diff.con commit Merge branch 'zj/mksh-columns-breakage' into lp/diffstat-with-graph (d4958d3)
   1/*
   2 * Copyright (C) 2005 Junio C Hamano
   3 */
   4#include "cache.h"
   5#include "quote.h"
   6#include "diff.h"
   7#include "diffcore.h"
   8#include "delta.h"
   9#include "xdiff-interface.h"
  10#include "color.h"
  11#include "attr.h"
  12#include "run-command.h"
  13#include "utf8.h"
  14#include "userdiff.h"
  15#include "sigchain.h"
  16#include "submodule.h"
  17#include "ll-merge.h"
  18
  19#ifdef NO_FAST_WORKING_DIRECTORY
  20#define FAST_WORKING_DIRECTORY 0
  21#else
  22#define FAST_WORKING_DIRECTORY 1
  23#endif
  24
  25static int diff_detect_rename_default;
  26static int diff_rename_limit_default = 400;
  27static int diff_suppress_blank_empty;
  28int diff_use_color_default = -1;
  29static const char *diff_word_regex_cfg;
  30static const char *external_diff_cmd_cfg;
  31int diff_auto_refresh_index = 1;
  32static int diff_mnemonic_prefix;
  33static int diff_no_prefix;
  34static int diff_stat_graph_width;
  35static int diff_dirstat_permille_default = 30;
  36static struct diff_options default_diff_options;
  37
  38static char diff_colors[][COLOR_MAXLEN] = {
  39        GIT_COLOR_RESET,
  40        GIT_COLOR_NORMAL,       /* PLAIN */
  41        GIT_COLOR_BOLD,         /* METAINFO */
  42        GIT_COLOR_CYAN,         /* FRAGINFO */
  43        GIT_COLOR_RED,          /* OLD */
  44        GIT_COLOR_GREEN,        /* NEW */
  45        GIT_COLOR_YELLOW,       /* COMMIT */
  46        GIT_COLOR_BG_RED,       /* WHITESPACE */
  47        GIT_COLOR_NORMAL,       /* FUNCINFO */
  48};
  49
  50static int parse_diff_color_slot(const char *var, int ofs)
  51{
  52        if (!strcasecmp(var+ofs, "plain"))
  53                return DIFF_PLAIN;
  54        if (!strcasecmp(var+ofs, "meta"))
  55                return DIFF_METAINFO;
  56        if (!strcasecmp(var+ofs, "frag"))
  57                return DIFF_FRAGINFO;
  58        if (!strcasecmp(var+ofs, "old"))
  59                return DIFF_FILE_OLD;
  60        if (!strcasecmp(var+ofs, "new"))
  61                return DIFF_FILE_NEW;
  62        if (!strcasecmp(var+ofs, "commit"))
  63                return DIFF_COMMIT;
  64        if (!strcasecmp(var+ofs, "whitespace"))
  65                return DIFF_WHITESPACE;
  66        if (!strcasecmp(var+ofs, "func"))
  67                return DIFF_FUNCINFO;
  68        return -1;
  69}
  70
  71static int parse_dirstat_params(struct diff_options *options, const char *params,
  72                                struct strbuf *errmsg)
  73{
  74        const char *p = params;
  75        int p_len, ret = 0;
  76
  77        while (*p) {
  78                p_len = strchrnul(p, ',') - p;
  79                if (!memcmp(p, "changes", p_len)) {
  80                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  81                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  82                } else if (!memcmp(p, "lines", p_len)) {
  83                        DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
  84                        DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
  85                } else if (!memcmp(p, "files", p_len)) {
  86                        DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
  87                        DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
  88                } else if (!memcmp(p, "noncumulative", p_len)) {
  89                        DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
  90                } else if (!memcmp(p, "cumulative", p_len)) {
  91                        DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
  92                } else if (isdigit(*p)) {
  93                        char *end;
  94                        int permille = strtoul(p, &end, 10) * 10;
  95                        if (*end == '.' && isdigit(*++end)) {
  96                                /* only use first digit */
  97                                permille += *end - '0';
  98                                /* .. and ignore any further digits */
  99                                while (isdigit(*++end))
 100                                        ; /* nothing */
 101                        }
 102                        if (end - p == p_len)
 103                                options->dirstat_permille = permille;
 104                        else {
 105                                strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%.*s'\n"),
 106                                            p_len, p);
 107                                ret++;
 108                        }
 109                } else {
 110                        strbuf_addf(errmsg, _("  Unknown dirstat parameter '%.*s'\n"),
 111                                    p_len, p);
 112                        ret++;
 113                }
 114
 115                p += p_len;
 116
 117                if (*p)
 118                        p++; /* more parameters, swallow separator */
 119        }
 120        return ret;
 121}
 122
 123static int git_config_rename(const char *var, const char *value)
 124{
 125        if (!value)
 126                return DIFF_DETECT_RENAME;
 127        if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
 128                return  DIFF_DETECT_COPY;
 129        return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
 130}
 131
 132/*
 133 * These are to give UI layer defaults.
 134 * The core-level commands such as git-diff-files should
 135 * never be affected by the setting of diff.renames
 136 * the user happens to have in the configuration file.
 137 */
 138int git_diff_ui_config(const char *var, const char *value, void *cb)
 139{
 140        if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
 141                diff_use_color_default = git_config_colorbool(var, value);
 142                return 0;
 143        }
 144        if (!strcmp(var, "diff.renames")) {
 145                diff_detect_rename_default = git_config_rename(var, value);
 146                return 0;
 147        }
 148        if (!strcmp(var, "diff.autorefreshindex")) {
 149                diff_auto_refresh_index = git_config_bool(var, value);
 150                return 0;
 151        }
 152        if (!strcmp(var, "diff.mnemonicprefix")) {
 153                diff_mnemonic_prefix = git_config_bool(var, value);
 154                return 0;
 155        }
 156        if (!strcmp(var, "diff.noprefix")) {
 157                diff_no_prefix = git_config_bool(var, value);
 158                return 0;
 159        }
 160        if (!strcmp(var, "diff.statgraphwidth")) {
 161                diff_stat_graph_width = git_config_int(var, value);
 162                return 0;
 163        }
 164        if (!strcmp(var, "diff.external"))
 165                return git_config_string(&external_diff_cmd_cfg, var, value);
 166        if (!strcmp(var, "diff.wordregex"))
 167                return git_config_string(&diff_word_regex_cfg, var, value);
 168
 169        if (!strcmp(var, "diff.ignoresubmodules"))
 170                handle_ignore_submodules_arg(&default_diff_options, value);
 171
 172        if (git_color_config(var, value, cb) < 0)
 173                return -1;
 174
 175        return git_diff_basic_config(var, value, cb);
 176}
 177
 178int git_diff_basic_config(const char *var, const char *value, void *cb)
 179{
 180        if (!strcmp(var, "diff.renamelimit")) {
 181                diff_rename_limit_default = git_config_int(var, value);
 182                return 0;
 183        }
 184
 185        if (userdiff_config(var, value) < 0)
 186                return -1;
 187
 188        if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
 189                int slot = parse_diff_color_slot(var, 11);
 190                if (slot < 0)
 191                        return 0;
 192                if (!value)
 193                        return config_error_nonbool(var);
 194                color_parse(value, var, diff_colors[slot]);
 195                return 0;
 196        }
 197
 198        /* like GNU diff's --suppress-blank-empty option  */
 199        if (!strcmp(var, "diff.suppressblankempty") ||
 200                        /* for backwards compatibility */
 201                        !strcmp(var, "diff.suppress-blank-empty")) {
 202                diff_suppress_blank_empty = git_config_bool(var, value);
 203                return 0;
 204        }
 205
 206        if (!strcmp(var, "diff.dirstat")) {
 207                struct strbuf errmsg = STRBUF_INIT;
 208                default_diff_options.dirstat_permille = diff_dirstat_permille_default;
 209                if (parse_dirstat_params(&default_diff_options, value, &errmsg))
 210                        warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
 211                                errmsg.buf);
 212                strbuf_release(&errmsg);
 213                diff_dirstat_permille_default = default_diff_options.dirstat_permille;
 214                return 0;
 215        }
 216
 217        if (!prefixcmp(var, "submodule."))
 218                return parse_submodule_config_option(var, value);
 219
 220        return git_default_config(var, value, cb);
 221}
 222
 223static char *quote_two(const char *one, const char *two)
 224{
 225        int need_one = quote_c_style(one, NULL, NULL, 1);
 226        int need_two = quote_c_style(two, NULL, NULL, 1);
 227        struct strbuf res = STRBUF_INIT;
 228
 229        if (need_one + need_two) {
 230                strbuf_addch(&res, '"');
 231                quote_c_style(one, &res, NULL, 1);
 232                quote_c_style(two, &res, NULL, 1);
 233                strbuf_addch(&res, '"');
 234        } else {
 235                strbuf_addstr(&res, one);
 236                strbuf_addstr(&res, two);
 237        }
 238        return strbuf_detach(&res, NULL);
 239}
 240
 241static const char *external_diff(void)
 242{
 243        static const char *external_diff_cmd = NULL;
 244        static int done_preparing = 0;
 245
 246        if (done_preparing)
 247                return external_diff_cmd;
 248        external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
 249        if (!external_diff_cmd)
 250                external_diff_cmd = external_diff_cmd_cfg;
 251        done_preparing = 1;
 252        return external_diff_cmd;
 253}
 254
 255static struct diff_tempfile {
 256        const char *name; /* filename external diff should read from */
 257        char hex[41];
 258        char mode[10];
 259        char tmp_path[PATH_MAX];
 260} diff_temp[2];
 261
 262typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
 263
 264struct emit_callback {
 265        int color_diff;
 266        unsigned ws_rule;
 267        int blank_at_eof_in_preimage;
 268        int blank_at_eof_in_postimage;
 269        int lno_in_preimage;
 270        int lno_in_postimage;
 271        sane_truncate_fn truncate;
 272        const char **label_path;
 273        struct diff_words_data *diff_words;
 274        struct diff_options *opt;
 275        int *found_changesp;
 276        struct strbuf *header;
 277};
 278
 279static int count_lines(const char *data, int size)
 280{
 281        int count, ch, completely_empty = 1, nl_just_seen = 0;
 282        count = 0;
 283        while (0 < size--) {
 284                ch = *data++;
 285                if (ch == '\n') {
 286                        count++;
 287                        nl_just_seen = 1;
 288                        completely_empty = 0;
 289                }
 290                else {
 291                        nl_just_seen = 0;
 292                        completely_empty = 0;
 293                }
 294        }
 295        if (completely_empty)
 296                return 0;
 297        if (!nl_just_seen)
 298                count++; /* no trailing newline */
 299        return count;
 300}
 301
 302static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
 303{
 304        if (!DIFF_FILE_VALID(one)) {
 305                mf->ptr = (char *)""; /* does not matter */
 306                mf->size = 0;
 307                return 0;
 308        }
 309        else if (diff_populate_filespec(one, 0))
 310                return -1;
 311
 312        mf->ptr = one->data;
 313        mf->size = one->size;
 314        return 0;
 315}
 316
 317/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
 318static unsigned long diff_filespec_size(struct diff_filespec *one)
 319{
 320        if (!DIFF_FILE_VALID(one))
 321                return 0;
 322        diff_populate_filespec(one, 1);
 323        return one->size;
 324}
 325
 326static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
 327{
 328        char *ptr = mf->ptr;
 329        long size = mf->size;
 330        int cnt = 0;
 331
 332        if (!size)
 333                return cnt;
 334        ptr += size - 1; /* pointing at the very end */
 335        if (*ptr != '\n')
 336                ; /* incomplete line */
 337        else
 338                ptr--; /* skip the last LF */
 339        while (mf->ptr < ptr) {
 340                char *prev_eol;
 341                for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
 342                        if (*prev_eol == '\n')
 343                                break;
 344                if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
 345                        break;
 346                cnt++;
 347                ptr = prev_eol - 1;
 348        }
 349        return cnt;
 350}
 351
 352static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
 353                               struct emit_callback *ecbdata)
 354{
 355        int l1, l2, at;
 356        unsigned ws_rule = ecbdata->ws_rule;
 357        l1 = count_trailing_blank(mf1, ws_rule);
 358        l2 = count_trailing_blank(mf2, ws_rule);
 359        if (l2 <= l1) {
 360                ecbdata->blank_at_eof_in_preimage = 0;
 361                ecbdata->blank_at_eof_in_postimage = 0;
 362                return;
 363        }
 364        at = count_lines(mf1->ptr, mf1->size);
 365        ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
 366
 367        at = count_lines(mf2->ptr, mf2->size);
 368        ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
 369}
 370
 371static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
 372                        int first, const char *line, int len)
 373{
 374        int has_trailing_newline, has_trailing_carriage_return;
 375        int nofirst;
 376        FILE *file = o->file;
 377
 378        if (o->output_prefix) {
 379                struct strbuf *msg = NULL;
 380                msg = o->output_prefix(o, o->output_prefix_data);
 381                assert(msg);
 382                fwrite(msg->buf, msg->len, 1, file);
 383        }
 384
 385        if (len == 0) {
 386                has_trailing_newline = (first == '\n');
 387                has_trailing_carriage_return = (!has_trailing_newline &&
 388                                                (first == '\r'));
 389                nofirst = has_trailing_newline || has_trailing_carriage_return;
 390        } else {
 391                has_trailing_newline = (len > 0 && line[len-1] == '\n');
 392                if (has_trailing_newline)
 393                        len--;
 394                has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
 395                if (has_trailing_carriage_return)
 396                        len--;
 397                nofirst = 0;
 398        }
 399
 400        if (len || !nofirst) {
 401                fputs(set, file);
 402                if (!nofirst)
 403                        fputc(first, file);
 404                fwrite(line, len, 1, file);
 405                fputs(reset, file);
 406        }
 407        if (has_trailing_carriage_return)
 408                fputc('\r', file);
 409        if (has_trailing_newline)
 410                fputc('\n', file);
 411}
 412
 413static void emit_line(struct diff_options *o, const char *set, const char *reset,
 414                      const char *line, int len)
 415{
 416        emit_line_0(o, set, reset, line[0], line+1, len-1);
 417}
 418
 419static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
 420{
 421        if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
 422              ecbdata->blank_at_eof_in_preimage &&
 423              ecbdata->blank_at_eof_in_postimage &&
 424              ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
 425              ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
 426                return 0;
 427        return ws_blank_line(line, len, ecbdata->ws_rule);
 428}
 429
 430static void emit_add_line(const char *reset,
 431                          struct emit_callback *ecbdata,
 432                          const char *line, int len)
 433{
 434        const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
 435        const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
 436
 437        if (!*ws)
 438                emit_line_0(ecbdata->opt, set, reset, '+', line, len);
 439        else if (new_blank_line_at_eof(ecbdata, line, len))
 440                /* Blank line at EOF - paint '+' as well */
 441                emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
 442        else {
 443                /* Emit just the prefix, then the rest. */
 444                emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
 445                ws_check_emit(line, len, ecbdata->ws_rule,
 446                              ecbdata->opt->file, set, reset, ws);
 447        }
 448}
 449
 450static void emit_hunk_header(struct emit_callback *ecbdata,
 451                             const char *line, int len)
 452{
 453        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
 454        const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
 455        const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
 456        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
 457        static const char atat[2] = { '@', '@' };
 458        const char *cp, *ep;
 459        struct strbuf msgbuf = STRBUF_INIT;
 460        int org_len = len;
 461        int i = 1;
 462
 463        /*
 464         * As a hunk header must begin with "@@ -<old>, +<new> @@",
 465         * it always is at least 10 bytes long.
 466         */
 467        if (len < 10 ||
 468            memcmp(line, atat, 2) ||
 469            !(ep = memmem(line + 2, len - 2, atat, 2))) {
 470                emit_line(ecbdata->opt, plain, reset, line, len);
 471                return;
 472        }
 473        ep += 2; /* skip over @@ */
 474
 475        /* The hunk header in fraginfo color */
 476        strbuf_add(&msgbuf, frag, strlen(frag));
 477        strbuf_add(&msgbuf, line, ep - line);
 478        strbuf_add(&msgbuf, reset, strlen(reset));
 479
 480        /*
 481         * trailing "\r\n"
 482         */
 483        for ( ; i < 3; i++)
 484                if (line[len - i] == '\r' || line[len - i] == '\n')
 485                        len--;
 486
 487        /* blank before the func header */
 488        for (cp = ep; ep - line < len; ep++)
 489                if (*ep != ' ' && *ep != '\t')
 490                        break;
 491        if (ep != cp) {
 492                strbuf_add(&msgbuf, plain, strlen(plain));
 493                strbuf_add(&msgbuf, cp, ep - cp);
 494                strbuf_add(&msgbuf, reset, strlen(reset));
 495        }
 496
 497        if (ep < line + len) {
 498                strbuf_add(&msgbuf, func, strlen(func));
 499                strbuf_add(&msgbuf, ep, line + len - ep);
 500                strbuf_add(&msgbuf, reset, strlen(reset));
 501        }
 502
 503        strbuf_add(&msgbuf, line + len, org_len - len);
 504        emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
 505        strbuf_release(&msgbuf);
 506}
 507
 508static struct diff_tempfile *claim_diff_tempfile(void) {
 509        int i;
 510        for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
 511                if (!diff_temp[i].name)
 512                        return diff_temp + i;
 513        die("BUG: diff is failing to clean up its tempfiles");
 514}
 515
 516static int remove_tempfile_installed;
 517
 518static void remove_tempfile(void)
 519{
 520        int i;
 521        for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
 522                if (diff_temp[i].name == diff_temp[i].tmp_path)
 523                        unlink_or_warn(diff_temp[i].name);
 524                diff_temp[i].name = NULL;
 525        }
 526}
 527
 528static void remove_tempfile_on_signal(int signo)
 529{
 530        remove_tempfile();
 531        sigchain_pop(signo);
 532        raise(signo);
 533}
 534
 535static void print_line_count(FILE *file, int count)
 536{
 537        switch (count) {
 538        case 0:
 539                fprintf(file, "0,0");
 540                break;
 541        case 1:
 542                fprintf(file, "1");
 543                break;
 544        default:
 545                fprintf(file, "1,%d", count);
 546                break;
 547        }
 548}
 549
 550static void emit_rewrite_lines(struct emit_callback *ecb,
 551                               int prefix, const char *data, int size)
 552{
 553        const char *endp = NULL;
 554        static const char *nneof = " No newline at end of file\n";
 555        const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
 556        const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
 557
 558        while (0 < size) {
 559                int len;
 560
 561                endp = memchr(data, '\n', size);
 562                len = endp ? (endp - data + 1) : size;
 563                if (prefix != '+') {
 564                        ecb->lno_in_preimage++;
 565                        emit_line_0(ecb->opt, old, reset, '-',
 566                                    data, len);
 567                } else {
 568                        ecb->lno_in_postimage++;
 569                        emit_add_line(reset, ecb, data, len);
 570                }
 571                size -= len;
 572                data += len;
 573        }
 574        if (!endp) {
 575                const char *plain = diff_get_color(ecb->color_diff,
 576                                                   DIFF_PLAIN);
 577                emit_line_0(ecb->opt, plain, reset, '\\',
 578                            nneof, strlen(nneof));
 579        }
 580}
 581
 582static void emit_rewrite_diff(const char *name_a,
 583                              const char *name_b,
 584                              struct diff_filespec *one,
 585                              struct diff_filespec *two,
 586                              struct userdiff_driver *textconv_one,
 587                              struct userdiff_driver *textconv_two,
 588                              struct diff_options *o)
 589{
 590        int lc_a, lc_b;
 591        const char *name_a_tab, *name_b_tab;
 592        const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
 593        const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
 594        const char *reset = diff_get_color(o->use_color, DIFF_RESET);
 595        static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
 596        const char *a_prefix, *b_prefix;
 597        char *data_one, *data_two;
 598        size_t size_one, size_two;
 599        struct emit_callback ecbdata;
 600        char *line_prefix = "";
 601        struct strbuf *msgbuf;
 602
 603        if (o && o->output_prefix) {
 604                msgbuf = o->output_prefix(o, o->output_prefix_data);
 605                line_prefix = msgbuf->buf;
 606        }
 607
 608        if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
 609                a_prefix = o->b_prefix;
 610                b_prefix = o->a_prefix;
 611        } else {
 612                a_prefix = o->a_prefix;
 613                b_prefix = o->b_prefix;
 614        }
 615
 616        name_a += (*name_a == '/');
 617        name_b += (*name_b == '/');
 618        name_a_tab = strchr(name_a, ' ') ? "\t" : "";
 619        name_b_tab = strchr(name_b, ' ') ? "\t" : "";
 620
 621        strbuf_reset(&a_name);
 622        strbuf_reset(&b_name);
 623        quote_two_c_style(&a_name, a_prefix, name_a, 0);
 624        quote_two_c_style(&b_name, b_prefix, name_b, 0);
 625
 626        size_one = fill_textconv(textconv_one, one, &data_one);
 627        size_two = fill_textconv(textconv_two, two, &data_two);
 628
 629        memset(&ecbdata, 0, sizeof(ecbdata));
 630        ecbdata.color_diff = want_color(o->use_color);
 631        ecbdata.found_changesp = &o->found_changes;
 632        ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
 633        ecbdata.opt = o;
 634        if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
 635                mmfile_t mf1, mf2;
 636                mf1.ptr = (char *)data_one;
 637                mf2.ptr = (char *)data_two;
 638                mf1.size = size_one;
 639                mf2.size = size_two;
 640                check_blank_at_eof(&mf1, &mf2, &ecbdata);
 641        }
 642        ecbdata.lno_in_preimage = 1;
 643        ecbdata.lno_in_postimage = 1;
 644
 645        lc_a = count_lines(data_one, size_one);
 646        lc_b = count_lines(data_two, size_two);
 647        fprintf(o->file,
 648                "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
 649                line_prefix, metainfo, a_name.buf, name_a_tab, reset,
 650                line_prefix, metainfo, b_name.buf, name_b_tab, reset,
 651                line_prefix, fraginfo);
 652        if (!o->irreversible_delete)
 653                print_line_count(o->file, lc_a);
 654        else
 655                fprintf(o->file, "?,?");
 656        fprintf(o->file, " +");
 657        print_line_count(o->file, lc_b);
 658        fprintf(o->file, " @@%s\n", reset);
 659        if (lc_a && !o->irreversible_delete)
 660                emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
 661        if (lc_b)
 662                emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
 663        if (textconv_one)
 664                free((char *)data_one);
 665        if (textconv_two)
 666                free((char *)data_two);
 667}
 668
 669struct diff_words_buffer {
 670        mmfile_t text;
 671        long alloc;
 672        struct diff_words_orig {
 673                const char *begin, *end;
 674        } *orig;
 675        int orig_nr, orig_alloc;
 676};
 677
 678static void diff_words_append(char *line, unsigned long len,
 679                struct diff_words_buffer *buffer)
 680{
 681        ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
 682        line++;
 683        len--;
 684        memcpy(buffer->text.ptr + buffer->text.size, line, len);
 685        buffer->text.size += len;
 686        buffer->text.ptr[buffer->text.size] = '\0';
 687}
 688
 689struct diff_words_style_elem {
 690        const char *prefix;
 691        const char *suffix;
 692        const char *color; /* NULL; filled in by the setup code if
 693                            * color is enabled */
 694};
 695
 696struct diff_words_style {
 697        enum diff_words_type type;
 698        struct diff_words_style_elem new, old, ctx;
 699        const char *newline;
 700};
 701
 702static struct diff_words_style diff_words_styles[] = {
 703        { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
 704        { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
 705        { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
 706};
 707
 708struct diff_words_data {
 709        struct diff_words_buffer minus, plus;
 710        const char *current_plus;
 711        int last_minus;
 712        struct diff_options *opt;
 713        regex_t *word_regex;
 714        enum diff_words_type type;
 715        struct diff_words_style *style;
 716};
 717
 718static int fn_out_diff_words_write_helper(FILE *fp,
 719                                          struct diff_words_style_elem *st_el,
 720                                          const char *newline,
 721                                          size_t count, const char *buf,
 722                                          const char *line_prefix)
 723{
 724        int print = 0;
 725
 726        while (count) {
 727                char *p = memchr(buf, '\n', count);
 728                if (print)
 729                        fputs(line_prefix, fp);
 730                if (p != buf) {
 731                        if (st_el->color && fputs(st_el->color, fp) < 0)
 732                                return -1;
 733                        if (fputs(st_el->prefix, fp) < 0 ||
 734                            fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
 735                            fputs(st_el->suffix, fp) < 0)
 736                                return -1;
 737                        if (st_el->color && *st_el->color
 738                            && fputs(GIT_COLOR_RESET, fp) < 0)
 739                                return -1;
 740                }
 741                if (!p)
 742                        return 0;
 743                if (fputs(newline, fp) < 0)
 744                        return -1;
 745                count -= p + 1 - buf;
 746                buf = p + 1;
 747                print = 1;
 748        }
 749        return 0;
 750}
 751
 752/*
 753 * '--color-words' algorithm can be described as:
 754 *
 755 *   1. collect a the minus/plus lines of a diff hunk, divided into
 756 *      minus-lines and plus-lines;
 757 *
 758 *   2. break both minus-lines and plus-lines into words and
 759 *      place them into two mmfile_t with one word for each line;
 760 *
 761 *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
 762 *
 763 * And for the common parts of the both file, we output the plus side text.
 764 * diff_words->current_plus is used to trace the current position of the plus file
 765 * which printed. diff_words->last_minus is used to trace the last minus word
 766 * printed.
 767 *
 768 * For '--graph' to work with '--color-words', we need to output the graph prefix
 769 * on each line of color words output. Generally, there are two conditions on
 770 * which we should output the prefix.
 771 *
 772 *   1. diff_words->last_minus == 0 &&
 773 *      diff_words->current_plus == diff_words->plus.text.ptr
 774 *
 775 *      that is: the plus text must start as a new line, and if there is no minus
 776 *      word printed, a graph prefix must be printed.
 777 *
 778 *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
 779 *      *(diff_words->current_plus - 1) == '\n'
 780 *
 781 *      that is: a graph prefix must be printed following a '\n'
 782 */
 783static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
 784{
 785        if ((diff_words->last_minus == 0 &&
 786                diff_words->current_plus == diff_words->plus.text.ptr) ||
 787                (diff_words->current_plus > diff_words->plus.text.ptr &&
 788                *(diff_words->current_plus - 1) == '\n')) {
 789                return 1;
 790        } else {
 791                return 0;
 792        }
 793}
 794
 795static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
 796{
 797        struct diff_words_data *diff_words = priv;
 798        struct diff_words_style *style = diff_words->style;
 799        int minus_first, minus_len, plus_first, plus_len;
 800        const char *minus_begin, *minus_end, *plus_begin, *plus_end;
 801        struct diff_options *opt = diff_words->opt;
 802        struct strbuf *msgbuf;
 803        char *line_prefix = "";
 804
 805        if (line[0] != '@' || parse_hunk_header(line, len,
 806                        &minus_first, &minus_len, &plus_first, &plus_len))
 807                return;
 808
 809        assert(opt);
 810        if (opt->output_prefix) {
 811                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 812                line_prefix = msgbuf->buf;
 813        }
 814
 815        /* POSIX requires that first be decremented by one if len == 0... */
 816        if (minus_len) {
 817                minus_begin = diff_words->minus.orig[minus_first].begin;
 818                minus_end =
 819                        diff_words->minus.orig[minus_first + minus_len - 1].end;
 820        } else
 821                minus_begin = minus_end =
 822                        diff_words->minus.orig[minus_first].end;
 823
 824        if (plus_len) {
 825                plus_begin = diff_words->plus.orig[plus_first].begin;
 826                plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
 827        } else
 828                plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
 829
 830        if (color_words_output_graph_prefix(diff_words)) {
 831                fputs(line_prefix, diff_words->opt->file);
 832        }
 833        if (diff_words->current_plus != plus_begin) {
 834                fn_out_diff_words_write_helper(diff_words->opt->file,
 835                                &style->ctx, style->newline,
 836                                plus_begin - diff_words->current_plus,
 837                                diff_words->current_plus, line_prefix);
 838                if (*(plus_begin - 1) == '\n')
 839                        fputs(line_prefix, diff_words->opt->file);
 840        }
 841        if (minus_begin != minus_end) {
 842                fn_out_diff_words_write_helper(diff_words->opt->file,
 843                                &style->old, style->newline,
 844                                minus_end - minus_begin, minus_begin,
 845                                line_prefix);
 846        }
 847        if (plus_begin != plus_end) {
 848                fn_out_diff_words_write_helper(diff_words->opt->file,
 849                                &style->new, style->newline,
 850                                plus_end - plus_begin, plus_begin,
 851                                line_prefix);
 852        }
 853
 854        diff_words->current_plus = plus_end;
 855        diff_words->last_minus = minus_first;
 856}
 857
 858/* This function starts looking at *begin, and returns 0 iff a word was found. */
 859static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
 860                int *begin, int *end)
 861{
 862        if (word_regex && *begin < buffer->size) {
 863                regmatch_t match[1];
 864                if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
 865                        char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
 866                                        '\n', match[0].rm_eo - match[0].rm_so);
 867                        *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
 868                        *begin += match[0].rm_so;
 869                        return *begin >= *end;
 870                }
 871                return -1;
 872        }
 873
 874        /* find the next word */
 875        while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
 876                (*begin)++;
 877        if (*begin >= buffer->size)
 878                return -1;
 879
 880        /* find the end of the word */
 881        *end = *begin + 1;
 882        while (*end < buffer->size && !isspace(buffer->ptr[*end]))
 883                (*end)++;
 884
 885        return 0;
 886}
 887
 888/*
 889 * This function splits the words in buffer->text, stores the list with
 890 * newline separator into out, and saves the offsets of the original words
 891 * in buffer->orig.
 892 */
 893static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
 894                regex_t *word_regex)
 895{
 896        int i, j;
 897        long alloc = 0;
 898
 899        out->size = 0;
 900        out->ptr = NULL;
 901
 902        /* fake an empty "0th" word */
 903        ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
 904        buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
 905        buffer->orig_nr = 1;
 906
 907        for (i = 0; i < buffer->text.size; i++) {
 908                if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
 909                        return;
 910
 911                /* store original boundaries */
 912                ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
 913                                buffer->orig_alloc);
 914                buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
 915                buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
 916                buffer->orig_nr++;
 917
 918                /* store one word */
 919                ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
 920                memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
 921                out->ptr[out->size + j - i] = '\n';
 922                out->size += j - i + 1;
 923
 924                i = j - 1;
 925        }
 926}
 927
 928/* this executes the word diff on the accumulated buffers */
 929static void diff_words_show(struct diff_words_data *diff_words)
 930{
 931        xpparam_t xpp;
 932        xdemitconf_t xecfg;
 933        mmfile_t minus, plus;
 934        struct diff_words_style *style = diff_words->style;
 935
 936        struct diff_options *opt = diff_words->opt;
 937        struct strbuf *msgbuf;
 938        char *line_prefix = "";
 939
 940        assert(opt);
 941        if (opt->output_prefix) {
 942                msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
 943                line_prefix = msgbuf->buf;
 944        }
 945
 946        /* special case: only removal */
 947        if (!diff_words->plus.text.size) {
 948                fputs(line_prefix, diff_words->opt->file);
 949                fn_out_diff_words_write_helper(diff_words->opt->file,
 950                        &style->old, style->newline,
 951                        diff_words->minus.text.size,
 952                        diff_words->minus.text.ptr, line_prefix);
 953                diff_words->minus.text.size = 0;
 954                return;
 955        }
 956
 957        diff_words->current_plus = diff_words->plus.text.ptr;
 958        diff_words->last_minus = 0;
 959
 960        memset(&xpp, 0, sizeof(xpp));
 961        memset(&xecfg, 0, sizeof(xecfg));
 962        diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
 963        diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
 964        xpp.flags = 0;
 965        /* as only the hunk header will be parsed, we need a 0-context */
 966        xecfg.ctxlen = 0;
 967        xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
 968                      &xpp, &xecfg);
 969        free(minus.ptr);
 970        free(plus.ptr);
 971        if (diff_words->current_plus != diff_words->plus.text.ptr +
 972                        diff_words->plus.text.size) {
 973                if (color_words_output_graph_prefix(diff_words))
 974                        fputs(line_prefix, diff_words->opt->file);
 975                fn_out_diff_words_write_helper(diff_words->opt->file,
 976                        &style->ctx, style->newline,
 977                        diff_words->plus.text.ptr + diff_words->plus.text.size
 978                        - diff_words->current_plus, diff_words->current_plus,
 979                        line_prefix);
 980        }
 981        diff_words->minus.text.size = diff_words->plus.text.size = 0;
 982}
 983
 984/* In "color-words" mode, show word-diff of words accumulated in the buffer */
 985static void diff_words_flush(struct emit_callback *ecbdata)
 986{
 987        if (ecbdata->diff_words->minus.text.size ||
 988            ecbdata->diff_words->plus.text.size)
 989                diff_words_show(ecbdata->diff_words);
 990}
 991
 992static void free_diff_words_data(struct emit_callback *ecbdata)
 993{
 994        if (ecbdata->diff_words) {
 995                diff_words_flush(ecbdata);
 996                free (ecbdata->diff_words->minus.text.ptr);
 997                free (ecbdata->diff_words->minus.orig);
 998                free (ecbdata->diff_words->plus.text.ptr);
 999                free (ecbdata->diff_words->plus.orig);
1000                if (ecbdata->diff_words->word_regex) {
1001                        regfree(ecbdata->diff_words->word_regex);
1002                        free(ecbdata->diff_words->word_regex);
1003                }
1004                free(ecbdata->diff_words);
1005                ecbdata->diff_words = NULL;
1006        }
1007}
1008
1009const char *diff_get_color(int diff_use_color, enum color_diff ix)
1010{
1011        if (want_color(diff_use_color))
1012                return diff_colors[ix];
1013        return "";
1014}
1015
1016static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1017{
1018        const char *cp;
1019        unsigned long allot;
1020        size_t l = len;
1021
1022        if (ecb->truncate)
1023                return ecb->truncate(line, len);
1024        cp = line;
1025        allot = l;
1026        while (0 < l) {
1027                (void) utf8_width(&cp, &l);
1028                if (!cp)
1029                        break; /* truncated in the middle? */
1030        }
1031        return allot - l;
1032}
1033
1034static void find_lno(const char *line, struct emit_callback *ecbdata)
1035{
1036        const char *p;
1037        ecbdata->lno_in_preimage = 0;
1038        ecbdata->lno_in_postimage = 0;
1039        p = strchr(line, '-');
1040        if (!p)
1041                return; /* cannot happen */
1042        ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1043        p = strchr(p, '+');
1044        if (!p)
1045                return; /* cannot happen */
1046        ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1047}
1048
1049static void fn_out_consume(void *priv, char *line, unsigned long len)
1050{
1051        struct emit_callback *ecbdata = priv;
1052        const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1053        const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1054        const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1055        struct diff_options *o = ecbdata->opt;
1056        char *line_prefix = "";
1057        struct strbuf *msgbuf;
1058
1059        if (o && o->output_prefix) {
1060                msgbuf = o->output_prefix(o, o->output_prefix_data);
1061                line_prefix = msgbuf->buf;
1062        }
1063
1064        if (ecbdata->header) {
1065                fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1066                strbuf_reset(ecbdata->header);
1067                ecbdata->header = NULL;
1068        }
1069        *(ecbdata->found_changesp) = 1;
1070
1071        if (ecbdata->label_path[0]) {
1072                const char *name_a_tab, *name_b_tab;
1073
1074                name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1075                name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1076
1077                fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1078                        line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1079                fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1080                        line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1081                ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1082        }
1083
1084        if (diff_suppress_blank_empty
1085            && len == 2 && line[0] == ' ' && line[1] == '\n') {
1086                line[0] = '\n';
1087                len = 1;
1088        }
1089
1090        if (line[0] == '@') {
1091                if (ecbdata->diff_words)
1092                        diff_words_flush(ecbdata);
1093                len = sane_truncate_line(ecbdata, line, len);
1094                find_lno(line, ecbdata);
1095                emit_hunk_header(ecbdata, line, len);
1096                if (line[len-1] != '\n')
1097                        putc('\n', ecbdata->opt->file);
1098                return;
1099        }
1100
1101        if (len < 1) {
1102                emit_line(ecbdata->opt, reset, reset, line, len);
1103                if (ecbdata->diff_words
1104                    && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1105                        fputs("~\n", ecbdata->opt->file);
1106                return;
1107        }
1108
1109        if (ecbdata->diff_words) {
1110                if (line[0] == '-') {
1111                        diff_words_append(line, len,
1112                                          &ecbdata->diff_words->minus);
1113                        return;
1114                } else if (line[0] == '+') {
1115                        diff_words_append(line, len,
1116                                          &ecbdata->diff_words->plus);
1117                        return;
1118                } else if (!prefixcmp(line, "\\ ")) {
1119                        /*
1120                         * Eat the "no newline at eof" marker as if we
1121                         * saw a "+" or "-" line with nothing on it,
1122                         * and return without diff_words_flush() to
1123                         * defer processing. If this is the end of
1124                         * preimage, more "+" lines may come after it.
1125                         */
1126                        return;
1127                }
1128                diff_words_flush(ecbdata);
1129                if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1130                        emit_line(ecbdata->opt, plain, reset, line, len);
1131                        fputs("~\n", ecbdata->opt->file);
1132                } else {
1133                        /*
1134                         * Skip the prefix character, if any.  With
1135                         * diff_suppress_blank_empty, there may be
1136                         * none.
1137                         */
1138                        if (line[0] != '\n') {
1139                              line++;
1140                              len--;
1141                        }
1142                        emit_line(ecbdata->opt, plain, reset, line, len);
1143                }
1144                return;
1145        }
1146
1147        if (line[0] != '+') {
1148                const char *color =
1149                        diff_get_color(ecbdata->color_diff,
1150                                       line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1151                ecbdata->lno_in_preimage++;
1152                if (line[0] == ' ')
1153                        ecbdata->lno_in_postimage++;
1154                emit_line(ecbdata->opt, color, reset, line, len);
1155        } else {
1156                ecbdata->lno_in_postimage++;
1157                emit_add_line(reset, ecbdata, line + 1, len - 1);
1158        }
1159}
1160
1161static char *pprint_rename(const char *a, const char *b)
1162{
1163        const char *old = a;
1164        const char *new = b;
1165        struct strbuf name = STRBUF_INIT;
1166        int pfx_length, sfx_length;
1167        int len_a = strlen(a);
1168        int len_b = strlen(b);
1169        int a_midlen, b_midlen;
1170        int qlen_a = quote_c_style(a, NULL, NULL, 0);
1171        int qlen_b = quote_c_style(b, NULL, NULL, 0);
1172
1173        if (qlen_a || qlen_b) {
1174                quote_c_style(a, &name, NULL, 0);
1175                strbuf_addstr(&name, " => ");
1176                quote_c_style(b, &name, NULL, 0);
1177                return strbuf_detach(&name, NULL);
1178        }
1179
1180        /* Find common prefix */
1181        pfx_length = 0;
1182        while (*old && *new && *old == *new) {
1183                if (*old == '/')
1184                        pfx_length = old - a + 1;
1185                old++;
1186                new++;
1187        }
1188
1189        /* Find common suffix */
1190        old = a + len_a;
1191        new = b + len_b;
1192        sfx_length = 0;
1193        while (a <= old && b <= new && *old == *new) {
1194                if (*old == '/')
1195                        sfx_length = len_a - (old - a);
1196                old--;
1197                new--;
1198        }
1199
1200        /*
1201         * pfx{mid-a => mid-b}sfx
1202         * {pfx-a => pfx-b}sfx
1203         * pfx{sfx-a => sfx-b}
1204         * name-a => name-b
1205         */
1206        a_midlen = len_a - pfx_length - sfx_length;
1207        b_midlen = len_b - pfx_length - sfx_length;
1208        if (a_midlen < 0)
1209                a_midlen = 0;
1210        if (b_midlen < 0)
1211                b_midlen = 0;
1212
1213        strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1214        if (pfx_length + sfx_length) {
1215                strbuf_add(&name, a, pfx_length);
1216                strbuf_addch(&name, '{');
1217        }
1218        strbuf_add(&name, a + pfx_length, a_midlen);
1219        strbuf_addstr(&name, " => ");
1220        strbuf_add(&name, b + pfx_length, b_midlen);
1221        if (pfx_length + sfx_length) {
1222                strbuf_addch(&name, '}');
1223                strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1224        }
1225        return strbuf_detach(&name, NULL);
1226}
1227
1228struct diffstat_t {
1229        int nr;
1230        int alloc;
1231        struct diffstat_file {
1232                char *from_name;
1233                char *name;
1234                char *print_name;
1235                unsigned is_unmerged:1;
1236                unsigned is_binary:1;
1237                unsigned is_renamed:1;
1238                uintmax_t added, deleted;
1239        } **files;
1240};
1241
1242static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1243                                          const char *name_a,
1244                                          const char *name_b)
1245{
1246        struct diffstat_file *x;
1247        x = xcalloc(sizeof (*x), 1);
1248        if (diffstat->nr == diffstat->alloc) {
1249                diffstat->alloc = alloc_nr(diffstat->alloc);
1250                diffstat->files = xrealloc(diffstat->files,
1251                                diffstat->alloc * sizeof(x));
1252        }
1253        diffstat->files[diffstat->nr++] = x;
1254        if (name_b) {
1255                x->from_name = xstrdup(name_a);
1256                x->name = xstrdup(name_b);
1257                x->is_renamed = 1;
1258        }
1259        else {
1260                x->from_name = NULL;
1261                x->name = xstrdup(name_a);
1262        }
1263        return x;
1264}
1265
1266static void diffstat_consume(void *priv, char *line, unsigned long len)
1267{
1268        struct diffstat_t *diffstat = priv;
1269        struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1270
1271        if (line[0] == '+')
1272                x->added++;
1273        else if (line[0] == '-')
1274                x->deleted++;
1275}
1276
1277const char mime_boundary_leader[] = "------------";
1278
1279static int scale_linear(int it, int width, int max_change)
1280{
1281        if (!it)
1282                return 0;
1283        /*
1284         * make sure that at least one '-' or '+' is printed if
1285         * there is any change to this path. The easiest way is to
1286         * scale linearly as if the alloted width is one column shorter
1287         * than it is, and then add 1 to the result.
1288         */
1289        return 1 + (it * (width - 1) / max_change);
1290}
1291
1292static void show_name(FILE *file,
1293                      const char *prefix, const char *name, int len)
1294{
1295        fprintf(file, " %s%-*s |", prefix, len, name);
1296}
1297
1298static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1299{
1300        if (cnt <= 0)
1301                return;
1302        fprintf(file, "%s", set);
1303        while (cnt--)
1304                putc(ch, file);
1305        fprintf(file, "%s", reset);
1306}
1307
1308static void fill_print_name(struct diffstat_file *file)
1309{
1310        char *pname;
1311
1312        if (file->print_name)
1313                return;
1314
1315        if (!file->is_renamed) {
1316                struct strbuf buf = STRBUF_INIT;
1317                if (quote_c_style(file->name, &buf, NULL, 0)) {
1318                        pname = strbuf_detach(&buf, NULL);
1319                } else {
1320                        pname = file->name;
1321                        strbuf_release(&buf);
1322                }
1323        } else {
1324                pname = pprint_rename(file->from_name, file->name);
1325        }
1326        file->print_name = pname;
1327}
1328
1329int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1330{
1331        struct strbuf sb = STRBUF_INIT;
1332        int ret;
1333
1334        if (!files) {
1335                assert(insertions == 0 && deletions == 0);
1336                return fputs(_(" 0 files changed\n"), fp);
1337        }
1338
1339        strbuf_addf(&sb,
1340                    Q_(" %d file changed", " %d files changed", files),
1341                    files);
1342
1343        /*
1344         * For binary diff, the caller may want to print "x files
1345         * changed" with insertions == 0 && deletions == 0.
1346         *
1347         * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1348         * is probably less confusing (i.e skip over "2 files changed
1349         * but nothing about added/removed lines? Is this a bug in Git?").
1350         */
1351        if (insertions || deletions == 0) {
1352                /*
1353                 * TRANSLATORS: "+" in (+) is a line addition marker;
1354                 * do not translate it.
1355                 */
1356                strbuf_addf(&sb,
1357                            Q_(", %d insertion(+)", ", %d insertions(+)",
1358                               insertions),
1359                            insertions);
1360        }
1361
1362        if (deletions || insertions == 0) {
1363                /*
1364                 * TRANSLATORS: "-" in (-) is a line removal marker;
1365                 * do not translate it.
1366                 */
1367                strbuf_addf(&sb,
1368                            Q_(", %d deletion(-)", ", %d deletions(-)",
1369                               deletions),
1370                            deletions);
1371        }
1372        strbuf_addch(&sb, '\n');
1373        ret = fputs(sb.buf, fp);
1374        strbuf_release(&sb);
1375        return ret;
1376}
1377
1378static void show_stats(struct diffstat_t *data, struct diff_options *options)
1379{
1380        int i, len, add, del, adds = 0, dels = 0;
1381        uintmax_t max_change = 0, max_len = 0;
1382        int total_files = data->nr;
1383        int width, name_width, graph_width, number_width = 4, count;
1384        const char *reset, *add_c, *del_c;
1385        const char *line_prefix = "";
1386        int extra_shown = 0;
1387        struct strbuf *msg = NULL;
1388
1389        if (data->nr == 0)
1390                return;
1391
1392        if (options->output_prefix) {
1393                msg = options->output_prefix(options, options->output_prefix_data);
1394                line_prefix = msg->buf;
1395        }
1396
1397        count = options->stat_count ? options->stat_count : data->nr;
1398
1399        reset = diff_get_color_opt(options, DIFF_RESET);
1400        add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1401        del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1402
1403        /*
1404         * Find the longest filename and max number of changes
1405         */
1406        for (i = 0; (i < count) && (i < data->nr); i++) {
1407                struct diffstat_file *file = data->files[i];
1408                uintmax_t change = file->added + file->deleted;
1409                if (!data->files[i]->is_renamed &&
1410                         (change == 0)) {
1411                        count++; /* not shown == room for one more */
1412                        continue;
1413                }
1414                fill_print_name(file);
1415                len = strlen(file->print_name);
1416                if (max_len < len)
1417                        max_len = len;
1418
1419                if (file->is_binary || file->is_unmerged)
1420                        continue;
1421                if (max_change < change)
1422                        max_change = change;
1423        }
1424        count = i; /* min(count, data->nr) */
1425
1426        /*
1427         * We have width = stat_width or term_columns() columns total.
1428         * We want a maximum of min(max_len, stat_name_width) for the name part.
1429         * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1430         * We also need 1 for " " and 4 + decimal_width(max_change)
1431         * for " | NNNN " and one the empty column at the end, altogether
1432         * 6 + decimal_width(max_change).
1433         *
1434         * If there's not enough space, we will use the smaller of
1435         * stat_name_width (if set) and 5/8*width for the filename,
1436         * and the rest for constant elements + graph part, but no more
1437         * than stat_graph_width for the graph part.
1438         * (5/8 gives 50 for filename and 30 for the constant parts + graph
1439         * for the standard terminal size).
1440         *
1441         * In other words: stat_width limits the maximum width, and
1442         * stat_name_width fixes the maximum width of the filename,
1443         * and is also used to divide available columns if there
1444         * aren't enough.
1445         */
1446
1447        if (options->stat_width == -1)
1448                width = term_columns() - options->output_prefix_length;
1449        else
1450                width = options->stat_width ? options->stat_width : 80;
1451
1452        if (options->stat_graph_width == -1)
1453                options->stat_graph_width = diff_stat_graph_width;
1454
1455        /*
1456         * Guarantee 3/8*16==6 for the graph part
1457         * and 5/8*16==10 for the filename part
1458         */
1459        if (width < 16 + 6 + number_width)
1460                width = 16 + 6 + number_width;
1461
1462        /*
1463         * First assign sizes that are wanted, ignoring available width.
1464         */
1465        graph_width = (options->stat_graph_width &&
1466                       options->stat_graph_width < max_change) ?
1467                options->stat_graph_width : max_change;
1468        name_width = (options->stat_name_width > 0 &&
1469                      options->stat_name_width < max_len) ?
1470                options->stat_name_width : max_len;
1471
1472        /*
1473         * Adjust adjustable widths not to exceed maximum width
1474         */
1475        if (name_width + number_width + 6 + graph_width > width) {
1476                if (graph_width > width * 3/8 - number_width - 6) {
1477                        graph_width = width * 3/8 - number_width - 6;
1478                        if (graph_width < 6)
1479                                graph_width = 6;
1480                }
1481
1482                if (options->stat_graph_width &&
1483                    graph_width > options->stat_graph_width)
1484                        graph_width = options->stat_graph_width;
1485                if (name_width > width - number_width - 6 - graph_width)
1486                        name_width = width - number_width - 6 - graph_width;
1487                else
1488                        graph_width = width - number_width - 6 - name_width;
1489        }
1490
1491        /*
1492         * From here name_width is the width of the name area,
1493         * and graph_width is the width of the graph area.
1494         * max_change is used to scale graph properly.
1495         */
1496        for (i = 0; i < count; i++) {
1497                const char *prefix = "";
1498                char *name = data->files[i]->print_name;
1499                uintmax_t added = data->files[i]->added;
1500                uintmax_t deleted = data->files[i]->deleted;
1501                int name_len;
1502
1503                if (!data->files[i]->is_renamed &&
1504                         (added + deleted == 0)) {
1505                        total_files--;
1506                        continue;
1507                }
1508                /*
1509                 * "scale" the filename
1510                 */
1511                len = name_width;
1512                name_len = strlen(name);
1513                if (name_width < name_len) {
1514                        char *slash;
1515                        prefix = "...";
1516                        len -= 3;
1517                        name += name_len - len;
1518                        slash = strchr(name, '/');
1519                        if (slash)
1520                                name = slash;
1521                }
1522
1523                if (data->files[i]->is_binary) {
1524                        fprintf(options->file, "%s", line_prefix);
1525                        show_name(options->file, prefix, name, len);
1526                        fprintf(options->file, "  Bin ");
1527                        fprintf(options->file, "%s%"PRIuMAX"%s",
1528                                del_c, deleted, reset);
1529                        fprintf(options->file, " -> ");
1530                        fprintf(options->file, "%s%"PRIuMAX"%s",
1531                                add_c, added, reset);
1532                        fprintf(options->file, " bytes");
1533                        fprintf(options->file, "\n");
1534                        continue;
1535                }
1536                else if (data->files[i]->is_unmerged) {
1537                        fprintf(options->file, "%s", line_prefix);
1538                        show_name(options->file, prefix, name, len);
1539                        fprintf(options->file, "  Unmerged\n");
1540                        continue;
1541                }
1542
1543                /*
1544                 * scale the add/delete
1545                 */
1546                add = added;
1547                del = deleted;
1548                adds += add;
1549                dels += del;
1550
1551                if (graph_width <= max_change) {
1552                        int total = add + del;
1553
1554                        total = scale_linear(add + del, graph_width, max_change);
1555                        if (total < 2 && add && del)
1556                                /* width >= 2 due to the sanity check */
1557                                total = 2;
1558                        if (add < del) {
1559                                add = scale_linear(add, graph_width, max_change);
1560                                del = total - add;
1561                        } else {
1562                                del = scale_linear(del, graph_width, max_change);
1563                                add = total - del;
1564                        }
1565                }
1566                fprintf(options->file, "%s", line_prefix);
1567                show_name(options->file, prefix, name, len);
1568                fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1569                                added + deleted ? " " : "");
1570                show_graph(options->file, '+', add, add_c, reset);
1571                show_graph(options->file, '-', del, del_c, reset);
1572                fprintf(options->file, "\n");
1573        }
1574        for (i = count; i < data->nr; i++) {
1575                uintmax_t added = data->files[i]->added;
1576                uintmax_t deleted = data->files[i]->deleted;
1577                if (!data->files[i]->is_renamed &&
1578                         (added + deleted == 0)) {
1579                        total_files--;
1580                        continue;
1581                }
1582                adds += added;
1583                dels += deleted;
1584                if (!extra_shown)
1585                        fprintf(options->file, "%s ...\n", line_prefix);
1586                extra_shown = 1;
1587        }
1588        fprintf(options->file, "%s", line_prefix);
1589        print_stat_summary(options->file, total_files, adds, dels);
1590}
1591
1592static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1593{
1594        int i, adds = 0, dels = 0, total_files = data->nr;
1595
1596        if (data->nr == 0)
1597                return;
1598
1599        for (i = 0; i < data->nr; i++) {
1600                if (!data->files[i]->is_binary &&
1601                    !data->files[i]->is_unmerged) {
1602                        int added = data->files[i]->added;
1603                        int deleted= data->files[i]->deleted;
1604                        if (!data->files[i]->is_renamed &&
1605                            (added + deleted == 0)) {
1606                                total_files--;
1607                        } else {
1608                                adds += added;
1609                                dels += deleted;
1610                        }
1611                }
1612        }
1613        if (options->output_prefix) {
1614                struct strbuf *msg = NULL;
1615                msg = options->output_prefix(options,
1616                                options->output_prefix_data);
1617                fprintf(options->file, "%s", msg->buf);
1618        }
1619        print_stat_summary(options->file, total_files, adds, dels);
1620}
1621
1622static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1623{
1624        int i;
1625
1626        if (data->nr == 0)
1627                return;
1628
1629        for (i = 0; i < data->nr; i++) {
1630                struct diffstat_file *file = data->files[i];
1631
1632                if (options->output_prefix) {
1633                        struct strbuf *msg = NULL;
1634                        msg = options->output_prefix(options,
1635                                        options->output_prefix_data);
1636                        fprintf(options->file, "%s", msg->buf);
1637                }
1638
1639                if (file->is_binary)
1640                        fprintf(options->file, "-\t-\t");
1641                else
1642                        fprintf(options->file,
1643                                "%"PRIuMAX"\t%"PRIuMAX"\t",
1644                                file->added, file->deleted);
1645                if (options->line_termination) {
1646                        fill_print_name(file);
1647                        if (!file->is_renamed)
1648                                write_name_quoted(file->name, options->file,
1649                                                  options->line_termination);
1650                        else {
1651                                fputs(file->print_name, options->file);
1652                                putc(options->line_termination, options->file);
1653                        }
1654                } else {
1655                        if (file->is_renamed) {
1656                                putc('\0', options->file);
1657                                write_name_quoted(file->from_name, options->file, '\0');
1658                        }
1659                        write_name_quoted(file->name, options->file, '\0');
1660                }
1661        }
1662}
1663
1664struct dirstat_file {
1665        const char *name;
1666        unsigned long changed;
1667};
1668
1669struct dirstat_dir {
1670        struct dirstat_file *files;
1671        int alloc, nr, permille, cumulative;
1672};
1673
1674static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1675                unsigned long changed, const char *base, int baselen)
1676{
1677        unsigned long this_dir = 0;
1678        unsigned int sources = 0;
1679        const char *line_prefix = "";
1680        struct strbuf *msg = NULL;
1681
1682        if (opt->output_prefix) {
1683                msg = opt->output_prefix(opt, opt->output_prefix_data);
1684                line_prefix = msg->buf;
1685        }
1686
1687        while (dir->nr) {
1688                struct dirstat_file *f = dir->files;
1689                int namelen = strlen(f->name);
1690                unsigned long this;
1691                char *slash;
1692
1693                if (namelen < baselen)
1694                        break;
1695                if (memcmp(f->name, base, baselen))
1696                        break;
1697                slash = strchr(f->name + baselen, '/');
1698                if (slash) {
1699                        int newbaselen = slash + 1 - f->name;
1700                        this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1701                        sources++;
1702                } else {
1703                        this = f->changed;
1704                        dir->files++;
1705                        dir->nr--;
1706                        sources += 2;
1707                }
1708                this_dir += this;
1709        }
1710
1711        /*
1712         * We don't report dirstat's for
1713         *  - the top level
1714         *  - or cases where everything came from a single directory
1715         *    under this directory (sources == 1).
1716         */
1717        if (baselen && sources != 1) {
1718                if (this_dir) {
1719                        int permille = this_dir * 1000 / changed;
1720                        if (permille >= dir->permille) {
1721                                fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1722                                        permille / 10, permille % 10, baselen, base);
1723                                if (!dir->cumulative)
1724                                        return 0;
1725                        }
1726                }
1727        }
1728        return this_dir;
1729}
1730
1731static int dirstat_compare(const void *_a, const void *_b)
1732{
1733        const struct dirstat_file *a = _a;
1734        const struct dirstat_file *b = _b;
1735        return strcmp(a->name, b->name);
1736}
1737
1738static void show_dirstat(struct diff_options *options)
1739{
1740        int i;
1741        unsigned long changed;
1742        struct dirstat_dir dir;
1743        struct diff_queue_struct *q = &diff_queued_diff;
1744
1745        dir.files = NULL;
1746        dir.alloc = 0;
1747        dir.nr = 0;
1748        dir.permille = options->dirstat_permille;
1749        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1750
1751        changed = 0;
1752        for (i = 0; i < q->nr; i++) {
1753                struct diff_filepair *p = q->queue[i];
1754                const char *name;
1755                unsigned long copied, added, damage;
1756                int content_changed;
1757
1758                name = p->two->path ? p->two->path : p->one->path;
1759
1760                if (p->one->sha1_valid && p->two->sha1_valid)
1761                        content_changed = hashcmp(p->one->sha1, p->two->sha1);
1762                else
1763                        content_changed = 1;
1764
1765                if (!content_changed) {
1766                        /*
1767                         * The SHA1 has not changed, so pre-/post-content is
1768                         * identical. We can therefore skip looking at the
1769                         * file contents altogether.
1770                         */
1771                        damage = 0;
1772                        goto found_damage;
1773                }
1774
1775                if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1776                        /*
1777                         * In --dirstat-by-file mode, we don't really need to
1778                         * look at the actual file contents at all.
1779                         * The fact that the SHA1 changed is enough for us to
1780                         * add this file to the list of results
1781                         * (with each file contributing equal damage).
1782                         */
1783                        damage = 1;
1784                        goto found_damage;
1785                }
1786
1787                if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1788                        diff_populate_filespec(p->one, 0);
1789                        diff_populate_filespec(p->two, 0);
1790                        diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1791                                               &copied, &added);
1792                        diff_free_filespec_data(p->one);
1793                        diff_free_filespec_data(p->two);
1794                } else if (DIFF_FILE_VALID(p->one)) {
1795                        diff_populate_filespec(p->one, 1);
1796                        copied = added = 0;
1797                        diff_free_filespec_data(p->one);
1798                } else if (DIFF_FILE_VALID(p->two)) {
1799                        diff_populate_filespec(p->two, 1);
1800                        copied = 0;
1801                        added = p->two->size;
1802                        diff_free_filespec_data(p->two);
1803                } else
1804                        continue;
1805
1806                /*
1807                 * Original minus copied is the removed material,
1808                 * added is the new material.  They are both damages
1809                 * made to the preimage.
1810                 * If the resulting damage is zero, we know that
1811                 * diffcore_count_changes() considers the two entries to
1812                 * be identical, but since content_changed is true, we
1813                 * know that there must have been _some_ kind of change,
1814                 * so we force all entries to have damage > 0.
1815                 */
1816                damage = (p->one->size - copied) + added;
1817                if (!damage)
1818                        damage = 1;
1819
1820found_damage:
1821                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1822                dir.files[dir.nr].name = name;
1823                dir.files[dir.nr].changed = damage;
1824                changed += damage;
1825                dir.nr++;
1826        }
1827
1828        /* This can happen even with many files, if everything was renames */
1829        if (!changed)
1830                return;
1831
1832        /* Show all directories with more than x% of the changes */
1833        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1834        gather_dirstat(options, &dir, changed, "", 0);
1835}
1836
1837static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1838{
1839        int i;
1840        unsigned long changed;
1841        struct dirstat_dir dir;
1842
1843        if (data->nr == 0)
1844                return;
1845
1846        dir.files = NULL;
1847        dir.alloc = 0;
1848        dir.nr = 0;
1849        dir.permille = options->dirstat_permille;
1850        dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1851
1852        changed = 0;
1853        for (i = 0; i < data->nr; i++) {
1854                struct diffstat_file *file = data->files[i];
1855                unsigned long damage = file->added + file->deleted;
1856                if (file->is_binary)
1857                        /*
1858                         * binary files counts bytes, not lines. Must find some
1859                         * way to normalize binary bytes vs. textual lines.
1860                         * The following heuristic assumes that there are 64
1861                         * bytes per "line".
1862                         * This is stupid and ugly, but very cheap...
1863                         */
1864                        damage = (damage + 63) / 64;
1865                ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1866                dir.files[dir.nr].name = file->name;
1867                dir.files[dir.nr].changed = damage;
1868                changed += damage;
1869                dir.nr++;
1870        }
1871
1872        /* This can happen even with many files, if everything was renames */
1873        if (!changed)
1874                return;
1875
1876        /* Show all directories with more than x% of the changes */
1877        qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1878        gather_dirstat(options, &dir, changed, "", 0);
1879}
1880
1881static void free_diffstat_info(struct diffstat_t *diffstat)
1882{
1883        int i;
1884        for (i = 0; i < diffstat->nr; i++) {
1885                struct diffstat_file *f = diffstat->files[i];
1886                if (f->name != f->print_name)
1887                        free(f->print_name);
1888                free(f->name);
1889                free(f->from_name);
1890                free(f);
1891        }
1892        free(diffstat->files);
1893}
1894
1895struct checkdiff_t {
1896        const char *filename;
1897        int lineno;
1898        int conflict_marker_size;
1899        struct diff_options *o;
1900        unsigned ws_rule;
1901        unsigned status;
1902};
1903
1904static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1905{
1906        char firstchar;
1907        int cnt;
1908
1909        if (len < marker_size + 1)
1910                return 0;
1911        firstchar = line[0];
1912        switch (firstchar) {
1913        case '=': case '>': case '<': case '|':
1914                break;
1915        default:
1916                return 0;
1917        }
1918        for (cnt = 1; cnt < marker_size; cnt++)
1919                if (line[cnt] != firstchar)
1920                        return 0;
1921        /* line[1] thru line[marker_size-1] are same as firstchar */
1922        if (len < marker_size + 1 || !isspace(line[marker_size]))
1923                return 0;
1924        return 1;
1925}
1926
1927static void checkdiff_consume(void *priv, char *line, unsigned long len)
1928{
1929        struct checkdiff_t *data = priv;
1930        int marker_size = data->conflict_marker_size;
1931        const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
1932        const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
1933        const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
1934        char *err;
1935        char *line_prefix = "";
1936        struct strbuf *msgbuf;
1937
1938        assert(data->o);
1939        if (data->o->output_prefix) {
1940                msgbuf = data->o->output_prefix(data->o,
1941                        data->o->output_prefix_data);
1942                line_prefix = msgbuf->buf;
1943        }
1944
1945        if (line[0] == '+') {
1946                unsigned bad;
1947                data->lineno++;
1948                if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1949                        data->status |= 1;
1950                        fprintf(data->o->file,
1951                                "%s%s:%d: leftover conflict marker\n",
1952                                line_prefix, data->filename, data->lineno);
1953                }
1954                bad = ws_check(line + 1, len - 1, data->ws_rule);
1955                if (!bad)
1956                        return;
1957                data->status |= bad;
1958                err = whitespace_error_string(bad);
1959                fprintf(data->o->file, "%s%s:%d: %s.\n",
1960                        line_prefix, data->filename, data->lineno, err);
1961                free(err);
1962                emit_line(data->o, set, reset, line, 1);
1963                ws_check_emit(line + 1, len - 1, data->ws_rule,
1964                              data->o->file, set, reset, ws);
1965        } else if (line[0] == ' ') {
1966                data->lineno++;
1967        } else if (line[0] == '@') {
1968                char *plus = strchr(line, '+');
1969                if (plus)
1970                        data->lineno = strtol(plus, NULL, 10) - 1;
1971                else
1972                        die("invalid diff");
1973        }
1974}
1975
1976static unsigned char *deflate_it(char *data,
1977                                 unsigned long size,
1978                                 unsigned long *result_size)
1979{
1980        int bound;
1981        unsigned char *deflated;
1982        git_zstream stream;
1983
1984        memset(&stream, 0, sizeof(stream));
1985        git_deflate_init(&stream, zlib_compression_level);
1986        bound = git_deflate_bound(&stream, size);
1987        deflated = xmalloc(bound);
1988        stream.next_out = deflated;
1989        stream.avail_out = bound;
1990
1991        stream.next_in = (unsigned char *)data;
1992        stream.avail_in = size;
1993        while (git_deflate(&stream, Z_FINISH) == Z_OK)
1994                ; /* nothing */
1995        git_deflate_end(&stream);
1996        *result_size = stream.total_out;
1997        return deflated;
1998}
1999
2000static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2001{
2002        void *cp;
2003        void *delta;
2004        void *deflated;
2005        void *data;
2006        unsigned long orig_size;
2007        unsigned long delta_size;
2008        unsigned long deflate_size;
2009        unsigned long data_size;
2010
2011        /* We could do deflated delta, or we could do just deflated two,
2012         * whichever is smaller.
2013         */
2014        delta = NULL;
2015        deflated = deflate_it(two->ptr, two->size, &deflate_size);
2016        if (one->size && two->size) {
2017                delta = diff_delta(one->ptr, one->size,
2018                                   two->ptr, two->size,
2019                                   &delta_size, deflate_size);
2020                if (delta) {
2021                        void *to_free = delta;
2022                        orig_size = delta_size;
2023                        delta = deflate_it(delta, delta_size, &delta_size);
2024                        free(to_free);
2025                }
2026        }
2027
2028        if (delta && delta_size < deflate_size) {
2029                fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2030                free(deflated);
2031                data = delta;
2032                data_size = delta_size;
2033        }
2034        else {
2035                fprintf(file, "%sliteral %lu\n", prefix, two->size);
2036                free(delta);
2037                data = deflated;
2038                data_size = deflate_size;
2039        }
2040
2041        /* emit data encoded in base85 */
2042        cp = data;
2043        while (data_size) {
2044                int bytes = (52 < data_size) ? 52 : data_size;
2045                char line[70];
2046                data_size -= bytes;
2047                if (bytes <= 26)
2048                        line[0] = bytes + 'A' - 1;
2049                else
2050                        line[0] = bytes - 26 + 'a' - 1;
2051                encode_85(line + 1, cp, bytes);
2052                cp = (char *) cp + bytes;
2053                fprintf(file, "%s", prefix);
2054                fputs(line, file);
2055                fputc('\n', file);
2056        }
2057        fprintf(file, "%s\n", prefix);
2058        free(data);
2059}
2060
2061static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2062{
2063        fprintf(file, "%sGIT binary patch\n", prefix);
2064        emit_binary_diff_body(file, one, two, prefix);
2065        emit_binary_diff_body(file, two, one, prefix);
2066}
2067
2068static void diff_filespec_load_driver(struct diff_filespec *one)
2069{
2070        /* Use already-loaded driver */
2071        if (one->driver)
2072                return;
2073
2074        if (S_ISREG(one->mode))
2075                one->driver = userdiff_find_by_path(one->path);
2076
2077        /* Fallback to default settings */
2078        if (!one->driver)
2079                one->driver = userdiff_find_by_name("default");
2080}
2081
2082int diff_filespec_is_binary(struct diff_filespec *one)
2083{
2084        if (one->is_binary == -1) {
2085                diff_filespec_load_driver(one);
2086                if (one->driver->binary != -1)
2087                        one->is_binary = one->driver->binary;
2088                else {
2089                        if (!one->data && DIFF_FILE_VALID(one))
2090                                diff_populate_filespec(one, 0);
2091                        if (one->data)
2092                                one->is_binary = buffer_is_binary(one->data,
2093                                                one->size);
2094                        if (one->is_binary == -1)
2095                                one->is_binary = 0;
2096                }
2097        }
2098        return one->is_binary;
2099}
2100
2101static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2102{
2103        diff_filespec_load_driver(one);
2104        return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2105}
2106
2107static const char *userdiff_word_regex(struct diff_filespec *one)
2108{
2109        diff_filespec_load_driver(one);
2110        return one->driver->word_regex;
2111}
2112
2113void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2114{
2115        if (!options->a_prefix)
2116                options->a_prefix = a;
2117        if (!options->b_prefix)
2118                options->b_prefix = b;
2119}
2120
2121struct userdiff_driver *get_textconv(struct diff_filespec *one)
2122{
2123        if (!DIFF_FILE_VALID(one))
2124                return NULL;
2125
2126        diff_filespec_load_driver(one);
2127        return userdiff_get_textconv(one->driver);
2128}
2129
2130static void builtin_diff(const char *name_a,
2131                         const char *name_b,
2132                         struct diff_filespec *one,
2133                         struct diff_filespec *two,
2134                         const char *xfrm_msg,
2135                         int must_show_header,
2136                         struct diff_options *o,
2137                         int complete_rewrite)
2138{
2139        mmfile_t mf1, mf2;
2140        const char *lbl[2];
2141        char *a_one, *b_two;
2142        const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2143        const char *reset = diff_get_color_opt(o, DIFF_RESET);
2144        const char *a_prefix, *b_prefix;
2145        struct userdiff_driver *textconv_one = NULL;
2146        struct userdiff_driver *textconv_two = NULL;
2147        struct strbuf header = STRBUF_INIT;
2148        struct strbuf *msgbuf;
2149        char *line_prefix = "";
2150
2151        if (o->output_prefix) {
2152                msgbuf = o->output_prefix(o, o->output_prefix_data);
2153                line_prefix = msgbuf->buf;
2154        }
2155
2156        if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2157                        (!one->mode || S_ISGITLINK(one->mode)) &&
2158                        (!two->mode || S_ISGITLINK(two->mode))) {
2159                const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2160                const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2161                show_submodule_summary(o->file, one ? one->path : two->path,
2162                                one->sha1, two->sha1, two->dirty_submodule,
2163                                del, add, reset);
2164                return;
2165        }
2166
2167        if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2168                textconv_one = get_textconv(one);
2169                textconv_two = get_textconv(two);
2170        }
2171
2172        diff_set_mnemonic_prefix(o, "a/", "b/");
2173        if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2174                a_prefix = o->b_prefix;
2175                b_prefix = o->a_prefix;
2176        } else {
2177                a_prefix = o->a_prefix;
2178                b_prefix = o->b_prefix;
2179        }
2180
2181        /* Never use a non-valid filename anywhere if at all possible */
2182        name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2183        name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2184
2185        a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2186        b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2187        lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2188        lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2189        strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2190        if (lbl[0][0] == '/') {
2191                /* /dev/null */
2192                strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2193                if (xfrm_msg)
2194                        strbuf_addstr(&header, xfrm_msg);
2195                must_show_header = 1;
2196        }
2197        else if (lbl[1][0] == '/') {
2198                strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2199                if (xfrm_msg)
2200                        strbuf_addstr(&header, xfrm_msg);
2201                must_show_header = 1;
2202        }
2203        else {
2204                if (one->mode != two->mode) {
2205                        strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2206                        strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2207                        must_show_header = 1;
2208                }
2209                if (xfrm_msg)
2210                        strbuf_addstr(&header, xfrm_msg);
2211
2212                /*
2213                 * we do not run diff between different kind
2214                 * of objects.
2215                 */
2216                if ((one->mode ^ two->mode) & S_IFMT)
2217                        goto free_ab_and_return;
2218                if (complete_rewrite &&
2219                    (textconv_one || !diff_filespec_is_binary(one)) &&
2220                    (textconv_two || !diff_filespec_is_binary(two))) {
2221                        fprintf(o->file, "%s", header.buf);
2222                        strbuf_reset(&header);
2223                        emit_rewrite_diff(name_a, name_b, one, two,
2224                                                textconv_one, textconv_two, o);
2225                        o->found_changes = 1;
2226                        goto free_ab_and_return;
2227                }
2228        }
2229
2230        if (o->irreversible_delete && lbl[1][0] == '/') {
2231                fprintf(o->file, "%s", header.buf);
2232                strbuf_reset(&header);
2233                goto free_ab_and_return;
2234        } else if (!DIFF_OPT_TST(o, TEXT) &&
2235            ( (!textconv_one && diff_filespec_is_binary(one)) ||
2236              (!textconv_two && diff_filespec_is_binary(two)) )) {
2237                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2238                        die("unable to read files to diff");
2239                /* Quite common confusing case */
2240                if (mf1.size == mf2.size &&
2241                    !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2242                        if (must_show_header)
2243                                fprintf(o->file, "%s", header.buf);
2244                        goto free_ab_and_return;
2245                }
2246                fprintf(o->file, "%s", header.buf);
2247                strbuf_reset(&header);
2248                if (DIFF_OPT_TST(o, BINARY))
2249                        emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2250                else
2251                        fprintf(o->file, "%sBinary files %s and %s differ\n",
2252                                line_prefix, lbl[0], lbl[1]);
2253                o->found_changes = 1;
2254        } else {
2255                /* Crazy xdl interfaces.. */
2256                const char *diffopts = getenv("GIT_DIFF_OPTS");
2257                xpparam_t xpp;
2258                xdemitconf_t xecfg;
2259                struct emit_callback ecbdata;
2260                const struct userdiff_funcname *pe;
2261
2262                if (must_show_header) {
2263                        fprintf(o->file, "%s", header.buf);
2264                        strbuf_reset(&header);
2265                }
2266
2267                mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2268                mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2269
2270                pe = diff_funcname_pattern(one);
2271                if (!pe)
2272                        pe = diff_funcname_pattern(two);
2273
2274                memset(&xpp, 0, sizeof(xpp));
2275                memset(&xecfg, 0, sizeof(xecfg));
2276                memset(&ecbdata, 0, sizeof(ecbdata));
2277                ecbdata.label_path = lbl;
2278                ecbdata.color_diff = want_color(o->use_color);
2279                ecbdata.found_changesp = &o->found_changes;
2280                ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2281                if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2282                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2283                ecbdata.opt = o;
2284                ecbdata.header = header.len ? &header : NULL;
2285                xpp.flags = o->xdl_opts;
2286                xecfg.ctxlen = o->context;
2287                xecfg.interhunkctxlen = o->interhunkcontext;
2288                xecfg.flags = XDL_EMIT_FUNCNAMES;
2289                if (DIFF_OPT_TST(o, FUNCCONTEXT))
2290                        xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2291                if (pe)
2292                        xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2293                if (!diffopts)
2294                        ;
2295                else if (!prefixcmp(diffopts, "--unified="))
2296                        xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2297                else if (!prefixcmp(diffopts, "-u"))
2298                        xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2299                if (o->word_diff) {
2300                        int i;
2301
2302                        ecbdata.diff_words =
2303                                xcalloc(1, sizeof(struct diff_words_data));
2304                        ecbdata.diff_words->type = o->word_diff;
2305                        ecbdata.diff_words->opt = o;
2306                        if (!o->word_regex)
2307                                o->word_regex = userdiff_word_regex(one);
2308                        if (!o->word_regex)
2309                                o->word_regex = userdiff_word_regex(two);
2310                        if (!o->word_regex)
2311                                o->word_regex = diff_word_regex_cfg;
2312                        if (o->word_regex) {
2313                                ecbdata.diff_words->word_regex = (regex_t *)
2314                                        xmalloc(sizeof(regex_t));
2315                                if (regcomp(ecbdata.diff_words->word_regex,
2316                                                o->word_regex,
2317                                                REG_EXTENDED | REG_NEWLINE))
2318                                        die ("Invalid regular expression: %s",
2319                                                        o->word_regex);
2320                        }
2321                        for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2322                                if (o->word_diff == diff_words_styles[i].type) {
2323                                        ecbdata.diff_words->style =
2324                                                &diff_words_styles[i];
2325                                        break;
2326                                }
2327                        }
2328                        if (want_color(o->use_color)) {
2329                                struct diff_words_style *st = ecbdata.diff_words->style;
2330                                st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2331                                st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2332                                st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
2333                        }
2334                }
2335                xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2336                              &xpp, &xecfg);
2337                if (o->word_diff)
2338                        free_diff_words_data(&ecbdata);
2339                if (textconv_one)
2340                        free(mf1.ptr);
2341                if (textconv_two)
2342                        free(mf2.ptr);
2343                xdiff_clear_find_func(&xecfg);
2344        }
2345
2346 free_ab_and_return:
2347        strbuf_release(&header);
2348        diff_free_filespec_data(one);
2349        diff_free_filespec_data(two);
2350        free(a_one);
2351        free(b_two);
2352        return;
2353}
2354
2355static void builtin_diffstat(const char *name_a, const char *name_b,
2356                             struct diff_filespec *one,
2357                             struct diff_filespec *two,
2358                             struct diffstat_t *diffstat,
2359                             struct diff_options *o,
2360                             int complete_rewrite)
2361{
2362        mmfile_t mf1, mf2;
2363        struct diffstat_file *data;
2364
2365        data = diffstat_add(diffstat, name_a, name_b);
2366
2367        if (!one || !two) {
2368                data->is_unmerged = 1;
2369                return;
2370        }
2371
2372        if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2373                data->is_binary = 1;
2374                data->added = diff_filespec_size(two);
2375                data->deleted = diff_filespec_size(one);
2376        }
2377
2378        else if (complete_rewrite) {
2379                diff_populate_filespec(one, 0);
2380                diff_populate_filespec(two, 0);
2381                data->deleted = count_lines(one->data, one->size);
2382                data->added = count_lines(two->data, two->size);
2383        }
2384
2385        else {
2386                /* Crazy xdl interfaces.. */
2387                xpparam_t xpp;
2388                xdemitconf_t xecfg;
2389
2390                if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2391                        die("unable to read files to diff");
2392
2393                memset(&xpp, 0, sizeof(xpp));
2394                memset(&xecfg, 0, sizeof(xecfg));
2395                xpp.flags = o->xdl_opts;
2396                xecfg.ctxlen = o->context;
2397                xecfg.interhunkctxlen = o->interhunkcontext;
2398                xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2399                              &xpp, &xecfg);
2400        }
2401
2402        diff_free_filespec_data(one);
2403        diff_free_filespec_data(two);
2404}
2405
2406static void builtin_checkdiff(const char *name_a, const char *name_b,
2407                              const char *attr_path,
2408                              struct diff_filespec *one,
2409                              struct diff_filespec *two,
2410                              struct diff_options *o)
2411{
2412        mmfile_t mf1, mf2;
2413        struct checkdiff_t data;
2414
2415        if (!two)
2416                return;
2417
2418        memset(&data, 0, sizeof(data));
2419        data.filename = name_b ? name_b : name_a;
2420        data.lineno = 0;
2421        data.o = o;
2422        data.ws_rule = whitespace_rule(attr_path);
2423        data.conflict_marker_size = ll_merge_marker_size(attr_path);
2424
2425        if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2426                die("unable to read files to diff");
2427
2428        /*
2429         * All the other codepaths check both sides, but not checking
2430         * the "old" side here is deliberate.  We are checking the newly
2431         * introduced changes, and as long as the "new" side is text, we
2432         * can and should check what it introduces.
2433         */
2434        if (diff_filespec_is_binary(two))
2435                goto free_and_return;
2436        else {
2437                /* Crazy xdl interfaces.. */
2438                xpparam_t xpp;
2439                xdemitconf_t xecfg;
2440
2441                memset(&xpp, 0, sizeof(xpp));
2442                memset(&xecfg, 0, sizeof(xecfg));
2443                xecfg.ctxlen = 1; /* at least one context line */
2444                xpp.flags = 0;
2445                xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2446                              &xpp, &xecfg);
2447
2448                if (data.ws_rule & WS_BLANK_AT_EOF) {
2449                        struct emit_callback ecbdata;
2450                        int blank_at_eof;
2451
2452                        ecbdata.ws_rule = data.ws_rule;
2453                        check_blank_at_eof(&mf1, &mf2, &ecbdata);
2454                        blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2455
2456                        if (blank_at_eof) {
2457                                static char *err;
2458                                if (!err)
2459                                        err = whitespace_error_string(WS_BLANK_AT_EOF);
2460                                fprintf(o->file, "%s:%d: %s.\n",
2461                                        data.filename, blank_at_eof, err);
2462                                data.status = 1; /* report errors */
2463                        }
2464                }
2465        }
2466 free_and_return:
2467        diff_free_filespec_data(one);
2468        diff_free_filespec_data(two);
2469        if (data.status)
2470                DIFF_OPT_SET(o, CHECK_FAILED);
2471}
2472
2473struct diff_filespec *alloc_filespec(const char *path)
2474{
2475        int namelen = strlen(path);
2476        struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2477
2478        memset(spec, 0, sizeof(*spec));
2479        spec->path = (char *)(spec + 1);
2480        memcpy(spec->path, path, namelen+1);
2481        spec->count = 1;
2482        spec->is_binary = -1;
2483        return spec;
2484}
2485
2486void free_filespec(struct diff_filespec *spec)
2487{
2488        if (!--spec->count) {
2489                diff_free_filespec_data(spec);
2490                free(spec);
2491        }
2492}
2493
2494void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2495                   unsigned short mode)
2496{
2497        if (mode) {
2498                spec->mode = canon_mode(mode);
2499                hashcpy(spec->sha1, sha1);
2500                spec->sha1_valid = !is_null_sha1(sha1);
2501        }
2502}
2503
2504/*
2505 * Given a name and sha1 pair, if the index tells us the file in
2506 * the work tree has that object contents, return true, so that
2507 * prepare_temp_file() does not have to inflate and extract.
2508 */
2509static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2510{
2511        struct cache_entry *ce;
2512        struct stat st;
2513        int pos, len;
2514
2515        /*
2516         * We do not read the cache ourselves here, because the
2517         * benchmark with my previous version that always reads cache
2518         * shows that it makes things worse for diff-tree comparing
2519         * two linux-2.6 kernel trees in an already checked out work
2520         * tree.  This is because most diff-tree comparisons deal with
2521         * only a small number of files, while reading the cache is
2522         * expensive for a large project, and its cost outweighs the
2523         * savings we get by not inflating the object to a temporary
2524         * file.  Practically, this code only helps when we are used
2525         * by diff-cache --cached, which does read the cache before
2526         * calling us.
2527         */
2528        if (!active_cache)
2529                return 0;
2530
2531        /* We want to avoid the working directory if our caller
2532         * doesn't need the data in a normal file, this system
2533         * is rather slow with its stat/open/mmap/close syscalls,
2534         * and the object is contained in a pack file.  The pack
2535         * is probably already open and will be faster to obtain
2536         * the data through than the working directory.  Loose
2537         * objects however would tend to be slower as they need
2538         * to be individually opened and inflated.
2539         */
2540        if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2541                return 0;
2542
2543        len = strlen(name);
2544        pos = cache_name_pos(name, len);
2545        if (pos < 0)
2546                return 0;
2547        ce = active_cache[pos];
2548
2549        /*
2550         * This is not the sha1 we are looking for, or
2551         * unreusable because it is not a regular file.
2552         */
2553        if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2554                return 0;
2555
2556        /*
2557         * If ce is marked as "assume unchanged", there is no
2558         * guarantee that work tree matches what we are looking for.
2559         */
2560        if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2561                return 0;
2562
2563        /*
2564         * If ce matches the file in the work tree, we can reuse it.
2565         */
2566        if (ce_uptodate(ce) ||
2567            (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2568                return 1;
2569
2570        return 0;
2571}
2572
2573static int populate_from_stdin(struct diff_filespec *s)
2574{
2575        struct strbuf buf = STRBUF_INIT;
2576        size_t size = 0;
2577
2578        if (strbuf_read(&buf, 0, 0) < 0)
2579                return error("error while reading from stdin %s",
2580                                     strerror(errno));
2581
2582        s->should_munmap = 0;
2583        s->data = strbuf_detach(&buf, &size);
2584        s->size = size;
2585        s->should_free = 1;
2586        return 0;
2587}
2588
2589static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2590{
2591        int len;
2592        char *data = xmalloc(100), *dirty = "";
2593
2594        /* Are we looking at the work tree? */
2595        if (s->dirty_submodule)
2596                dirty = "-dirty";
2597
2598        len = snprintf(data, 100,
2599                       "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2600        s->data = data;
2601        s->size = len;
2602        s->should_free = 1;
2603        if (size_only) {
2604                s->data = NULL;
2605                free(data);
2606        }
2607        return 0;
2608}
2609
2610/*
2611 * While doing rename detection and pickaxe operation, we may need to
2612 * grab the data for the blob (or file) for our own in-core comparison.
2613 * diff_filespec has data and size fields for this purpose.
2614 */
2615int diff_populate_filespec(struct diff_filespec *s, int size_only)
2616{
2617        int err = 0;
2618        if (!DIFF_FILE_VALID(s))
2619                die("internal error: asking to populate invalid file.");
2620        if (S_ISDIR(s->mode))
2621                return -1;
2622
2623        if (s->data)
2624                return 0;
2625
2626        if (size_only && 0 < s->size)
2627                return 0;
2628
2629        if (S_ISGITLINK(s->mode))
2630                return diff_populate_gitlink(s, size_only);
2631
2632        if (!s->sha1_valid ||
2633            reuse_worktree_file(s->path, s->sha1, 0)) {
2634                struct strbuf buf = STRBUF_INIT;
2635                struct stat st;
2636                int fd;
2637
2638                if (!strcmp(s->path, "-"))
2639                        return populate_from_stdin(s);
2640
2641                if (lstat(s->path, &st) < 0) {
2642                        if (errno == ENOENT) {
2643                        err_empty:
2644                                err = -1;
2645                        empty:
2646                                s->data = (char *)"";
2647                                s->size = 0;
2648                                return err;
2649                        }
2650                }
2651                s->size = xsize_t(st.st_size);
2652                if (!s->size)
2653                        goto empty;
2654                if (S_ISLNK(st.st_mode)) {
2655                        struct strbuf sb = STRBUF_INIT;
2656
2657                        if (strbuf_readlink(&sb, s->path, s->size))
2658                                goto err_empty;
2659                        s->size = sb.len;
2660                        s->data = strbuf_detach(&sb, NULL);
2661                        s->should_free = 1;
2662                        return 0;
2663                }
2664                if (size_only)
2665                        return 0;
2666                fd = open(s->path, O_RDONLY);
2667                if (fd < 0)
2668                        goto err_empty;
2669                s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2670                close(fd);
2671                s->should_munmap = 1;
2672
2673                /*
2674                 * Convert from working tree format to canonical git format
2675                 */
2676                if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2677                        size_t size = 0;
2678                        munmap(s->data, s->size);
2679                        s->should_munmap = 0;
2680                        s->data = strbuf_detach(&buf, &size);
2681                        s->size = size;
2682                        s->should_free = 1;
2683                }
2684        }
2685        else {
2686                enum object_type type;
2687                if (size_only) {
2688                        type = sha1_object_info(s->sha1, &s->size);
2689                        if (type < 0)
2690                                die("unable to read %s", sha1_to_hex(s->sha1));
2691                } else {
2692                        s->data = read_sha1_file(s->sha1, &type, &s->size);
2693                        if (!s->data)
2694                                die("unable to read %s", sha1_to_hex(s->sha1));
2695                        s->should_free = 1;
2696                }
2697        }
2698        return 0;
2699}
2700
2701void diff_free_filespec_blob(struct diff_filespec *s)
2702{
2703        if (s->should_free)
2704                free(s->data);
2705        else if (s->should_munmap)
2706                munmap(s->data, s->size);
2707
2708        if (s->should_free || s->should_munmap) {
2709                s->should_free = s->should_munmap = 0;
2710                s->data = NULL;
2711        }
2712}
2713
2714void diff_free_filespec_data(struct diff_filespec *s)
2715{
2716        diff_free_filespec_blob(s);
2717        free(s->cnt_data);
2718        s->cnt_data = NULL;
2719}
2720
2721static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2722                           void *blob,
2723                           unsigned long size,
2724                           const unsigned char *sha1,
2725                           int mode)
2726{
2727        int fd;
2728        struct strbuf buf = STRBUF_INIT;
2729        struct strbuf template = STRBUF_INIT;
2730        char *path_dup = xstrdup(path);
2731        const char *base = basename(path_dup);
2732
2733        /* Generate "XXXXXX_basename.ext" */
2734        strbuf_addstr(&template, "XXXXXX_");
2735        strbuf_addstr(&template, base);
2736
2737        fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2738                        strlen(base) + 1);
2739        if (fd < 0)
2740                die_errno("unable to create temp-file");
2741        if (convert_to_working_tree(path,
2742                        (const char *)blob, (size_t)size, &buf)) {
2743                blob = buf.buf;
2744                size = buf.len;
2745        }
2746        if (write_in_full(fd, blob, size) != size)
2747                die_errno("unable to write temp-file");
2748        close(fd);
2749        temp->name = temp->tmp_path;
2750        strcpy(temp->hex, sha1_to_hex(sha1));
2751        temp->hex[40] = 0;
2752        sprintf(temp->mode, "%06o", mode);
2753        strbuf_release(&buf);
2754        strbuf_release(&template);
2755        free(path_dup);
2756}
2757
2758static struct diff_tempfile *prepare_temp_file(const char *name,
2759                struct diff_filespec *one)
2760{
2761        struct diff_tempfile *temp = claim_diff_tempfile();
2762
2763        if (!DIFF_FILE_VALID(one)) {
2764        not_a_valid_file:
2765                /* A '-' entry produces this for file-2, and
2766                 * a '+' entry produces this for file-1.
2767                 */
2768                temp->name = "/dev/null";
2769                strcpy(temp->hex, ".");
2770                strcpy(temp->mode, ".");
2771                return temp;
2772        }
2773
2774        if (!remove_tempfile_installed) {
2775                atexit(remove_tempfile);
2776                sigchain_push_common(remove_tempfile_on_signal);
2777                remove_tempfile_installed = 1;
2778        }
2779
2780        if (!one->sha1_valid ||
2781            reuse_worktree_file(name, one->sha1, 1)) {
2782                struct stat st;
2783                if (lstat(name, &st) < 0) {
2784                        if (errno == ENOENT)
2785                                goto not_a_valid_file;
2786                        die_errno("stat(%s)", name);
2787                }
2788                if (S_ISLNK(st.st_mode)) {
2789                        struct strbuf sb = STRBUF_INIT;
2790                        if (strbuf_readlink(&sb, name, st.st_size) < 0)
2791                                die_errno("readlink(%s)", name);
2792                        prep_temp_blob(name, temp, sb.buf, sb.len,
2793                                       (one->sha1_valid ?
2794                                        one->sha1 : null_sha1),
2795                                       (one->sha1_valid ?
2796                                        one->mode : S_IFLNK));
2797                        strbuf_release(&sb);
2798                }
2799                else {
2800                        /* we can borrow from the file in the work tree */
2801                        temp->name = name;
2802                        if (!one->sha1_valid)
2803                                strcpy(temp->hex, sha1_to_hex(null_sha1));
2804                        else
2805                                strcpy(temp->hex, sha1_to_hex(one->sha1));
2806                        /* Even though we may sometimes borrow the
2807                         * contents from the work tree, we always want
2808                         * one->mode.  mode is trustworthy even when
2809                         * !(one->sha1_valid), as long as
2810                         * DIFF_FILE_VALID(one).
2811                         */
2812                        sprintf(temp->mode, "%06o", one->mode);
2813                }
2814                return temp;
2815        }
2816        else {
2817                if (diff_populate_filespec(one, 0))
2818                        die("cannot read data blob for %s", one->path);
2819                prep_temp_blob(name, temp, one->data, one->size,
2820                               one->sha1, one->mode);
2821        }
2822        return temp;
2823}
2824
2825/* An external diff command takes:
2826 *
2827 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2828 *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2829 *
2830 */
2831static void run_external_diff(const char *pgm,
2832                              const char *name,
2833                              const char *other,
2834                              struct diff_filespec *one,
2835                              struct diff_filespec *two,
2836                              const char *xfrm_msg,
2837                              int complete_rewrite)
2838{
2839        const char *spawn_arg[10];
2840        int retval;
2841        const char **arg = &spawn_arg[0];
2842
2843        if (one && two) {
2844                struct diff_tempfile *temp_one, *temp_two;
2845                const char *othername = (other ? other : name);
2846                temp_one = prepare_temp_file(name, one);
2847                temp_two = prepare_temp_file(othername, two);
2848                *arg++ = pgm;
2849                *arg++ = name;
2850                *arg++ = temp_one->name;
2851                *arg++ = temp_one->hex;
2852                *arg++ = temp_one->mode;
2853                *arg++ = temp_two->name;
2854                *arg++ = temp_two->hex;
2855                *arg++ = temp_two->mode;
2856                if (other) {
2857                        *arg++ = other;
2858                        *arg++ = xfrm_msg;
2859                }
2860        } else {
2861                *arg++ = pgm;
2862                *arg++ = name;
2863        }
2864        *arg = NULL;
2865        fflush(NULL);
2866        retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2867        remove_tempfile();
2868        if (retval) {
2869                fprintf(stderr, "external diff died, stopping at %s.\n", name);
2870                exit(1);
2871        }
2872}
2873
2874static int similarity_index(struct diff_filepair *p)
2875{
2876        return p->score * 100 / MAX_SCORE;
2877}
2878
2879static void fill_metainfo(struct strbuf *msg,
2880                          const char *name,
2881                          const char *other,
2882                          struct diff_filespec *one,
2883                          struct diff_filespec *two,
2884                          struct diff_options *o,
2885                          struct diff_filepair *p,
2886                          int *must_show_header,
2887                          int use_color)
2888{
2889        const char *set = diff_get_color(use_color, DIFF_METAINFO);
2890        const char *reset = diff_get_color(use_color, DIFF_RESET);
2891        struct strbuf *msgbuf;
2892        char *line_prefix = "";
2893
2894        *must_show_header = 1;
2895        if (o->output_prefix) {
2896                msgbuf = o->output_prefix(o, o->output_prefix_data);
2897                line_prefix = msgbuf->buf;
2898        }
2899        strbuf_init(msg, PATH_MAX * 2 + 300);
2900        switch (p->status) {
2901        case DIFF_STATUS_COPIED:
2902                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2903                            line_prefix, set, similarity_index(p));
2904                strbuf_addf(msg, "%s\n%s%scopy from ",
2905                            reset,  line_prefix, set);
2906                quote_c_style(name, msg, NULL, 0);
2907                strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2908                quote_c_style(other, msg, NULL, 0);
2909                strbuf_addf(msg, "%s\n", reset);
2910                break;
2911        case DIFF_STATUS_RENAMED:
2912                strbuf_addf(msg, "%s%ssimilarity index %d%%",
2913                            line_prefix, set, similarity_index(p));
2914                strbuf_addf(msg, "%s\n%s%srename from ",
2915                            reset, line_prefix, set);
2916                quote_c_style(name, msg, NULL, 0);
2917                strbuf_addf(msg, "%s\n%s%srename to ",
2918                            reset, line_prefix, set);
2919                quote_c_style(other, msg, NULL, 0);
2920                strbuf_addf(msg, "%s\n", reset);
2921                break;
2922        case DIFF_STATUS_MODIFIED:
2923                if (p->score) {
2924                        strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2925                                    line_prefix,
2926                                    set, similarity_index(p), reset);
2927                        break;
2928                }
2929                /* fallthru */
2930        default:
2931                *must_show_header = 0;
2932        }
2933        if (one && two && hashcmp(one->sha1, two->sha1)) {
2934                int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2935
2936                if (DIFF_OPT_TST(o, BINARY)) {
2937                        mmfile_t mf;
2938                        if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2939                            (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2940                                abbrev = 40;
2941                }
2942                strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2943                            find_unique_abbrev(one->sha1, abbrev));
2944                strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2945                if (one->mode == two->mode)
2946                        strbuf_addf(msg, " %06o", one->mode);
2947                strbuf_addf(msg, "%s\n", reset);
2948        }
2949}
2950
2951static void run_diff_cmd(const char *pgm,
2952                         const char *name,
2953                         const char *other,
2954                         const char *attr_path,
2955                         struct diff_filespec *one,
2956                         struct diff_filespec *two,
2957                         struct strbuf *msg,
2958                         struct diff_options *o,
2959                         struct diff_filepair *p)
2960{
2961        const char *xfrm_msg = NULL;
2962        int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2963        int must_show_header = 0;
2964
2965        if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2966                pgm = NULL;
2967        else {
2968                struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2969                if (drv && drv->external)
2970                        pgm = drv->external;
2971        }
2972
2973        if (msg) {
2974                /*
2975                 * don't use colors when the header is intended for an
2976                 * external diff driver
2977                 */
2978                fill_metainfo(msg, name, other, one, two, o, p,
2979                              &must_show_header,
2980                              want_color(o->use_color) && !pgm);
2981                xfrm_msg = msg->len ? msg->buf : NULL;
2982        }
2983
2984        if (pgm) {
2985                run_external_diff(pgm, name, other, one, two, xfrm_msg,
2986                                  complete_rewrite);
2987                return;
2988        }
2989        if (one && two)
2990                builtin_diff(name, other ? other : name,
2991                             one, two, xfrm_msg, must_show_header,
2992                             o, complete_rewrite);
2993        else
2994                fprintf(o->file, "* Unmerged path %s\n", name);
2995}
2996
2997static void diff_fill_sha1_info(struct diff_filespec *one)
2998{
2999        if (DIFF_FILE_VALID(one)) {
3000                if (!one->sha1_valid) {
3001                        struct stat st;
3002                        if (!strcmp(one->path, "-")) {
3003                                hashcpy(one->sha1, null_sha1);
3004                                return;
3005                        }
3006                        if (lstat(one->path, &st) < 0)
3007                                die_errno("stat '%s'", one->path);
3008                        if (index_path(one->sha1, one->path, &st, 0))
3009                                die("cannot hash %s", one->path);
3010                }
3011        }
3012        else
3013                hashclr(one->sha1);
3014}
3015
3016static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3017{
3018        /* Strip the prefix but do not molest /dev/null and absolute paths */
3019        if (*namep && **namep != '/') {
3020                *namep += prefix_length;
3021                if (**namep == '/')
3022                        ++*namep;
3023        }
3024        if (*otherp && **otherp != '/') {
3025                *otherp += prefix_length;
3026                if (**otherp == '/')
3027                        ++*otherp;
3028        }
3029}
3030
3031static void run_diff(struct diff_filepair *p, struct diff_options *o)
3032{
3033        const char *pgm = external_diff();
3034        struct strbuf msg;
3035        struct diff_filespec *one = p->one;
3036        struct diff_filespec *two = p->two;
3037        const char *name;
3038        const char *other;
3039        const char *attr_path;
3040
3041        name  = p->one->path;
3042        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3043        attr_path = name;
3044        if (o->prefix_length)
3045                strip_prefix(o->prefix_length, &name, &other);
3046
3047        if (DIFF_PAIR_UNMERGED(p)) {
3048                run_diff_cmd(pgm, name, NULL, attr_path,
3049                             NULL, NULL, NULL, o, p);
3050                return;
3051        }
3052
3053        diff_fill_sha1_info(one);
3054        diff_fill_sha1_info(two);
3055
3056        if (!pgm &&
3057            DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3058            (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3059                /*
3060                 * a filepair that changes between file and symlink
3061                 * needs to be split into deletion and creation.
3062                 */
3063                struct diff_filespec *null = alloc_filespec(two->path);
3064                run_diff_cmd(NULL, name, other, attr_path,
3065                             one, null, &msg, o, p);
3066                free(null);
3067                strbuf_release(&msg);
3068
3069                null = alloc_filespec(one->path);
3070                run_diff_cmd(NULL, name, other, attr_path,
3071                             null, two, &msg, o, p);
3072                free(null);
3073        }
3074        else
3075                run_diff_cmd(pgm, name, other, attr_path,
3076                             one, two, &msg, o, p);
3077
3078        strbuf_release(&msg);
3079}
3080
3081static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3082                         struct diffstat_t *diffstat)
3083{
3084        const char *name;
3085        const char *other;
3086        int complete_rewrite = 0;
3087
3088        if (DIFF_PAIR_UNMERGED(p)) {
3089                /* unmerged */
3090                builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3091                return;
3092        }
3093
3094        name = p->one->path;
3095        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3096
3097        if (o->prefix_length)
3098                strip_prefix(o->prefix_length, &name, &other);
3099
3100        diff_fill_sha1_info(p->one);
3101        diff_fill_sha1_info(p->two);
3102
3103        if (p->status == DIFF_STATUS_MODIFIED && p->score)
3104                complete_rewrite = 1;
3105        builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3106}
3107
3108static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3109{
3110        const char *name;
3111        const char *other;
3112        const char *attr_path;
3113
3114        if (DIFF_PAIR_UNMERGED(p)) {
3115                /* unmerged */
3116                return;
3117        }
3118
3119        name = p->one->path;
3120        other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3121        attr_path = other ? other : name;
3122
3123        if (o->prefix_length)
3124                strip_prefix(o->prefix_length, &name, &other);
3125
3126        diff_fill_sha1_info(p->one);
3127        diff_fill_sha1_info(p->two);
3128
3129        builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3130}
3131
3132void diff_setup(struct diff_options *options)
3133{
3134        memcpy(options, &default_diff_options, sizeof(*options));
3135
3136        options->file = stdout;
3137
3138        options->line_termination = '\n';
3139        options->break_opt = -1;
3140        options->rename_limit = -1;
3141        options->dirstat_permille = diff_dirstat_permille_default;
3142        options->context = 3;
3143
3144        options->change = diff_change;
3145        options->add_remove = diff_addremove;
3146        options->use_color = diff_use_color_default;
3147        options->detect_rename = diff_detect_rename_default;
3148
3149        if (diff_no_prefix) {
3150                options->a_prefix = options->b_prefix = "";
3151        } else if (!diff_mnemonic_prefix) {
3152                options->a_prefix = "a/";
3153                options->b_prefix = "b/";
3154        }
3155}
3156
3157int diff_setup_done(struct diff_options *options)
3158{
3159        int count = 0;
3160
3161        if (options->output_format & DIFF_FORMAT_NAME)
3162                count++;
3163        if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3164                count++;
3165        if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3166                count++;
3167        if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3168                count++;
3169        if (count > 1)
3170                die("--name-only, --name-status, --check and -s are mutually exclusive");
3171
3172        /*
3173         * Most of the time we can say "there are changes"
3174         * only by checking if there are changed paths, but
3175         * --ignore-whitespace* options force us to look
3176         * inside contents.
3177         */
3178
3179        if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3180            DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3181            DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3182                DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3183        else
3184                DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3185
3186        if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3187                options->detect_rename = DIFF_DETECT_COPY;
3188
3189        if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3190                options->prefix = NULL;
3191        if (options->prefix)
3192                options->prefix_length = strlen(options->prefix);
3193        else
3194                options->prefix_length = 0;
3195
3196        if (options->output_format & (DIFF_FORMAT_NAME |
3197                                      DIFF_FORMAT_NAME_STATUS |
3198                                      DIFF_FORMAT_CHECKDIFF |
3199                                      DIFF_FORMAT_NO_OUTPUT))
3200                options->output_format &= ~(DIFF_FORMAT_RAW |
3201                                            DIFF_FORMAT_NUMSTAT |
3202                                            DIFF_FORMAT_DIFFSTAT |
3203                                            DIFF_FORMAT_SHORTSTAT |
3204                                            DIFF_FORMAT_DIRSTAT |
3205                                            DIFF_FORMAT_SUMMARY |
3206                                            DIFF_FORMAT_PATCH);
3207
3208        /*
3209         * These cases always need recursive; we do not drop caller-supplied
3210         * recursive bits for other formats here.
3211         */
3212        if (options->output_format & (DIFF_FORMAT_PATCH |
3213                                      DIFF_FORMAT_NUMSTAT |
3214                                      DIFF_FORMAT_DIFFSTAT |
3215                                      DIFF_FORMAT_SHORTSTAT |
3216                                      DIFF_FORMAT_DIRSTAT |
3217                                      DIFF_FORMAT_SUMMARY |
3218                                      DIFF_FORMAT_CHECKDIFF))
3219                DIFF_OPT_SET(options, RECURSIVE);
3220        /*
3221         * Also pickaxe would not work very well if you do not say recursive
3222         */
3223        if (options->pickaxe)
3224                DIFF_OPT_SET(options, RECURSIVE);
3225        /*
3226         * When patches are generated, submodules diffed against the work tree
3227         * must be checked for dirtiness too so it can be shown in the output
3228         */
3229        if (options->output_format & DIFF_FORMAT_PATCH)
3230                DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3231
3232        if (options->detect_rename && options->rename_limit < 0)
3233                options->rename_limit = diff_rename_limit_default;
3234        if (options->setup & DIFF_SETUP_USE_CACHE) {
3235                if (!active_cache)
3236                        /* read-cache does not die even when it fails
3237                         * so it is safe for us to do this here.  Also
3238                         * it does not smudge active_cache or active_nr
3239                         * when it fails, so we do not have to worry about
3240                         * cleaning it up ourselves either.
3241                         */
3242                        read_cache();
3243        }
3244        if (options->abbrev <= 0 || 40 < options->abbrev)
3245                options->abbrev = 40; /* full */
3246
3247        /*
3248         * It does not make sense to show the first hit we happened
3249         * to have found.  It does not make sense not to return with
3250         * exit code in such a case either.
3251         */
3252        if (DIFF_OPT_TST(options, QUICK)) {
3253                options->output_format = DIFF_FORMAT_NO_OUTPUT;
3254                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3255        }
3256
3257        return 0;
3258}
3259
3260static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3261{
3262        char c, *eq;
3263        int len;
3264
3265        if (*arg != '-')
3266                return 0;
3267        c = *++arg;
3268        if (!c)
3269                return 0;
3270        if (c == arg_short) {
3271                c = *++arg;
3272                if (!c)
3273                        return 1;
3274                if (val && isdigit(c)) {
3275                        char *end;
3276                        int n = strtoul(arg, &end, 10);
3277                        if (*end)
3278                                return 0;
3279                        *val = n;
3280                        return 1;
3281                }
3282                return 0;
3283        }
3284        if (c != '-')
3285                return 0;
3286        arg++;
3287        eq = strchr(arg, '=');
3288        if (eq)
3289                len = eq - arg;
3290        else
3291                len = strlen(arg);
3292        if (!len || strncmp(arg, arg_long, len))
3293                return 0;
3294        if (eq) {
3295                int n;
3296                char *end;
3297                if (!isdigit(*++eq))
3298                        return 0;
3299                n = strtoul(eq, &end, 10);
3300                if (*end)
3301                        return 0;
3302                *val = n;
3303        }
3304        return 1;
3305}
3306
3307static int diff_scoreopt_parse(const char *opt);
3308
3309static inline int short_opt(char opt, const char **argv,
3310                            const char **optarg)
3311{
3312        const char *arg = argv[0];
3313        if (arg[0] != '-' || arg[1] != opt)
3314                return 0;
3315        if (arg[2] != '\0') {
3316                *optarg = arg + 2;
3317                return 1;
3318        }
3319        if (!argv[1])
3320                die("Option '%c' requires a value", opt);
3321        *optarg = argv[1];
3322        return 2;
3323}
3324
3325int parse_long_opt(const char *opt, const char **argv,
3326                   const char **optarg)
3327{
3328        const char *arg = argv[0];
3329        if (arg[0] != '-' || arg[1] != '-')
3330                return 0;
3331        arg += strlen("--");
3332        if (prefixcmp(arg, opt))
3333                return 0;
3334        arg += strlen(opt);
3335        if (*arg == '=') { /* sticked form: --option=value */
3336                *optarg = arg + 1;
3337                return 1;
3338        }
3339        if (*arg != '\0')
3340                return 0;
3341        /* separate form: --option value */
3342        if (!argv[1])
3343                die("Option '--%s' requires a value", opt);
3344        *optarg = argv[1];
3345        return 2;
3346}
3347
3348static int stat_opt(struct diff_options *options, const char **av)
3349{
3350        const char *arg = av[0];
3351        char *end;
3352        int width = options->stat_width;
3353        int name_width = options->stat_name_width;
3354        int graph_width = options->stat_graph_width;
3355        int count = options->stat_count;
3356        int argcount = 1;
3357
3358        arg += strlen("--stat");
3359        end = (char *)arg;
3360
3361        switch (*arg) {
3362        case '-':
3363                if (!prefixcmp(arg, "-width")) {
3364                        arg += strlen("-width");
3365                        if (*arg == '=')
3366                                width = strtoul(arg + 1, &end, 10);
3367                        else if (!*arg && !av[1])
3368                                die("Option '--stat-width' requires a value");
3369                        else if (!*arg) {
3370                                width = strtoul(av[1], &end, 10);
3371                                argcount = 2;
3372                        }
3373                } else if (!prefixcmp(arg, "-name-width")) {
3374                        arg += strlen("-name-width");
3375                        if (*arg == '=')
3376                                name_width = strtoul(arg + 1, &end, 10);
3377                        else if (!*arg && !av[1])
3378                                die("Option '--stat-name-width' requires a value");
3379                        else if (!*arg) {
3380                                name_width = strtoul(av[1], &end, 10);
3381                                argcount = 2;
3382                        }
3383                } else if (!prefixcmp(arg, "-graph-width")) {
3384                        arg += strlen("-graph-width");
3385                        if (*arg == '=')
3386                                graph_width = strtoul(arg + 1, &end, 10);
3387                        else if (!*arg && !av[1])
3388                                die("Option '--stat-graph-width' requires a value");
3389                        else if (!*arg) {
3390                                graph_width = strtoul(av[1], &end, 10);
3391                                argcount = 2;
3392                        }
3393                } else if (!prefixcmp(arg, "-count")) {
3394                        arg += strlen("-count");
3395                        if (*arg == '=')
3396                                count = strtoul(arg + 1, &end, 10);
3397                        else if (!*arg && !av[1])
3398                                die("Option '--stat-count' requires a value");
3399                        else if (!*arg) {
3400                                count = strtoul(av[1], &end, 10);
3401                                argcount = 2;
3402                        }
3403                }
3404                break;
3405        case '=':
3406                width = strtoul(arg+1, &end, 10);
3407                if (*end == ',')
3408                        name_width = strtoul(end+1, &end, 10);
3409                if (*end == ',')
3410                        count = strtoul(end+1, &end, 10);
3411        }
3412
3413        /* Important! This checks all the error cases! */
3414        if (*end)
3415                return 0;
3416        options->output_format |= DIFF_FORMAT_DIFFSTAT;
3417        options->stat_name_width = name_width;
3418        options->stat_graph_width = graph_width;
3419        options->stat_width = width;
3420        options->stat_count = count;
3421        return argcount;
3422}
3423
3424static int parse_dirstat_opt(struct diff_options *options, const char *params)
3425{
3426        struct strbuf errmsg = STRBUF_INIT;
3427        if (parse_dirstat_params(options, params, &errmsg))
3428                die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3429                    errmsg.buf);
3430        strbuf_release(&errmsg);
3431        /*
3432         * The caller knows a dirstat-related option is given from the command
3433         * line; allow it to say "return this_function();"
3434         */
3435        options->output_format |= DIFF_FORMAT_DIRSTAT;
3436        return 1;
3437}
3438
3439int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3440{
3441        const char *arg = av[0];
3442        const char *optarg;
3443        int argcount;
3444
3445        /* Output format options */
3446        if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3447                options->output_format |= DIFF_FORMAT_PATCH;
3448        else if (opt_arg(arg, 'U', "unified", &options->context))
3449                options->output_format |= DIFF_FORMAT_PATCH;
3450        else if (!strcmp(arg, "--raw"))
3451                options->output_format |= DIFF_FORMAT_RAW;
3452        else if (!strcmp(arg, "--patch-with-raw"))
3453                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3454        else if (!strcmp(arg, "--numstat"))
3455                options->output_format |= DIFF_FORMAT_NUMSTAT;
3456        else if (!strcmp(arg, "--shortstat"))
3457                options->output_format |= DIFF_FORMAT_SHORTSTAT;
3458        else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3459                return parse_dirstat_opt(options, "");
3460        else if (!prefixcmp(arg, "-X"))
3461                return parse_dirstat_opt(options, arg + 2);
3462        else if (!prefixcmp(arg, "--dirstat="))
3463                return parse_dirstat_opt(options, arg + 10);
3464        else if (!strcmp(arg, "--cumulative"))
3465                return parse_dirstat_opt(options, "cumulative");
3466        else if (!strcmp(arg, "--dirstat-by-file"))
3467                return parse_dirstat_opt(options, "files");
3468        else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3469                parse_dirstat_opt(options, "files");
3470                return parse_dirstat_opt(options, arg + 18);
3471        }
3472        else if (!strcmp(arg, "--check"))
3473                options->output_format |= DIFF_FORMAT_CHECKDIFF;
3474        else if (!strcmp(arg, "--summary"))
3475                options->output_format |= DIFF_FORMAT_SUMMARY;
3476        else if (!strcmp(arg, "--patch-with-stat"))
3477                options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3478        else if (!strcmp(arg, "--name-only"))
3479                options->output_format |= DIFF_FORMAT_NAME;
3480        else if (!strcmp(arg, "--name-status"))
3481                options->output_format |= DIFF_FORMAT_NAME_STATUS;
3482        else if (!strcmp(arg, "-s"))
3483                options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3484        else if (!prefixcmp(arg, "--stat"))
3485                /* --stat, --stat-width, --stat-name-width, or --stat-count */
3486                return stat_opt(options, av);
3487
3488        /* renames options */
3489        else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3490                 !strcmp(arg, "--break-rewrites")) {
3491                if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3492                        return error("invalid argument to -B: %s", arg+2);
3493        }
3494        else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3495                 !strcmp(arg, "--find-renames")) {
3496                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3497                        return error("invalid argument to -M: %s", arg+2);
3498                options->detect_rename = DIFF_DETECT_RENAME;
3499        }
3500        else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3501                options->irreversible_delete = 1;
3502        }
3503        else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3504                 !strcmp(arg, "--find-copies")) {
3505                if (options->detect_rename == DIFF_DETECT_COPY)
3506                        DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3507                if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3508                        return error("invalid argument to -C: %s", arg+2);
3509                options->detect_rename = DIFF_DETECT_COPY;
3510        }
3511        else if (!strcmp(arg, "--no-renames"))
3512                options->detect_rename = 0;
3513        else if (!strcmp(arg, "--relative"))
3514                DIFF_OPT_SET(options, RELATIVE_NAME);
3515        else if (!prefixcmp(arg, "--relative=")) {
3516                DIFF_OPT_SET(options, RELATIVE_NAME);
3517                options->prefix = arg + 11;
3518        }
3519
3520        /* xdiff options */
3521        else if (!strcmp(arg, "--minimal"))
3522                DIFF_XDL_SET(options, NEED_MINIMAL);
3523        else if (!strcmp(arg, "--no-minimal"))
3524                DIFF_XDL_CLR(options, NEED_MINIMAL);
3525        else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3526                DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3527        else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3528                DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3529        else if (!strcmp(arg, "--ignore-space-at-eol"))
3530                DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3531        else if (!strcmp(arg, "--patience"))
3532                DIFF_XDL_SET(options, PATIENCE_DIFF);
3533        else if (!strcmp(arg, "--histogram"))
3534                DIFF_XDL_SET(options, HISTOGRAM_DIFF);
3535
3536        /* flags options */
3537        else if (!strcmp(arg, "--binary")) {
3538                options->output_format |= DIFF_FORMAT_PATCH;
3539                DIFF_OPT_SET(options, BINARY);
3540        }
3541        else if (!strcmp(arg, "--full-index"))
3542                DIFF_OPT_SET(options, FULL_INDEX);
3543        else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3544                DIFF_OPT_SET(options, TEXT);
3545        else if (!strcmp(arg, "-R"))
3546                DIFF_OPT_SET(options, REVERSE_DIFF);
3547        else if (!strcmp(arg, "--find-copies-harder"))
3548                DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3549        else if (!strcmp(arg, "--follow"))
3550                DIFF_OPT_SET(options, FOLLOW_RENAMES);
3551        else if (!strcmp(arg, "--color"))
3552                options->use_color = 1;
3553        else if (!prefixcmp(arg, "--color=")) {
3554                int value = git_config_colorbool(NULL, arg+8);
3555                if (value < 0)
3556                        return error("option `color' expects \"always\", \"auto\", or \"never\"");
3557                options->use_color = value;
3558        }
3559        else if (!strcmp(arg, "--no-color"))
3560                options->use_color = 0;
3561        else if (!strcmp(arg, "--color-words")) {
3562                options->use_color = 1;
3563                options->word_diff = DIFF_WORDS_COLOR;
3564        }
3565        else if (!prefixcmp(arg, "--color-words=")) {
3566                options->use_color = 1;
3567                options->word_diff = DIFF_WORDS_COLOR;
3568                options->word_regex = arg + 14;
3569        }
3570        else if (!strcmp(arg, "--word-diff")) {
3571                if (options->word_diff == DIFF_WORDS_NONE)
3572                        options->word_diff = DIFF_WORDS_PLAIN;
3573        }
3574        else if (!prefixcmp(arg, "--word-diff=")) {
3575                const char *type = arg + 12;
3576                if (!strcmp(type, "plain"))
3577                        options->word_diff = DIFF_WORDS_PLAIN;
3578                else if (!strcmp(type, "color")) {
3579                        options->use_color = 1;
3580                        options->word_diff = DIFF_WORDS_COLOR;
3581                }
3582                else if (!strcmp(type, "porcelain"))
3583                        options->word_diff = DIFF_WORDS_PORCELAIN;
3584                else if (!strcmp(type, "none"))
3585                        options->word_diff = DIFF_WORDS_NONE;
3586                else
3587                        die("bad --word-diff argument: %s", type);
3588        }
3589        else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3590                if (options->word_diff == DIFF_WORDS_NONE)
3591                        options->word_diff = DIFF_WORDS_PLAIN;
3592                options->word_regex = optarg;
3593                return argcount;
3594        }
3595        else if (!strcmp(arg, "--exit-code"))
3596                DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3597        else if (!strcmp(arg, "--quiet"))
3598                DIFF_OPT_SET(options, QUICK);
3599        else if (!strcmp(arg, "--ext-diff"))
3600                DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3601        else if (!strcmp(arg, "--no-ext-diff"))
3602                DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3603        else if (!strcmp(arg, "--textconv"))
3604                DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3605        else if (!strcmp(arg, "--no-textconv"))
3606                DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3607        else if (!strcmp(arg, "--ignore-submodules")) {
3608                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3609                handle_ignore_submodules_arg(options, "all");
3610        } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3611                DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3612                handle_ignore_submodules_arg(options, arg + 20);
3613        } else if (!strcmp(arg, "--submodule"))
3614                DIFF_OPT_SET(options, SUBMODULE_LOG);
3615        else if (!prefixcmp(arg, "--submodule=")) {
3616                if (!strcmp(arg + 12, "log"))
3617                        DIFF_OPT_SET(options, SUBMODULE_LOG);
3618        }
3619
3620        /* misc options */
3621        else if (!strcmp(arg, "-z"))
3622                options->line_termination = 0;
3623        else if ((argcount = short_opt('l', av, &optarg))) {
3624                options->rename_limit = strtoul(optarg, NULL, 10);
3625                return argcount;
3626        }
3627        else if ((argcount = short_opt('S', av, &optarg))) {
3628                options->pickaxe = optarg;
3629                options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3630                return argcount;
3631        } else if ((argcount = short_opt('G', av, &optarg))) {
3632                options->pickaxe = optarg;
3633                options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3634                return argcount;
3635        }
3636        else if (!strcmp(arg, "--pickaxe-all"))
3637                options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3638        else if (!strcmp(arg, "--pickaxe-regex"))
3639                options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3640        else if ((argcount = short_opt('O', av, &optarg))) {
3641                options->orderfile = optarg;
3642                return argcount;
3643        }
3644        else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3645                options->filter = optarg;
3646                return argcount;
3647        }
3648        else if (!strcmp(arg, "--abbrev"))
3649                options->abbrev = DEFAULT_ABBREV;
3650        else if (!prefixcmp(arg, "--abbrev=")) {
3651                options->abbrev = strtoul(arg + 9, NULL, 10);
3652                if (options->abbrev < MINIMUM_ABBREV)
3653                        options->abbrev = MINIMUM_ABBREV;
3654                else if (40 < options->abbrev)
3655                        options->abbrev = 40;
3656        }
3657        else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3658                options->a_prefix = optarg;
3659                return argcount;
3660        }
3661        else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3662                options->b_prefix = optarg;
3663                return argcount;
3664        }
3665        else if (!strcmp(arg, "--no-prefix"))
3666                options->a_prefix = options->b_prefix = "";
3667        else if (opt_arg(arg, '\0', "inter-hunk-context",
3668                         &options->interhunkcontext))
3669                ;
3670        else if (!strcmp(arg, "-W"))
3671                DIFF_OPT_SET(options, FUNCCONTEXT);
3672        else if (!strcmp(arg, "--function-context"))
3673                DIFF_OPT_SET(options, FUNCCONTEXT);
3674        else if (!strcmp(arg, "--no-function-context"))
3675                DIFF_OPT_CLR(options, FUNCCONTEXT);
3676        else if ((argcount = parse_long_opt("output", av, &optarg))) {
3677                options->file = fopen(optarg, "w");
3678                if (!options->file)
3679                        die_errno("Could not open '%s'", optarg);
3680                options->close_file = 1;
3681                return argcount;
3682        } else
3683                return 0;
3684        return 1;
3685}
3686
3687int parse_rename_score(const char **cp_p)
3688{
3689        unsigned long num, scale;
3690        int ch, dot;
3691        const char *cp = *cp_p;
3692
3693        num = 0;
3694        scale = 1;
3695        dot = 0;
3696        for (;;) {
3697                ch = *cp;
3698                if ( !dot && ch == '.' ) {
3699                        scale = 1;
3700                        dot = 1;
3701                } else if ( ch == '%' ) {
3702                        scale = dot ? scale*100 : 100;
3703                        cp++;   /* % is always at the end */
3704                        break;
3705                } else if ( ch >= '0' && ch <= '9' ) {
3706                        if ( scale < 100000 ) {
3707                                scale *= 10;
3708                                num = (num*10) + (ch-'0');
3709                        }
3710                } else {
3711                        break;
3712                }
3713                cp++;
3714        }
3715        *cp_p = cp;
3716
3717        /* user says num divided by scale and we say internally that
3718         * is MAX_SCORE * num / scale.
3719         */
3720        return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3721}
3722
3723static int diff_scoreopt_parse(const char *opt)
3724{
3725        int opt1, opt2, cmd;
3726
3727        if (*opt++ != '-')
3728                return -1;
3729        cmd = *opt++;
3730        if (cmd == '-') {
3731                /* convert the long-form arguments into short-form versions */
3732                if (!prefixcmp(opt, "break-rewrites")) {
3733                        opt += strlen("break-rewrites");
3734                        if (*opt == 0 || *opt++ == '=')
3735                                cmd = 'B';
3736                } else if (!prefixcmp(opt, "find-copies")) {
3737                        opt += strlen("find-copies");
3738                        if (*opt == 0 || *opt++ == '=')
3739                                cmd = 'C';
3740                } else if (!prefixcmp(opt, "find-renames")) {
3741                        opt += strlen("find-renames");
3742                        if (*opt == 0 || *opt++ == '=')
3743                                cmd = 'M';
3744                }
3745        }
3746        if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3747                return -1; /* that is not a -M, -C nor -B option */
3748
3749        opt1 = parse_rename_score(&opt);
3750        if (cmd != 'B')
3751                opt2 = 0;
3752        else {
3753                if (*opt == 0)
3754                        opt2 = 0;
3755                else if (*opt != '/')
3756                        return -1; /* we expect -B80/99 or -B80 */
3757                else {
3758                        opt++;
3759                        opt2 = parse_rename_score(&opt);
3760                }
3761        }
3762        if (*opt != 0)
3763                return -1;
3764        return opt1 | (opt2 << 16);
3765}
3766
3767struct diff_queue_struct diff_queued_diff;
3768
3769void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3770{
3771        if (queue->alloc <= queue->nr) {
3772                queue->alloc = alloc_nr(queue->alloc);
3773                queue->queue = xrealloc(queue->queue,
3774                                        sizeof(dp) * queue->alloc);
3775        }
3776        queue->queue[queue->nr++] = dp;
3777}
3778
3779struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3780                                 struct diff_filespec *one,
3781                                 struct diff_filespec *two)
3782{
3783        struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3784        dp->one = one;
3785        dp->two = two;
3786        if (queue)
3787                diff_q(queue, dp);
3788        return dp;
3789}
3790
3791void diff_free_filepair(struct diff_filepair *p)
3792{
3793        free_filespec(p->one);
3794        free_filespec(p->two);
3795        free(p);
3796}
3797
3798/* This is different from find_unique_abbrev() in that
3799 * it stuffs the result with dots for alignment.
3800 */
3801const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3802{
3803        int abblen;
3804        const char *abbrev;
3805        if (len == 40)
3806                return sha1_to_hex(sha1);
3807
3808        abbrev = find_unique_abbrev(sha1, len);
3809        abblen = strlen(abbrev);
3810        if (abblen < 37) {
3811                static char hex[41];
3812                if (len < abblen && abblen <= len + 2)
3813                        sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3814                else
3815                        sprintf(hex, "%s...", abbrev);
3816                return hex;
3817        }
3818        return sha1_to_hex(sha1);
3819}
3820
3821static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3822{
3823        int line_termination = opt->line_termination;
3824        int inter_name_termination = line_termination ? '\t' : '\0';
3825        if (opt->output_prefix) {
3826                struct strbuf *msg = NULL;
3827                msg = opt->output_prefix(opt, opt->output_prefix_data);
3828                fprintf(opt->file, "%s", msg->buf);
3829        }
3830
3831        if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3832                fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3833                        diff_unique_abbrev(p->one->sha1, opt->abbrev));
3834                fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3835        }
3836        if (p->score) {
3837                fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3838                        inter_name_termination);
3839        } else {
3840                fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3841        }
3842
3843        if (p->status == DIFF_STATUS_COPIED ||
3844            p->status == DIFF_STATUS_RENAMED) {
3845                const char *name_a, *name_b;
3846                name_a = p->one->path;
3847                name_b = p->two->path;
3848                strip_prefix(opt->prefix_length, &name_a, &name_b);
3849                write_name_quoted(name_a, opt->file, inter_name_termination);
3850                write_name_quoted(name_b, opt->file, line_termination);
3851        } else {
3852                const char *name_a, *name_b;
3853                name_a = p->one->mode ? p->one->path : p->two->path;
3854                name_b = NULL;
3855                strip_prefix(opt->prefix_length, &name_a, &name_b);
3856                write_name_quoted(name_a, opt->file, line_termination);
3857        }
3858}
3859
3860int diff_unmodified_pair(struct diff_filepair *p)
3861{
3862        /* This function is written stricter than necessary to support
3863         * the currently implemented transformers, but the idea is to
3864         * let transformers to produce diff_filepairs any way they want,
3865         * and filter and clean them up here before producing the output.
3866         */
3867        struct diff_filespec *one = p->one, *two = p->two;
3868
3869        if (DIFF_PAIR_UNMERGED(p))
3870                return 0; /* unmerged is interesting */
3871
3872        /* deletion, addition, mode or type change
3873         * and rename are all interesting.
3874         */
3875        if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3876            DIFF_PAIR_MODE_CHANGED(p) ||
3877            strcmp(one->path, two->path))
3878                return 0;
3879
3880        /* both are valid and point at the same path.  that is, we are
3881         * dealing with a change.
3882         */
3883        if (one->sha1_valid && two->sha1_valid &&
3884            !hashcmp(one->sha1, two->sha1) &&
3885            !one->dirty_submodule && !two->dirty_submodule)
3886                return 1; /* no change */
3887        if (!one->sha1_valid && !two->sha1_valid)
3888                return 1; /* both look at the same file on the filesystem. */
3889        return 0;
3890}
3891
3892static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3893{
3894        if (diff_unmodified_pair(p))
3895                return;
3896
3897        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3898            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3899                return; /* no tree diffs in patch format */
3900
3901        run_diff(p, o);
3902}
3903
3904static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3905                            struct diffstat_t *diffstat)
3906{
3907        if (diff_unmodified_pair(p))
3908                return;
3909
3910        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3911            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3912                return; /* no useful stat for tree diffs */
3913
3914        run_diffstat(p, o, diffstat);
3915}
3916
3917static void diff_flush_checkdiff(struct diff_filepair *p,
3918                struct diff_options *o)
3919{
3920        if (diff_unmodified_pair(p))
3921                return;
3922
3923        if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3924            (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3925                return; /* nothing to check in tree diffs */
3926
3927        run_checkdiff(p, o);
3928}
3929
3930int diff_queue_is_empty(void)
3931{
3932        struct diff_queue_struct *q = &diff_queued_diff;
3933        int i;
3934        for (i = 0; i < q->nr; i++)
3935                if (!diff_unmodified_pair(q->queue[i]))
3936                        return 0;
3937        return 1;
3938}
3939
3940#if DIFF_DEBUG
3941void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3942{
3943        fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3944                x, one ? one : "",
3945                s->path,
3946                DIFF_FILE_VALID(s) ? "valid" : "invalid",
3947                s->mode,
3948                s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3949        fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3950                x, one ? one : "",
3951                s->size, s->xfrm_flags);
3952}
3953
3954void diff_debug_filepair(const struct diff_filepair *p, int i)
3955{
3956        diff_debug_filespec(p->one, i, "one");
3957        diff_debug_filespec(p->two, i, "two");
3958        fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3959                p->score, p->status ? p->status : '?',
3960                p->one->rename_used, p->broken_pair);
3961}
3962
3963void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3964{
3965        int i;
3966        if (msg)
3967                fprintf(stderr, "%s\n", msg);
3968        fprintf(stderr, "q->nr = %d\n", q->nr);
3969        for (i = 0; i < q->nr; i++) {
3970                struct diff_filepair *p = q->queue[i];
3971                diff_debug_filepair(p, i);
3972        }
3973}
3974#endif
3975
3976static void diff_resolve_rename_copy(void)
3977{
3978        int i;
3979        struct diff_filepair *p;
3980        struct diff_queue_struct *q = &diff_queued_diff;
3981
3982        diff_debug_queue("resolve-rename-copy", q);
3983
3984        for (i = 0; i < q->nr; i++) {
3985                p = q->queue[i];
3986                p->status = 0; /* undecided */
3987                if (DIFF_PAIR_UNMERGED(p))
3988                        p->status = DIFF_STATUS_UNMERGED;
3989                else if (!DIFF_FILE_VALID(p->one))
3990                        p->status = DIFF_STATUS_ADDED;
3991                else if (!DIFF_FILE_VALID(p->two))
3992                        p->status = DIFF_STATUS_DELETED;
3993                else if (DIFF_PAIR_TYPE_CHANGED(p))
3994                        p->status = DIFF_STATUS_TYPE_CHANGED;
3995
3996                /* from this point on, we are dealing with a pair
3997                 * whose both sides are valid and of the same type, i.e.
3998                 * either in-place edit or rename/copy edit.
3999                 */
4000                else if (DIFF_PAIR_RENAME(p)) {
4001                        /*
4002                         * A rename might have re-connected a broken
4003                         * pair up, causing the pathnames to be the
4004                         * same again. If so, that's not a rename at
4005                         * all, just a modification..
4006                         *
4007                         * Otherwise, see if this source was used for
4008                         * multiple renames, in which case we decrement
4009                         * the count, and call it a copy.
4010                         */
4011                        if (!strcmp(p->one->path, p->two->path))
4012                                p->status = DIFF_STATUS_MODIFIED;
4013                        else if (--p->one->rename_used > 0)
4014                                p->status = DIFF_STATUS_COPIED;
4015                        else
4016                                p->status = DIFF_STATUS_RENAMED;
4017                }
4018                else if (hashcmp(p->one->sha1, p->two->sha1) ||
4019                         p->one->mode != p->two->mode ||
4020                         p->one->dirty_submodule ||
4021                         p->two->dirty_submodule ||
4022                         is_null_sha1(p->one->sha1))
4023                        p->status = DIFF_STATUS_MODIFIED;
4024                else {
4025                        /* This is a "no-change" entry and should not
4026                         * happen anymore, but prepare for broken callers.
4027                         */
4028                        error("feeding unmodified %s to diffcore",
4029                              p->one->path);
4030                        p->status = DIFF_STATUS_UNKNOWN;
4031                }
4032        }
4033        diff_debug_queue("resolve-rename-copy done", q);
4034}
4035
4036static int check_pair_status(struct diff_filepair *p)
4037{
4038        switch (p->status) {
4039        case DIFF_STATUS_UNKNOWN:
4040                return 0;
4041        case 0:
4042                die("internal error in diff-resolve-rename-copy");
4043        default:
4044                return 1;
4045        }
4046}
4047
4048static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4049{
4050        int fmt = opt->output_format;
4051
4052        if (fmt & DIFF_FORMAT_CHECKDIFF)
4053                diff_flush_checkdiff(p, opt);
4054        else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4055                diff_flush_raw(p, opt);
4056        else if (fmt & DIFF_FORMAT_NAME) {
4057                const char *name_a, *name_b;
4058                name_a = p->two->path;
4059                name_b = NULL;
4060                strip_prefix(opt->prefix_length, &name_a, &name_b);
4061                write_name_quoted(name_a, opt->file, opt->line_termination);
4062        }
4063}
4064
4065static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4066{
4067        if (fs->mode)
4068                fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4069        else
4070                fprintf(file, " %s ", newdelete);
4071        write_name_quoted(fs->path, file, '\n');
4072}
4073
4074
4075static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4076                const char *line_prefix)
4077{
4078        if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4079                fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4080                        p->two->mode, show_name ? ' ' : '\n');
4081                if (show_name) {
4082                        write_name_quoted(p->two->path, file, '\n');
4083                }
4084        }
4085}
4086
4087static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4088                        const char *line_prefix)
4089{
4090        char *names = pprint_rename(p->one->path, p->two->path);
4091
4092        fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4093        free(names);
4094        show_mode_change(file, p, 0, line_prefix);
4095}
4096
4097static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4098{
4099        FILE *file = opt->file;
4100        char *line_prefix = "";
4101
4102        if (opt->output_prefix) {
4103                struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4104                line_prefix = buf->buf;
4105        }
4106
4107        switch(p->status) {
4108        case DIFF_STATUS_DELETED:
4109                fputs(line_prefix, file);
4110                show_file_mode_name(file, "delete", p->one);
4111                break;
4112        case DIFF_STATUS_ADDED:
4113                fputs(line_prefix, file);
4114                show_file_mode_name(file, "create", p->two);
4115                break;
4116        case DIFF_STATUS_COPIED:
4117                fputs(line_prefix, file);
4118                show_rename_copy(file, "copy", p, line_prefix);
4119                break;
4120        case DIFF_STATUS_RENAMED:
4121                fputs(line_prefix, file);
4122                show_rename_copy(file, "rename", p, line_prefix);
4123                break;
4124        default:
4125                if (p->score) {
4126                        fprintf(file, "%s rewrite ", line_prefix);
4127                        write_name_quoted(p->two->path, file, ' ');
4128                        fprintf(file, "(%d%%)\n", similarity_index(p));
4129                }
4130                show_mode_change(file, p, !p->score, line_prefix);
4131                break;
4132        }
4133}
4134
4135struct patch_id_t {
4136        git_SHA_CTX *ctx;
4137        int patchlen;
4138};
4139
4140static int remove_space(char *line, int len)
4141{
4142        int i;
4143        char *dst = line;
4144        unsigned char c;
4145
4146        for (i = 0; i < len; i++)
4147                if (!isspace((c = line[i])))
4148                        *dst++ = c;
4149
4150        return dst - line;
4151}
4152
4153static void patch_id_consume(void *priv, char *line, unsigned long len)
4154{
4155        struct patch_id_t *data = priv;
4156        int new_len;
4157
4158        /* Ignore line numbers when computing the SHA1 of the patch */
4159        if (!prefixcmp(line, "@@ -"))
4160                return;
4161
4162        new_len = remove_space(line, len);
4163
4164        git_SHA1_Update(data->ctx, line, new_len);
4165        data->patchlen += new_len;
4166}
4167
4168/* returns 0 upon success, and writes result into sha1 */
4169static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4170{
4171        struct diff_queue_struct *q = &diff_queued_diff;
4172        int i;
4173        git_SHA_CTX ctx;
4174        struct patch_id_t data;
4175        char buffer[PATH_MAX * 4 + 20];
4176
4177        git_SHA1_Init(&ctx);
4178        memset(&data, 0, sizeof(struct patch_id_t));
4179        data.ctx = &ctx;
4180
4181        for (i = 0; i < q->nr; i++) {
4182                xpparam_t xpp;
4183                xdemitconf_t xecfg;
4184                mmfile_t mf1, mf2;
4185                struct diff_filepair *p = q->queue[i];
4186                int len1, len2;
4187
4188                memset(&xpp, 0, sizeof(xpp));
4189                memset(&xecfg, 0, sizeof(xecfg));
4190                if (p->status == 0)
4191                        return error("internal diff status error");
4192                if (p->status == DIFF_STATUS_UNKNOWN)
4193                        continue;
4194                if (diff_unmodified_pair(p))
4195                        continue;
4196                if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4197                    (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4198                        continue;
4199                if (DIFF_PAIR_UNMERGED(p))
4200                        continue;
4201
4202                diff_fill_sha1_info(p->one);
4203                diff_fill_sha1_info(p->two);
4204                if (fill_mmfile(&mf1, p->one) < 0 ||
4205                                fill_mmfile(&mf2, p->two) < 0)
4206                        return error("unable to read files to diff");
4207
4208                len1 = remove_space(p->one->path, strlen(p->one->path));
4209                len2 = remove_space(p->two->path, strlen(p->two->path));
4210                if (p->one->mode == 0)
4211                        len1 = snprintf(buffer, sizeof(buffer),
4212                                        "diff--gita/%.*sb/%.*s"
4213                                        "newfilemode%06o"
4214                                        "---/dev/null"
4215                                        "+++b/%.*s",
4216                                        len1, p->one->path,
4217                                        len2, p->two->path,
4218                                        p->two->mode,
4219                                        len2, p->two->path);
4220                else if (p->two->mode == 0)
4221                        len1 = snprintf(buffer, sizeof(buffer),
4222                                        "diff--gita/%.*sb/%.*s"
4223                                        "deletedfilemode%06o"
4224                                        "---a/%.*s"
4225                                        "+++/dev/null",
4226                                        len1, p->one->path,
4227                                        len2, p->two->path,
4228                                        p->one->mode,
4229                                        len1, p->one->path);
4230                else
4231                        len1 = snprintf(buffer, sizeof(buffer),
4232                                        "diff--gita/%.*sb/%.*s"
4233                                        "---a/%.*s"
4234                                        "+++b/%.*s",
4235                                        len1, p->one->path,
4236                                        len2, p->two->path,
4237                                        len1, p->one->path,
4238                                        len2, p->two->path);
4239                git_SHA1_Update(&ctx, buffer, len1);
4240
4241                if (diff_filespec_is_binary(p->one) ||
4242                    diff_filespec_is_binary(p->two)) {
4243                        git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4244                        git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4245                        continue;
4246                }
4247
4248                xpp.flags = 0;
4249                xecfg.ctxlen = 3;
4250                xecfg.flags = 0;
4251                xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4252                              &xpp, &xecfg);
4253        }
4254
4255        git_SHA1_Final(sha1, &ctx);
4256        return 0;
4257}
4258
4259int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4260{
4261        struct diff_queue_struct *q = &diff_queued_diff;
4262        int i;
4263        int result = diff_get_patch_id(options, sha1);
4264
4265        for (i = 0; i < q->nr; i++)
4266                diff_free_filepair(q->queue[i]);
4267
4268        free(q->queue);
4269        DIFF_QUEUE_CLEAR(q);
4270
4271        return result;
4272}
4273
4274static int is_summary_empty(const struct diff_queue_struct *q)
4275{
4276        int i;
4277
4278        for (i = 0; i < q->nr; i++) {
4279                const struct diff_filepair *p = q->queue[i];
4280
4281                switch (p->status) {
4282                case DIFF_STATUS_DELETED:
4283                case DIFF_STATUS_ADDED:
4284                case DIFF_STATUS_COPIED:
4285                case DIFF_STATUS_RENAMED:
4286                        return 0;
4287                default:
4288                        if (p->score)
4289                                return 0;
4290                        if (p->one->mode && p->two->mode &&
4291                            p->one->mode != p->two->mode)
4292                                return 0;
4293                        break;
4294                }
4295        }
4296        return 1;
4297}
4298
4299static const char rename_limit_warning[] =
4300"inexact rename detection was skipped due to too many files.";
4301
4302static const char degrade_cc_to_c_warning[] =
4303"only found copies from modified paths due to too many files.";
4304
4305static const char rename_limit_advice[] =
4306"you may want to set your %s variable to at least "
4307"%d and retry the command.";
4308
4309void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4310{
4311        if (degraded_cc)
4312                warning(degrade_cc_to_c_warning);
4313        else if (needed)
4314                warning(rename_limit_warning);
4315        else
4316                return;
4317        if (0 < needed && needed < 32767)
4318                warning(rename_limit_advice, varname, needed);
4319}
4320
4321void diff_flush(struct diff_options *options)
4322{
4323        struct diff_queue_struct *q = &diff_queued_diff;
4324        int i, output_format = options->output_format;
4325        int separator = 0;
4326        int dirstat_by_line = 0;
4327
4328        /*
4329         * Order: raw, stat, summary, patch
4330         * or:    name/name-status/checkdiff (other bits clear)
4331         */
4332        if (!q->nr)
4333                goto free_queue;
4334
4335        if (output_format & (DIFF_FORMAT_RAW |
4336                             DIFF_FORMAT_NAME |
4337                             DIFF_FORMAT_NAME_STATUS |
4338                             DIFF_FORMAT_CHECKDIFF)) {
4339                for (i = 0; i < q->nr; i++) {
4340                        struct diff_filepair *p = q->queue[i];
4341                        if (check_pair_status(p))
4342                                flush_one_pair(p, options);
4343                }
4344                separator++;
4345        }
4346
4347        if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4348                dirstat_by_line = 1;
4349
4350        if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4351            dirstat_by_line) {
4352                struct diffstat_t diffstat;
4353
4354                memset(&diffstat, 0, sizeof(struct diffstat_t));
4355                for (i = 0; i < q->nr; i++) {
4356                        struct diff_filepair *p = q->queue[i];
4357                        if (check_pair_status(p))
4358                                diff_flush_stat(p, options, &diffstat);
4359                }
4360                if (output_format & DIFF_FORMAT_NUMSTAT)
4361                        show_numstat(&diffstat, options);
4362                if (output_format & DIFF_FORMAT_DIFFSTAT)
4363                        show_stats(&diffstat, options);
4364                if (output_format & DIFF_FORMAT_SHORTSTAT)
4365                        show_shortstats(&diffstat, options);
4366                if (output_format & DIFF_FORMAT_DIRSTAT)
4367                        show_dirstat_by_line(&diffstat, options);
4368                free_diffstat_info(&diffstat);
4369                separator++;
4370        }
4371        if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4372                show_dirstat(options);
4373
4374        if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4375                for (i = 0; i < q->nr; i++) {
4376                        diff_summary(options, q->queue[i]);
4377                }
4378                separator++;
4379        }
4380
4381        if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4382            DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4383            DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4384                /*
4385                 * run diff_flush_patch for the exit status. setting
4386                 * options->file to /dev/null should be safe, becaue we
4387                 * aren't supposed to produce any output anyway.
4388                 */
4389                if (options->close_file)
4390                        fclose(options->file);
4391                options->file = fopen("/dev/null", "w");
4392                if (!options->file)
4393                        die_errno("Could not open /dev/null");
4394                options->close_file = 1;
4395                for (i = 0; i < q->nr; i++) {
4396                        struct diff_filepair *p = q->queue[i];
4397                        if (check_pair_status(p))
4398                                diff_flush_patch(p, options);
4399                        if (options->found_changes)
4400                                break;
4401                }
4402        }
4403
4404        if (output_format & DIFF_FORMAT_PATCH) {
4405                if (separator) {
4406                        putc(options->line_termination, options->file);
4407                        if (options->stat_sep) {
4408                                /* attach patch instead of inline */
4409                                fputs(options->stat_sep, options->file);
4410                        }
4411                }
4412
4413                for (i = 0; i < q->nr; i++) {
4414                        struct diff_filepair *p = q->queue[i];
4415                        if (check_pair_status(p))
4416                                diff_flush_patch(p, options);
4417                }
4418        }
4419
4420        if (output_format & DIFF_FORMAT_CALLBACK)
4421                options->format_callback(q, options, options->format_callback_data);
4422
4423        for (i = 0; i < q->nr; i++)
4424                diff_free_filepair(q->queue[i]);
4425free_queue:
4426        free(q->queue);
4427        DIFF_QUEUE_CLEAR(q);
4428        if (options->close_file)
4429                fclose(options->file);
4430
4431        /*
4432         * Report the content-level differences with HAS_CHANGES;
4433         * diff_addremove/diff_change does not set the bit when
4434         * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4435         */
4436        if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4437                if (options->found_changes)
4438                        DIFF_OPT_SET(options, HAS_CHANGES);
4439                else
4440                        DIFF_OPT_CLR(options, HAS_CHANGES);
4441        }
4442}
4443
4444static void diffcore_apply_filter(const char *filter)
4445{
4446        int i;
4447        struct diff_queue_struct *q = &diff_queued_diff;
4448        struct diff_queue_struct outq;
4449        DIFF_QUEUE_CLEAR(&outq);
4450
4451        if (!filter)
4452                return;
4453
4454        if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4455                int found;
4456                for (i = found = 0; !found && i < q->nr; i++) {
4457                        struct diff_filepair *p = q->queue[i];
4458                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4459                             ((p->score &&
4460                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4461                              (!p->score &&
4462                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4463                            ((p->status != DIFF_STATUS_MODIFIED) &&
4464                             strchr(filter, p->status)))
4465                                found++;
4466                }
4467                if (found)
4468                        return;
4469
4470                /* otherwise we will clear the whole queue
4471                 * by copying the empty outq at the end of this
4472                 * function, but first clear the current entries
4473                 * in the queue.
4474                 */
4475                for (i = 0; i < q->nr; i++)
4476                        diff_free_filepair(q->queue[i]);
4477        }
4478        else {
4479                /* Only the matching ones */
4480                for (i = 0; i < q->nr; i++) {
4481                        struct diff_filepair *p = q->queue[i];
4482
4483                        if (((p->status == DIFF_STATUS_MODIFIED) &&
4484                             ((p->score &&
4485                               strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4486                              (!p->score &&
4487                               strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4488                            ((p->status != DIFF_STATUS_MODIFIED) &&
4489                             strchr(filter, p->status)))
4490                                diff_q(&outq, p);
4491                        else
4492                                diff_free_filepair(p);
4493                }
4494        }
4495        free(q->queue);
4496        *q = outq;
4497}
4498
4499/* Check whether two filespecs with the same mode and size are identical */
4500static int diff_filespec_is_identical(struct diff_filespec *one,
4501                                      struct diff_filespec *two)
4502{
4503        if (S_ISGITLINK(one->mode))
4504                return 0;
4505        if (diff_populate_filespec(one, 0))
4506                return 0;
4507        if (diff_populate_filespec(two, 0))
4508                return 0;
4509        return !memcmp(one->data, two->data, one->size);
4510}
4511
4512static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4513{
4514        int i;
4515        struct diff_queue_struct *q = &diff_queued_diff;
4516        struct diff_queue_struct outq;
4517        DIFF_QUEUE_CLEAR(&outq);
4518
4519        for (i = 0; i < q->nr; i++) {
4520                struct diff_filepair *p = q->queue[i];
4521
4522                /*
4523                 * 1. Entries that come from stat info dirtiness
4524                 *    always have both sides (iow, not create/delete),
4525                 *    one side of the object name is unknown, with
4526                 *    the same mode and size.  Keep the ones that
4527                 *    do not match these criteria.  They have real
4528                 *    differences.
4529                 *
4530                 * 2. At this point, the file is known to be modified,
4531                 *    with the same mode and size, and the object
4532                 *    name of one side is unknown.  Need to inspect
4533                 *    the identical contents.
4534                 */
4535                if (!DIFF_FILE_VALID(p->one) || /* (1) */
4536                    !DIFF_FILE_VALID(p->two) ||
4537                    (p->one->sha1_valid && p->two->sha1_valid) ||
4538                    (p->one->mode != p->two->mode) ||
4539                    diff_populate_filespec(p->one, 1) ||
4540                    diff_populate_filespec(p->two, 1) ||
4541                    (p->one->size != p->two->size) ||
4542                    !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4543                        diff_q(&outq, p);
4544                else {
4545                        /*
4546                         * The caller can subtract 1 from skip_stat_unmatch
4547                         * to determine how many paths were dirty only
4548                         * due to stat info mismatch.
4549                         */
4550                        if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4551                                diffopt->skip_stat_unmatch++;
4552                        diff_free_filepair(p);
4553                }
4554        }
4555        free(q->queue);
4556        *q = outq;
4557}
4558
4559static int diffnamecmp(const void *a_, const void *b_)
4560{
4561        const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4562        const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4563        const char *name_a, *name_b;
4564
4565        name_a = a->one ? a->one->path : a->two->path;
4566        name_b = b->one ? b->one->path : b->two->path;
4567        return strcmp(name_a, name_b);
4568}
4569
4570void diffcore_fix_diff_index(struct diff_options *options)
4571{
4572        struct diff_queue_struct *q = &diff_queued_diff;
4573        qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4574}
4575
4576void diffcore_std(struct diff_options *options)
4577{
4578        if (options->skip_stat_unmatch)
4579                diffcore_skip_stat_unmatch(options);
4580        if (!options->found_follow) {
4581                /* See try_to_follow_renames() in tree-diff.c */
4582                if (options->break_opt != -1)
4583                        diffcore_break(options->break_opt);
4584                if (options->detect_rename)
4585                        diffcore_rename(options);
4586                if (options->break_opt != -1)
4587                        diffcore_merge_broken();
4588        }
4589        if (options->pickaxe)
4590                diffcore_pickaxe(options);
4591        if (options->orderfile)
4592                diffcore_order(options->orderfile);
4593        if (!options->found_follow)
4594                /* See try_to_follow_renames() in tree-diff.c */
4595                diff_resolve_rename_copy();
4596        diffcore_apply_filter(options->filter);
4597
4598        if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4599                DIFF_OPT_SET(options, HAS_CHANGES);
4600        else
4601                DIFF_OPT_CLR(options, HAS_CHANGES);
4602
4603        options->found_follow = 0;
4604}
4605
4606int diff_result_code(struct diff_options *opt, int status)
4607{
4608        int result = 0;
4609
4610        diff_warn_rename_limit("diff.renamelimit",
4611                               opt->needed_rename_limit,
4612                               opt->degraded_cc_to_c);
4613        if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4614            !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4615                return status;
4616        if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4617            DIFF_OPT_TST(opt, HAS_CHANGES))
4618                result |= 01;
4619        if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4620            DIFF_OPT_TST(opt, CHECK_FAILED))
4621                result |= 02;
4622        return result;
4623}
4624
4625int diff_can_quit_early(struct diff_options *opt)
4626{
4627        return (DIFF_OPT_TST(opt, QUICK) &&
4628                !opt->filter &&
4629                DIFF_OPT_TST(opt, HAS_CHANGES));
4630}
4631
4632/*
4633 * Shall changes to this submodule be ignored?
4634 *
4635 * Submodule changes can be configured to be ignored separately for each path,
4636 * but that configuration can be overridden from the command line.
4637 */
4638static int is_submodule_ignored(const char *path, struct diff_options *options)
4639{
4640        int ignored = 0;
4641        unsigned orig_flags = options->flags;
4642        if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4643                set_diffopt_flags_from_submodule_config(options, path);
4644        if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4645                ignored = 1;
4646        options->flags = orig_flags;
4647        return ignored;
4648}
4649
4650void diff_addremove(struct diff_options *options,
4651                    int addremove, unsigned mode,
4652                    const unsigned char *sha1,
4653                    const char *concatpath, unsigned dirty_submodule)
4654{
4655        struct diff_filespec *one, *two;
4656
4657        if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4658                return;
4659
4660        /* This may look odd, but it is a preparation for
4661         * feeding "there are unchanged files which should
4662         * not produce diffs, but when you are doing copy
4663         * detection you would need them, so here they are"
4664         * entries to the diff-core.  They will be prefixed
4665         * with something like '=' or '*' (I haven't decided
4666         * which but should not make any difference).
4667         * Feeding the same new and old to diff_change()
4668         * also has the same effect.
4669         * Before the final output happens, they are pruned after
4670         * merged into rename/copy pairs as appropriate.
4671         */
4672        if (DIFF_OPT_TST(options, REVERSE_DIFF))
4673                addremove = (addremove == '+' ? '-' :
4674                             addremove == '-' ? '+' : addremove);
4675
4676        if (options->prefix &&
4677            strncmp(concatpath, options->prefix, options->prefix_length))
4678                return;
4679
4680        one = alloc_filespec(concatpath);
4681        two = alloc_filespec(concatpath);
4682
4683        if (addremove != '+')
4684                fill_filespec(one, sha1, mode);
4685        if (addremove != '-') {
4686                fill_filespec(two, sha1, mode);
4687                two->dirty_submodule = dirty_submodule;
4688        }
4689
4690        diff_queue(&diff_queued_diff, one, two);
4691        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4692                DIFF_OPT_SET(options, HAS_CHANGES);
4693}
4694
4695void diff_change(struct diff_options *options,
4696                 unsigned old_mode, unsigned new_mode,
4697                 const unsigned char *old_sha1,
4698                 const unsigned char *new_sha1,
4699                 const char *concatpath,
4700                 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4701{
4702        struct diff_filespec *one, *two;
4703
4704        if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4705            is_submodule_ignored(concatpath, options))
4706                return;
4707
4708        if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4709                unsigned tmp;
4710                const unsigned char *tmp_c;
4711                tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4712                tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4713                tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4714                        new_dirty_submodule = tmp;
4715        }
4716
4717        if (options->prefix &&
4718            strncmp(concatpath, options->prefix, options->prefix_length))
4719                return;
4720
4721        one = alloc_filespec(concatpath);
4722        two = alloc_filespec(concatpath);
4723        fill_filespec(one, old_sha1, old_mode);
4724        fill_filespec(two, new_sha1, new_mode);
4725        one->dirty_submodule = old_dirty_submodule;
4726        two->dirty_submodule = new_dirty_submodule;
4727
4728        diff_queue(&diff_queued_diff, one, two);
4729        if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4730                DIFF_OPT_SET(options, HAS_CHANGES);
4731}
4732
4733struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4734{
4735        struct diff_filepair *pair;
4736        struct diff_filespec *one, *two;
4737
4738        if (options->prefix &&
4739            strncmp(path, options->prefix, options->prefix_length))
4740                return NULL;
4741
4742        one = alloc_filespec(path);
4743        two = alloc_filespec(path);
4744        pair = diff_queue(&diff_queued_diff, one, two);
4745        pair->is_unmerged = 1;
4746        return pair;
4747}
4748
4749static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4750                size_t *outsize)
4751{
4752        struct diff_tempfile *temp;
4753        const char *argv[3];
4754        const char **arg = argv;
4755        struct child_process child;
4756        struct strbuf buf = STRBUF_INIT;
4757        int err = 0;
4758
4759        temp = prepare_temp_file(spec->path, spec);
4760        *arg++ = pgm;
4761        *arg++ = temp->name;
4762        *arg = NULL;
4763
4764        memset(&child, 0, sizeof(child));
4765        child.use_shell = 1;
4766        child.argv = argv;
4767        child.out = -1;
4768        if (start_command(&child)) {
4769                remove_tempfile();
4770                return NULL;
4771        }
4772
4773        if (strbuf_read(&buf, child.out, 0) < 0)
4774                err = error("error reading from textconv command '%s'", pgm);
4775        close(child.out);
4776
4777        if (finish_command(&child) || err) {
4778                strbuf_release(&buf);
4779                remove_tempfile();
4780                return NULL;
4781        }
4782        remove_tempfile();
4783
4784        return strbuf_detach(&buf, outsize);
4785}
4786
4787size_t fill_textconv(struct userdiff_driver *driver,
4788                     struct diff_filespec *df,
4789                     char **outbuf)
4790{
4791        size_t size;
4792
4793        if (!driver || !driver->textconv) {
4794                if (!DIFF_FILE_VALID(df)) {
4795                        *outbuf = "";
4796                        return 0;
4797                }
4798                if (diff_populate_filespec(df, 0))
4799                        die("unable to read files to diff");
4800                *outbuf = df->data;
4801                return df->size;
4802        }
4803
4804        if (driver->textconv_cache && df->sha1_valid) {
4805                *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4806                                          &size);
4807                if (*outbuf)
4808                        return size;
4809        }
4810
4811        *outbuf = run_textconv(driver->textconv, df, &size);
4812        if (!*outbuf)
4813                die("unable to read files to diff");
4814
4815        if (driver->textconv_cache && df->sha1_valid) {
4816                /* ignore errors, as we might be in a readonly repository */
4817                notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4818                                size);
4819                /*
4820                 * we could save up changes and flush them all at the end,
4821                 * but we would need an extra call after all diffing is done.
4822                 * Since generating a cache entry is the slow path anyway,
4823                 * this extra overhead probably isn't a big deal.
4824                 */
4825                notes_cache_write(driver->textconv_cache);
4826        }
4827
4828        return size;
4829}