0' && ch <= '9' ) {
3406                        if ( scale < 100000 ) {
3407                                scale *= 10;
3408                                num = (num*10) + (ch-'0');
3409                        }
3410                } else {
3411                        break;
3412                }
3413                cp++;
3414        }
3415        *cp_p = cp;
3416
3417        /* user says num divided by scale and we say internally that
3418         * is MAX_SCORE * num / scale.
3419         */
3420        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3421}
3422
3423static int diff_scoreopt_parse(const char *opt)
3424{
3425        int opt1, opt2, cmd;
3426
3427        if (*opt++ != '-')
3428                return -1;
3429        cmd = *opt++;
3430        if (cmd == '-') {
3431                /* convert the long-form arguments into short-form versions */
3432                if (!prefixcmp(opt, "break-rewrites")) {
3433                        opt += strlen("break-rewrites");
3434                        if (*opt == 0 || *opt++ == '=')
3435                                cmd = 'B';
3436                } else if (!prefixcmp(opt, "find-copies")) {
3437                        opt += strlen("find-copies");
3438                        if (*opt == 0 || *opt++ == '=')
3439                                cmd = 'C';
3440                } else if (!prefixcmp(opt, "find-renames")) {
3441                        opt += strlen("find-renames");
3442                        if (*opt == 0 || *opt++ == '=')
3443                                cmd = 'M';
3444                }
3445        }
3446        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3447                return -1; /* that is not a -M, -C nor -B option */
3448
3449        opt1 = parse_rename_score(&opt);
3450        if (cmd != 'B')
3451                opt2 = 0;
3452        else {
3453                if (*opt == 0)
3454                        opt2 = 0;
3455                else if (*opt != '/')
3456                        return -1; /* we expect -B80/99 or -B80 */
3457                else {
3458                        opt++;
3459                        opt2 = parse_rename_score(&opt);
3460                }
3461        }
3462        if (*opt != 0)
3463                return -1;
3464        return opt1 | (opt2 << 16);
3465}
3466
3467struct diff_queue_struct diff_queued_diff;
3468
3469void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3470{
3471        if (queue->alloc <= queue->nr) {
3472                queue->alloc = alloc_nr(queue->alloc);
3473                queue->queue = xrealloc(queue->queue,
3474                                        sizeof(dp) * queue->alloc);
3475        }
3476        queue->queue[queue->nr++] = dp;
3477}
3478
3479struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3480                                 struct diff_filespec *one,
3481                                 struct diff_filespec *two)
3482{
3483        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3484        dp->one = one;
3485        dp->two = two;
3486        if (queue)
3487                diff_q(queue, dp);
3488        return dp;
3489}
3490
3491void diff_free_filepair(struct diff_filepair *p)
3492{
3493        free_filespec(p->one);
3494        free_filespec(p->two);
3495        free(p);
3496}
3497
3498/* This is different from find_unique_abbrev() in that
3499 * it stuffs the result with dots for alignment.
3500 */
3501const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3502{
3503        int abblen;
3504        const char *abbrev;
3505        if (len == 40)
3506                return sha1_to_hex(sha1);
3507
3508        abbrev = find_unique_abbrev(sha1, len);
3509        abblen = strlen(abbrev);
3510        if (abblen < 37) {
3511                static char hex[41];
3512                if (len < abblen && abblen <= len + 2)
3513                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3514                else
3515                        sprintf(hex, "%s...", abbrev);
3516                return hex;
3517        }
3518        return sha1_to_hex(sha1);
3519}
3520
3521static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3522{
3523        int line_termination = opt->line_termination;
3524        int inter_name_termination = line_termination ? '\t' : '\0';
3525        if (opt->output_prefix) {
3526                struct strbuf *msg = NULL;
3527                msg = opt->output_prefix(opt, opt->output_prefix_data);
3528                fprintf(opt->file, "%s", msg->buf);
3529        }
3530
3531        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3532                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3533                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3534                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3535        }
3536        if (p->score) {
3537                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3538                        inter_name_termination);
3539        } else {
3540                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3541        }
3542
3543        if (p->status == DIFF_STATUS_COPIED ||
3544            p->status == DIFF_STATUS_RENAMED) {
3545                const char *name_a, *name_b;
3546                name_a = p->one->path;
3547                name_b = p->two->path;
3548                strip_prefix(opt->prefix_length, &name_a, &name_b);
3549                write_name_quoted(name_a, opt->file, inter_name_termination);
3550                write_name_quoted(name_b, opt->file, line_termination);
3551        } else {
3552                const char *name_a, *name_b;
3553                name_a = p->one->mode ? p->one->path : p->two->path;
3554                name_b = NULL;
3555                strip_prefix(opt->prefix_length, &name_a, &name_b);
3556                write_name_quoted(name_a, opt->file, line_termination);
3557        }
3558}
3559
3560int diff_unmodified_pair(struct diff_filepair *p)
3561{
3562        /* This function is written stricter than necessary to support
3563         * the currently implemented transformers, but the idea is to
3564         * let transformers to produce diff_filepairs any way they want,
3565         * and filter and clean them up here before producing the output.
3566         */
3567        struct diff_filespec *one = p->one, *two = p->two;
3568
3569        if (DIFF_PAIR_UNMERGED(p))
3570                return 0; /* unmerged is interesting */
3571
3572        /* deletion, addition, mode or type change
3573         * and rename are all interesting.
3574         */
3575        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3576            DIFF_PAIR_MODE_CHANGED(p) ||
3577            strcmp(one->path, two->path))
3578                return 0;
3579
3580        /* both are valid and point at the same path.  that is, we are
3581         * dealing with a change.
3582         */
3583        if (one->sha1_valid && two->sha1_valid &&
3584            !hashcmp(one->sha1, two->sha1) &&
3585            !one->dirty_submodule && !two->dirty_submodule)
3586                return 1; /* no change */
3587        if (!one->sha1_valid && !two->sha1_valid)
3588                return 1; /* both look at the same file on the filesystem. */
3589        return 0;
3590}
3591
3592static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3593{
3594        if (diff_unmodified_pair(p))
3595                return;
3596
3597        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3598            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3599                return; /* no tree diffs in patch format */
3600
3601        run_diff(p, o);
3602}
3603
3604static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3605                            struct diffstat_t *diffstat)
3606{
3607        if (diff_unmodified_pair(p))
3608                return;
3609
3610        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3611            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3612                return; /* no useful stat for tree diffs */
3613
3614        run_diffstat(p, o, diffstat);
3615}
3616
3617static void diff_flush_checkdiff(struct diff_filepair *p,
3618                struct diff_options *o)
3619{
3620        if (diff_unmodified_pair(p))
3621                return;
3622
3623        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3624            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3625                return; /* nothing to check in tree diffs */
3626
3627        run_checkdiff(p, o);
3628}
3629
3630int diff_queue_is_empty(void)
3631{
3632        struct diff_queue_struct *q = &diff_queued_diff;
3633        int i;
3634        for (i = 0; i < q->nr; i++)
3635                if (!diff_unmodified_pair(q->queue[i]))
3636                        return 0;
3637        return 1;
3638}
3639
3640#if DIFF_DEBUG
3641void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3642{
3643        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3644                x, one ? one : "",
3645                s->path,
3646                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3647                s->mode,
3648                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3649        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3650                x, one ? one : "",
3651                s->size, s->xfrm_flags);
3652}
3653
3654void diff_debug_filepair(const struct diff_filepair *p, int i)
3655{
3656        diff_debug_filespec(p->one, i, "one");
3657        diff_debug_filespec(p->two, i, "two");
3658        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3659                p->score, p->status ? p->status : '?',
3660                p->one->rename_used, p->broken_pair);
3661}
3662
3663void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3664{
3665        int i;
3666        if (msg)
3667                fprintf(stderr, "%s\n", msg);
3668        fprintf(stderr, "q->nr = %d\n", q->nr);
3669        for (i = 0; i < q->nr; i++) {
3670                struct diff_filepair *p = q->queue[i];
3671                diff_debug_filepair(p, i);
3672        }
3673}
3674#endif
3675
3676static void diff_resolve_rename_copy(void)
3677{
3678        int i;
3679        struct diff_filepair *p;
3680        struct diff_queue_struct *q = &diff_queued_diff;
3681
3682        diff_debug_queue("resolve-rename-copy", q);
3683
3684        for (i = 0; i < q->nr; i++) {
3685                p = q->queue[i];
3686                p->status = 0; /* undecided */
3687                if (DIFF_PAIR_UNMERGED(p))
3688                        p->status = DIFF_STATUS_UNMERGED;
3689                else if (!DIFF_FILE_VALID(p->one))
3690                        p->status = DIFF_STATUS_ADDED;
3691                else if (!DIFF_FILE_VALID(p->two))
3692                        p->status = DIFF_STATUS_DELETED;
3693                else if (DIFF_PAIR_TYPE_CHANGED(p))
3694                        p->status = DIFF_STATUS_TYPE_CHANGED;
3695
3696                /* from this point on, we are dealing with a pair
3697                 * whose both sides are valid and of the same type, i.e.
3698                 * either in-place edit or rename/copy edit.
3699                 */
3700                else if (DIFF_PAIR_RENAME(p)) {
3701                        /*
3702                         * A rename might have re-connected a broken
3703                         * pair up, causing the pathnames to be the
3704                         * same again. If so, that's not a rename at
3705                         * all, just a modification..
3706                         *
3707                         * Otherwise, see if this source was used for
3708                         * multiple renames, in which case we decrement
3709                         * the count, and call it a copy.
3710                         */
3711                        if (!strcmp(p->one->path, p->two->path))
3712                                p->status = DIFF_STATUS_MODIFIED;
3713                        else if (--p->one->rename_used > 0)
3714                                p->status = DIFF_STATUS_COPIED;
3715                        else
3716                                p->status = DIFF_STATUS_RENAMED;
3717                }
3718                else if (hashcmp(p->one->sha1, p->two->sha1) ||
3719                         p->one->mode != p->two->mode ||
3720                         p->one->dirty_submodule ||
3721                         p->two->dirty_submodule ||
3722                         is_null_sha1(p->one->sha1))
3723                        p->status = DIFF_STATUS_MODIFIED;
3724                else {
3725                        /* This is a "no-change" entry and should not
3726                         * happen anymore, but prepare for broken callers.
3727                         */
3728                        error("feeding unmodified %s to diffcore",
3729                              p->one->path);
3730                        p->status = DIFF_STATUS_UNKNOWN;
3731                }
3732        }
3733        diff_debug_queue("resolve-rename-copy done", q);
3734}
3735
3736static int check_pair_status(struct diff_filepair *p)
3737{
3738        switch (p->status) {
3739        case DIFF_STATUS_UNKNOWN:
3740                return 0;
3741        case 0:
3742                die("internal error in diff-resolve-rename-copy");
3743        default:
3744                return 1;
3745        }
3746}
3747
3748static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3749{
3750        int fmt = opt->output_format;
3751
3752        if (fmt & DIFF_FORMAT_CHECKDIFF)
3753                diff_flush_checkdiff(p, opt);
3754        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
3755                diff_flush_raw(p, opt);
3756        else if (fmt & DIFF_FORMAT_NAME) {
3757                const char *name_a, *name_b;
3758                name_a = p->two->path;
3759                name_b = NULL;
3760                strip_prefix(opt->prefix_length, &name_a, &name_b);
3761                write_name_quoted(name_a, opt->file, opt->line_termination);
3762        }
3763}
3764
3765static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
3766{
3767        if (fs->mode)
3768                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
3769        else
3770                fprintf(file, " %s ", newdelete);
3771        write_name_quoted(fs->path, file, '\n');
3772}
3773
3774
3775static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
3776                const char *line_prefix)
3777{
3778        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
3779                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
3780                        p->two->mode, show_name ? ' ' : '\n');
3781                if (show_name) {
3782                        write_name_quoted(p->two->path, file, '\n');
3783                }
3784        }
3785}
3786
3787static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
3788                        const char *line_prefix)
3789{
3790        char *names = pprint_rename(p->one->path, p->two->path);
3791
3792        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
3793        free(names);
3794        show_mode_change(file, p, 0, line_prefix);
3795}
3796
3797static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
3798{
3799        FILE *file = opt->file;
3800        char *line_prefix = "";
3801
3802        if (opt->output_prefix) {
3803                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
3804                line_prefix = buf->buf;
3805        }
3806
3807        switch(p->status) {
3808        case DIFF_STATUS_DELETED:
3809                fputs(line_prefix, file);
3810                show_file_mode_name(file, "delete", p->one);
3811                break;
3812        case DIFF_STATUS_ADDED:
3813                fputs(line_prefix, file);
3814                show_file_mode_name(file, "create", p->two);
3815                break;
3816        case DIFF_STATUS_COPIED:
3817                fputs(line_prefix, file);
3818                show_rename_copy(file, "copy", p, line_prefix);
3819                break;
3820        case DIFF_STATUS_RENAMED:
3821                fputs(line_prefix, file);
3822                show_rename_copy(file, "rename", p, line_prefix);
3823                break;
3824        default:
3825                if (p->score) {
3826                        fprintf(file, "%s rewrite ", line_prefix);
3827                        write_name_quoted(p->two->path, file, ' ');
3828                        fprintf(file, "(%d%%)\n", similarity_index(p));
3829                }
3830                show_mode_change(file, p, !p->score, line_prefix);
3831                break;
3832        }
3833}
3834
3835struct patch_id_t {
3836        git_SHA_CTX *ctx;
3837        int patchlen;
3838};
3839
3840static int remove_space(char *line, int len)
3841{
3842        int i;
3843        char *dst = line;
3844        unsigned char c;
3845
3846        for (i = 0; i < len; i++)
3847                if (!isspace((c = line[i])))
3848                        *dst++ = c;
3849
3850        return dst - line;
3851}
3852
3853static void patch_id_consume(void *priv, char *line, unsigned long len)
3854{
3855        struct patch_id_t *data = priv;
3856        int new_len;
3857
3858        /* Ignore line numbers when computing the SHA1 of the patch */
3859        if (!prefixcmp(line, "@@ -"))
3860                return;
3861
3862        new_len = remove_space(line, len);
3863
3864        git_SHA1_Update(data->ctx, line, new_len);
3865        data->patchlen += new_len;
3866}
3867
3868/* returns 0 upon success, and writes result into sha1 */
3869static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
3870{
3871        struct diff_queue_struct *q = &diff_queued_diff;
3872        int i;
3873        git_SHA_CTX ctx;
3874        struct patch_id_t data;
3875        char buffer[PATH_MAX * 4 + 20];
3876
3877        git_SHA1_Init(&ctx);
3878        memset(&data, 0, sizeof(struct patch_id_t));
3879        data.ctx = &ctx;
3880
3881        for (i = 0; i < q->nr; i++) {
3882                xpparam_t xpp;
3883                xdemitconf_t xecfg;
3884                mmfile_t mf1, mf2;
3885                struct diff_filepair *p = q->queue[i];
3886                int len1, len2;
3887
3888                memset(&xpp, 0, sizeof(xpp));
3889                memset(&xecfg, 0, sizeof(xecfg));
3890                if (p->status == 0)
3891                        return error("internal diff status error");
3892                if (p->status == DIFF_STATUS_UNKNOWN)
3893                        continue;
3894                if (diff_unmodified_pair(p))
3895                        continue;
3896                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3897                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3898                        continue;
3899                if (DIFF_PAIR_UNMERGED(p))
3900                        continue;
3901
3902                diff_fill_sha1_info(p->one);
3903                diff_fill_sha1_info(p->two);
3904                if (fill_mmfile(&mf1, p->one) < 0 ||
3905                                fill_mmfile(&mf2, p->two) < 0)
3906                        return error("unable to read files to diff");
3907
3908                len1 = remove_space(p->one->path, strlen(p->one->path));
3909                len2 = remove_space(p->two->path, strlen(p->two->path));
3910                if (p->one->mode == 0)
3911                        len1 = snprintf(buffer, sizeof(buffer),
3912                                        "diff--gita/%.*sb/%.*s"
3913                                        "newfilemode%06o"
3914                                        "---/dev/null"
3915                                        "+++b/%.*s",
3916                                        len1, p->one->path,
3917                                        len2, p->two->path,
3918                                        p->two->mode,
3919                                        len2, p->two->path);
3920                else if (p->two->mode == 0)
3921                        len1 = snprintf(buffer, sizeof(buffer),
3922                                        "diff--gita/%.*sb/%.*s"
3923                                        "deletedfilemode%06o"
3924                                        "---a/%.*s"
3925                                        "+++/dev/null",
3926                                        len1, p->one->path,
3927                                        len2, p->two->path,
3928                                        p->one->mode,
3929                                        len1, p->one->path);
3930                else
3931                        len1 = snprintf(buffer, sizeof(buffer),
3932                                        "diff--gita/%.*sb/%.*s"
3933                                        "---a/%.*s"
3934                                        "+++b/%.*s",
3935                                        len1, p->one->path,
3936                                        len2, p->two->path,
3937                                        len1, p->one->path,
3938                                        len2, p->two->path);
3939                git_SHA1_Update(&ctx, buffer, len1);
3940
3941                if (diff_filespec_is_binary(p->one) ||
3942                    diff_filespec_is_binary(p->two)) {
3943                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
3944                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
3945                        continue;
3946                }
3947
3948                xpp.flags = 0;
3949                xecfg.ctxlen = 3;
3950                xecfg.flags = 0;
3951                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
3952                              &xpp, &xecfg);
3953        }
3954
3955        git_SHA1_Final(sha1, &ctx);
3956        return 0;
3957}
3958
3959int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
3960{
3961        struct diff_queue_struct *q = &diff_queued_diff;
3962        int i;
3963        int result = diff_get_patch_id(options, sha1);
3964
3965        for (i = 0; i < q->nr; i++)
3966                diff_free_filepair(q->queue[i]);
3967
3968        free(q->queue);
3969        DIFF_QUEUE_CLEAR(q);
3970
3971        return result;
3972}
3973
3974static int is_summary_empty(const struct diff_queue_struct *q)
3975{
3976        int i;
3977
3978        for (i = 0; i < q->nr; i++) {
3979                const struct diff_filepair *p = q->queue[i];
3980
3981                switch (p->status) {
3982                case DIFF_STATUS_DELETED:
3983                case DIFF_STATUS_ADDED:
3984                case DIFF_STATUS_COPIED:
3985                case DIFF_STATUS_RENAMED:
3986                        return 0;
3987                default:
3988                        if (p->score)
3989                                return 0;
3990                        if (p->one->mode && p->two->mode &&
3991                            p->one->mode != p->two->mode)
3992                                return 0;
3993                        break;
3994                }
3995        }
3996        return 1;
3997}
3998
3999static const char rename_limit_warning[] =
4000"inexact rename detection was skipped due to too many files.";
4001
4002static const char degrade_cc_to_c_warning[] =
4003"only found copies from modified paths due to too many files.";
4004
4005static const char rename_limit_advice[] =
4006"you may want to set your %s variable to at least "
4007"%d and retry the command.";
4008
4009void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4010{
4011        if (degraded_cc)
4012                warning(degrade_cc_to_c_warning);
4013        else if (needed)
4014                warning(rename_limit_warning);
4015        else
4016                return;
4017        if (0 < needed && needed < 32767)
4018                warning(rename_limit_advice, varname, needed);
4019}
4020
4021void diff_flush(struct diff_options *options)
4022{
4023        struct diff_queue_struct *q = &diff_queued_diff;
4024        int i, output_format = options->output_format;
4025        int separator = 0;
4026
4027        /*
4028         * Order: raw, stat, summary, patch
4029         * or:    name/name-status/checkdiff (other bits clear)
4030         */
4031        if (!q->nr)
4032                goto free_queue;
4033
4034        if (output_format & (DIFF_FORMAT_RAW |
4035                             DIFF_FORMAT_NAME |
4036                             DIFF_FORMAT_NAME_STATUS |
4037                             DIFF_FORMAT_CHECKDIFF)) {
4038                for (i = 0; i < q->nr; i++) {
4039                        struct diff_filepair *p = q->queue[i];
4040                        if (check_pair_status(p))
4041                                flush_one_pair(p, options);
4042                }
4043                separator++;
4044        }
4045
4046        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
4047                struct diffstat_t diffstat;
4048
4049                memset(&diffstat, 0, sizeof(struct diffstat_t));
4050                for (i = 0; i < q->nr; i++) {
4051                        struct diff_filepair *p = q->queue[i];
4052                        if (check_pair_status(p))
4053                                diff_flush_stat(p, options, &diffstat);
4054                }
4055                if (output_format & DIFF_FORMAT_NUMSTAT)
4056                        show_numstat(&diffstat, options);
4057                if (output_format & DIFF_FORMAT_DIFFSTAT)
4058                        show_stats(&diffstat, options);
4059                if (output_format & DIFF_FORMAT_SHORTSTAT)
4060                        show_shortstats(&diffstat, options);
4061                free_diffstat_info(&diffstat);
4062                separator++;
4063        }
4064        if (output_format & DIFF_FORMAT_DIRSTAT)
4065                show_dirstat(options);
4066
4067        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4068                for (i = 0; i < q->nr; i++) {
4069                        diff_summary(options, q->queue[i]);
4070                }
4071                separator++;
4072        }
4073
4074        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4075            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4076            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4077                /*
4078                 * run diff_flush_patch for the exit status. setting
4079                 * options->file to /dev/null should be safe, becaue we
4080                 * aren't supposed to produce any output anyway.
4081                 */
4082                if (options->close_file)
4083                        fclose(options->file);
4084                options->file = fopen("/dev/null", "w");
4085                if (!options->file)
4086                        die_errno("Could not open /dev/null");
4087                options->close_file = 1;
4088                for (i = 0; i < q->nr; i++) {
4089                        struct diff_filepair *p = q->queue[i];
4090                        if (check_pair_status(p))
4091                                diff_flush_patch(p, options);
4092                        if (options->found_changes)
4093                                break;
4094                }
4095        }
4096
4097        if (output_format & DIFF_FORMAT_PATCH) {
4098                if (separator) {
4099                        putc(options->line_termination, options->file);
4100                        if (options->stat_sep) {
4101                                /* attach patch instead of inline */
4102                                fputs(options->stat_sep, options->file);
4103                        }
4104                }
4105
4106                for (i = 0; i < q->nr; i++) {
4107                        struct diff_filepair *p = q->queue[i];
4108                        if (check_pair_status(p))
4109                                diff_flush_patch(p, options);
4110                }
4111        }
4112
4113        if (output_format & DIFF_FORMAT_CALLBACK)
4114                options->format_callback(q, options, options->format_callback_data);
4115
4116        for (i = 0; i < q->nr; i++)
4117                diff_free_filepair(q->queue[i]);
4118free_queue:
4119        free(q->queue);
4120        DIFF_QUEUE_CLEAR(q);
4121        if (options->close_file)
4122                fclose(options->file);
4123
4124        /*
4125         * Report the content-level differences with HAS_CHANGES;
4126         * diff_addremove/diff_change does not set the bit when
4127         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4128         */
4129        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4130                if (options->found_changes)
4131                        DIFF_OPT_SET(options, HAS_CHANGES);
4132                else
4133                        DIFF_OPT_CLR(options, HAS_CHANGES);
4134        }
4135}
4136
4137static void diffcore_apply_filter(const char *filter)
4138{
4139        int i;
4140        struct diff_queue_struct *q = &diff_queued_diff;
4141        struct diff_queue_struct outq;
4142        DIFF_QUEUE_CLEAR(&outq);
4143
4144        if (!filter)
4145                return;
4146
4147        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4148                int found;
4149                for (i = found = 0; !found && i < q->nr; i++) {
4150                        struct diff_filepair *p = q->queue[i];
4151                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4152                             ((p->score &&
4153                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4154                              (!p->score &&
4155                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4156                            ((p->status != DIFF_STATUS_MODIFIED) &&
4157                             strchr(filter, p->status)))
4158                                found++;
4159                }
4160                if (found)
4161                        return;
4162
4163                /* otherwise we will clear the whole queue
4164                 * by copying the empty outq at the end of this
4165                 * function, but first clear the current entries
4166                 * in the queue.
4167                 */
4168                for (i = 0; i < q->nr; i++)
4169                        diff_free_filepair(q->queue[i]);
4170        }
4171        else {
4172                /* Only the matching ones */
4173                for (i = 0; i < q->nr; i++) {
4174                        struct diff_filepair *p = q->queue[i];
4175
4176                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4177                             ((p->score &&
4178                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4179                              (!p->score &&
4180                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4181                            ((p->status != DIFF_STATUS_MODIFIED) &&
4182                             strchr(filter, p->status)))
4183                                diff_q(&outq, p);
4184                        else
4185                                diff_free_filepair(p);
4186                }
4187        }
4188        free(q->queue);
4189        *q = outq;
4190}
4191
4192/* Check whether two filespecs with the same mode and size are identical */
4193static int diff_filespec_is_identical(struct diff_filespec *one,
4194                                      struct diff_filespec *two)
4195{
4196        if (S_ISGITLINK(one->mode))
4197                return 0;
4198        if (diff_populate_filespec(one, 0))
4199                return 0;
4200        if (diff_populate_filespec(two, 0))
4201                return 0;
4202        return !memcmp(one->data, two->data, one->size);
4203}
4204
4205static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4206{
4207        int i;
4208        struct diff_queue_struct *q = &diff_queued_diff;
4209        struct diff_queue_struct outq;
4210        DIFF_QUEUE_CLEAR(&outq);
4211
4212        for (i = 0; i < q->nr; i++) {
4213                struct diff_filepair *p = q->queue[i];
4214
4215                /*
4216                 * 1. Entries that come from stat info dirtiness
4217                 *    always have both sides (iow, not create/delete),
4218                 *    one side of the object name is unknown, with
4219                 *    the same mode and size.  Keep the ones that
4220                 *    do not match these criteria.  They have real
4221                 *    differences.
4222                 *
4223                 * 2. At this point, the file is known to be modified,
4224                 *    with the same mode and size, and the object
4225                 *    name of one side is unknown.  Need to inspect
4226                 *    the identical contents.
4227                 */
4228                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4229                    !DIFF_FILE_VALID(p->two) ||
4230                    (p->one->sha1_valid && p->two->sha1_valid) ||
4231                    (p->one->mode != p->two->mode) ||
4232                    diff_populate_filespec(p->one, 1) ||
4233                    diff_populate_filespec(p->two, 1) ||
4234                    (p->one->size != p->two->size) ||
4235                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4236                        diff_q(&outq, p);
4237                else {
4238                        /*
4239                         * The caller can subtract 1 from skip_stat_unmatch
4240                         * to determine how many paths were dirty only
4241                         * due to stat info mismatch.
4242                         */
4243                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4244                                diffopt->skip_stat_unmatch++;
4245                        diff_free_filepair(p);
4246                }
4247        }
4248        free(q->queue);
4249        *q = outq;
4250}
4251
4252static int diffnamecmp(const void *a_, const void *b_)
4253{
4254        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4255        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4256        const char *name_a, *name_b;
4257
4258        name_a = a->one ? a->one->path : a->two->path;
4259        name_b = b->one ? b->one->path : b->two->path;
4260        return strcmp(name_a, name_b);
4261}
4262
4263void diffcore_fix_diff_index(struct diff_options *options)
4264{
4265        struct diff_queue_struct *q = &diff_queued_diff;
4266        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4267}
4268
4269void diffcore_std(struct diff_options *options)
4270{
4271        if (options->skip_stat_unmatch)
4272                diffcore_skip_stat_unmatch(options);
4273        if (!options->found_follow) {
4274                /* See try_to_follow_renames() in tree-diff.c */
4275                if (options->break_opt != -1)
4276                        diffcore_break(options->break_opt);
4277                if (options->detect_rename)
4278                        diffcore_rename(options);
4279                if (options->break_opt != -1)
4280                        diffcore_merge_broken();
4281        }
4282        if (options->pickaxe)
4283                diffcore_pickaxe(options);
4284        if (options->orderfile)
4285                diffcore_order(options->orderfile);
4286        if (!options->found_follow)
4287                /* See try_to_follow_renames() in tree-diff.c */
4288                diff_resolve_rename_copy();
4289        diffcore_apply_filter(options->filter);
4290
4291        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4292                DIFF_OPT_SET(options, HAS_CHANGES);
4293        else
4294                DIFF_OPT_CLR(options, HAS_CHANGES);
4295
4296        options->found_follow = 0;
4297}
4298
4299int diff_result_code(struct diff_options *opt, int status)
4300{
4301        int result = 0;
4302
4303        diff_warn_rename_limit("diff.renamelimit",
4304                               opt->needed_rename_limit,
4305                               opt->degraded_cc_to_c);
4306        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4307            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4308                return status;
4309        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4310            DIFF_OPT_TST(opt, HAS_CHANGES))
4311                result |= 01;
4312        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4313            DIFF_OPT_TST(opt, CHECK_FAILED))
4314                result |= 02;
4315        return result;
4316}
4317
4318/*
4319 * Shall changes to this submodule be ignored?
4320 *
4321 * Submodule changes can be configured to be ignored separately for each path,
4322 * but that configuration can be overridden from the command line.
4323 */
4324static int is_submodule_ignored(const char *path, struct diff_options *options)
4325{
4326        int ignored = 0;
4327        unsigned orig_flags = options->flags;
4328        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4329                set_diffopt_flags_from_submodule_config(options, path);
4330        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4331                ignored = 1;
4332        options->flags = orig_flags;
4333        return ignored;
4334}
4335
4336void diff_addremove(struct diff_options *options,
4337                    int addremove, unsigned mode,
4338                    const unsigned char *sha1,
4339                    const char *concatpath, unsigned dirty_submodule)
4340{
4341        struct diff_filespec *one, *two;
4342
4343        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4344                return;
4345
4346        /* This may look odd, but it is a preparation for
4347         * feeding "there are unchanged files which should
4348         * not produce diffs, but when you are doing copy
4349         * detection you would need them, so here they are"
4350         * entries to the diff-core.  They will be prefixed
4351         * with something like '=' or '*' (I haven't decided
4352         * which but should not make any difference).
4353         * Feeding the same new and old to diff_change()
4354         * also has the same effect.
4355         * Before the final output happens, they are pruned after
4356         * merged into rename/copy pairs as appropriate.
4357         */
4358        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4359                addremove = (addremove == '+' ? '-' :
4360                             addremove == '-' ? '+' : addremove);
4361
4362        if (options->prefix &&
4363            strncmp(concatpath, options->prefix, options->prefix_length))
4364                return;
4365
4366        one = alloc_filespec(concatpath);
4367        two = alloc_filespec(concatpath);
4368
4369        if (addremove != '+')
4370                fill_filespec(one, sha1, mode);
4371        if (addremove != '-') {
4372                fill_filespec(two, sha1, mode);
4373                two->dirty_submodule = dirty_submodule;
4374        }
4375
4376        diff_queue(&diff_queued_diff, one, two);
4377        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4378                DIFF_OPT_SET(options, HAS_CHANGES);
4379}
4380
4381void diff_change(struct diff_options *options,
4382                 unsigned old_mode, unsigned new_mode,
4383                 const unsigned char *old_sha1,
4384                 const unsigned char *new_sha1,
4385                 const char *concatpath,
4386                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4387{
4388        struct diff_filespec *one, *two;
4389
4390        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4391            is_submodule_ignored(concatpath, options))
4392                return;
4393
4394        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4395                unsigned tmp;
4396                const unsigned char *tmp_c;
4397                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4398                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4399                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4400                        new_dirty_submodule = tmp;
4401        }
4402
4403        if (options->prefix &&
4404            strncmp(concatpath, options->prefix, options->prefix_length))
4405                return;
4406
4407        one = alloc_filespec(concatpath);
4408        two = alloc_filespec(concatpath);
4409        fill_filespec(one, old_sha1, old_mode);
4410        fill_filespec(two, new_sha1, new_mode);
4411        one->dirty_submodule = old_dirty_submodule;
4412        two->dirty_submodule = new_dirty_submodule;
4413
4414        diff_queue(&diff_queued_diff, one, two);
4415        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4416                DIFF_OPT_SET(options, HAS_CHANGES);
4417}
4418
4419struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4420{
4421        struct diff_filepair *pair;
4422        struct diff_filespec *one, *two;
4423
4424        if (options->prefix &&
4425            strncmp(path, options->prefix, options->prefix_length))
4426                return NULL;
4427
4428        one = alloc_filespec(path);
4429        two = alloc_filespec(path);
4430        pair = diff_queue(&diff_queued_diff, one, two);
4431        pair->is_unmerged = 1;
4432        return pair;
4433}
4434
4435static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4436                size_t *outsize)
4437{
4438        struct diff_tempfile *temp;
4439        const char *argv[3];
4440        const char **arg = argv;
4441        struct child_process child;
4442        struct strbuf buf = STRBUF_INIT;
4443        int err = 0;
4444
4445        temp = prepare_temp_file(spec->path, spec);
4446        *arg++ = pgm;
4447        *arg++ = temp->name;
4448        *arg = NULL;
4449
4450        memset(&child, 0, sizeof(child));
4451        child.use_shell = 1;
4452        child.argv = argv;
4453        child.out = -1;
4454        if (start_command(&child)) {
4455                remove_tempfile();
4456                return NULL;
4457        }
4458
4459        if (strbuf_read(&buf, child.out, 0) < 0)
4460                err = error("error reading from textconv command '%s'", pgm);
4461        close(child.out);
4462
4463        if (finish_command(&child) || err) {
4464                strbuf_release(&buf);
4465                remove_tempfile();
4466                return NULL;
4467        }
4468        remove_tempfile();
4469
4470        return strbuf_detach(&buf, outsize);
4471}
4472
4473size_t fill_textconv(struct userdiff_driver *driver,
4474                     struct diff_filespec *df,
4475                     char **outbuf)
4476{
4477        size_t size;
4478
4479        if (!driver || !driver->textconv) {
4480                if (!DIFF_FILE_VALID(df)) {
4481                        *outbuf = "";
4482                        return 0;
4483                }
4484                if (diff_populate_filespec(df, 0))
4485                        die("unable to read files to diff");
4486                *outbuf = df->data;
4487                return df->size;
4488        }
4489
4490        if (driver->textconv_cache && df->sha1_valid) {
4491                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4492                                          &size);
4493                if (*outbuf)
4494                        return size;
4495        }
4496
4497        *outbuf = run_textconv(driver->textconv, df, &size);
4498        if (!*outbuf)
4499                die("unable to read files to diff");
4500
4501        if (driver->textconv_cache && df->sha1_valid) {
4502                /* ignore errors, as we might be in a readonly repository */
4503                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4504                                size);
4505                /*
4506                 * we could save up changes and flush them all at the end,
4507                 * but we would need an extra call after all diffing is done.
4508                 * Since generating a cache entry is the slow path anyway,
4509                 * this extra overhead probably isn't a big deal.
4510                 */
4511                notes_cache_write(driver->textconv_cache);
4512        }
4513
4514        return size;
4515}